.. V... . win... .. c . ... ... ....1. ._ ...w.. .u .....v. . . , 4: ._ . .v . . . . . « 5.? C23. .... #91.» .-. _ .2... V . a .0 .0. 0 .. ‘3 . ‘th96 2.. . “0 ... w. .02 .V ~31 ... “....uxfikluxflt ...-0‘ r 4.951%. . . .... . $2.21.. . 4 .. .... .. Erika: .1040 ... .. .... LI..." 0...”. J . V L . . um I . M... ._ . . .. (,4... '1'. .31.»... 1.....1. . 1 .41 _ 1 1.2.... .. ....-. .. 36.3.... .870...» 3.0.110". . . I! t . .... t. .835. i . . o. . .m : . 00... . “I 2 . J . , . I 1.1!. t, 32:. . . n 0 u ” 4.4.0”. 0 00 ’0... . . . .... 3 . o. . V .. .1 . ...I...:......\ 2:! . . I. .1 ... . 2.3.2... .r ....L . . . .....v...Iv:.a..I... .... .... .30 0 . . ...-...... 8...... ...: .0. . .01 .0... .59.! 3...... .9... r .. V l . $.11 It... 4 , I L J. . 391...... r . 1 11h]. ... . . 21.x... 0.... 0.. .. . . ._ .... .-.... a... . ... .... .4.... .18....0 . .‘3 ... .. . ... 0.... ... 3.. ..4...... .....x!....I: ...: .0... 0.. .0... a... . .0L3‘11t ... v.3... II. O I o. .30... I; 4 4. ~. . Lo. , . ...). I .0... fiftd. Jouaiamr." .0 .0 ... m3... .T..r0. 0. 28..V..... o ......I 3.... .u 0951....6‘0! . 4.. ....v. ...! . . 0.. 3.3:. a . o .0 .0. 70.1.00! .30. C .8 ..I31 ..I ...-0.0 5* 9.be ..I. 0 v0: fi\fhv£ , j «I a. .nuéxnwfl . I. .3 l .J . .-...1 .).. : a. I .93.... ....v .....0........... ...... 1..., .... .. .... ...... 4 4.3.... - .. ...; .o ... .Ir:a:..n .0. .....Q J! 2.8... C. .1. .. .. 0:9»? 5...... lotlsai... I ...-0 ...-g VF. , , it: o v ...... o . r 90. . 3: ...... .. ’ . r 0.. .I .... . I. ....0‘. n4 . . .. 3... 2.3.... 1... 0‘... I ......3 .0: ... ... . P:.«.fi.~.. flfln. .... ....Vt..0.q~.0t$....3. O b vin\ ~60. - D I . 33....‘740: . . I .u .Iv. .. c? ....v .. «0. , . 3800.. .. .....035‘..I. -. “003.10.- 0.. 3.09.. 30.1300». ...-01.4.638‘0“ . 3,... .31 La! .0! j ‘Ulll! | . I a . . I a.” ...r . Z .. .. r13. . . ... . .a .... :0 .. ... ..- :r..: 00... ......I...‘........J. A .65.... .0. ... 85... .... :00. .... .1. 0... z 0. .‘ti...$.~ . .0 .l... LA. . ...—0‘. .0...‘» .o"? 0 0.3!...Jnfldui ...50. La.- . . ..., ...! . 8?... ... ...III. 0.0.3.8 Ji..0¢o0¢ 0-00 .0 :3: I. L 0 .dU‘ODSS 0. d an. O 0 .... L . . fit— , . 1 fl... .... ...\. . . . ... e ......I . L... .. 3. .(t .2; ...00 .... Mu ...-... . .0. .0 I.. . .. . ...-...... .kIO" .’ I. .Nh... 0.0.: .23.. 07: . 3 ' .~X‘.Lu w. L. L. ...-.a.:. .... 1.2.1.0.. ...-.0. .058! 00...... ... a I. “8:79 .... v a . 00.. ....3... .8. .. .. .....0. 03.3.0 .1 9‘ I08 0.. 0..¢.I You...” ... l... L k... . ...... 6:. ... .. ... ... ..n J ...:..U‘.. . .51-... . .... 3...... .33.... . ... $13-40... .. ...“... .0. .....50 . I4... 3.. I5 00.! .w J.I.06 J. .. up 0 .0 .3. I. .I .1337..vl.4.. .u 0.. ...0 . I.PL.. .l.!... 1.... . . ...... .04 . . If!" «....0. a o (.01.. .. 0... £0.10: ‘0 (. 05:83:31....4. .. .. . ...3‘ not 0 003.00 .é ..I I... 40‘ 00,0 .0 .l . 9... k0! . . (c....fl.»lv ..., . 3L3 . .31.. _ . . : .. . 4.. . . S01]. . . . ..l . .o. 1... . .... . . I32... 3...... .001 its .. I .... .....- ...: ...... 2.6..) I... . £3......:I.0 n.... .. . .....I0 9 .0 z .0. i I A V), .. J .V . . . «$60! 0.... ...pru...lo.?5.£ .. .0 ....I .205... .... ...?33. ... ............. v.1... 23...... f . 2.1.0.0 .3121... .00 8.03.0.3... t .6 fladaall’. 0 ... .303 U‘Ih... 40.0- ”0 8". I 0 Q ‘0. 1 .6.~r 63.0 ...”. . :2. 'K 9. . .a 01.... . . . P. I . .....u... ..J u ..5 .0. I... 6.. .1 I 2:. I. .. ......10‘...000. f... «...... Id. ...... J..l...I . 0 . ...-..- ...40 c...“ ‘L . 5...; 1.0.10... .00. r. _ . In... .8. I. .023 féJaJ 33:005.... ....J. .01.. .1. .....a ....r. .....00. .... 2.1.3.1....rt... ....f 0 .... f... t. a. ...0 .. . . 1.0153“... 638...... . . . 83.1.. 01.08.043.40 1.00932‘ ‘z to: . I .0 .. .. .( ouOi 0! . ... ...l . . ..rr...0..o.. 314.0... .. ..‘PJ...30.....30’ 9!..3 33911.8 .. ...0\. .300 .0. .88 8:80.... .1: . .00 -. .....0 00-- 0...... .... . . .5003 0.0L 0L... 1.00. .. ... 3.0. f 2“}... I I .u. . u .l: ......IQJ .. I. .....c‘! ... ..."..— 0. .. . ... t 0..3....0..00.G..6|7-!> . ......l. . . . ..0 In o. 0.... . ...Z... . I .. .JQIS ...: ‘03:: .. i 30.04 . .F . I. . . 0.. v . 0’... 1‘0. .0 ' a. 0 00'... .0. J...I..‘.0I..05.... .3 3‘. ‘.3V. .. . ..4Qllnof o . .0... .9... 0..-0.v .l. ‘43,..." 00.0‘ . 0.... 03...... 0...... 0 3......0. -. ‘.4..3!0¢.0 . .0003... .' (~33... .. 3.3.3. .r. p............. Z . I. ... . ... . .. ...: v0.22. ...}...fw. 1'...:8..r.05 . .3: ... .0. .... I... 1.9.35. . ...: I: o! . I. .... I .. 3.1.. J ... ..- .. . V V . . ,I : 3.032339. .0. .... ....w u... .-...2... ...: . ...0 0 ...-9 ...fi... 0.0. 03:93.... I .. 3.... 30,321.09..- 2...»...0l .0 . ....O'Ilqt.‘ 450...... . . .00 o .. 1.4.3.05”... ... . l.- ..I: d , .I I3 .0..f..00v.8'« 40—35303... .3.... 4.. .... .. . ....00 . .. sE......“0-.... 0.... ... .. ...0'3... .... 3.)..600. ; ...... I 1.0%.... 003...! 0... ... I. ‘ D . I . 1130......- JFJ4 ....l I ..l. 0 u . ... . 1'.v,.p. 000...! 0 .13v .. . .I u. . ~. . . i . .".l!.o.:u 0 ...! ‘.... .0 . 0 ...! ...: r .3 0..|...0-.0I..00.—’..0fl.1.0;. ...... 5.51237...- ...... 0. :05... .00 . 5.0-0 . 5.0 o 0.0;: . - 7.01.0003. 731...... . ...: .. (I... .... . 90. :J... . . I . ...-.....I .1111: . 3... .00. ....r .I.......¢..4..I0v. . ... . ..88..t..! a .. 3.0.0.7.! .00? .l. .... .v f... .4 l. ....0....0.....‘: 0070.. It 00.6.2 2...... ... . A... 3.1.... ...... . . r: ...: .19... ...r.......:. ......9) r... ...... ......I.i..4... .... at :I 3:133:33 . . ......» .At -. .............7.. 12183....) . :10 51...... 4.1.5.1.:3133.7: .... I x ...O... .. .... . I... .133... .. .. .ov . ... . .. .....1 .. ... . I... .. ...; ...... . 3.0.2.0 3.... ... .... .513... . £31... .1 . ....IJI . 3.30.103 .6 ...? . 301......5331011JIL .21.!510: ... 0.1 .02030’... . .r. I. 1! uf......2 .I......'... .p 4.. .... ...-1.12.... ... v .... :0 0&0... ...... 4 0 . . .t . ;‘ ...: . 3 . . . 0 I .16....330440 . ... . ... . 0 05' 3.03. 0 . . .3. .4.... ......V. .. ‘23:»... . I :3. ...-(0...... I... . 0 0.7.1.!!! .3. I. £13.21 IG‘.V.I.0. 9.8330 3.0 .II. .0...l.. ..., ...... .3 .300... . .....EvIJQQIC ‘4‘ f . I .1 0 0 :30 .c ..!?.:.X.! .00 . r‘l;7 ....c ..X. i T0.‘0.-..:031... ~.nu0..o0...020.0 0. ...-.00.:‘03 . . 0., ... ... I... on 0. 0 .0... .6. . Ir. .0 .0... .l .... h. .‘c.... ”0.! .33.. 0... o"...3.-..33:4 ..00.J.oo0.o 0 ...-c0.- 05...0I.00..£...050 0 ..H . 00.....v 0'0 .- . 00.070.de 3 0 o. . 0O 0 . 0:00. I l..—0w 0 . § 00.0.. ...-0...... .....3158 00......‘L000200: . . u ..0 0‘ ...:3’} .00. ...O. 0": .0, .A 0 r ....u.......... .u .....70. _. ... ...-t .....0 ,.I0....0I.. 3...... ....I...Ot..... 3820!&801..83..J..$io..:.0 .... L 2.00] 11033 ... ... 32.0.. . 00.81030... ...-03 ......th 0.01 . o .0. ...: ‘ . . o .... : . .. . . 1 324.0. .. . . ....w .. .3..b..00.l..:... 3.... .C .. . ... 0.. . . . _ 30...-.. . . $13.30.. ...08 ...k-b...-O 3...... .43...0..9.......1.l.l '0, 4 m... O..!...~I.I...I.c0.0: . . . . ..0 1'5 . .0. . . i .. ..0. . 0 . . _¢..I.J .D 4 3. . ......3C .... 30:91.01 0... .- .0..~Ii-0.0‘ ... ...Ol’iltl .0600 .I‘nwt....0.d .005'"I. .. on) 1.0.1.000 tn... . .‘ ‘30. ..U‘ 0 .0 30 ...... I. ....t. l .0. .0 I. .0 . . 0. J. . .0... o .w...,. 1. 4’ V . r 0.. w ..0.. . 6.... .3... .. . :7 .0... . .. ...«n . .... 300:. c. . 1... . . .. 0 .... .. ... . . .0. . ... . .....ff... . 30... .0 ......-.- ...!llut0..l .L..Il.... .1... .0...I : . r .3 I, 4'00- . ...‘I. .9: 331.09., I I1... I... 0.00.0. ...... . 1.7.03.0...02 \ .0..I. ‘50-‘..0.90...I.:I.0£...-000.50... . $8000.00... ,..0I0......\. .16006I o o .,. .f . .L “...-I... .0..9..J..1..? . o. L . . ‘01... . 00.1 .. . I. . 0...... 92.0.2: .... . . II 0.0.: .123 . ...?ppli00a...lt.0!.0 .. 09.10 ... . . .. 9 .. .....8‘3L‘...7.:0 .03 30.30! . no.1 p. .I ... .0... ..vgia.. 0...'0.0.-,......0. .y ..5..:.. .-.-.3023). . . r .. . , 0... 2.90.830 ...III...‘ 0. . . . ....‘0 ...... ...0 v. I. 310‘ ...! Vt . .....a . I. 0...: .140.- .l .?.Jir‘,...ullil . [0’33 ...... ... 301...... 2... luv.» . .... 4.5.4.5..“ U..r...m......ghm.amv.mm.rhfl.hsgw . III .... . ....f.r~.3..tot .. ...}...1390. 05 O.\.0"I‘|.O. . $0.2. .. I 13.“..90’32’4 S .- 12.. ... 3.71.3. 9.9.9.. In...t..¢:..t..1 045.0459»... .Iflihfl. .53....011JIKL (1.1.... .0 Iain... (2.0!: ...... 0.0 0.0 .3 {352‘ di‘!!..§ 30. 00...... .4 ..3 .0 o... . $4. _ (... . ...... 0 ....10..I..: 3...... ...l .33... 0 .31... 492......0 [0.3.0312r I ... A ..4; L... . 0‘63 ...-... ... J..Ild....!b.0.ol.0)i.0u .... . Lb?» 10.30....” a3..: '38.. .0 . 9.8.3.... . ... . 19.. o0..;..<..$ . Z . Z... 1 .,.Al.... 0...! ”0.00... .50..I . D .... .12....I6Y531.39.0315. II. V} ha . ... .. l 3103?..092325 ......3. 0. ....0 .. VI... 0 ..‘I‘. ..500 4 3 10540.0...0 0.3.. 0. ., .I. ..h . . ... ...-.00....01... ... .... I.) . ... .2! ..55; I....!. 22......- V V 0%..I4nfn ..3, 0.1.0.5. ....0 I.» O.“ . I ...... f...) .4 .o . ,. v 03......2 . I. . ...0. .30 . ... I0. ..0 . 30...!!! 668:0.233 0.. ......I: I. .00....2... 3IS1:‘II.. r0..VQ’ I30.) .0 '10. ......Q C.»..,.0...... .I..- 3?... .0; . ... .. 20:00.3'. .0. ..... F01. . . 0...... 20'33‘0 . .63.}... I213... ... . . 0....18’2..I‘7..1, I .0 u :3 0.3.1.... I . ..2 V? ... o. 0 e ..I .40... 0. ...00 ...-.8.- .. ... #3:... A23 .0 .....au.. ...-6‘ .... 0 ...-..pfill. .Iy 3- .l q .... . ... .....4. .41. 0!! ....L. ...... 0...: . . .l\ 01.0-00.30: V . ..l0.13 .6 0.. .00. 0|... 000.0..00I0..: ‘03:... ....0... 0. a 0L .90....InQQI gob]. 0-1... 3.- 3.. . .0 0 ... ...; . .53... . . . .. .....904 . . 00. . ..0 ... .00. .1... .... :1... 0:00... $00.22!? 0-0....)..~.l..-. .....00 w... a: H.000. .. l a” D .. I 1.3..-. . .00.. 00. w: 0.0..L2188l0‘........192.. 4 . . . ll........8 4 .0 ... ...-v. ..t..~.4...4323.0)£..l. O)...0I0. - .‘251 82...... A .‘J...u.0.0.l~l.d522 3.01 «0.. 30......2: :0. 0‘: II: 0 . .OII..«... ‘6. u but... . . . ... 0.. . 01%....318......0.8 .3. n. .5! 00!... 10-0 3.1.... 200.0 . . I .101'33 I. ...v ...-In I}. ... ’iv 70’ oi. - . . .... .. .‘ 5’0. 0. .0... . . . .V . ......o. ..P ...... . r. . 0 .. i . yo... I. 69.. .0128; .-.-3:05 ...: .. ...I .... . 8..-...923I ..O . .o 1 0... .9: ..Ib...I.. 0.00.. 0-. ...!il .... .... 3‘I...’ ..., .. 003.. .0.00...$.I¢ll.00 0.3;. . 9‘. A . 3'34. :3 . .00 I» .. ’f: Invnuo‘iiltaiI v .u .0001 .0 . :01... .20....3.203.033.03393:0031I 1:01...“ 0.0.0.0.... 3.,— Il . £0113... 3... I33... 25...? to! . .0. .... . .10 J. .... . ...Ii... 10.0.3.- . . r319. {00' . ... 11....- .I I. ...‘... 00" 0:00}. I‘ . . V . . . 0.0%}..11 ...).Iooi ; 6. 4.010: . . .o 3‘ ..0. .V . It. ...}..l....\l..,.. [:1 0 .... .... , 0‘00. .4 ... . . . . r. 3 000.0 . ..0.. . , . . ... ..rt .00....‘0.’ 1“... .. ..00'.. .0... 00 . 1’ . I ... ..Iv .- _ . 00.000 ...: 0 ll. 0.- .... . v. 0.01 . or 10.4.! ...-.... . .0 0 1.....10.’ 4 _ . , ...!JWSO1VV .vt’ .0... (A. II . . ‘. - .- 0. 0.. 7...! . ...) ... . .IID: _ . . . i.ao.i..blkql+ it... 3 ‘1 vb ...... 4 .. . . . ...-x... .. . ...-...} .5 :11 000 0.3. . ... u _ . . . . 5100.309} ...00.......‘ ‘3 : .. . n A . ... . o _ ..............: . . .. .. _ , .V . .OIIIoI..§..£ 3......91 .. .. 0 . ...... .0 ....t 2.000... 09.0.3.3: . . . . . . . . . . . .I.p....... . ..l “...-.3. ..0... .....I 1.01.0.0....0.. . . . . . . . . .. _ u . 0.. i=0. ‘i 0. ......» . ... .. . . . V . . . .0.\ .- 0 . ...- . 0‘ 0 ... 00‘ .00 . , u . . n . .1! III-..» :0. ..Y‘.a.b 0‘ . . ‘ I. .- gulzoo" ‘4 .1. ‘H‘ 0”“... ivfl'tfkg.‘ 0””!3 1M- 0. .3. 0.:- D I. ..0. .000 A . I. . 3 0h...) 3. ..Ioabzhflcfi.l.izhr:.fit . “If .. H0... 1...... .9..€II38..I. 3.. I I .0 {010.00. 0.33:31I100010‘. pith... . I00 . ... .0 ..d- 12.2...) 3..) 5.00410030530150‘00011 0 LIQ..IIVO ... 0.. . ...... o .0... ... . 01.4.20! ... 0.0.0.0...0: 1...... I.si&..’.allo...0.zvf . 0.0.‘.{0\o,.0000(0-‘I..0\. .00- .4.. 0110 . ..00..‘Q—.d' 0.0 0.01.2.0... 10 1’.0.a..DI’.90.‘ 0 ..PO..000.-4 I . I dealifil. 3:190:00. fio.‘ .....I «.....Iivi!lt..c 3.33.00. 11‘. . gag-...“.If‘bqatéz . 1.1.0 . ... . . 1.0 . .1003 :0... "Mittfilzfil‘ .0. . 100...... . ...... .01. 1.13.771... .I. 0.....0 .63 ...0...IO...Q..000¢-.00.3. .00 .O.lf...-. 0.. 0...): . .. . V. . ..000I0510. I I . , .1 . .. .II 4 4 .I. .0. : I}... .. .... 0.’\.40.(U0.\o.t_..‘0v . . O v... .. .. 4kg) ...) I 5... 6.00.13.03.22: , 'l0...‘ 0!. L 0-0... . . ‘Iu..o’.. I91 .00. .005. a u]... 00:"... . .. ... .n...0.....fl.1 :- \U .00. .20... 3| . (I01. 0.0““ .30... ....l . c . . . 0. 4.. I. ....00I.00...| IL:.3..320~...NI0.I.0.20.I!I v .1 i3!!! )3! I! o . .31. . .... . ._ ....0ilfi 0.0 ...._....J.l..u..\s....‘jo.’$!. o 20‘ . ..‘..‘1:Ith m0 ..o. A . v I .. . .0“.0I0...0I.0.0 v 3.2.4... 00.93.5033,..IU‘ so ..2."l3..313 .3?.80.0.SI:$“V ling”. $3 ‘I o . o. .. . ..V ..P o . o lo. .. ... .. .. .. ..I I. . . .f . . . . . . .. .. ..grafinra3o...t\.l «...-0.0 .31... Liit!‘.if:‘..030.0l0 0.10.2..‘36 o. ...... :3. 5.0..IPOUJ ... of 03.003... :0 3.7:... .8 O :0) .. ...! c .8: 0.. I: $2-2“. \ ...!!! .1 6.... 051.10.033.30... it’b.‘ -l& J . .l . . .. . ...: . ..0v _ 0 . v. .0. .. . 10.... ... . I . .' . ... oi‘v....1!0...1ws. 8.1.5.533 .....301. . ii... 509:4 .0 tQIg .31 It .. .. Oil-.... .... ‘.J..9I.....90I0.1§Ilo .I..lv..-..\.‘ . . 0. : . . .. .I . 150.200 .30.'.Ia.9l‘c o .... I; I0 .- . ...2 .. 2 0.0..I. 0\.v.0..v0I... ‘4‘ .0 ...0.. .VI.‘ . .a . 0 0‘0. 3.....- 4.. ..0 0 . .... I... ... .. . .0. 8d .0350 I333... . ‘93....3f8o1. tin.tz\ua5.kufl. I..." ...... 03‘ c0: . o unvastunu ’3I234".2.3:x?. I. l. 0..”0». 7 .9 I9: In) 0: q. ,0. .v .. ...: o. .0... z .0 ‘00; 0 _ . 0.. 5‘ ... 3‘; .00.». ...0: ...-“3.. 00:00.30 \Jv . {'IO.IV...U~..L’.CI ’13 v. 0" 0Q...-00l.p.£. (0.)..- .0. 0.010 0 . 0. i .i‘2530.‘ ... I... .I .1 0. I . ”..I .v Lgrvt...’ sfiusfiql“9t .II 05...? I ,0. I. 0. 1‘s...:z’0. I 33!. . V I. ...... I. ...:l!....... .....0 332’: .2...on .3”. 0... Soon-£30.35. I. ‘34. .00.. fl , nut. .‘0: 4|...o..0.t..D.-......I. .I. 000"...£~... t!0.§&.r 030....310. I“Io.. .l‘V1053. 0000.}.‘1-0' 0:31 . 4. . 4 0 0...... .....I ....PO....:,1.0O oltvoni. ..!.Cl$0....‘.llnl..49'il it...) Icky-T053. .30! ..0 J. z . . . . . . . . . . . V . . . 0.) 00308.0... 00...»...1’14 «45...... I. .....0... ’00....‘0. 1.01.520. 3.9. €0.10 . .g1ouhnu‘ S . .. . . . . . 4 . A 4 .. . . t. .9: 0.. .l 1...... L..I¢....')..).‘.......33I0ulo IGS‘II.‘.~T&I£$G ......OIIJEI’, ‘0‘33138I :1 l . v . . . . . . . . : . . 3.4:...003 4%.. 3......‘003r...’ 3.....30. ...-0.... v.¢!.v!‘v......¢ lS"?1I‘Ioi9‘l-‘IIII‘|OU‘ .0. g I. 1...... r . . . .. _ . . . I... .03. L. 8 I r..... . . .- . '21-‘90. 3.3:.:II)~.OOCQIIR010‘.I. IRW¥QI§I .. . . I . . Ina-0.00....0.!.’3f“’l‘.O.IJI.I-zo -009... 0‘00“. Iii.“ .. . . U....o....3.. ..’.._i8’|tl"tl{390.0utn:l.!‘l {1‘ gigl‘l 0* ..0 . n . . . . V . . ’08!.£Ioo. 0.1 . 4.0.... I 0.. 30.....8... ‘ .0 . 0H. ... . . . . . . . .. . . . In. ..0‘Io.’$.0.v0.20 It’ll. . .00. . . . .. . . .. . . . . . . .. .. 0 . . . . . . . . ._ . . . . . .1 .07 . ..... I... . . I. . ... .. . . . . . . . n .. .... .0... V .....- ... .I.. 4 . . . . . . v0. .. .4 . .. 0 . .. .. . .. .. . 0 . . . ._ - 3". a ' {to 1“... III-0': 0 O is r": ,l K II! 0M!b1la.’000u1l00:0 II00.I"0"0)~. ‘. .‘v‘o000‘l Hvtn .. 'I D 1.13 II... If .I . {Ir-3.13.6.0}! ..t’tofiel .31?! 91...! 3...}. . :2‘8.‘:Lf’§. ’3.I3QII ...-001....“ 1".."005 ‘IS . I. fin-0...}. ... o. . ...)....LQLQ¢¢..C.0!. $375.09....3. it-.. 0.00 .. gulcéfil: . b I ...(00: ... $W’I‘itl‘g . ... .... 0 ‘. .... . . J .07“. 0-4-0 O‘.‘.o0.0.¢0l0r01¢.0. . ‘."H .Il‘n ‘O’Dtr0.“sln‘0 ..I‘Qout...- .Io-II . 7 ...-£4...‘c...u.0.0.x.’001.§lor'30.0 .Ic' . V‘u’x‘),0oll“"’l ‘0‘. . ......38 051’..;¢...0..0I..0v\. . .13799 . . . ‘ 5...! {I000 I... Q‘. . ...-... .. . . . . tit-{.1001£33531?!OI... I... . . It!!’.;.:gl!‘ be .. . . 0.... . . ..00! L . .. . . . . . 000:.0I‘I'II..00Q00I:.0 I‘..O‘I . if‘i . 3...! . V 0 . I 0 ..0.0 0 .10. .n. v. . . ... .0. O. ., -.IPII.O..0 Igufgo ..0 . . . . . o. 0.. .3, 900...... ... . ...l) .R ‘ . . . .. . 30.. . 4. . , ... . . . . ... . ...... 2.9....0... 0. 2033:... ..SI‘...‘ . . . .0. . 00V . . . o . . ..cl. .... u.) ...: 7.32.... ..0 . 4 .II'... 0.0 ..X‘L .... .20} 12’ tact-$00!.Ufllttlfai.’ .x‘ffi0't”. . . . I..- .0 . . ... . O , . , .. .A . 4 . 4 . .4.. ... . . . . . . I .4. , 3...! 0. ....IA . . . .I ... 0.5.3.0.}? 0. .Ioo..8o(o..$I.0.r..I0’0...’00..II.0.II) 5.3.00.0 ’95,“.1.).0»I.‘o.”c.¢.o’o' . . .4 .. 4.. .. r : .00 0 .. ... _ ... .. a .. 0 . ... . .. {3533...I‘tz ... . . u (.90.?! «0.930.... 3 a €¢I...I.IZI..Q.I0.-..II‘.J. -01... 9-9. C’.1-.~.‘0‘00I.IO,)O. .Ih33t030I-FI. . ... . . 0 .. . , . . . .... _ . . ... .-. . . 1...! .. ... ..oQ .. . .. - .... 47.4 .... .3! I... . ..II. 1.1Illl . 1K0...S$ll‘.’.ll'2“003.\. . 'a v I n .. .. . .. . . . ... ...!I. . 0 . ... .....0 I o . . . .0 0 . .0 . .0...- n. . .'0.Av’,..0.o\0l.\‘.32..‘n 0.01.0} ’0‘... ’o.0{l‘.El-0‘l..-..l 3h“ . . . . . . . . . . . .v. .o I .... ....1: .0. .1 0 . .cI... . 0 J. 990.01.32.953oqiso‘59 .00 if!» .‘l’tou.h93l‘!fl:lo."ib’0§. 0.. ..I ... . .. I ... .. ...... ..t... :.. 0...... ..... IlaI: 2500.053... . . .. J u . ......I. 0 ..v: r .v. .0 V..O. 0.0 .. 0 .9... .... .2. .. . 1.. ... 394.021.30.100...— I..I... ...... .... .v . .4.. b I I . 0 0 V.... . . . .71.I. . ... . . , . I...... 1.0 . . . I $ .0 . . t b. . 2.3.. r . I}: A... .006. ... ......Oco......§0.! . o. 0. . . .I .. .00 .D .0» ‘ - ... o. . . . v; I v ... .00, 5030.}? I... 0.....qhnv‘: Io...o§tu. . . . ... . p . . . I. . . v.3. . . . o . .v . o a :0! ll . 11.»? .00....70..I00-.!\I.flrlo. . . . I. 0| .. . . 0. I. . .. 3. . . 0 .. . 11:51:31,21...’ .4. 1.0.0:... 0 .0. .0. .....3: . ...... .1 ....I ...0 o. . . .. . I I , .o ..c ...0-..0.....:. . V .l 0. ... .. 0.... 0 - ..oA-.v.o 0.... II ..o. ..'.|..-... .. ...‘...- v!.Ioio.. ...... I... . o .. . . 0.. .c c. . . ... . .-..- . In’flc..-22..l...’0‘$ov 00...: ... .Iu ..0 ‘5 ‘ 900$...v’0‘..l: . In. 0......» .00... ...0 .. . ... . . .. .. I . I . .. . 0. ..‘o .0}... . 0.. . 1.70.010. I... .0 .Y.’. .. 80:0 .v...‘:.tt. I: 23......00. .I l..'0.a-O. IO... . 70 ..Q. . o ..000 I... ... 5.. 06." '05,...‘r‘. .0. .....l. I . .Il .. . .-. .. ... 00 ......0 0D .09. 0.5. _ 0- I03£I0t..l.. ...}... 9 .00. ...! I: ..l . .. . 0.00 .0. g.. .05).: 0...... ...‘0. .....‘v . ’0 0.0 It, ... o 03 . . . . . 7‘ ... .0. ..If 5.. . .0317... .63?! . 7.30.. . . ... .o . A... . ..-. 0 I... ,0 u. a. , . . .4 . . . . .. .0 . 0:... .. . . 3...... . .... a lo . v . ..‘I.I1ro0...... ..... ..t I .. . ‘00,... .0... .0. ‘0 . .. . J. . . on. 007.0. .. 1.: 100.0! O.Io.C4 ..90..r. (70¢ ... ..... .. .0 SL0. . - .... ... ..00‘...... In... . .. .. 0. .0 i ... . ... .0....a.0.....0. 0 . _. I ...-n . . . . I . . . 0 0...... i... . .. v .....0.I . 0:1. V: . . 5-. .- ..». . . . . I- I... . .. . ..I; ..o. :0le 0. A -..... 1 '0. .. ..0: t.l..‘.-... lt‘...:. «7..., . I... ... . . .. u. .l...0. . ...I. a. . . . r . .. l... ... _ .. I... .. ..0 .. ton? . u '0. ..II...0I‘ . V0. . A. .. u . v .10. . I o. .. . .I ‘ wt 01.. _ . » .... . .0. l0 . .. .. Z .‘.O.... III ...... .0..\ . . . ... 0.. . .-. . .4. . u . . _ . , .. .. I v . .03... I 0 . ... . . .. . . 0 . . ... . . ...... . o o ,0 ...... .v _ . . . . .. . .. . . . .. . 4a 1....0 V ‘0I. .c.v....0.ol.o‘.. . . ... 2 30...... 0 0'0. .... . ’30. 0. 0‘... .!'O ..’I . . | 0.:‘000 4: . .I I . .. . . . ..0... ~. . I 0 .0 ...: . . .I I... 0- ...0 . 65.11. ...-0. o . . I . . .10 ... . ._ . .0 3. , n1 . . ... . . ...o. . . .. .t... . . . . . _ . . J0 I...) l .. 3 . .. .. .. .V . 0.0.. .. . V. I; 0.1.v. . . .. . ..0 0 ...—.... I... .I .0. .. . ... , . . o . 9.6 . .. . ... .. . .0 ,3 . a ...a II. V .0 ...... . . 0 A o I.. w ‘1 . '0. . . .0. . I. .. 20.. a ....l’..00.'i0,u $.I U... . .5001. ..I-OO-bo Ion. . u. . , I... p . .. 0.. :I. . _. ,. .Ir. '1 . .0 0,. .... 0 .MI ID‘I.I..0 0......tv'0.‘ (.I.v .. I ) .‘..0.'.. .‘O ' In. .4.- .0 00. .90.... . 620.01 '0... -..! . .0 I ......90.... .I c ....o A . . . . . . .. . '01,... I .10- ....... .3100... ... ..0 o .. O... ‘ .. O 0.. 3.2.... I .. . . . ... . . . -6091? .0 . ... . I.0.... . ., IVSQII. It..0. ’i.0lvtc‘(.( . I...II\¢IO0‘.". 27.9.0.0-I 9.0 300‘ . .00.. 0 I .90.'00A:.000o00.0.. . . .10 000.. 1.4 v0.p.1...000.l.05. . . .....I.II! ...-09.0.3.1."10. 33‘ .. 0 0. '0'. 0.0.1.0»! .. , v0. .0 0.0 I...I.IJ . .I... .O.‘ I: t I... . - 0‘..' IV. 0 3 I I'Id‘. 0.. {II-... 5.200 . o 0.0 .0.Ol .0. .00...V.. I . ... ..o 00 .0 . ... .0.¢.O.l..o Q‘I...’ I I 0 u u .I . .0 0. II. I V I..- I 0‘0: ..IOOI'0I0.. 9. l.- .00 , o ... ......no! 0 00. .30.}! . . I . v! Q. t. 0 I3... .00....1...¢..-..O. ... . «a. 0‘ . 3.3.9.-.. . .. .I 0 .0..4‘.I0.. I ‘ ...-.....OIIOIIIOAQ .q ..- O09 .\l‘00 9.0 I ‘1_I.\.. 0V '. .0 vii? ..0... . . . . . .. u . . no.1.- u...0. !?0:.I.. .0. I v.0, . Q v. a. ..L.. 5,0. 1' 0213.2 .0! 83'.-. 70.! u 4 . 0‘.‘.O‘0I§0 It,“ I 3.16.10 0 30.?! ’ ’0.‘!’ ..0I [13"0: ..‘.I.DI.Q. 0‘ .2321... 0 . . :0 o.- p . Q t. -..-A Si‘l."\‘.0,‘0 it! ..1..’\"OQI. c O; I 00.10.01... 0,. .-.O- .01 I.‘!’. .’.'C...I_¢ 010.0 0.! Oil 0'. 0.- .30.... ......Ionr 0.00, Q.Io.'?2?.. I I.0.0 0.. ‘1 ’I . . . .. . . .. ... , . 0. .. . . . . (.00 3.0.0.... . . , . . I... I0.....I10..0 A . u .. . . . .. . . . .. . p [I ..I . ..'. . 0 . . ..vc. .I..0..0. . . . . .0. o . ... I . . . . . . .I . 0 .ufi. Al 4 0 .ob 0.. II o ...-.0.. .. . .0. . o4 . .4 O . 0.. Y!“ 0 0 O t 00 .04 {3.0.0. 0 .. I . 4. . . . I v0 00.... . t. .10» a I. . .0. .. ...V0 Io. . II ...I .... I. . . .. I . _ . . 9.. . . V . . .9 . .0.o .. 0 1. av .. I. o ... . ...0 0.. . ,0 O ..0IQ0A . .... .n‘. 0 II I. O . u .4... ...... 0.7.00 .... .. 5.10,..0“. . . ... . ..c. . . ...I O. I '0‘... ... . 0" .0 ....t .l .0 O~ . . . . ICI .. O. . pl . ..II . .I.00 0. . . . ,.. .01....33...‘ n! . .. .....5. I .O . 2v. I‘ . .. .. . .. . . .. . ... . . . o ... . .V. I . . . .0... 0 . v. n . . I 0... .. ..0? 2 . 0. . I... 0 .0. I: ...v .0 I. O l. t ..-" I ...-I. l 0.. (1, .0 . .O.. . ... ... v . . .. ,u .. . 4 . .. .. ...!Ii .. .. - .. . l. . u . ... 0.09. . . .. . . .0 .... I O o .......I_ . .. «.00.... ... Vv.u..0 c . . .. I0. 0 101....IIIQ. .. y . . . . . II rt. a n . . . ... I .¢ I I .00. I i ... .II. . . O... ... 0.10.0.-. 3 6. ‘. - ....0. . . . . L. . . . . . o. . . . _ .0 . V V . ....Ol....,\....... . . .. .00. .0I... ,O.Io.ocvan II . _ ...-4‘...’0...I.0 I... I ...f. . . .0600... ..n-... . . . 0.. . I ... :00 .00!» I .‘ . .0 .o I y. I a-.-..-0O.I .1 . 0 ,.. .0...‘ .. o. .. .....I. .... I III. 40.5.3003..r'0..‘03...9.0. .. . . . .. ... 0 ...... .. . . . . . 0 A ...: ‘.100 ... l tot: L . .. 00... . ... 0 '0 .40 61...... 4.0.. ....b. ...I. 00. 4 C ...4 ..3 ...; 1 T . .I. . £0,000 0. . .¢. . to .03. . 4. . _ . I. . n O . I; . ... . . . .. v. 1 0 I . . . .....o . v P - . :. . . . A . '. 3 . . u q . . I ..... IV'IAIOI. 0 000'. I y 0 n a I V -0 ' . .. .9 I . A o .r . ‘00 ...I ..........‘I ... ... I .. ..OIO.I . . . 13.0 o . ...-.3 - :d\\ 05.0 ...0 t. 4a.)...v.‘$.. Q. 9.... .0 . ... -. .. .... . v 0. 00.. ...? ... I. I. . _ I. .. . , .4 v .‘0 U. . 0 . . Iv. c .v. ‘00.. .. .l (O c .' 0.-. . 0 v I. u . . .. I )5 ...?4 .9 . . 4 4 . . . . - . .035. .. . y.. .9“. 3.333tz “5.; . , 0 . . .... 32.0.. . . .. 30.0.0... ”.53.... ...... 0L... ...... ....rt 03.310.53.005 03..L..0.fl..n0.01;~.”uu.f (0.5.. ..l.‘3&00......0$.u‘a .0 4‘. “- ué‘: ‘0 ... 0.0 .0 ‘.o . .. Unit”). 0.1“... v. 0.. I3‘I30 0 J o :00. 0...... ... 0-.‘0..’.....0 i. 9.... .V. a4 .. .. .. .... 4n . o . v .I. . 0- ....0 to. 0'0. ..........0 09000...- I .0.- _ 0 . u. v . 0| . 0 mi? Ikiur::..f W... 0.8.” ....onu‘. £.W.....l I .u . Q. I .. . I.\l'|0-I0.‘I.0OC ... '0‘ 3.0:. a .. . I ‘0 . o It I I I . . . I351. ‘1009 C... 0 am». ..3 '0 0 ..100‘5‘! 10.... ’2. .40.: .II.I1.Q .l . 00 . .12.... .... .0 0.\’- O..u0.ll£" I. 1‘.II r . .... an t0 .0 . . .1... 50 z.....-...-.. It.90.I.... ... 03.50....3'30003293 4 _. ... ...O. 0.. II.- . ..0.. .. .0! . . I. .4. Cl . .0 I. I D . 1000. 0'90 0 I 0 #0:, 00! II I . I ll C U3 LIBRARY Michigan State University This is to certify that the dissertation entitled SPECIATION IN THE WESTERN NORTH AMERICAN WILDFLOWER GENUS MIMULUS presented by JAMES MICHAEL SOBEL has been accepted towards fulfillment of the requirements for the PhD degree in Plant Biology Ecology, Evolutionary Biology and Behavior \\ w W Major Br’ofessor’s Signature (7 / l /IQ Date MSU is an Affirmative Action/Equal Opportunity Employer PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE 5/08 KzlProyAchres/CIRCIDateDue.indd .___..—...-__. ._.._.-...___.. -.i t__ __.. H..- _,_._ __. __.-.h~_._.....‘ ‘“ SPECIATION IN THE WESTERN NORTH AMERICAN WILDF LOWER GENUS MIMUL US By James Michael Sobel A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Plant Biology Ecology. Evolutionary Biology and Behavior 2010 ABSTRACT SPECIATION IN THE WESTERN NORTH AMERICAN WILDFLOWER GENUS MIMUL US By James Michael Sobel A central task of evolutionary biologists is to identify the forms of reproductive isolation that are most important in speciation. The presented study approaches this issue by examining the strength of multiple forms of reproductive isolation across a group of recently diverged species pairs spanning the western North American wildflower genus, Mimulus. Chapter 1, Unification of Methods for Estimating the Strength of Individual Reproductive Barriers, reviews methods that have been used by previous authors for calculating the strength of individual reproductive barriers and presents a simple linear explanation for unifying the disparate approaches. Chapter 2, Ecogeographic Isolation Among Recently Diverged Species Pairs in the Genus Mimulus, presents a method for separating the intrinsic biological effects of differences in geographic range for estimating the strength of ecogeographic isolation using an ecological niche modeling approach. This work examines the strength of ecogeographic isolation for 12 recently diverged species pairs in the genus, and shows that this barrier is commonly very strong, with an average strength of 0.64 to 0.69 depending on which threshold of habitat suitability is employed. Because isolating barriers act sequentially throughout the life history of organisms, this first-to-act barrier is therefore responsible for the majority of isolation experienced by recently diverged species. Chapter 3, entitled The Evolution of Reproductive Isolation Across the Genus Mimulus, provides estimates of the strength of other forms of reproductive isolation that can act upon the gene flow remaining after the effect of ecogeographic isolation. Additional reproductive barriers examined include temporal isolation, isolation due to relative seed set, intrinsic postzygotic isolation due to relative hybrid viability, and relative hybrid fertility. Individually, intrinsic postzygotic isolation due to inviability is highly variable but relatively strong. Temporal isolation, seed set isolation showed moderate strength, and relative hybrid fertility was relatively weak. When viewed in light of the linear sequential stages at which reproductive barriers act in nature, ecogeographic isolation dominates the relative contribution of barriers to total isolation, while other forms including postzygotic isolation due to hybrid inviability prevents only a limited amount of gene flow. Several species pairs show evidence of incomplete total reproductive isolation, and it is unclear whether this represents cases of incomplete speciation or the effect of unmeasured barriers such as pollinator isolation or extrinsic postzygotic isolation. In Chapter 4, Contrasting Patterns of Introgression in Two Pairs of Mimulus Species, an introgression analysis of neutral molecular sequences is presented to corroborate the measured values of isolation for two species pairs representing relatively low and high reproductive isolation estimates. The species pair with relatively high reproductive isolation measured in the lab (M. constrictus and M. whitneyi) shows strong evidence of molecular sequence divergence, while the species pair with lower estimates of isolation (M. bicolor and M. filicaulis) shows evidence consistent with rampant gene exchange. Taken together, these analyses suggest that measurements of isolation estimated in the lab accurately represent effects on gene flow in nature, and ecogeographic isolation is of primary importance in the origin of Mimulus species. ACKNOWLEDGMENTS This dissertation could not have been completed without the support of many people and programs. I would like to thank the members of my guidance committee, Doug Schemske, Jeff Conner, Kay Gross, Rich Lenski, and Alan Prather, for providing advice through all stages of this project, reading drafts of chapters, and providing helpful comments. I would like to especially thank Jeff Conner and Doug Schemske. Jeff provided encouragement and mentorship both before and during graduate school, and has been a great influence on me. Doug has provided immeasurable amounts of encouragement and guidance through all stages of my dissertation. He has taught me a tremendous amount, and I truly could not have asked for a better mentor. Other professors at MSU have also been very supportive, especially Gary Mittelbach, Pete Murphy, and Barb Sears. Interactions with many colleagues have helped shape my thoughts on this project. Previous graduate students in the Schemske lab have improved the quality of this work tremendously through helpful discussions and criticisms including, Kathleen Kay, Amy Angert, Grace Chen, Jeff Evans, Lorna Watt, and Emily Dittmar. In particular, my interaction with Grace on many aspects of the work presented in Chapter 1 has been a very fruitful and enjoyable collaboration. Fellow graduate students in other labs have also helped shape ideas and provided encouragement including, Orlando Alvarez-Fuentes, Frances Knapczyck, Chris Gaukler, Mike Grillo, Emily Grman, Chris Hamm, Kane Keller, Lauren Kinsman, Jason Kilgore, Melissa Kjelvik, Raffica LaRosa, Jason Martina, Todd Robinson, Angela Roles, Anne Royer, and Heather Sahli. I am also grateful for the iv insight I gained through discussions with other researchers utilizing Mimulus as a model organism including Toby Bradshaw, Andrea Case, Arielle Cooley, Lila F ishman, Nick Griffin, Dena Grossenbacher, Chris Ivey, David Lowry, Noland Martin, Jason Sexton, Matt Streisfeld, Andrea Sweigart, John Willis, and Carrie Wu. I am indebted to Schemske lab technician, Mark Hammond for all of his assistance. His advice, support, and comradery have been greatly valued throughout this project. Much of the work presented would not have been possible without the assistance of many undergrads, including Christina Cooper, Emily Hall, Anna Kilbourn, Christine Leduc, Liana Nichols, Lindsay Petroff, Joshua Rilko, Maria Roberts, and Julia Smith. It is very difficult to single out especially valuable members of this list because so many of them have contributed incredible amounts of time and energy to the project. Interacting with them has been one of the true joys of my graduate school experience. I would also like to again thank my advisor, Doug Schemske, for providing the financial means to hire this incredible team of people. Several people provided valuable assistance in establishing the field component of this work. In particular, Steve Schoenig contributed much time and information that helped get me started in finding species of Mimulus in the mountains of California, and Joe Callizo was also incredibly generous with his time and knowledge of the plants of Napa and Lake counties. I spent time in the field with several other excellent botanists, including Paul Beardsley, Paul Aigner, and Cathy Koehler, and their advice and information helped considerably. In addition, I would like to thank the many National Forest and National Park Service botanists who assisted with permits and collection information. Funding for this work is gratefully acknowledged. The National Science Foundation provided support through both a l-'/2 year GK-12 fellowship and a 2-year NCEAS graduate fellowship. In addition, NSF provided research support in the form of a FIBR grant awarded to John Willis (Doug Schemske co-PI; DBI-O328636) and a Doctoral Dissertation Improvement Grant (Doug Schemske & J. Sobel, co-PI’s; DEB-0808447). In addition, several units within Michigan State University provided funds either for travel or stipends, including the Department of Plant Biology, Ecology, Evolutionary Biology & Behavior Program, Kellogg Biological Station, College of Natural Science, and the Graduate School. In particular, the Plant Biology Department’s Paul Taylor Travel Fund made much of the travel to field sites and meetings possible. Many former teachers and professors were inspirational in guiding me towards a life in science, and I would like to take this opportunity to formally thank them for their efforts. Dale Rosene and Gloria Wheeler were instrumental in introducing me to science in 8th and 9'h grades, and their passion for their subjects first sparked my interest in science in general and biology in particular. During my undergraduate work, several professors and scientists also served to inspire continued work in the subject, including Steve Bertman, Jeff Conner, Woody Ehrle, David Karowe, and Steve Roberds. Finally, I am deeply grateful to the love and support from my family during this project. My parents, Mike and Cindy Sobel, and grandparents, Andrew and Kathleen Sobel & Jim and Judy Huggett, have always encouraged and supported my education, for which I am eternally thankful. Most of all, I want to thank my wife, Darcy Sobel, and son, Luke Sobel, for their love, encouragement, and patience throughout this project. vi TABLE OF CONTENTS LIST OF TABLES .............................................................................................................. ix LIST OF FIGURES ............................................................................................................ xi CHAPTER 1 UNIFICATION OF METHODS FOR ESTIMATING THE STRENGTH OF INDIVIDUAL REPRODUCTIVE BARRIERS .................................................................. 1 Introduction .............................................................................................................. 3 Currently Used Methods .......................................................................................... 5 A Simple Linear Solution ...................................................................................... 10 Accommodation of Unequal Null Expectations .................................................... 15 Reproductive Barriers that Affect Co-Occurrence ................................................ 17 Summary ................................................................................................................ 21 Tables ..................................................................................................................... 23 Figures .................................................................................................................... 24 Literature Cited ...................................................................................................... 29 CHAPTER 2 ECOGEOGRAPHIC ISOLATION AMONG RECENTLY DIVERGED SPECIES PAIRS IN THE GENUS MIMULUS ................................................................................. 32 Introduction ............................................................................................................ 34 Materials & Methods ............................................................................................. 38 Results .................................................................................................................... 48 Discussion .............................................................................................................. 52 Summary ................................................................................................................ 57 Tables ..................................................................................................................... 59 Figures .................................................................................................................... 64 Literature Cited .................................................................................................... 131 CHAPTER 3 THE EVOLUTION OF REPRODUCTIVE ISOLATION ACROSS THE GENUS MIMULUS ........................................................................................................................ 137 Introduction .......................................................................................................... 140 Materials & Methods ........................................................................................... 145 Results .................................................................................................................. 155 Discussion ............................................................................................................ 161 Summary .............................................................................................................. 171 Tables ................................................................................................................... 172 Figures .................................................................................................................. 176 Literature Cited .................................................................................................... 192 vii CHAPTER 4 CONTRASTING PATTERNS OF INTROGRESSION IN TWO PAIRS OF MIMUL US SPECIES .......................................................................................................................... 197 Introduction .......................................................................................................... 200 Materials & Methods ........................................................................................... 203 Results .................................................................................................................. 209 Discussion ............................................................................................................ 2 1 2 Summary .............................................................................................................. 215 Tables ................................................................................................................... 216 Figures .................................................................................................................. 225 Literature Cited .................................................................................................... 227 viii LIST OF TABLES Table 1.1. Alternative methods for calculating reproductive isolation indexes ................................................................................................................... 23 Table 2.1. Summary of Mimulus species pairs used and methods employed for ecological niche modeling ....................................................................................................... 59 Table 2.2. Environmental layers used in ecological niche modeling ................................ 60 Table 2.3. Correlation coefficients among environmental layers used in ecological niche modeling ................................................................................................................ 61 Table 2.4. Results of one-way MANOVA analysis on environmental variables between Mimulus species pairs ............................................................................................ 62 Table 2.5. Ecogeographic reproductive isolation between species pairs in Mimulus ........ 63 Table 3.1. Mimulus species pairs used in comparative study of reproductive isolation ................................................................................................................ 172 Table 3.2. Estimates of the strength of individual reproductive barriers in pairs of species in Mimulus ........................................................................................................... 173 Table 3.3. Relative parent and F l viability in Mimulus species pairs ............................. 174 Table 3.4. Comparison of individual barrier strength and relative contributions to total reproductive isolation in recently diverged species of Mimulus .......................... 175 Table 4.1A. Molecular markers sequenced for analysis of introgression within Mimulus bicolor / M. filicaulis species pair ........................................................................ 216 Table 4.1B. Molecular markers sequenced for analysis of introgression within Mimulus constrictus / M. whitneyi species pair .................................................................. 217 Table 4.2. Primer sequences used to amplify MgSTS markers in this study ................... 218 Table 4.3A. Results of polymorphism analysis using SITES software for Mimulus bicolor / M. filicaulis species pair ........................................................................ 219 Table 4.38. Results of polymorphism analysis using SITES software for Mimulus constrictus / M. whitneyi species pair .................................................................. 220 Table 4.4A. Pairwise FST values from molecular polymorphism analysis performed in SITES for Mimulus bicolor / M. filicaulis species pair ....................................... 221 ix Table 4.4B. Pairwise F ST values from molecular polymorphism analysis performed in SITES for Mimulus constrictus / M. whitneyi species pair .................................. 222 Table 4.5A. Estimates of Tajima’s D and the mutation rate parameter 0 (4Nu) per base pair for Mimulus bicolor / M. filicaulis species pair ............................................ 223 Table 4.53. Estimates of Tajima’s D and the mutation rate parameter 0 (4Nu) per base pair for Mimulus constrictus / M. whitneyi species pair ...................................... 224 LIST OF FIGURES Figure 1.1. Relationship between the probability of heterospecific gene flow and the measure of reproductive isolation obtained through three commonly used methods ................................................................................................................. 24 Figure 1.2. Diagram showing similarities between pre- and postzygotic isolation index measures of equal magnitude ................................................................................. 25 Figure 1.3. The effect of changes in expected frequency of hybrid formation on the value of reproductive isolation using equation R16 ......................................................... 26 Figure 1.4. Hypothetical data showing the relationship between three methods for calculating the strength of temporal isolation ........................................................ 28 Figure 2.1. Phylogenetic relationships among Mimulus species in this study ................... 64 Figure 2.2A. Ecological niche model of Mimulus androsaceus ........................................ 65 Figure 2.28. Ecological niche model of Mimulus shevockii ............................................. 66 Figure 2.2C. Ecological niche model of Mimulus angustatus ........................................... 67 Figure 2.2D. Ecological niche model of Mimulus pulchellus ............................................ 68 Figure 2.2E. Ecological niche model of Mimulus bicolor ................................................. 69 Figure 2.2F. Ecological niche model of Mimulusfilicaulis ............................................... 70 Figure 2.2G. Ecological niche model of Mimulus bigelovii .............................................. 71 Figure 2.2H. Ecological niche model of Mimulus bolanderi ............................................. 72 Figure 2.21. Ecological niche model of Mimulus brevipes ................................................ 73 Figure 2.2J. Ecological niche model of Mimulusjohnstonii ............................................. 74 Figure 2.2K. Ecological niche model of Mimulus cardinalis ............................................ 75 Figure 2.2L. Ecological niche model of Mimulus Iewz’sii .................................................. 76 Figure 2.2M. Ecological niche model of Mimulus constrictus .......................................... 77 Figure 2.2N. Ecological niche model of Mimulus whitneyi ............................................... 78 Figure 2.20. Ecological niche model of Mimulus cusickii ................................................ 79 xi Figure 2.2F. Ecological niche model of Mimulus nanus ................................................... 80 Figure 2.2Q. Ecological niche model of Mimulus douglasii ............................................. 81 Figure 2.2R. Ecological niche model of Mimulus kelloggii .............................................. 82 Figure 2.28. Ecological niche model of Mimulusfloribundus ........................................ ..83 Figure 2.2T. Ecological niche model of Mimulus norrisii ................................................. 84 Figure 2.2U. Ecological niche model of Mimulus gracilipes ............................................ 85 Figure 2.2V. Ecological niche model of Mimulus palmeri ................................................ 86 Figure 2.2W. Ecological niche model of Mimulus parryi ................................................. 87 Figure 2.2X. Ecological niche model of Mimulus rupicola ............................................... 88 Figure 2.3A. Variation in environmental variables between the species pair Mimulus androsaceus and M shevockii ............................................................................... 90 Figure 2.3B. Variation in environmental variables between the species pair Mimulus angustatus and M. pulchellus ................................................................................. 92 Figure 2.3C. Variation in environmental variables between the species pair Mimulus bicolor and M. filicaulis ......................................................................................... 94 Figure 2.3D. Variation in environmental variables between the species pair Mimulus bigelovii and M. bolanderi ..................................................................................... 96 Figure 2.3B. Variation in environmental variables between the species pair Mimulus brevipes and M. johnstonii ..................................................................................... 98 Figure 2.3F. Variation in environmental variables between the species pair Mimulus cardinalis and M. lewisii ...................................................................................... 100 Figure 2.3G. Variation in environmental variables between the species pair Mimulus constrictus and M. whitneyi ................................................................................. 102 Figure 2.3H. Variation in environmental variables between the species pair Mimulus cusickii and M. nanus ........................................................................................... 104 Figure 2.31. Variation in environmental variables between the species pair Mimulus douglasii and M kelloggii .................................................................................... 106 xii Figure 2.3J. Variation in environmental variables between the species pair Mimulus floribundus and M. norrisii .................................................................................. 108 Figure 2.3K. Variation in environmental variables between the species pair Mimulus gracilipes and M. palmeri .................................................................................... 110 Figure 2.3L. Variation in environmental variables between the species pair Mimulus parryi and M. rupicola ......................................................................................... 112 Figure 2.4A Identity tests using Schoener’s D statistic ................................................... 114 Figure 2.4B Identity tests using 1 statistic ........................................................................ 116 Figure 2.5A. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus androsaceus and M shevockii ....................................................... 117 Figure 2.53. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus angustatus and M. pulchellus ........................................................ 118 Figure 2.5C. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus bicolor and M filicaulis ................................................................. 119 Figure 2.5D. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus bigelovii and M bolanderi ............................................................. 120 Figure 2.5E. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus brevipes and M. johnstonii ............................................................. 121 Figure 2.5F. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus cardinalis and M Iewisii ................................................................ 122 Figure 2.5G. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus constrictus and M whitneyi ........................................................... 123 Figure 2.5H. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus cusickii and M nanus .................................................................... 124 Figure 2.51. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus douglasii and M kelloggii ............................................................. 125 Figure 2.5J. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulusfloribundus and M norrisii ............................................................ 126 Figure 2.5K. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus gracilipes and M palmeri .............................................................. 127 xiii Figure 2.5L. Overlay plot showing the amount of ecogeographic overlap in the species pair Mimulus parryi and M rupicola ................................................................... 128 Figure 2.6. Estimates of ecogeographic isolation in Mimulus obtained using threshold 2 regressed on sequence divergence between the species pairs .............................. 129 Figure 2.7. Relationship between niche similarity and ecogeographic isolation ............. 130 Figure 3.1. Recently diverged pairs of species in the genus Mimulus used in this study ..................................................................................................................... 176 Figure 3.2A-D. Differences in flowering time based on herbarium collection for the species pairs Mimulus angustatus /pulchellus, M bicolor /filicaulis, M brevipes /johnst0nii, M constrictus / whitneyi .................................................................. 177 Figure 3.2E-H. Differences in flowering time based on herbarium collection for the species pairs Mimulus cusickii /' nanus, M douglasii / kelloggii, M floribundus / norrisii, M gracilipes /palmeri ........................................................................... 178 Figure 3.3A-I. Comparisons of flowering time differences between Mimulus species under common garden conditions in the greenhouse ........................................... 179 Figure 3.4A-C. Postmating reproductive isolation in the species pairs Mimulus angustatus /pulchellus, M bicolor /filicaulis, and M brevipes /johnst0nii ........................ 180 Figure 3.4D-F. Postmating reproductive isolation in the species pairs Mimulus constrictus / whitneyi, M cusickii / nanus, and M douglasii / kelloggii ................................ 181 Figure 3.4G-I. Postmating reproductive isolation in the species pairs Mimulusfloribundus / norrisii, M gracilipes /palmeri, and M jungermanm'oides / washingtonensis ................................................................................................... 1 82 Figure 3.5A-B. Relative pollen fertility measures in species pairs Mimulus angustatus / pulchellus and M bicolor /filicaulis ................................................................... 183 Figure 3.5C-E. Relative pollen fertility measures in species pairs Mimulus douglasii / kelloggii and individual species M constrictus and M floribundus ................... 184 Figure 3.5F-G. Relative pollen fertility measures in species pairs Mimulus gracilipes / palmeri and M jungermanm'oides / washingtonensis .......................................... 185 Figure 3.6. Individual estimates of reproductive isolation across the genus Mimulus....186 Figure 3.7A. Individual reproductive barrier strengths in species pairs of Mimulus ....... 188 xiv Figure 3.73. Relative contribution of reproductive barriers to total reproductive isolation in species pairs of Mimulus .................................................................................. 190 Figure 3.8. Relationship between total reproductive isolation of species pairs in Mimulus and genetic distance ............................................................................................. 191 Figure 4.1A. Distribution of populations sampled for molecular analysis in Mimulus bicolor and M filicaulis. ..................................................................................... 225 Figure 4.13. Distribution of populations sampled for molecular analysis in Mimulus constrictus and M whitneyi ................................................................................. 226 XV CHAPTER 1: UNIFICATION OF METHODS FOR ESTIMATING THE STRENGTH OF INDIVIDUAL REPRODUCTIVE BARRIERS CHAPTER 1: UNIFICATION OF METHODS FOR ESTIMATING THE STRENGTH OF INDIVIDUAL REPRODUCTIVE BARRIERS Under the biological species concept, understanding the evolution of reproductive isolation is tantamount to describing the origin of species. A major goal of speciation research is to identify which of the many forms of reproductive isolation contribute most to total isolation at the time of speciation. In order to achieve this goal, the strength of multiple forms of isolation must be compared in an equivalent manner. Despite this necessity, there exists a wide diversity of methods that have been employed to estimate isolation, falling into several mathematically non-equivalent categories. This has resulted in comparisons between forms of isolation and between taxa in which the measure of isolation is not directly comparable. In this study, a simple linear formulation for the isolation index with the most biological support is given. This method for estimating isolation provides three distinct advantages: 1) it is directly related to gene flow between population, 2) it is symmetrical about the origin, such that measures of disassortative mating and heterosis are comparable to measures of isolation in the positive range, and 3) it is equivalent between broad categories of reproductive isolation. This linear formulation can be adopted in cases where expected amounts of con— and heterospecific gene flow differ under a null model and can be adjusted for use in all forms of isolation. Introduction Adherents of the biological species concept (Mayr 1942) have developed a research program into the origin of species based upon understanding the evolution of barriers to reproduction (Coyne and Orr 2004). Within this framework, the process of speciation involves the accumulation of barriers to gene exchange leading to complete reproductive isolation (Coyne and Orr 2004). It has been long recognized (Dobzhansky 1937; Mayr 1942; Poulton 1908) that reproductive barriers can take many forms from ecological isolation to intrinsic hybrid inviability. These barriers are often separated into those that operate before the formation of a hybrid (prezygotic) and those that act once a hybrid has formed (postzygotic), and there is longstanding debate about which of these broad categories of isolating barriers are most important in speciation (Schemske 2000; Coyne and Orr 2004; Rice and Hostert 1993). Early work on the nature of reproductive barriers typically consisted of gathering data that showed that barriers existed (e. g. Dobzhansky 1938) without necessarily estimating a metric that related the barrier strength to an amount of gene flow reduced by it. However, a central task of speciation studies is to identify which forms of reproductive isolation are the most important in the process (Coyne and Orr 2004); therefore, some method for comparison is clearly necessary. In their highly influential study in Drosophila, Coyne and Orr (1989; 1997) used data from the literature on 171 interspecific hybridization attempts, and developed a metric that related the relative frequency of con- and heterospecific matings to reproductive isolation. This work provided evidence that the evolution of prezygotic isolation outpaces postzygotic due to differences in sympatric taxa and that Haldane’s rule is a common feature of postzygotic isolation evolution. A surge of interest in calculating the strength of reproductive isolation resulted (e.g. Bolnick and Near 2005; Coyne and Orr 1989; 1997; Dopman et a1. 2010; Matsubayashi and Katakura 2009; Mendelson 2003; Moyle et a1. 2004; Presgraves 2002; Price and Bouvier 2002; Ramsey et al. 2003; Sasa et al. 1998), and most authors adopted the mathematical formulation provided by Coyne and Orr (1989). However, Martin and Willis (2007) recently challenged whether the isolation coefficients estimated by these methods are truly representative of the declines in gene flow experienced by species. Among other concerns, they raise issues about whether pre- and postzygotic isolation are being calculated equivalently and propose methods for accounting for different null expectations of gene flow due to unequal population or gamete abundances. While these issues have been appreciated by some (6. g. Matsubayashi and Katakura 2009), others have failed to adopt the proposed methods (e. g. Dopman et a1. 2010), creating substantial confusion over the correct methods for analyzing isolation. Given the importance of these measurements, the lack of consensus on the methods used to calculate indexes of isolation is problematic. As a result, isolation indexes are used inconsistently, even within a single study. Clearly, identifying the most appropriate indexes for each form of isolation would be of great benefit, both when assessing the relative strength of individual barriers in single species pair studies and when assessing the strength of isolation among groups in a comparative context. Our goal in this paper is to review the most common methods used by previous authors, present a simplified view of the alternative with the most mathematical and biological justification, and provide examples to illustrate how to utilize this method for both ideal and realistic datasets. Ultimately, we hope this unification of methods will allow researchers the opportunity to compare measures of reproductive isolation among disparate forms of isolation and/or taxa. Currently Used Methods The purpose of calculating the strength of a reproductive isolating barrier is to estimate how much gene flow is (or is potentially) reduced by a barrier (Coyne and Orr 2004). In most cases, previous workers have used equations to describe reproductive isolation that range from 0 when there is no isolation to 1 when there is complete isolation (Table 1.1). However, as long as a metric satisfies this one requirement, the relationship between the strength of a barrier and the amount of isolation attributed to it has been essentially unexplored. Because pre- and postzygotic isolation have traditionally been approached in non-equivalent ways (Martin and Willis 2007), we will begin by reviewing these separately. Prezygotic Isolation Metrics of prezygotic isolation attempt to predict how a barrier will affect the probability of heterospecific zygote formation. Sexual isolation due to mating preferences has been widely studied as a potential agent of isolation (Coyne et al. 2005; Coyne and Orr 1997; Ehrman 1965; Matsubayashi and Katakura 2009; Mendelson 2003; Tilley et al. 1990). One equation used to describe the relationship between mating preferences and isolation is presented in Coyne and Orr (1989) as: frequency of heterospecific matings R] = l 9 frequency of conspecific matings which we will henceforward simplify as: H R] - 1— —, 1 1 C ( ) where H represents the frequency of heterospecific matings and C represents the frequency of conspecific matings. This equation (or minor deviations from it) is the most commonly used approach to estimating the contribution of mating preferences to reproductive isolation (Table 1.1). The metric indeed results in an isolation index that equals 1 when sexual preferences insure no heterospecific mating and zero when there are no preferences. However, the relationship is not linear between 0 and 1, and ranges to -oo in cases of disassortative mating (Figure 1.1A). As an example, in a choice mating trial, if species X females mate heterospecifically 25% of the time and conspecifically 75% of the time, there is clearly a preference for within species matings, and equation 1 would result in an isolation index of 0.67. Alternatively, if females of species X actually prefer males of species 0 resulting in the reversal of the above frequencies (25% conspecific mating and 75% heterospecific mating), then equation 1 would give an isolation index of -2. Given that the departure from random mating preferences are equivalent, the effect on gene flow would be identical, but in opposite directions. Obviously, this isolation index does not capture this symmetry, and the resulting indexes are difficult to ascribe to the probability of gene flow. This creates situations in which instances of disassortative mating have to be removed from consideration (or are replaced with a zero) because they are not directly comparable to measures within the normal bounds. While disassortative mating is relatively rare in interspecific studies (Coyne & Orr 1989 note at least 5 examples in Drosophila), a method that can accommodate its effects on gene flow is warranted for situations where it may be more common, such as studies of pre-speciation population divergence. Another equation used for estimating the strength of prezygotic isolation appears in Ramsey et al. (2003) for calculating the strength of reproductive isolation due to pollinator preferences on monkeyflowers Mimulus lewisii and M cardinalis. They present an equation using the relative frequency of cross-species foraging bouts to all visits to estimate isolation: number of cross - species foraging bouts RI =1- total number of foraging bouts which we will henceforward simplify into: R12 =1- (2) C+H' R12 can range from 0 (all heterospecific gene flow) to 0.5 (random mating) to 1 (no heterospecific gene flow) (Figure 1.18). The biological interpretation of R12 may be very different from the same R11 value, presenting some problems. For instance, when R11 = 0.5, heterospecific mating is half of conspecific mating, while an R12 = 0.5 results from hetero- and conspecific mating being equivalent. Clearly, a situation that doesn’t deviate from a null expectation under random mating is most appropriately considered to have no isolation. Yet another form of the equation used to calculate prezygotic isolation is presented in Mendelson’s (2003) study of reproductive isolation across fish in the genus Etheostoma. Sexual isolation was calculated as: # conspecific spawning events - # heterospecific spawning events RI . total # spawning events This equation was referenced from Stalker’s (1942) study of sexual isolation in the Drosophila virilis complex where the equation was presented as: % conspecific females inseminated— % alien females inseminated RI - . % conspecrfrc females inseminated + % alien females msemmated We will simplify this expression into: C-H. (3) This index ranges from -1 (all heterospecific mating) to 0 (random mating) to 1 (no heterospecific mating), which is mathematically easy to understand; yet the biological meaning of the ratio between the numerator and the denominator is not made clear. The symmetrical nature of this form of the equation presents significant benefits, and we will return to its derivation in the following section. Postzygotic Isolation Isolation indexes of postzygotic isolation estimate the amount of gene flow reduced by the inviability or sterility of hybrids (either through intrinsic or extrinsic means). As pointed out by Martin & Willis (2007), one of the most serious difficulties in comparing pre— and postzygotic isolation within studies is that the two isolation indexes are commonly non-equivalent. For example, Coyne & Orr (1989) counted the number of instances of complete sterility or complete inviability in a reciprocal cross and divided by 4 to make a metric that could only be 0.25, 0.5, 0.75, or 1. Others have attempted to calculate postzygotic isolation in a more quantitative way, with equations that are algebraically equivalent or similar to one of the three presented thus far. For example, many studies (Table 1.1) calculate postzygotic isolation as: average fitness of offspring from heterospecific cross R1 = 1 - average fitness of offspring from conspecific cross which is mathematically equivalent to R1,, Using this form of the equation presents difficulties when considering cases in which hybrids outperform the parents. Heterosis is relatively common among recently diverged species (e.g. Taylor et al. 2009), but the value that RI 1 gives in these instances is not proportional to the amount of gene flow that may be facilitated by the phenomenon. For example, if a plant of species X makes twice as many seeds when mated interspecifically to species 0, the isolation calculated by R11 would be -1. However, if the seed sets are reversed (intraspecific matings produce twice as many seeds), R11 would yield an isolation coefficient of 0.5. This is analogous to the problem discussed above for cases of disassortative mating in prezygotic measures, and can present a considerable challenge when making comparisons of the isolation strength or combining multiple forms of isolation into composite metrics (as in Lowry et al. 2008). Additional studies use equations that can be equivalent to R13 under certain circumstances. For example, Palmer and Feldman (2009) present this equation: 2 x average fitness of hybrid offspring sum of average fitnesses from two allopatric populations. RI=I— While this form of the equation is the same as the simple linear solution discussed below, the authors did not provide a justification for the equation used, and it is not made clear what is gained by using this form of the equation. The variety of isolation indexes available makes it difficult to compare the strength of isolation among disparate taxa. Sufficient data on reproductive isolation exist to allow researchers to approach higher order questions about the processes that increase the rate of reproductive isolation evolution (e.g. Funk et al. 2006; Lowry et al. 2008). However, if different forms of reproductive isolation indexes are used for different barriers or taxa, it is not possible to combine the data into global analyses. Within individual studies that measure more than one reproductive barrier, the goal is often to test hypotheses concerning which form of isolation evolves at a faster rate (e.g. Coyne and Orr 1989; 1997). The relative rate of isolation evolution among barriers is an actively debated subject in evolutionary biology; however, it is impossible to legitimately compare different forms of isolation if they are calculated by different measures. Clearly, a unification of methods would benefit those trying to make such comparisons. A Simple Linear Solution As described above, most would agree that metrics of reproductive isolation should be scaled to 0 when no isolation exists, and 1 when isolation is complete, but the manner in which this is achieved can greatly affect the resulting interpretation of the value obtained. The most straightforward way to achieve this goal is to use a simple linear equation that describes the relationship between the probability of gene flow past a specific barrier and the reproductive isolation index with the following conditions: I) if 10 the probability of interspecific gene flow is 0, reproductive isolation is 1, 11) if the probability of gene flow is 0.5 (e.g. random mating), reproductive isolation is 0, 111) if the probability of gene flow is l (e.g. complete disassortative mating), reproductive isolation is -1. These conditions are satisfied by a line with y-intercept of 1 and slope of -2 (Figure 1.1C), producing the linear equation: = -2x + 1 (4) where y is the reproductive isolation index and x is the probability of heterospecific gene flow. The probability of heterospecific gene flow can be estimated in several ways, but the simplest approach is to calculate the relative proportion of heterospecific to total gene . Substituting for x gives our simplest flow with the following term: 4. proposed method for calculating reproductive isolation as: * H RI5=1-2 (0”). (5) As alluded to above, this equation is algebraically equivalent to equation 3 under certain circumstances, and has been used in different forms in several previous treatments of isolation (e.g. Mendelson 2003; Palmer and Feldman 2009, Table 1.1). However, the previous uses of this metric did not describe the purpose of the reformulation, and failed to demonstrate its advantages over previous versions. Consider a case of prezygotic isolation in which a mating study showed that females of one species mated conspecifically 90% of the time and heterospecifically 10%. Equation R11 gives a reproductive isolation index for those females of 0.89 because conspecific matings represent 8/9 that of heterospecific. Equation R12 gives a 11 reproductive isolation index of 0.9 because conspecific matings represent 90% of overall matings. However, assuming both species are equally abundant in a mating trial, the null expectation is that random mating would produce a 50:50 ratio of heterospecific to conspecific mating. Using our proposed R15 equation, the isolation index is instead 0.8, which represents the average of the proportional increase in conspecific mating and decrease in heterospecific mating compared to this random expectation (a conspecific mating frequency of 0.9 is an 80% increase over the random expectation of 0.5, and a heterospecific frequency of 0.1 is an 80% reduction of 0.5). There are several major advantages to using this simple linear formulation for estimating the strength of reproductive isolation. The first is that the index calculated can be directly related to gene flow as it represents the relative under-representation of heterospecific reproduction and over-representation of conspecific mating relative to expectations under random mating; i.e. how isolated a species is based on a specific barrier. In addition to being directly applicable to gene flow, another major advantage is that R15 gives a meaningful value when heterospecific gene flow is favored. Previous studies have eliminated these data or set their values equal to zero because the potentially large negative values are not directly comparable to measures of isolation from the typical part of the range (Coyne and Orr 1989; 1997). This is not necessary when using R15_ This can be illustrated by imagining a case of disassortative mating in which females prefer males of the opposite species. In an extreme example, if the mating preferences used above were reversed such that females of species X prefer males of species 0 and 12 mate heterospecifically 90% of the time and conspecifically only 10%, R11 would yield a value of -8. The proposed method using R15 would instead give a value of -0.8. Unlike the value obtained by R11, this index has biological significance. The negative value indicates that the preference trait facilitates gene flow rather than confers isolation, and the magnitude denotes that there is 80% more heterospecific mating (i.e. gene flow) than expected by chance. This property allows the inclusion of these data when comparing to data from the positive range, and facilitates creating composite metrics that estimate the combined strength of multiple forms of isolation (e. g. Ramsey et al. 2003). The third advantage of using the proposed R15 is that the same equation can be applied to both pre- and postzygotic isolation barriers and produce comparable values. While the explanation for applying this equation to prezygotic is reasonably intuitive, it may seem less so for postzygotic isolation. For example, if species X has relative fitness of l and hybrids that result from mating females of species X to males of species 0 have half the fitness, R15 would result in an isolation coefficient for females of species X of 0.33. The biological reasoning for this result becomes clear when imagining a simplified scenario where postzygotic isolation is the only barrier present and manifests as a difference in hybrid survival rates. Under random mating, females of species X will mate with males of species 0 with a frequency of 0.5. With a hypothetical starting population of 20 X females, 10 will mate with other X males while 10 will mate with species 0 males. If each of those matings result in one offspring, the next generation will have 10 species X offspring and 10 X/O hybrids. If half the X/O hybrids die as a result of postzygotic isolation, the total population consists of 15 individuals, 5 of which are 13 hybrids. A null expectation for 15 offspring of species X females is that 7.5 will be species X and 7.5 will beX/O hybrids; therefore, the 5 viable hybrids that are observed represents a reduction in gene flow of 1/3 relative to the random expectation (and 10 conspecific offspring represents 1/3 more than the null expectation of 7.5). To illustrate that measures of pre- and postzygotic isolation using the proposed method of R15 have equivalent effects on gene flow, Figure 1.2 provides a pair of hypothetical scenarios in which a population begins with 10 females of species 0. Figure 1.2A shows the effect of a mating preference that results in conspecific mating with a frequency of 0.9 and heterospecific frequency of 0.1. Using R15, the reproductive isolation value obtained from this hypothetical scenario is 0.8. If each female produces 10 offspring, after reproduction there should be 90 species 0 offspring and 10 O/X hybrids based on these preferences. We impose postzygotic isolation of 0.9 on both conspecific and heterospecific offspring (the exact value is irrelevant as long as the two classes of offspring are equal). This results in a population with 81 species 0 offspring and 9 O/X hybrids. We can again see the biological implication of a reproductive isolation value of 0.8 in the fact that there are 80% fewer heterospecific offspring than expected by chance. Figure 1.2B shows that the same amount of postzygotic isolation will result in an equivalent effect on gene flow. In this scenario the same number of females of species 0 begin with no mating isolation, such that the offspring produced in the next generation are 50 species 0 and 50 O/X hybrids. To impose postzygotic isolation of 0.8, we set the survival rate of conspecific offspring to 0.9 and the survival of heterospecific offspring to 0.1. After passing through this selective filter, the population consists of 45 species 0 offspring and 5 O/X hybrids. The isolation coefficient of 0.8 again makes biological sense 14 because in a population of 50 offspring of species 0 females, the expectation under a null model is 25 of each offspring class. The coefficient of 0.8 represents both the proportional increase in conspecific offspring and the proportional decrease in heterospecific offspring. Again this direct estimate of the reduction of heterospecific offspring is directly related to the decline in gene flow experienced compared to the null expectation. Accommodation of Unegual Null Expectations In the above examples, the simplifying assumption of equivalent null expectations of con- and heterospecific gene flow allowed the utilization of R15. However, as Martin & Willis (2007) argue, it is sometimes necessary to scale indexes of reproductive isolation to different null expectations, such as when there are consistent differences in relative abundances or gamete numbers between species. Using a similar convention to that proposed by Martin & Willis (2007), the observed frequencies of con- and heterospecific gene flow can be made proportional to their individual expected frequencies: observed C/expected C and observed H/expected H. The average over-representation of conspecific and under-representation of heterospecific can be calculated by simply dividing the sum by two: (observed C / expected C) + (observed H / expected H) 2 . To calculate isolation, the relative proportion of the term (observed H/expected H) to this average is subtracted from 1: 15 1 (observed H / expected H) (observed C / expected C) + (observed H / expected H) ' 2 This equation can be rearranged to give: 2(0bserved H / expected H) (observed C / expected C) + (observed H / expected H) . R16 =1— (6) This form is easily related to our simplified R15 version, and indeed, if expected H and expected C are equivalent, the equation is identical to R15. The probability of is elaborated to accommodate the unequal heterospecific gene flow from R15 of H C + H observed H / expected H oberserved C/expected C + oberved H/expected H expected values as . The R16 form of the equation results in a non-linear relationship between the reproductive isolation index and observed frequency of heterospecific gene flow for all values of expected H and C except 0.5, but is still bounded by 1 and -l at the extremes of complete isolation and disassortative mating/heterosis (Figure 1.3). The line passes through the x-axis at the expected value of heterospecific gene flow (i.e. reproductive isolation equals zero when the observed H equals the expected H). As an example, if species X and species 0 differ in relative abundance in areas of sympatry such that on average, the frequency of species X is 0.6 and the frequency of species 0 is 0.4, then there is no longer an expectation of a 50:50 ratio in hetero:conspecific mating under a random expectation. Instead, even if there are no mating preferences, species X will mate conspefically 60% of the time by chance. Therefore, if species X is observed mating heterospecifically 20% of the time and 16 conspecifically 80%, there is reproductive isolation due to mating preference (R16 = 0.45), but it is milder than would be calculated without taking the altered null expectation into consideration (R15 = 0.6). Reproductive barriers that affect co-occurrence The isolating barriers discussed so far have been based on traits that can both limit (reproductive isolation) or facilitate gene flow (e. g. disassortative mating, heterosis). However, forms of isolation that affect the potential for co-occurrence in either time or space are generally based on traits that can not act to facilitate heterospecific gene flow, and are therefore bound by 0 (no isolation) and 1 (complete isolation) rather than -1 and 1. These include ecogeographic isolation, temporal isolation, and some forms of pollinator isolation data, such as lists of shared and non-shared pollinators. These barriers have some unique features that affect the calculation of isolation. For example, the distribution of flowering times for species X may be completely separate from species 0, resulting in complete reproductive isolation. This is equivalent to a mating isolation situation in which species X mates conspecifically 100% of the time. Flowering time could also overlap completely, which would result in zero reproductive isolation. This is equivalent to a situation in which no mating preferences exist, and therefore random mating occurs. The difference between mating isolation and temporal isolation is that mating preferences can occur that would actually facilitate gene flow (i.e. disassortative mating), while there is no equivalent ability for flowering time traits to increase gene flow beyond random mating. 17 While the simple form of equation R14 still applies, calculating the probability of heterospecific gene flow for barriers that affect co-occurrence is less straightforward than above. In these instances there are both shared regions of a distribution and non-shared regions. Within the shared region the expectation of heterospecific gene flow is 0.5 (or can be altered to a different expectation when necessary), while in the non-shared region, the expected heterospecific gene flow is 0. Therefore, the probability of heterospecific gene flow is equal to the relative proportion of half the shared overlap area to the total area occupied, and can be substituted for x in equation 4 to yield: 1/2 Shared R1 = l — 2( , Unshared + Shared which simplifies to: Shared . (7) Unshared + Shared R17=I— It is necessary to use this form of the equation under any circumstances in which there are shared and non-shared portions of a distribution in which interactions can occur. Using equation R17 assumes that the shared region of overlap between species results in a 50:50 ratio of hetero:conspecific mating. This may be nearly true in many cases as relative abundances may be close to 50% each when averaged across the entire shared region of the distribution. However, estimates of reproductive isolation can be refined when more specific data about relative abundance of each species can be collected (Martin and Willis 2007). Using temporal isolation of flowering plants as an example, when each species can be assumed to be of similar abundance overall, the probability of heterospecific gene flow on day i can be calculated by estimating the proportion of species A flowering on day i to 18 its total abundance throughout the growing season (Aj/Atotal) and multiplying by the relative abundance of the heterospecific species on day i (Bi/(A,+B,-)). Summing these products across all days of the growing season of species A and substituting into the probability of heterospcific from equation R14, the isolation index can be calculated by: Bi .- . (8) Atotal xAi + Bi R18=l- —2.(2 This equation can be further complicated by incorporating different global expected frequencies of each species based on consistent differences in relative abundance. In these cases, the terms involving both heterospecific and conspecific mating must be scaled by the non-equivalent expected values, resulting in: observed H expected H observed C observed H + expected C expected H 21 Ai x Bi ) IAtotal Ai +Bi R19 = l _ 2 Btotal ((Atotal + Btotal) (9) Ai Ai .( ) ( Bl l Atotal xAi + Bi 21A Atotal xAi + Bi Atotal “Atom! "' Btotal)+ Btotal /( Atotal + Btotal) R19 = l - 2 ; which can be expressed as: This equation is very similar to the one employed by Martin & Willis (2007); however, their base equation was equivalent to the form shown in R1,, which we argue should be replaced with our base of R14. Figure 1.4 gives a simple example of data that can be calculated using these methods. Figure 1.4A represents the data that are most available to researchers as the total number of days in which there was co-occurrence and non co-occurrence of sexually 19 mature individuals. In the example, species X mates on days 1 through 20, while species 0 mates on days 15 through 34. This results in an overlap of 6 days in which both species co-occur. Using these simplest data available and form of the equation (R17), an isolation index of 0.7 would be reached for species X. If the abundance of individuals on a given date are available (e. g. number of reproductive adults on each day), a more precise estimate can be achieved using R18 or R19, depending on the equivalency of the overall abundance of the two species. When the overall abundances of the two species are the same, temporal isolation for species X would be calculated as R18 = 0.81 (Figure 1.4B). However, if species 0 is twice as abundant as species X, then the global expectation of heterospecific mating under the null model is no longer 50%. Instead, in reference to species X, we expect that overall it will mate heterospecifically 2/3 of the time if mating is random; i.e., expected H = 0.67 and expected C = 0.33. Using R19, we most accurately represent isolation as 0.87 (Figure 1.4C). Geographic isolation is another form of isolation that affects co-occurrence of species, and should be estimated using equations R17, R18 or R19 depending on the data available. Geographic isolation can be incorporated into the biological species concept (Mayr 1942) by estimating the portion of geographic isolation that is due to intrinsic biological differences between organisms (Sobel et al. 2010). This form of isolation is very similar mathematically to the temporal case presented above, with only some additional complexity added by an extra dimension. As above, for simplification it is assumed that there is a shared region of overlap in which species can interbreed freely and a region of non-overlap in which no interbreeding occurs (though in reality this may 20 occur infrequently due to long distance dispersal). As an example, consider two species in which geographic distributions are based on intrinsic genetic differences, such that ecogeographic isolation exists (Sobel et al. 2010). For simplification, species X starts with 1000 individuals spread evenly across its geographic distribution, and each individual produces one offspring. If the ecogeographic extent of species X overlaps species 0 by 20% (i.e. 1/5 of the ecogeographic range of species X is shared with 0 and 4/5 is unshared), then after reproduction, 800 offspring will be species X from the offspring produced in the unshared portion of the range. In the shared portion random mating would produce 100 species X offspring and 100 X/O hybrids. In total 100 offspring are X/O hybrids, representing a 80% reduction in heterospecific offspring relative to a null expectation, and 900 offspring are species X, representing an 80% increase in conspecific offspring. Therefore, isolation is equal to 0.8, which is the value obtained when using R17. Of course, the assumption that all individuals will be evenly distributed across the range of a species will often be violated. Therefore, to employ equations R18 or R19, relative abundances across multiple contact zones can be incorporated into the calculation. Summag; Many would agree that one of the ultimate goals of speciation research is to uncover which forms of reproductive isolation are most important to the process. Ideally, we want to know what the relative contributions of each barrier to total reproductive isolation are at the exact moment of speciation. Different researchers may approach this problem in various ways, by either trying to understand the evolution of specific forms of 21 reproductive isolation through time or by focusing on differences among the strength of barriers experienced by recently diverged species. Regardless, estimating the strength of reproductive isolation is an important part of any approach to understanding the origin of species, but no single method for making the calculations necessary has emerged. We here provide a simple linear solution with many advantages for estimating the strength of individual barriers. While the proposed method does not differ significantly over much of the useable range of the metric, it is both simple to understand (in its base form) and easily related to differences in the probability of heterospecific gene flow. It can also be used in cases that depart from the general 50:50 null expectation, allowing the most accurate measures of isolation to be estimated. The proposed method also enables the incorporation of values below zero that previously had been discarded because of the asymmetry of previous equations. It therefore allows researchers the opportunity to make comparisons among forms of isolation and taxa by providing metrics that have equivalent effects on gene flow. R14 provides a framework in which to place estimates of the probability of heterospecific gene flow, but calculating this probability can sometimes be challenging. The estimate for each form of isolation can be refined further by adjusting the equation to non-equivalent null expectations when appropriate data exist. (R16, R19). Differences from equivalency can occur by a number of factors including differences in relative abundance, gamete abundance, mating frequency, etc. By selecting the appropriate form of these equations, future workers can be assured of producing estimates of isolation that are as accurate as possible while allowing direct comparisons with other estimates. 22 SOON 300000505 90385 $033K 03338.? 3.58003. .30 .x. .30 30 coo—moEv :EI n 030me05 $038 03858,? _Emm0000m .30 o\.. I $0338 0330—? 3380003 .30 .x. I E ASON WE I 308000.385 w8308v And Iv 00830 80388 000:: $030.8 .305 I 5 I E .50 30 MES—«C 030M555 038 0360588000: 3:? @0308 .305 E €030=008 0003 .8803 00.88 23 ‘3 008m 80% .30 8038508 AoooN 0803: E ... 8308:0985 30:00 0088 23 .3 3003 £003 .30 803.8508 I 5 I E SOON 008205 nE I Ab=B338008V 80303805 03330—50 025 0.3 03335 88083 0w80>m I _ I E 300 .8835 03095585 335.350 2.53 23 .30 $083 0w80>m x N E I A038 0033303355 @8 3393 .30 003.8508 I G AoooN 0.0085 E I 030385 000m 2.53 .30 003.8508 5 E SOON 30x82 EE m I 35303 3.555 , N I E 3:0 0:05 _ 030w>N3m05 Ame—«8 085 0.3.55 :0_toa08\m0_08 3393 0.3.5.3 00300505 I o CboNv 03:3 EE I A030 .083 $30305 @0308 03603080: 303005x0\30>.~0m30 I _ I E 08“ 03.82 0309305 %0308 03605000030: 003005x0\00>00m30 8:85 99: ”$2 <2 2a 0:53., area: :0 05.. 2:80 88888 _ a who .2 .80 .o I E AMOONV E + U I I N @5033 00:83—03 3:03 w8w805 .30 .8388 5803 I .50 30 58805 E 5 E 030335 0803 w8w805 880% - 80.6 50 .8988 5 E 352 E + D m. 308000.308 @0385 &:308 0830080: .30 00388 + $0308 03.0080: 50 .8388 ..a 30 .0508 EN 5 E 0309305 @8308 03500030: .30 0038:: I @8308 0353080: .30 .8388 E :02 Mowa N I 5 I NE 900000505 @8385 $0308 036058803 .30 50880005 I _ I E to 000 08508 E 030me05 @0308 03500580030: .30 5080:5003 80:83:00 803.83 000000.305 28mg? meg—0mm .30 053. 803835 .0035009800 03 8058 U 0:33 3030 80300305 $353.30 $8308 036058.830: .wdv 003500588030: 50 300500 8a 03 80.38 E @8808 003208 330530.52 330—0030 .55 80.308 330832 A; 03.? 23 ,_ H “\“ — R11 - 1- _— UN) ‘\“‘ C . -‘ ‘x‘ H 0 ~\ ----- R1, -1- _— ‘~“ -' 4'” ‘~\ 2H m _ ‘\~~ _ R15 :- l — — C ‘‘‘‘ +H .2 ‘s“ iii to \“ '5 N - .9 ° 2 -..: o I I I I I I I I 0 8 In 0 “l — I— Q, c? 0) CE I!) o- — 0 IO N. _ C? -1 I I I I I I I I I O 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Probability of heterospecific gene flow Figure 1.1. Relationship between the probability of heterospecific gene flow and the measure of reproductive isolation obtained through three commonly used methods. R11 ranges from 1 at complete isolation, through 0 at random mating (probability of gene flow = 0.5), and to -00 when gene flow is facilitated. R12 ranges from 1 at complete isolation to 0.5 at random mating and 0 when gene flow is facilitate. R15 (and R13) range from 1 at complete isolation, O at random mating, and -1 when gene flow is facilitated. 24 10$ Species O 0.8 0 m in . . . i 3:: atign freq. conspecific mating = 0.9 freq. conspecmc mating = 0.5 heterospecific = 0.1 heterospecific = 0.5 reproduction 90 Species O 50 Species O 10 cm hybrids 50 01x hybrids O 0.8 postzygotic . . . . isolation survnval consp. offspring = 0.9 survuval consp. offspring = 0.9 survival hetero. offspring = 0.9 survival hetero. offspring = 0.1 result 81 Species O 45 Species X 9 CM hybrids 5 0/x hybrids 80% more conspecific offspring 80% more conspecific offspring than expected by chance: than expected by chance: impact on 90 total offspring, null exp. is 50 total offspring, null expectation is gene flow 45 species 0 I45 O/X hybrids 25 species 0 / 25 O/X hybrids 45*0.8=36 25*O.8=20 (=excess # consp. offspring (excess # conspecific offspring or reduction of heterospecific) or reduction of heterospecific) Figure 1.2. Diagram showing similarities between pre- and postzygotic isolation index measures of equal magnitude (R1 = 0.8). In both A & B, the population begins with ten females, and each produces 10 offspring. A) Species 0 females have strong mating preferences, mating with conspecific males with a frequency of 0.9 and heterospecific males of species X with a frequency of 0.1. Of the 100 offspring produced, 90 are pure species 0 and 10 are O/X hybrids. The isolation coefficient of 0.8 represents the proportional decrease in heterospecific offspring below a null expectation. B) Species 0 females have no mating preferences, but do exhibit strong postzygotic isolation such that the survival rate of conspecific offspring is 0.9 while the heterospecific offspring have a lower survival rate of 0.1. This results in an equivalent effect on gene flow with 80% fewer heterospecific offspring (i.e. gene flow) than would have been produced under the null model. 25 0.75 0.5 l 0.9 0.8 0.25 0 0.5 0.4 -0.25 0.2 Reproductive isolation —0.5 J Exp H = 0.1 -0.75 ‘T l T l i i i i i i 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Observed frequency of heterospecific gene flow Figure 1.3. The effect of changes in expected frequency of hybrid formation on the value of reproductive isolation using equation R16. Each line represents a different expectation of hybrid formation (due to differences in relative abundance for example) ranging from 0.1 to 0.9 (labeled below each line). The point at which the expected H line intersects the x-axis is the point at which observed heterospecific gene flow equals expected heterospecific gene flow; therefore reproductive isolation equals zero. The thick line represents the point of equal expectation of con— and heterospecific gene flow, which is equivalent to R15 from Figure 1.1. 26 Figure 1.4. Hypothetical data showing the relationship between three methods for calculating the strength of temporal isolation. Days of reproduction is on the x-axis and number of reproductive individuals is on the y-axis (mirrored). A) Only data on days of reproduction is available. Species X has sexually mature individuals on days 1 to 20 while species 0 is mature on days 15 to 34. Given the data available, the most simple equation (R17) is used, yielding an isolation coefficient of 0.7 for species X. B) If data on relative abundance throughout the mating season are available, and both species have roughly equivalent relative numbers, the estimate of isolation can be refined by using equation R18, giving an isolation index of 0.81 for species X. C) When species X and 0 exhibit consistent differences in relative abundance, the null expectation under random mating can be adjusted using equation R19, resulting in an isolation coefficient of 0.87. 27 20‘ 10" species X —_ _r l _ species 0 hu— 10" 20‘ 30‘ 40" 50‘ 60‘ 28 Literature Cited Ackermann, M., M. Achatz, and M. Weigend. 2008. Hybridization and crossability in Caiophora (Loasaceae subfam. Loasoideae): Are interfertile species and inbred populations results of a recent radiation? American Journal of Botany 95:1109- 1121. Bolnick, D. I., and T. J. Near. 2005. Tempo of hybrid inviability in centrarchid fishes (Teleostei: Centrarchidae). Evolution 59: 1 754-1767. Bono, J. M., and T. A. Markow. 2009. Post-zygotic isolation in cactophilic Drosophila: larval viability and adult life-history traits of D. mojavensis/D. arizonae hybrids. Journal of Evolutionary Biology 22:1387-1395. Brock, M. T. 2009. Prezygotic barriers to gene flow between T araxacum ceratophorum and the invasive Taraxacum officinale (Asteraceae). Oecologia 161:241-251. Cayol, J. P., J. Vilardi, E. Rial, and M. T. Vera. 1999. New indices and method to measure the sexual compatibility and mating performance of Ceratitis capitata (Diptera : Tephritidae) laboratory-reared strains under field cage conditions. Journal of Economic Entomology 92:140-145. Coyne, J. A., S. Elwyn, and E. Rolan-Alvarez. 2005. Impact of experimental design on Drosophila sexual isolation studies: Direct effects and comparison to field hybridization data. Evolution 59:2588-2601. Coyne, J. A., and H. A. Orr. 1989. Patterns of speciation in Drosophila. Evolution 43:362-381. —. 1997. "Patterns of speciation in Drosophila" revisited. Evolution 51:295-303. —. 2004, Speciation. Sunderland, MA, Sinauer Associates. Dobzhansky, T. 1937, Genetics and the Origin of Species. New York, Columbia University Press. Dobzhansky, T., and P. C. Koller. 1938. An experimental study of sexual isolation in Drosophila. Biologisches Zentralblatt 58:589-607. Dopman, E. B., P. S. Robbins, and A. Seaman. 2010. Components of reproductive isolation between North American pheromone strains of the European corn borer. Evolution 64:881-902. Ehrman, L. 1965. Direct observation of sexual isolation between allopatric and between sympatric strains of the different Drosophila paulistorum Races. Evolution 19:459-464. 29 Funk, D. J., P. Nosil, and W. J. Etges. 2006. Ecological divergence exhibits consistently positive associations with reproductive isolation across disparate taxa. Proceedings of the National Academy of Sciences of the United States of America 10313209-3213. Hosken, D. J., O. Y. Martin, S. Wigby, T. Chapman, and D. J. Hodgson. 2009. Sexual conflict and reproductive isolation in flies. Biology Letters 5:697-699. Lowry, D. B., J. L. Modliszewski, K. M. Wright, C. A. Wu, and J. H. Willis. 2008. The strength and genetic basis of reproductive isolating barriers in flowering plants. Philosophical Transactions of the Royal Society B-Biological Sciences 363:3009- 3021. Martin, N. H., and J. H. Willis. 2007. Ecological divergence associated with mating system causes nearly complete reproductive isolation between sympatric Mimulus species. Evolution 61:68-82. Matsubayashi, K. W., and H. Katakura. 2009. Contribution of multiple isolating barriers to reproductive isolation between a pair of phytophagous ladybird. beetles. Evolution 63: 2563- 2580. Mayr, E. 1942, Systematics and the origin of species. New York, Columbia University Press. Mendelson, T. C. 2003. Sexual isolation evolves faster than hybrid inviability in a diverse and sexually dimorphic genus of fish (Percidae: Etheostoma). Evolution 57:317- 327. Moyle, L. C., M. S. Olson, and P. Tiffin. 2004. Patterns of reproductive isolation in three angiosperm genera. Evolution 58: 1 195-1208. Palmer, M. E., and M. W. F eldman. 2009. Dynamics of hybrid incompatibility in gene networks in a constant environment. Evolution 63:418-431. Poulton, E. G. 1908. What is a species? Pages 46-94 Essays on Evolution 1889-1907. London, Oxford at the Clarendon Press. Presgraves, D. C. 2002. Patterns of postzygotic isolation in Lepidoptera. Evolution 56:1168-1183. Price, T. D., and M. M. Bouvier. 2002. The evolution of F -1 postzygotic incompatibilities in birds. Evolution 56:2083-2089. 30 Ramsey, J ., H. D. Bradshaw, and D. W. Schemske. 2003. Components of reproductive isolation between the monkeyflowers Mimulus lewisii and M. cardinalis (Phrymaceae). Evolution 57: 1520-1534. Rice, W. R., and E. E. Hostert. 1993. Laboratory experiments on speciation: What have we learned in 40 years? Evolution 47:1637-1653. Ruane, L. G. 2009. Post-pollination processes and non-random mating among compatible mates. Evolutionary Ecology Research 11:1031-1051. Sasa, M. M., P. T. Chippindale, and N. A. Johnson. 1998. Patterns of postzygotic isolation in frogs. Evolution 52: 1 81 1-1820. Schemske, D. W. 2000. Understanding the origin of species. Evolution 54:1069-1073. Sobel, J. M., G. F. Chen, L. R. Watt, and D. W. Schemske. 2010. The biology of speciation. Evolution 642295-315. Stalker, H. D. 1942. Sexual isolation studies in the species complex Drosophila virilis. Genetics 27:238-257. Takami, Y., N. Nagata, M. Sasabe, and T. Sota. 2007. Asymmetry in reproductive isolation and its effect on directional mitochondrial introgression in the parapatric ground beetles Carabus yamato and C. albrechti. Population Ecology 49:337-346. Taylor, S. J ., M. Arnold, and N. H. Martin. 2009. The genetic architecture of reproductive isolation in Louisiana irises: Hybrid fitness in nature. Evolution 63:2581-2594. Tilley, S. G., P. A. Verrell, and S. J. Arnold. 1990. Correspondence between sexual isolation and allozyme differentiation: A test in the salamander Desmognathus ochrophaeus. Proceedings of the National Academy of Sciences of the United States of America 87:2715-2719. 31 CHAPTER 2: ECOGEOGRAPHIC ISOLATION AMONG RECENTLY DIVERGED SPECIES PAIRS IN THE GENUS MIMUL US 32 CHAPTER 2: ECOGEOGRAPHIC ISOLATION AMONG RECENTLY DIVERGED SPECIES PAIRS IN THE GENUS MIM UL US Abstract Despite a long history of examining the geographic context of speciation, evolutionary biologists rarely consider differences in geographic range as a legitimate form of reproductive isolation. This stems mainly from the fact that geographic ranges are both the product of historic and ecological processes. The presented work provides a method for estimating the ecological aspects of geographic isolation, bringing this potentially important form of isolation under the umbrella of the biological species concept. Ecological niche modeling was used to evaluate the potential ranges of twelve pairs of species in the genus Mimulus. There was substantial variation in which environmental variables used to construct niche models were most important, with aspects of geology and rainfall providing the most explanatory power. Examination of differences in niche models generated for species pairs showed that recently diverged species of Mimulus differ significantly in the niches they occupy. These differences in niche translate into strong ecogeographic isolation, with an average strength of at least 0.64. Given the position of this barrier in the order of sequentially acting reproductive barriers, this value leads to the conclusion that ecogeographic isolation will be the highest contributor to total isolation among all other forms of reproductive isolation. 33 Introduction Identifying the mechanisms of speciation is an area of considerable importance in evolutionary biology, and one of the most debated topics in speciation biology is the relative impact of different forms of reproductive isolation (Coyne and Orr 2004). This debate typically simplifies into an argument of whether barriers that act prior to fertilization (prezygotic) or after fertilization (postzygotic) are more important (Coyne and Orr 2004; Poulton 1908; Rice and Hostert 1993; Schemske 2000). Despite the fact that a complete inventory of potential reproductive barriers has been recognized since the modern synthesis (Dobzhansky 1937; Mayr 1942), several forms of reproductive isolation have been essentially ignored in this debate. In particular, the role of geographic separation in limiting gene flow has been largely ignored, with very few estimates of its potential impact on gene flow (but see Kay 2006; Kirkpatrick and Ravigne 2002; Ramsey et al. 2003). Although its effects on gene flow are rarely measured as an actual prezygotic barrier, the geography of speciation is far from an unexplored topic. Jordan first recognized the role of geography in generating species formulating the rule, “Given any species, in any region, the nearest related species is not to be found in the same region nor in a remote region, but in a neighboring district separated fi'om the first by a barrier of some sort or at least by a belt of country, the breadth of which gives the effect of a barrier. " (Jordan 1908; Jordan and Kellogg 1907, p. 120) Indeed, much of the speciation research in the past several decades has been squarely focused on whether Jordan’s rule is correct, asking if allopatry is a requirement for speciation (Bolnick 2004; Bolnick and Fitzpatrick 2007; Coyne and Orr 2004; Feder et al. 34 2005; Mayr 1947; Nosil 2008; Turelli et al. 2001; Via 2001). Some authors have measured the degree of overlap in geographic space occupied by recently diverged species (Barraclough and Vogler 2000; Fitzpatrick and Turelli 2006), but almost exclusively within the context of determining whether allopatric or sympatric speciation is most common. Despite this substantial attention, most workers have treated allopatric separation itself as a completely historical process with very little regard for how the biology of organisms might be involved in creating or maintaining allopatry (Wiens 2004). This has resulted in geographic isolation being considered a potent part of speciation, but mostly as a force that allows the evolution of additional forms of isolation rather than as a legitimate form of isolation itself. Historically, several authors pointed out the need to consider geographic isolation as a barrier per se, though often with some trepidation. For example, Stebbins notes that Jordan’s rule is often correct, but adds that closely related species “... are usually separated from their relatives by ecological barriers as well as by geographic ones (Stebbins 1950, p. 197).” Dobzhansky also noted that, “... the occupation of separate areas by two species may be due not only to the fact that they have developed there, but also to the presence of physiological characteristics that make each species attached to the environment...(Dobzhansky 1937, p. 231).” These views suggest a revision to Jordan’s rule that might read, closely related species are most often near each other but separated by a geological barrier, and also exhibit ecological divergence that may enhance the fidelity of each to its home range. As argued in Sobel et al. (2010), an examination of the definition of reproductive barriers under the Biological Species Concept (BSC) makes it possible to reconcile issues 35 involved in treating geographic isolation as a legitimate prezygotic barrier. The BSC identifies reproductive barriers as biological differences between populations that actually or potentially limit gene flow (Mayr I942; Mayr 1963; Mayr 1984). Therefore, the aspects of geographic isolation that are based on historical factors and not biological differences are not legitimate barriers, but geographic separation that arises (or is maintained) by intrinsic biological differences between populations can and should be measured as a component of prezygotic isolation, which is referred to as ecogeographic isolation (Sobel et al. 2010). Several advances have made it possible to explore the relationship between history and biology in deconstructing their relative role in geographic isolation. The first is the availability of species—level molecular phylogenetic hypotheses for many groups of organisms (Barraclough and Vogler 2000; Graham et al. 2004b), providing an objective identification of species pairs that have recently diverged. Another is GPS (geographic positioning system) technology for collecting detailed spatial information on the distribution of species, and the computing power required to manage large databases of geographic information (GIS technology). Combining these advances with the availability of museum and herbaria records of species collections (Graham et al. 2004a), the ability to catalog and analyze the spatial patterns of speciation have grown immensely in recent years. Ecological niche modeling has emerged as a powerful method for examining the climatic tolerances and geographic ranges of species (Ortega-Huerta and Peterson 2008), identifying suitable habitat for species in under-sampled regions (Jarnevich et al. 2006), and predicting the outcome of species invasions (Peterson 2003). In addition, it has been 36 recently adopted by evolutionary biologists (Kozak et al. 2008), and has been applied to questions of whether ecological niche conservatism (Kozak and Wiens 2006) or divergence (Nakazato et al. 2008; Nakazato et al. 2010; Warren et a1. 2008) are general components of speciation. This method has the potential to also aid in deconstructing the historical and biological components that determine geographic separation between species (Sobel et a1. 2010). Because ecological niche models show both where a species does live and where it can live, it provides an estimate of both the actual and potential range of a species. By overlaying individual niche models of a pair of closely related species, it is therefore possible to simultaneously assess the actual and potential gene flow limited by the geographic range occupied by species. Indeed, the overlap of two species’ spatially projected ecological niche model provides an estimate of how much physical space in a region that two species are capable of co-occurring over. This provides an estimate of the degree to which adaptation to different habitats results in actually and/or potentially geographically distinct ranges, and allows a measure of geographic isolation relevant under the BSC called ecogeographic isolation. In the study presented, ecological niche modeling was applied to a diverse group of western North American wildflowers in the genus Mimulus. Ecological niche models were developed for 12 pairs of species that represent recent diverged speciation events across the genus, and overlap in niche models was used to estimate the relative contribution of this potentially important barrier. 37 Materials & Methods Study system The genus Mimulus consists of approximately 120 species of wildflowers, approximately 75% of which are found in western North America (Grant 1924). The genus exhibits a broad diversity in interesting ecological and evolutionary traits including: growth form (Lowry et al. 2008), mating systems (F ishman et al. 2002; Willis 1993), ploidy levels (Beardsley et a1. 2004; Vickery 1995), pollination syndromes (Bradshaw and Schemske 2003; Bradshaw et al. 1995; Schemske and Bradshaw 1999; Streisfeld and Kohn 2005), edaphic specialization (Macnair 1983), and climatic tolerances (Angert and Schemske 2005). Previous phylogenetic work has elaborated relationships at broad taxonomic levels (Beardsley and Olmstead 2002), and a species- level phylogeny is available with nearly complete coverage of the North American species (Beardsley et al. 2004). A rich history of ecological and genetic work (Hiesey et al. 1971; Kiang 1972; Vickery 1959; Vickery 1964), combined with more recent advances in developing genomic resources (Wu et al. 2008) has elevated the genus to a model system in ecology and evolutionary biology, especially in studies of speciation (Martin and Willis 2007; Ramsey et al. 2003). Selection of species pairs The phylogeny presented in Beardsley et al. (2004) was used to aid species selection. For inclusion, species pairs exhibited recent speciation events and had high bootstrap support fiom the molecular phylogenetic hypothesis presented. In order to verify that the species pairs selected for inclusion were more closely related to each other 38 than to any other species in the study, sequence data (trnL/F, ITS, E TS) presented in Beardsley et al. (2004) were reanalyzed. A maximum likelihood phylogenetic analysis was performed in PHYLIP (Felsenstein 2005), and the majority rule consensus tree supports the relationships drawn from the genus-wide phylogeny (Figure 2.1). In most cases, these species pairs represent each other’s closest extant relative; however, in some cases alignment with species used in a broader study of reproductive barriers (Sobel, Chapter 3) necessitated choosing species pairs where another species may be more closely related to one of the included species. However, each pair of species used is more closely related to each other than any other species in the study (Figure 2.1), ensuring phylogenetic independence of reproductive isolation measured between each pair (F elsenstein 1985). In total, 12 pairs of closely related species were selected for inclusion (Table 2.1), spanning the phylogenetic breadth of the genus with members fi'om traditional morphological taxonomic sections Eunanus, Erythranthe, Oenoe, and Paradanthus (Grant 1924). Collection of georeferenced localifl Qua Georeferenced locality data were collected for each species using the Jepson Online Interchange for California Floristics (UC-JEPSZOO4). This database consists of specimen records of California plants compiled from seventeen different herbaria across the state, providing an unparalleled collection of Mimulus records. Recent accessions in the database are often georeferenced, but a large number of specimen records from older collections lack this detail. Using a combination of cartographic resources including Google Earth (http://earth.goog1e.com) and National Geographic TOPO! software 39 (National Geographic TOPO! Maps, San Francisco, CA; http://maps.nationalgeographic.com/topo), latitude and longitude coordinates were added to samples without this data, resulting in over 2000 georeferenced accessions at an average of 84 specimens per species (Table 2.1). Selection and manipulation of environmental layers Eight climatic variables were selected from the publicly available dataset WORLDCLIM (http://worldclim.org) that represented a mix of both temperature and precipitation variables (Table 2.2). These climatic layers consist of a grid of 1km2 resolution pixels that span the globe. Variables were selected to minimize correlations among climatic layers within the study area; however, correlations remain high among some of the variables (all are <0.9, Table 2.3). In addition, a geological map of California was obtained from the US. Geologic Survey (http://usgs. gov). This layer consists of a qualitative assessment of basic geologic parent material broken into 51 categories (e. g. sandstone, granodiorite, serpentine, etc.). All environmental variables were processed and visualized using ArcGIS 9.2 (ESRI, Redlands, California, USA). The original polygon geological shapefile was converted to a raster layer with the same spatial extent and resolution as the WORLDCLIM climatic variables using the ‘Polygon to Raster’ conversion tool within ArcGIS. Cells that contained multiple values were assigned using the maximum area criterion. Each WORLDCLIM layer was clipped to limit the spatial extent of analysis to the state of California in order to align with the geologic map. Limiting analysis to California is also justifiable given that the vast majority of accessions in the Jepson Interchange are from collections within the state, and 40 extrapolating to outside regions could prove problematic. Prior to analysis, any species occurrence data collected under different coordinate systems (e. g. NAD27, UTM, etc.) were converted to a common datum (WGS84) to match the environmental layers. Ecological niche modeling Ecological niche models (ENMs) were constructed for each species using Maxent software (version 3.3.1; http://cs.princeton.edu/~shapire/maxent) (Phillips et al. 2006). Maxent utilizes a machine-learning algorithm which employs the general principal that when estimating an unknown probability distribution, the best approximation will satisfy all known constraints on the distribution without imposing unfounded ones (Jaynes 195 7). When applied to modeling the distribution of species, the approach therefore finds the probability distribution of maximum uniformity (entropy) while satisfying known constraints imposed by the observed distribution of species and environmental variables across the study area. The approach has been consistently shown to offer improved ability to accurately predict the ranges of species over previously employed methods such as GARP or GLMs (Elith et al. 2006). It is especially useful in situations where information on species occurrences is readily available but data on absences is not (Elith et al. 2006; Phillips et al. 2006), and has been utilized in a number of recent speciation studies using museum and herbarium records (e. g. Kozak and Wiens 2006; Nakazato et a1. 2010). Common species For species with greater than 30 records, Maxent ENMs were generated using default settings of 500 maximum iterations and convergence threshold of 0.0000]. The 41 random test percentage was set to 50%, so that a random half of the data was used to train the model and the other half was used to test the model. The raw Maxent output is a grid of pixels with the same spatial extent as the inputted environmental variables, where each pixel in the grid is assigned a non-negative probability which sums to 1 over all pixels. Given the large spatial extent of analysis (state of California), the values in each pixel are extremely small. Therefore, a cumulative representation was selected that results in each pixel carrying a value equal to the sum of that pixel and all other pixels of equal or lesser probability. The sum is multiplied by 100 to give a percentage, which can be thought of as a relative suitability score for each pixel. Jackknifes were performed in Maxent to measure the relative importance of each environmental variable to the predicted model. In addition, the threshold independent test indicator, AUC (area under the curve) of the ROC (receiver operator characteristic), was recorded as an indicator of Maxent performance in predicting niches. Another test of Maxent niche modeling performance was omission rates of test samples based on delineated thresholds of the cumulative probability value (representing the cutoff for suitable and unsuitable habitat). Establishing threshold levels appropriate for producing binary suitable/unsuitable maps for species is an area of active research (Liu et al. 2005). Different threshold criteria have been shown to be optimal under different conditions, so to assess the performance of Maxent in producing Mimulus ENMs, two thresholds were established and the relative success rate of each was compared. The first threshold represents a maximized sum of the sensitivity and specificity values (Liu et al. 2005), which has been shown to perform relatively well compared to other methods and has been employed in recent studies of speciation (e. g. 42 Nakazato et al. 2010). This threshold attempts to balance the incidence of true positives (test species occurrence points assigned to suitable habitat) with the incidence of correctly identified negatives. Maxent simulates negative calls based on a random sampling of the background of the area of study, so correct negative calls are not actually based on known absences. The second threshold used is a cutoff of the 10th percentile of cumulative probability values that occur in pixels occupied by species occurrence records. While arbitrary, this cutoff has been shown to be an effective predictor of distributions (Pearson et al. 2007) and has also been used in recent studies of speciation (e. g. Kozak and Wiens 2006). In both cases, threshold performance was assessed by examining the rate of omission of test samples from the trained model. Rare species Species with highly restricted geographic ranges are common in Mimulus; therefore, there are several species in this study with very few collection records (Table 2.1). One of the benefits of using the Maxent platform is that it has been shown to be effective at very low sample sizes (Papes and Gaubert 2007; Pearson et al. 2007; Wisz et al. 2008), with reasonable performance with samples as few as 5 (Pearson et al. 2007). However, the 50% train/test method described above is not appropriate for these situations due to the limited number of samples. Instead following methods described by Pearson et al. (2007) a ‘jackknife minus 1’ cross validation technique was used for any species with fewer than 30 accessions. In brief, this method involves dividing the species occurrence data into N (# of occurrence points) subsamples, removing a single occurrence for each group. The N-l portion of each group was used as training data and 43 the single occurrence removed was used to test the model (repeated N times so that every point is removed from one subsample). The success rate of this method was assessed by setting a threshold and recording the number of times the model successfully predicted suitable habitat for the test occurrence point. Two thresholds were selected for this analysis. The first is a conservative measure based on the minimum cumulative probability value present in the training data. The second is a subjective cutoff of the 10th percentile of the values present in the training data. Pearson et al (2007) provides a method and software for using the area of suitable habitat predicted under each threshold in a chi-square test for significant departure from randomness, and this method was used for statistical validation of model performance. As in the common species analysis presented above, AUC of the ROC (averaged across all subsample runs) was recorded as an indicator of Maxent performance in predicting niches. Environmental variables were jackknifed in Maxent to estimate the relative contribution of environmental variables to each ecological niche model. Comparing niche models between recently diverged species pairs Comparisons were made between species pairs by both contrasting the ecological niche model output of species pairs and by examining the environmental data associated with pixels containing species occurrence records. Ecological niche models were compared using the software ENMTools (http://enmtools.blogspot.com) (Warren et al. 2008). For each species pair, the niche overlap tool was used to first calculate a value of ecological niche similarity based upon both Schoener’s D (Schoener 1968) and the 1 statistic described in Warren et al. (2008). The niche identity tool was used to randomly 44 assign species identity to the list of known localities for a species pair, and a Maxent ecological niche model was generated for each replicate. This was performed 100 times and a Schoener’s D and 1 statistic were calculated for each iteration, generating a null distribution of niche overlap. Schoener’s D and 1 statistics were compared to this null distribution to assess whether ecological niche models in species pairs of Mimulus differed from a chance expectation. ENMTools is not currently capable of handling categorical environmental variables, so this analysis was performed on only the WORLDCLIM set of climatic layers (Table 2.2). Additional tests to examine the differences between species pairs were performed by extracting the environmental variable values associated with pixels containing species records. For each species, the ‘Extraction-Sample’ tool within the Spatial Analyst package of ArcGIS was used to compile these data. One-way MANOVA (Wilks A) were conducted in R (R Development Core Team 2010) on the continuous environmental data extracted for each species pair to test for differences among the observed variables. Post- hoc analyses were performed on data from each pair with an a corrected for multiple comparisons using a Bonferroni adjustment (or = 0.00556). Because many environmental variables were not normally distributed within species samples, the non-parametric Mann-Whitney U was employed to test for differences between the continuous WORLDCLIM variables. The categorical geological variable was tested using a contingency table analysis. Mann-Whitney U and contingency analyses were performed in JMP 8.0 (SAS Institute, Cary, North Carolina, USA). 45 Calculation of reproductive isolation In order to calculate overlap in ecological niche models suitable for estimating reproductive isolation, threshold values need to be established so that overlap in binary suitable/unsuitable habitat can be assessed. Niche models generated in Maxent were loaded into ArcGIS, and the ‘Raster Calculator’ within the Spatial Analyst package was used to create binary maps of habitat. This was achieved by constructing an expression equivalent to [ecological niche model for species X] > threshold value for species X. This was done for both threshold values described above for testing the performance of niche modeling in Maxent. Threshold 1 is the maximized sum of sensitivity and specificity for ‘common’ species and is the minimum cumulative value present in training data for ‘rare’ species. Threshold 2 is the 10th percentile of cumulative probability values present within the training data for both common and rare species. The output of the raster calculator is a new raster layer with a value of 1 assigned to every pixel of suitable habitat and a 0 assigned to pixels of unsuitable habitat. This was done separately for both species 1 and 2 of a given pair. Species 2 of the pair was added to itself using the raster calculator to obtain a new raster layer identical to the original except each pixel of suitable habitat was assigned a value of 2 and each unsuitable pixel retained a value of 0. To calculate the area of overlap in suitable habitat between species pairs, the doubled threshold raster layer of species 2 was added to the single threshold layer of species 1. The result is a raster layer with values of 0- unsuitable habitat, 1- suitable habitat for species 1, 2- suitable habitat for species 2, and 3- suitable habitat for both species 1 and 2. This was performed for all species pairs in the study, and the total number of pixels (area in kmz) of species 1, total number of pixels of species 2, and number of shared pixels was recorded. Reproductive 46 isolation was calculated using the methods described in Chapter 1 using equation R17. Reproductive isolation was calculated asymmetrically for each species as: Shared Area x + y Rlspecies x = 1' - (Shared Area x + y) + Unshared Area x Reproductive isolation for both species was recorded, and an average between the two species was calculated. Isolation was estimated using both methods of threshold value definition giving a range of potential isolation depending on the conservativeness of the threshold employed. Genetic distance and degree of overlap in ecological niche models Estimates of sequence divergence were extracted from the phylogenetic analysis presented in Figure 2.1. Under the assumption that substitutions occur at a clock-like rate, the level of sequence divergence can be used as a proxy for the amount of time that has passed since divergence in each species pair. In order to examine if time since divergence affects the amount of overlap species pairs experience, a correlation was performed using JMP 8.0 (SAS Institute, Cary, North Carolina, USA). Congation between niche differentiation and ecogeographic isolation Because the geographic consequence of differences in ecological niche can vary tremendously depending on the geographic arrangement of suitable habitat (Sobel et al. 2010), the relationship between niche differentiation and ecogeographic isolation is expected to vary. To test for this, a correlation was performed in JMP 8.0 (SAS Institute, Cary, North Carolina, USA) between the levels of niche differentiation calculated in the 47 Schoener’s D identity test (Figures 2.3A) and the level of ecogeographic isolation calculated for each species obtained using the 10th percentile threshold (Table 2.5). M Ecological niche models Table 2.1 provides a summary of data from constructing the ecological niche models. Maxent performed reasonably well in producing models with an average AUC score of 0.95 (lowest was 0.85). For common species, threshold 1 (maximized sensitivity + specificity) correctly placed test occurrence points in suitable habitat an average of 89.7% (i6.5), and binomial probabilities at this threshold departed from the null expectation for all species (p < 10'1 1). At threshold 2 (10th percentile of training data), 75.6% (i194) of test points were correctly placed in suitable habitat and binomial probabilities at this threshold were better than null as well (p < 10.1 I). For rare species, threshold 1 (minimtun training presence) predicted the test point accurately 86.9% (i9.3), and Pearson et al’s (2007) method of comparing this prediction rate to null resulted in p < 0.0001 in all cases. Threshold 2 performed slightly better on average with a success rate of 82.7% (21:72), and p < 0.0001 in all cases as well. Figure 2.2A-X shows the continuous cumulative probability maps re-projected onto the spatial extent of the study area. The shading relates to the values of the cumulative probability distribution with lighter shades and white representing the highest-ranking suitable habitat. Figures are presented in same order as Table 2.1 such that species pairs are adjacent. Ecological niche models were predicted by Maxent with high confidence (Table 2.1), indicating that the environmental variables included in this study have significant 48 impacts on the geographic ranges of species in the genus Mimulus. The environmental variables that were most important in building ecological niche models varied considerably across species (see circles in Figure 2.2A-L). In general, precipitation layers (B1012, 15, l6, 17) had higher relative contributions than temperature with an average of 12.5% (i139, range 0, 64.7) averaged across all four layers, versus 5.6% ($9.6, range 0, 67.2) for the temperature layers (B101, 4, 5, 6). Of the continuous climatic variables, B1012 (mean annual precipitation) had the highest average contribution with 15.4% (£5.89, range 001-512). The categorical geology variable had the highest relative contribution with and average of 27.7% ($13.8, range 10, 57.2). It was the most important variable in 8 of the 24 species and was the second most important variable in 10 additional species. Comparison of niche models of species pairs MANOVA analysis revealed that environmental variables differed between species pairs (Table 2.4), indicating that substantial environmental differences between species. Figure 2.2A-L illustrates these differences and provides the results of post-hoe Mann-Whitney U (for continuous WORLDCLIM variables) and contingency table analysis (for categorical geology variable). While the analysis in the preceding section provides an indication of which variables were most important to building individual niche models, these analyses allow the identification of which layers differ significantly between species pairs. As with the relative contributions of layers presented above, the variables that are different between species exhibited considerable variation; however, all 49 species exhibited some niche differentiation with at least two environmental variables significantly different in all cases. Identity tests performed within ENMTools (Warren et al. 2008) corroborated the pattern of strong niche divergence. In the analysis of Schoener’s D, the actual measured value of D for each species pair was significantly different from the null distribution generated from randomizing species identities (p < 0.05 in all cases) (Figures 2.3A). In the analysis of the 1 niche identity statistic (Warren et al. 2008), all but one species pair also showed niches were less similar than expected under the null (Figures 2.3B). Substantial niche differentiation was detected between recently diverged species pairs in the genus, providing a link between speciation and ecological divergence that has been argued to be common in nature (Schluter 2001; Sobel et al. 2010). Niche divergence was detected by the MANOVA analysis in all species pairs (Table 2.4), and identity tests (Warren et al. 2008) indicated that species pairs had significantly different niches with the exception of species pair Mimulus cusickii / nanus in the 1 statistic test (Figures 2.3 A&B). However, M cusickii is a rare species with only 7 collection records used from the Jepson Online Interchange, and this lack of sample size most likely reduced the ability of the test to detect differences. Reproductive isolation Given that the above analyses showed significant divergence in ecological niche for closely related species pairs in Mimulus, the geographic consequence of these niche differences gives an estimate of the degree of ecogeographic reproductive isolation between these species. Figure 2.5A-L illustrates the amount of ecogeographic isolation 50 experienced between species pairs. These maps were generated based on the threshold 2 criterion described above because it was common to analyses of both common and rare species. Estimates of reproductive isolation obtained through measuring shared and non- shared areas for each species are given in Table 2.5, with values obtained using both threshold 1 and 2 presented. Results are given both asymmetrically (reproductive isolation experienced by species 2 with respect to species 2) and as an average composite of both species asymmetric scores. Reproductive isolation was quite strong in many cases with an average value of 0.64 to 0.69 depending on which threshold was used. Genetic distance and degree of overlap in ecological niche models Correlation analysis between ecogeographic isolation and sequence divergence showed no significant relationship (Figure 2.6; p = 0.9188). However, because these species pairs represent recent speciation events, the range of sequence divergence over which a relationship could be found was quite low. Correfirtion between niche differentiation and ecogeographic isolation Figure 2.7 shows the relationship between the degree of ecogeographic isolation and a measure of niche similarity, Schoener’s D. Regression analysis shows a significant negative correlation between the two variables (slope = -0.946, p=0.006) indicating that ecogeographic isolation rs highest in specres With the least Slmllal' niches. However, r rs only 0.547, revealing that variation in the spatial arrangement of suitable habitat results in significant variation in how much isolation results from niche differences. 51 Discussion Ecological niche modeling provides support for divergence in niches Strong levels of niche divergence between close relatives is consistent with previous studies using similar methodologies (e.g. Graham et al. 2004b; Nakazato et al. 2010), though others have argued that conservation of niches is also common (Kozak and Wiens 2006; Peterson et al. 1999). The ways in which species pairs differed was quite idiosyncratic. While precipitation level variables were the most important in constructing niche models for individual species, these were not any more likely to be significantly different in the posthoc tests than the temperature variables (Figure 2.2A-L). Many of the species pairs that exhibited strong niche divergence occupy different altitudes where both temperature and precipitation vary considerably (e.g. M. cardinalis/lewisii, M angustatus/pulchellus, M constrictus/whitneyi). As an example, Mimulus cardinalis and M. lewisii were significantly different (p < 0.0001) at every environmental variable used to construct the ecological niche models (Figure 2.2F), exhibiting one of the highest levels of niche divergence in the species pairs represented in this study. This corroborates extensive work into the nature of niche differences in this species pair using demographic (Angert 2006), reciprocal transplant (Angert and Schemske 2005), and experimental evolution approaches (Angert et al. 2008). Alternatively, very few species showed evidence of weak niche divergence. In one example, Mimulus gracilipes and M palmeri only differed significantly for two variables related to precipitation (B1012: average annual precip. and BlOl6: precip. of wettest quarter). However, as long as any aspects of niche are significantly different, a geographic consequence may be associated leading to reproductive isolation. 52 Correspondence between ecological niche models and intrinsic biological differences While indicative of biological differences, results from ecological niche modeling should be interpreted with caution (Guisan and Thuiller 2005). Differences among environmental variables at collection sites only indicate that environmental conditions are different in the places in which a species currently lives. For a given species, regions identified as suitable habitat may indeed be used to predict other regions with similar environmental conditions, as in predicting invasive species potential (e. g. Peterson 2003). However, for regions identified as unsuitable through ecological niche modeling, we only know that the species does not live in that type of habitat, not necessarily that it isn’t able to. This problem is especially acute for uncommon species with highly restricted ranges. Species with restricted ranges tend to have niche models with very tight parameter values, exhibiting only a very small range of potential habitat (e. g. Mimulusfilicaulis, Figure 2.2F). This may result from very tight ecological requirements of these species, or they may have small geographic ranges for other reasons, such as purely historical factors (Sexton et al. 2009). Even when geographic ranges do accurately represent the intrinsic tolerances of a species, it is not necessarily true that all environmental variables that appear important in niche modeling actually are. A species may actually be geographically restricted by only a subset of the variables that appear important from niche modeling. For example, Mimulus norrisii occurs only on limestone outcrops of the southern Sierra Nevada. This edaphic endemic has a highly restricted range based upon the occurrence of limestone outcrops in this region. The ecological niche modeling (Figure 2.2T, Figure 2.21) shows 53 that this species occurs over a very narrow range of climatic conditions, but that may simply be related to the range over which limestone outcrops exist. If limestone occurred across a wider range of climatic variation, M norrisii may or may not occupy a larger geographic and environmental range. In order to substantiate the findings of these ecological niche models, reciprocal transplants or growth chamber experiments are necessary. Because many species within the genus Mimulus are rare and potentially threatened (Beardsley et al. 2004), reciprocal transplants may be difficult or impossible in the majority of species pairs presented. Angert & Schemske (2005) performed reciprocal transplants between habitats with Mimulus cardinalis and M Iewisii, finding that each species performed considerably better in its own range than in the reciprocal environment. This indicates that the high levels of differentiation seen in the niche modeling presented (Figure 2.2K & L, Figure 2.2F) are based on intrinsic biological differences. However, in order to decouple which environmental variables are responsible, it would be necessary to vary each independently in 3 grth chamber experiment. Ecogeographic reproductive isolation Species can occupy a distribution both because of ecological and historical and historical factors (Endler 1982; Thorpe et al. 2008), and decoupling these two factors has been a major impediment to incorporating estimates of geographic isolation into reproductive isolation measures (Sobel et al. 2010). An example of this can be seen in the ecological niche model of Mimulusjohnstonii. M johnstonii occurs over a relatively small geographic area primarily within the Transverse Range of southern California. 54 Ecological niche modeling reveals that conditions similar to its home range also exist in the southern Sierra Nevada and northern Coast ranges, but it does not inhabit these areas most likely due to historic reasons (Figure 2.2J). This illustrates that the ecological niche modeling process does not simply resolve where a species does live, but also shows where a species could live. In this way, comparing overlap in niche models is more akin to examining intrinsic biological differences than simply calculating the overlap in species ranges currently occupied (Sobel et al. 2010). This allows a much more accurate depiction of the isolation that is both potentially and actually acting, making these estimates relevant under the biological species concept (Mayr 1942). Few estimates of ecogeographic isolation have been made, even though it is recognized as a potentially important form of reproductive isolation (Coyne and Orr 2004), perhaps due to a lack of methods needed to decouple historical and ecological processes. Ramsey et al. (2003) estimated ecogeographic isolation in Mimulus cardinalis and M lewisii at 0.587. While this is considerably lower than the estimates provided here (0.906 -- 0.970, Table 2.5), the previous method was much coarser, making a conservative estimate at best. As stressed by Ramsey et al. (2003) the sequential nature of reproductive barriers gives ecogeographic isolation enormous potential in preventing gene flow, as it is the first barrier to act. Within this linear sequential series methodology for calculating reproductive isolation, later acting barriers can only prevent gene flow that has not already been stopped by previously acting barriers. Because ecogeographic isolation gets the opportunity to act first, it only requires a value of 0.5 to ensure that no other barrier could possibly prevent more gene flow. This is true in 8 out of the 12 species pairs analyzed using the threshold 1 criterion, and for 11 out 12 species pairs 55 using threshold 2 (Table 2.5). And in total, the average ecogeographic isolation among all species pairs is above this level (0.64 to 0.69). This suggests that ecogeographic isolation must account for a sizeable proportion of the total isolation experienced by these recently diverged species, and therefore, it must be considered one of the most important barriers preventing gene flow in nature in this genus. Improving the accaracy of ecogeographic isolation estimates One of the biggest limitations to ecological niche modeling at present is the resolution of environmental variables available. The WORLDCLIM climatic layers . . . 2 . . employed in this study provrde 1km resolution, but many features relevant to specres distributions vary at smaller spatial scales. The geology layer presents particular issues for this study in giving a more conservative measure of niche boundaries than is really experienced. For example, both Mimulus norrisii and M rupicola are limestone endemics that occur in different regions (M norrisii in the southern Sierra and M rupicola in limestone cliffs around Death Valley). However, limestone often occurs as relatively small irregular outcrops surrounded by different geologic formations. For this reason, many of the occurrence points for these two species were in pixels that were classified as non-limestone areas even though all collections of these species were taken from limestone formations. If finer resolution geologic variable layers had been used, the ecological niche maps presented in Figures 2.2T and X would show far less suitable habitat, and instead only pixels with limestone present would be potential habitat for either species. As a result, estimates of ecogeographic isolation would be considerably higher than reported. 56 The issue of low resolution is likely to be most apparent in study regions with abrupt changes in elevation such as the Sierra Nevada. In fluctuating topography, climatic variables may vary considerably over a single square kilometer, making it possible for species occurrence records to be found in what appears to be unsuitable habitat due to pockets of microspatial climatic differences. Increased spatial resolution would also allow some assessment of microspatial habitat isolation. The overlapping threshold ecological niche models presented in Figures 2.5A-L show that many species have some degree of ecogeographic overlap. However, while some of these species pairs do co- occur over spatial scales of square kilometers, at microspatial scales, it is extremely uncommon to find two recently diverged species in close contact in the field (J. Sobel, pers. obs). This suggests that even with the large values of ecogeographic isolation presented here, differences in ecological niche occupied likely results in even more reproductive isolation than reported. If finer resolution variables were to become available in the future, microspatial habitat isolation should be estimated with caution, as the average dispersal distance of organisms may be greater than the grain size at those scales. Summau Understanding the evolution of reproductive isolation is a primary goal in evolutionary biology, and many would agree that a measure of which forms of isolation are most important at the time of speciation is crucial to this. Given its position in the linear sequence of potential reproductive barriers, ecogeographic isolation has enormous potential to play an important role. However, ecogeographic isolation has been largely 57 ignored in past studies of reproductive barriers, so estimates of its strength are sorely needed. The ecological niche maps and analyses in the presented work shows that ecogeographic isolation is of great importance in the genus Mimulus, with the certainty that it will be the highest contributor to total isolation for many of the pairs under study. This finding challenges the historical ignorance of the strength of this barrier, and suggests that additional studies are needed to assess the ubiquity of this pattern in other organisms, and should direct future studies of the traits and genetics underlying isolation in the future. 58 2.5 33m 85% £2 82o : 8555? ads 2 Sea .5352 282E 2 2.28 25.2 853 mm; 33 : 8585? ads _ 355 0.25202 .223 2 2.3 35.8 25.8 25.8 83 we 5% massage? Sam 523 2 85$ 83% 228 82% .83 w 8585? ads _ 8&8 acids 85:85 2 223 92.2. 245 22.9, $2. 5 855% ads _ See 555%.. 5E8 2 2.52 SEN 2.2% $3: as: am 5% 25355:? 85m assesses 2 2.25 $2.2 3.3 ES 22. we 5% 5585:5292 3% .2322 2 2.25 22 2.5: as.” a? 2 5% 35855:? 3% 52.585 2 2.58 atom 22% $52 23 ms 5% 3533555 3% ass: 2 2.5 24.3 €me 24. a 23 s 8585s, ass _ 2.58 55202 Essa 2 2.3 5.2 2.5 2 3 use 5 5% massage? 85m .552... 2 25E Sim 2.8V 23 235 E 5% 5585:5322 Sam senses... 2 net 20mm 2.8V 82: N35 3 5% 5825552 25% .5...st 2 263 5.2 2.5 23». Sad 42 5% massages: Saw 52528 2 6.82 832 2.5 2.4.8 $3 8 5% 3355:? 85... 52252.2 2 2.5 $5.2 see 3 2.: $2 Rm 5% 58355:? 3% secs 2 Gas Kn: Gas ~32 523 3 5% 55355:? 3% $228 2 9va as. _N 2.5: Seam 82 NS 5% massages: 85m 5553 2 2.25 25mm fine ass 32. mm 85523 ads 2 858 55282 538% 2 8% $22 2.2: 58.: 53 o: 5% 5583555 San beds 2 8.5: 5.2 22$ Beam 3% a 5% massages: 85m 9.23532 2 55 new: so: 23 R2 mm 5% 5:83:55 3% message 2 2.5 $34 2.25 an.” 8.2 2 8525 ads 2 858 55:02 58.65. 2 2.8V SE... 253 3.: 35o me 5% massage? Sam assesses 2 Adam mmooosmv GEM mmooosmv UD< 2 N 20:353. g 2232:. Sac “mob @0508 2382528 388 H2882 36on .wE—ouoE 23m: _aonBooo Sm coho—9:0 £50508 can com: when 86on “3:22: we EEG—2m .~.N 033—. 59 Table 2.2. Environmental layers used in ecological niche modeling. Layer Biol Bio4 BioS Bio6 Bi012 BiolS Biol6 Biol 7 Geology Description Annual mean temperature (C) Temperature seasonality (st. dev.) Maximum temperature of warmest month (C) Minimum temperature of coldest month (C) Annual mean precipitation (mm) Precipitation seasonality (coefficient of variation) Precipitation of wettest quarter (Inn!) Precipitation of driest quarter (mm) Geologic map of California 60 Source WORLDCLIM (http://worldclim.org) WORLDCLIM (http://worldclim.org) WORLDCLIM (http://worldclim.org) WORLDCLIM (http://worldclim.org) WORLDCLIM (http://worldclim.org) WORLDCLIM (http://worldclim.org) WORLDCLIM (http://worldclim.org) WORLDCLIM (http://worldclim.org) US. Geological Survey (http://www.usgs.g0\Q 3:5 mm_ md- 2036 ommmd wmawd on 55.2 . hmmwd- 2:6- woomd oovmd- waned- mmhmd- oSNd- awamdr whomd NVmod- wommd- w God- bwomd- an: .o- ovwcd $36. 336. 3 8.2 2.36- $3.2 $3.2 nummd N. ~05 0 ~05 205 N ~05 005 mOHm v05 "05 .wE—oeofi one? _momwofioo 5 com: mu??— acoficoherco macaw 3220508 doze—2.80 .M.N “...—uh. 0 ~05 m ~05 $05 005 mOHm v05 ~05 61 Table 2.4. Results of one-way MANOVA analysis on environmental variables between Mimulus species pairs. Species pair Wilks’ 7» Degrees of F p Freedom M androsaceus/shevockii 0.488 1, 49 6.417 < 0.0001 M angustatus/pulchellus 0.341 1, 66 15.97 < 0.0001 M bicolor/filicaulis 0.792 1, 153 5.024 < 0.0001 M bigelovii/bolanderi 0.106 1, 243 256.7 < 0.0001 M brevipes/johnstonii 0.390 1, 276 53.90 < 0.0001 M cardinalis/lewisii 0.296 1, 310 92.18 < 0.0001 M constrictus/whitneyi 0.493 1, 134 17.24 < 0.0001 M cusickii/nanus 0.552 1, 70 7.10 < 0.0001 M douglasii/kelloggii 0.636 1, 183 13.066 < 0.0001 M floribundus/norrisii 0.933 1, 327 2.9352 0.0035 M gracilipes/palmeri 0.414 1, 98 17.317 < 0.0001 M parryi/rupicola 0.038 1, 13 40.92 <0.0001 62 222.22 mmwflw $2.2 mmww $2.22 222%me 222.22 Mmmwwm 2mm...“ W22 22 22.22 Mwmw 82.2 MMMMM2 22:2 wwmuw 2222.2 WWW ..mwmhfi m E... mm a... was 2.... ”Mm. 2.... ”mm: ...hwmmmmm 22 22. 22 22. 222 22 22 22 .222 222.22 . ”WWW 222.2 mm: $2.22 “Mum 82.22 WMWHMM ...Mmmfiw m 22 22 22 .. 22 22 .22 22 .22... 2...... “MM“ 2.. mm... 3.... MM” .22 22mm... 2.2.2.”..me W22 22 22. 22 22 .2222 22:2 222me 28.2 wwwm 222.22 mmmw 28.22 mm Mfififi ”22 s... Mmmuw .2. mm... 2.... mm a... WNW”? 32.2%...“ 222 a 22522 222522 E 222522 222522 a owa2o>< 22222229: moan toga moan 280.2. 2 092.8322 E22222: moan 22222.5 822 228.2. 220on N 2.2.5.2222 2 2222282222 522225.222 5 2.223 828% 58323 232282 038260.52 oEQfiwoowoom .m.~ 0.2—ah. 63 100 M- [00 100 rupicola ~.M. parryii T'M° cusickii 100 100 100 v—M. nanus 7 2 M whitneyi 9 7 M. constricms F_M bolanderi 82 91 —M bigeloveii _{—M. johnstonii 100 —M brevipes —M. pulchellus l ()0 M. angustatus 100 —M kelloggii M douglasii 100 100 100 M. cardinalis M. lewisii M filicaulis 100 M bicolor 100 [M' shevockii 100 100 mat— — 0.01 substitutions per site lM androsaceus __[M palmeri 100 M. gracilipes M. jungermannioides M norrisii M. floribundus M ringens - Oenoe — E unamls _ Eunanm IL Oenoe ll Erythranthe _ Paradanthus — - Outgroup Figure 2.1. Phylogenetic relationships among Mimulus species in this study. Majority rule consensus tree obtained by maximum likelihood analysis in PHYLIP (Felsenstein 2005) showing relationships among Mimulus species (with M ringens as outgroup). Sequence data (trnL/F, ITS, and E TS) were the same as used in Beardsley et al. (2004), trimmed to only include species in this study. Labeled brackets indicate traditional delineations from morphological taxonomy (Grant 1924). Branch lengths are proportional to average substitutions per site, and numbers on branches indicate the number of times out of 100 that the partition of species separated by that branch occurred among all trees. o M. androsaceus - 0% cumulative probability score - 10% - 20% - 30% - 40% j 50% 60% 70% 80% 90% Figure 2.2A. Ecological niche model of Mimulus androsaceus. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M androsaceus. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 65 a M. shevockii - 0% cumulative probability score - 10% - 20% - 30% - 40% C: 50% 60% 70% 80% 90% Figure 2.28. Ecological niche model of Mimulus shevockii. Ecological niche model generated in Maxent using the ‘jackknife minus 1’ cross validation procedure for the species, Mimulus shevockii. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 66 M. angustatus 0% (cumulative probability score) 10% 20% 30% 40% 50% , 60% 70% 80% 90% 2.1 Figure 2.2C. Ecological niche model of Mimulus angustatus. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. angustatus. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots Show the locations of collections used. 67 a M. pu/che/Ius - 0% cumulative probability score - 10% - 20% - 30% - 40% ’._.: 50% 60% 70% 80% 90% Figure 2.2D. Ecological niche model of Mimulus pulchellus. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. pulchellus. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 68 o M. bicolor - 0% cumulative probability score - 10% - 20% - 30% - 40% L: 50% ‘ 60% 70% 80% 90% Figure 2.2E. Ecological niche model of Mimulus bicolor. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. bicolor. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 69 o M. filicaulis - 0% cumulative probability score - 10% - 20% - 30% - 40% l: 50% 60% 70% 80% 90% Figure 2.2F. Ecological niche model of Mimulusfilicaulis. Ecological niche model generated in Maxent using the ‘jackknife minus 1’ cross validation procedure for the species, M. filicaulis. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 70 o M. bigelovii 0% cumulative probability score 10% 20% 30% 40% 50% 60% 70% 80% 90% Figure 2.2G. Ecological niche model of Mimulus bigelovii. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. bigelovii. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 71 a M. bo/anderi - 0% cumulative probability score - 10% - 20% - 30% - 40% _'::i 50% 7 . 60% 70% 80% 90% Figure 2.2H. Ecological niche model of Mimulus bolanderi. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. bolanderi. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 72 o M. brevipes - 0% cumulative probability score - 10% - 20% - 30% - 40% T 50% 60% 70% 80% 90% Figure 2.21. Ecological niche model of Mimulus brevipes. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. brevipes. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 73 o M. johnstonii - 0% cumulative probability score - 10% - 20% - 30% - 40% :3 50% 60% 70% 80% 90% Figure 2.2J. Ecological niche model of Mimulusjohnstonii. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. johnstonii. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 74 o M. cardinalis - 0% cumulative probability score - 10% - 20% - 30% - 40% ': 50% 60% 70% 80% 90% Figure 2.2K. Ecological niche model of Mimulus cardinalis. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. cardinalis. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 75 a M. lewisii - 0% cumulative probability score - 10% - 20% - 30% - 40% C: 50% , 60% 70% 80% 90% Figure 2.2L. Ecological niche model of Mimulus lewisii. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. lewisii. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 76 M. constrictus 0% cumulative probability score 10% 20% 30% 40% -. 50% 60% 70% 80% 90% l j l _i Figure 2.2M. Ecological niche model of Mimulus constrictus. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. constrictus. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 77 o M. whitneyi - 0% cumulative probability score - 10% - 20% - 30% - 40% .L.‘ 50% 60% 70% 80% 90% Figure 2.2N. Ecological niche model of Mimulus whitneyi. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. whitneyi. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 78 o M. cusickii - 0% cumulative probability score - 10% - 20% - 30% - 40% 1:1 50% I 60% 70% 80% 90% Figure 2.20. Ecological niche model of Mimulus cusickii. Ecological niche model generated in Maxent using the ‘jackknife minus 1’ cross validation procedure for the species, M. cusickii. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 79 o M. nanus - 0% cumulative probability score - 10% - 20% - 30% - 40% '—_:i 50% i I 60% 70% 80% 90% Figure 2.2P. Ecological niche model of Mimulus nanus. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. nanus. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 80 o M. doug/asii - 0% cumulative probability score - 10% - 20% - 30% - 40% S 50% fl 7.. 60% ' 70% 80% 90% Figure 2.2Q. Ecological niche model of Mimulus douglasii. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. douglasii. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 81 u M. kelloggii - 0% cumulative probability score - 10% - 20% - 30% - 40% : 50% 60% 70% 80% 90% Figure 2.2R. Ecological niche model of Mimulus kelloggii. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. kelloggii. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 82 o M. floribundus - 0% cumulative probability score - 10% - 20% - 30% - 40% $37!] 50% i 60% 70% 80% 90% o o. 'o 'v “-5“ o r g \ b '0” ‘g... __ p '. .3 .. O ’0 Figure 2.28. Ecological niche model of Mimulusfloribundus. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. floribundus. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 83 o M. norrisii - 0% cumulative probability score 10% - - 20% - - 30% 40% ; 50% ,, 60% i 70% 80% 90% Figure 2.2T. Ecological niche model of Mimulus norrisii. Ecological niche model generated in Maxent using the ‘jackknife minus 1’ cross validation procedure for the species, M. norrisii. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 84 o M. gracilipes - 0% cumulative probability score - 10% - 20% - 30% - 40% T73 50% ; 60% 70% 80% 90% Figure 2.2U. Ecological niche model of Mimulus gracilipes. Ecological niche model generated in Maxent using the ‘jackknife minus 1’ cross validation procedure for the species, M. gracilipes. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 85 o M. palmeri - 0% cumulative probability score - 10% - 20% - 30% - 40% E 50% f -j 60% 70% , 80% 90% Figure 2.2V. Ecological niche model of Mimulus palmeri. Ecological niche model generated in Maxent using the 50% training/testing procedure for the species, M. palmeri. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 86 o M. parryi - 0% cumulative probability score - 10% - 20% - 30% - 40% :__.—.l 50% . 60% 70% 80% 90% Figure 2.2W. Ecological niche model of Mimulus parryi. Ecological niche model generated in Maxent using the ‘jackknife minus 1’ cross validation procedure for the species, M parryi. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 87 o M. rupicola - 0% cumulative probability score - 10% - 20% - 30% - 40% .L: 50% 60% , 70% 80% Figure 2.2x. Ecological niche model of Mimulus rupicola. Ecological niche model generated in Maxent using the ‘jackknife minus 1’ cross validation procedure for the species, M. rupicola. Shading corresponds to output of cumulative probability values multiplied by 100. Lighter areas indicate areas of highest habitat suitability for the species, and dots show the locations of collections used. 88 Figure 2.3A. Variation in environmental variables between the species pair Mimulus androsaceus and M. shevockii. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 89 les pair Mimulus tion of each variable inuous variables values for each le, geology. boxes y within the damsel ental variable are ed significant Geology O in 61 o {E '-2 2 _l O m _2 Q _.[ O a: 33 ——= .9 9. EB ———l O m .25 >01 :36 . sol at- ...o 9 1 N m 0% O l— OLD %- -- 2% 9 3'; "l 0" .22 rye-u .334 $3 5 04‘ Q as —l 0 ‘33 S: '5 [ll at 3% a 2 a a a 5% 2 é o 6; uopnqmuog waxed JO '(UD) U0!121!d!331d '(3) BJHIEJadLuai Precipitation (cm) Temperature (C) Figure 2.38. Variation in environmental variables between the species pair Mimulus angustatus and M. pulchellus. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 91 :5 pair .l/imu/us on of each variable nuous variables values for each le. geology, boxes / within the dataset :ntal variable are d significant _i l l 0 A :: 37 r i _[ § 2 0 0 g- «33% £73? ~93". T or“ 5.935 fi8'9 a53- Eag 3 a"? i 0 .~. .536 2‘05? *3 E —l 3 3 C ‘ s r a a, a z 2‘ El :3, 8 8 i e . . - '- 8 O T f i a s s e a 2 é a '5 uo unqmuog waxed JO ’(un) uonetgdpaud ‘(3) ainieraduial 92 Precipitation (cm) Temperature (C) Figure 2.3C. Variation in environmental variables between the species pair Mimulus bicolor and M. filicaulis. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 93 .So‘oou A63 832.935 m PCB 0.03 203 .205 g 232382. Fir mOE v05 POE O 005 O 3565 ”no 5506055 u 5 £385“ .2 I 3:060:20 N0 .8785 H .0 c283 .2 D . ov ,, ,8 Tom uonnqinuog tuaaiad JO '(un) uoneigdpaid '(3) ainieiadtua J. 94 Figure 2.3D. Variation in environmental variables between the species pair Mimulus bigelovii and M bolanderi. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 95 3339 2:3 co_§_a_oma 9 938383 T :05 lift. 0 203 203 N u 03 003 MOE V03 .03 a 2.93:8 ”Nu 9:060:80 ” G .5353 .2 I accumucaumoézsaau6.2.3283: D 11 .2 ON ..om oe 10m ...2 so. uonnqgnuog waxed JO '(un) uoueigdpaid '(3) ainieiadtua J. 96 Figure 2.3B. Variation in environmental variables between the species pair Mimulus brevipes and M. johnstonii. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 97 AEov cozmzfiuoi g 238chth _ a _ _ .9- 292000 k n 03 on 03 m FOE N n 02¢ 33 MOE 62¢ .03 int 0 o o o o o o o .-o. 0 cm A .2 J1 o l 3 tom A .. ton +8 r_L 2.93:8”an 9:050:895.228323 .2 I E2>==m ”Nu 6529” 5 623.33 .2 D . .8 02 uonnqgnuog waxed JO ‘(ui3) uoneiidpaid '(3) BJDIEJadUJal 98 Figure 2.3F. Variation in environmental variables between the species pair Mimulus cardinalis and M. lewisii. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 99 .3286 A63 coznguoi fl k ~03 on 03 :03 _JIW I. O :03 g EBSmQEwH 7:- _ .003 MOB V03 .03 . .4: q 87.3% ”no “£55056 ” 5 ....a.§& .2 . 5:323 flu .9229 u .0 .2323 .2 D roP om 6m 0». :Om co YON cm :8 -8. uogmqgnuog waxed JO ’(uD) uoueigdpaid ’(3) BJDIEladlual 100 Figure 2.3G. Variation in environmental variables between the species pair Mimulus constrictus and M. whitneyi. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 101 30.80 .83 8.5335 g 23.23th mug =03 {at “~03 _ «83 I 1.. 33 35 83 8a 0 O O ...—L lfiL Eco... N0 9205956 u G ..SoSE! .2 I 323 .E:_>:__~ “No 9:060:29 5 «38.538 .2 D .. JIOPI to. . .oN 110M tom .65 ..om -8— uonnqgnuog waxed JO '(UJD) uoneigdpaid ‘0) ainieiaduia J. 102 Figure 2.3H. Variation in environmental variables between the species pair Mimulus cusickii and M. nanus. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 103 .3236 AEUV 5:83:85 _ 5 PCB 203 whoa 203 in 85 .63 Ba g 239388. O i¢—.I 3565 ”mu .Btofiocem “ 5 Gate: .2 I E:_>:__m "No 9.93:8 n 5 €3.33 .2 D .. A: ON .0? ..0m ton om 8.. uonnqgnuog iuaarad JO '(uD) uoneudpaid '(3) BJHIEJadLual 104 Figure 2.31. Variation in environmental variables between the species pair Mimulus douglasii and M. kelloggii. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 105 E293 AEov cosmgafioi — =03 ifu O O 203 m POE g 2323th _ 1 d - op- «.03 003 mOE v05 _.O_m o o :2 ..om o 41 :8 :8 tom [fl .TS ten [1 3_moccm”~0 .ocoumucmmupu.....mu§~« .2 I :8 mac—29: 6:2ng ”Nu 6:03.29. ” .0 9:33:03 .2 _H— a8 -62 uogmqmuog wand 10 ‘(tu3) uogzexgdpaid '(3) aimeiadwa J. 106 Figure 2.3.]. Variation in environmental variables between the species pair Mimulus floribundus and M. norrisii. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 107 A3390 AEov 532532“. 0V 232095» n FOE 905 m FOE J _ _ N FOE 005 no; V03 FOE op- l IE 0 O O O O .- a 8986: "No 6538 u 6 ....anc .2 I 822...? ”mu 6:850:80 ” 5 £233.20: .2 D . 0m 110V 0.... .oo .. on .om r: uogmqgnuog waxed 10 '(w:) uoneudpald '(3) aimeladwai 108 Figure 2.3K. Variation in environmental variables between the species pair Mimulus gracilipes and M. palmeri. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 109 5.36 AEuV cozmufitmi _ n FOE 20% m POE {l O _ N FOE fio. a g 9.328th 003 hem ‘03 ~03 fiO o 3__~co:~u.3_.o_uoc29GEéuqS I a? N0 9:33:23 " 5 «3.535 .2 _H. ..OPI to— .ON iron . ,ov lrom ..oo Arc“ .28 18— uounqmuoy waxed JO '(LID) uognngdpard '(3) amwadwal 110 Figure 2.3L. Variation in environmental variables between the species pair Mimulus parryi and M. rupicola. Circles represent the relative contribution of each variable to the ecological niche models presented in Figures 2.2A-X. For continuous variables from the WORLDCLIM dataset (Table 2.2), boxes represent average values for each variable with standard deviation error bars. For the categorical variable, geology, boxes represent the frequency of the first and second most common category within the dataset, and their identities are noted in the legend. Labels for each environmental variable are bold/italicized when Mann-Whitney U or contingency analysis revealed significant differences between species. 111 O. 2323th r _L. 33.0. 33 53 , Ire—... 29 8:539: _ J .4363 to: 20,3 20.: 22a 14W 0 O l I1 323:35 N0 9.93:8 u G .38qu .2 I . 5333:: "NU 9.93:9. ” 5 €93 .2 D . :2 :ON ..cm .-om ion 6o 8. uoginqgnuog waxed JO '(un) uoueudpaid '(3) almeladwal 112 Figure 2.4A. Identity tests using Schoener’s D. Results of ‘Identity’ tests implemented in ENMTools (httpz/lenmtools.blogspot.com) (Warren et al. 2008). Histograms represent null distributions of Schoener’s D values for each species pair based on randomizing species identities associated with environmental variables used in ecological niche modeling. The distributions consist of 100 pseudoreplicates each, and the arrows indicate where the actual value for Schoener’s D between species pairs falls. In all cases, actual values are significantly different from a null expectation at on = 0.05, indicating that ecological niches of these species pairs are more dissimilar than expected by chance. 113 S s u U. "I. d S r t! u I." ei . .m w, .3 "u m 9 mun .Wr im .3 Uni 1 dwl [991 b.53 ”m4 Ya] I. mmo. .3 .gmo. who. .mmo. «mo. "No. .3 www 3 ,u..n.w dfiw fl.n.m wmw an -3 MMP o MMP MMP MMP MMP MMP no la -md m6 5o rod T 5 no Luzo U I. S . H. an” ”V0 w w .I.” .M .5 ...VO 0k . a” .5 in. S d . SC rmo IE I” Vd em .! nmo mol 1 whl Iou mn .msl my! -. .0 WO Ivymo 9k0 o m WM wmo .mMO .mo nhO. . nu0. k” io r00. 090. I a.<~< Ad a.D..< b.fi. bb. b..J< xl.< L6 MMP mo MMP MM MM MMP MMP no 114 Figure 2.48. Identity tests using 1 statistic. Results of ‘Identity’ tests implemented in ENMTools (http://enmtools.blogspot.com) (Warren et al. 2008). Histograms represent null distributions of 1 values (see Warren et al 2008 for details) for each species pair based on randomizing species identities associated with environmental variables used in ecological niche modeling. The distributions consist of 100 pseudoreplicates each, and the arrows indicate where the actual value for 1 between species pairs falls. In 11 of the 12 species pairs actual values were significantly different from a null expectation at a = 0.05, indicating that ecological niches of these species pairs are more dissimilar than expected by chance. 115 H , H ad ad ad ad ad d ad ad rmd md 5 1d m i .3 .ed fl i d .w rd 5 kw. .. "u T. mm. . n: .. Vin -. m . nn .3 ks #0 I9 mo uq. :3 me .3 .w.o mo 5.".1 u .KUM H. 901 Hmflz .. CM). 1 {Km m h M .~d w m o H~d w w M ~d w m w “Nd m I M n~d r WO. d T m m r... w. < er C. n. = rad d. k. < H.o fl. n = $6 0.. . _. -Hd . r < Hd MMP T MMP r MMP MMP . MMP I MMP To To o .o .o o H H H dd ad ad ad ad ad ad “Ad ”ad od mod .dd m md s nmd .3 c ”u «d w m d i .n .ed .u ..d R. IV; m d md m w d r m nmd .w Ana. Iv-md m. m d d . nmd mot mht mwt . mm! m imt mumt . .m M M d MW M ~d .c .m M .3 .9! M Ad w h 0. d r m 0 L3 1 d 1 I I O 0 a e 0 T a. J < Hd 0. D.. < H6 .0. f. < #6 b. .w. < #6 .w. .1. < Hd x I. < 3.0 MMP o MMP o MMP to MMP no MMP d MMP To 116 - unsuitable suitable for M. shevocki/ - suitable for M. androsaceus ' suitable for both species Figure 2.5A. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus androsaceus and M shevockii. The continuous ecological niche models shown in Figure 2.2A and B were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M androsaceus alone, suitable to M. shevockii alone, and suitable to both species. 117 - unsuitable ' _ suitable for M. pu/che/lus - suitable for M. angustatus I: suitable for both species Figure 2.58. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus angustatus and M pulchellus. The continuous ecological niche models shown in Figure 2.2C and D were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M angustatus alone, suitable to M. pulchellus alone, and suitable to both species. 118 - unsuitable suitable for M. filicaulis - suitable for M. bicolor V d ‘ suitable for both species Figure 2.5C. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus bicolor and M. filicaulis. The continuous ecological niche models shown in Figure 2.2E and F were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M bicolor alone, suitable to M filicaulis alone, and suitable to both species. 119 - unsuitable , * suitable for M. bo/anden' - suitable for M. bigelovii . suitable for both species Figure 2.5D. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus bigelovii and M bolanderi. The continuous ecological niche models shown in Figure 2.2G and H were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M bigelovii alone, suitable to M bolanderi alone, and suitable to both species. 120 - unsuitable ' suitable for M. johnstonii - suitable for M. brevipes . suitable for both species Figure 2.5E. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus brevipes and M johnstonii. The continuous ecological niche models shown in Figure 2.21 and 1 were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M brevipes alone, suitable to M johnstonii alone, and suitable to both species. 121 - unsuitable 7 suitable for M. lewisii - suitable for M. cardinalis 7:3: suitable for both species Figure 2.5F. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus cardinalis and M lewisii. The continuous ecological niche models shown in Figure 2.2K and L were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M cardinalis alone, suitable to M lewisii alone, and suitable to both species. 122 - unsuitable suitable for M. whitney/ - suitable for M. constrictus *. suitable for both species i _ __. Figure 2.5G. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus constrictus and M whitneyi. The continuous ecological niche models shown in Figure 2.2M and N were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M constrictus alone, suitable to M whitneyi alone, and suitable to both species. 123 "1H“ ’3‘? - unsuitable 3‘ i ‘ suitable for M. nanus 1 - suitable for M. cusickii ' A 7.1 suitable for both species L. Figure 2.5H. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus cusickii and M nanus. The continuous ecological niche models shown in Figure 2.20 and P were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M cusickii alone, suitable to M nanus alone, and suitable to both species. 124 - unsuitable suitable for M. kelloggii - suitable for M. douglasii' L: suitable for both species Figure 2.51. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus douglasii and M kelloggii. The continuous ecological niche models shown in Figure 2.2Q and R were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M douglasii alone, suitable to M kelloggii alone, and suitable to both species. 125 - unsuitable suitable for M. norrisii - suitable for M. f/oribundus ““- suitable for both species Figure 2.5J. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulusfloribundus and M norrisii. The continuous ecological niche models shown in Figure 2.28 and T were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M floribundus alone, suitable to M norrisii alone, and suitable to both species. 126 - unsuitable 7:3 suitable for M. palmen' - suitable for M. gracilipes l 1 suitable for both species Figure 2.5K. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus gracilipes and M palmeri. The continuous ecological niche models shown in Figure 2.2U and V were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M gracilipes alone, suitable to M palmeri alone, and suitable to both species. 127 - unsuitable suitable for M. rupicola - suitable for M. parryi suitable for both species Figure 2.5L. Overlay plot showing amount of ecogeographic overlap in the species pair Mimulus parryi and M rupicola. The continuous ecological niche models shown in Figure 2.2W and X were made into binary suitable/unsuitable maps using the 10th percentile of training data method described in the Methods section. Data from the two species were overlayed onto a single map in order to show habitat that is unsuitable to both species, suitable to M parryi alone, suitable to M rupicola alone, and suitable to both species. 128 Ecogeographic isolation .0 .0 .0 .0 .0 .0 .0 2" w .5 U! (h \I on O H H 1 l 1 l I l L l O O O O 0 .0 N l .0 H I O 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 Sequence divergence (avg # subst./site) Figure 2.6. Estimates of ecogeographic isolation in Mimulus obtained using threshold 2 regressed on sequence divergence between the species pairs (average number of substitutions per site from phylogenetic analysis). Correlation analysis indicates no relationship between ecogeographic isolation and sequence divergence (p = 0.9188). Note that 11 out of 12 species pairs exhibit ecogeographic isolation above 0.5 (dashed line), the point at which this barrier will contribute more to total reproductive isolation on a relative scale than any other barrier. The lone species pair (M kelloggii / M douglasii) that does not exceed 0.5 is both the species pair with the highest sequence divergence, and one member of the pair, M douglasii is highly selfing (J. Sobel, unpublished). 129 0.4- Ecogeographic Isolation 0.3- 0.2- 0.1- ! f l I l 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 Schoener'sD Figure 2.7. Relationship between niche similarity and ecogeographic isolation. Regression analysis of ecogeographic isolation on the measure of niche identity, Schoener’s D. While a significant linear relationship exists (r2 = 0.547; p = 0.006), the analysis illustrates that the ecogeographic consequence of niche differences can be relatively variable. 130 Literature Cited Angert, A. L. 2006. Demography of central and marginal populations of monkeyflowers (Mimulus cardinalis and M lewisii). Ecology 87:2014-2025. Angert, A. L., H. D. Bradshaw, and D. W. Schemske. 2008. Using experimental evolution to investigate geographic range limits in monkeyflowers. Evolution 62:2660-2675. Angert, A. L., and D. W. Schemske. 2005. The evolution of species' distributions: Reciprocal transplants across the elevation ranges of Mimulus cardinalis and M lewisii. Evolution 59: 1671-1684. Barraclough, T. G., and A. P. Vogler. 2000. Detecting the geographical pattern of speciation from species-level phylogenies. American Naturalist 155:419-434. Beardsley, P. M., and R. G. Olmstead. 2002. Redefining Phrymaceae: The placement of Mimulus, tribe Mimuleae and Phryma. American Journal of Botany 89:1093- 1102. Beardsley, P. M., S. E. Schoenig, J. B. Whittall, and R. G. Olmstead. 2004. Patterns of evolution in Western North American Mimulus (Phrymaceae). American Journal of Botany 91 :474-489. Bolnick, D. I. 2004. Waiting for sympatric speciation. Evolution 58:895-899. Bolnick, D. 1., and B. M. Fitzpatrick. 2007. Sympatric speciation: Models and empirical evidence. Annual Review of Ecology Evolution and Systematics 38:459-487. Bradshaw, H. D., and D. W. Schemske. 2003. Allele substitution at a flower colour locus produces a pollinator shift in monkeyflowers. Nature 426: 176-1 78. Bradshaw, H. D., S. M. Wilbert, K. G. Otto, and D. W. Schemske. 1995. Genetic mapping of floral traits associated with reproductive isolation in monkeyflowers (Mimulus). Nature 376:762-765. Coyne, J. A., and H. A. Orr. 2004, Speciation. Sunderland, MA, Sinauer Associates. Dobzhansky, T. 1937, Genetics and the Origin of Species. New York, Columbia University Press. Elith, J., C. H. Graham, R P. Anderson, M. Dudik, S. Perrier, A. Guisan, R. J. Hijmans et al. 2006. Novel methods improve prediction of species' distributions from occurrence data. Ecography 29: 129-15 1. 131 Endler, J. A. 1982. Problems in distinguishing historical from ecological factors in biogeography. American Zoologist 22:441-452. Feder, J. L., X. F. Xie, J. Rull, S. Velez, A. Forbes, B. Leung, H. Dambroski et al. 2005. Mayr, Dobzhansky, and Bush and the complexities of sympatric speciation in Rhagoletis. Proceedings of the National Academy of Sciences of the United States of America 10226573-6580. F elsenstein, J. 1985. Phylogenies and the comparative method. American Naturalist 125:1-15. —. 2005.PHYLIP (Phylogeny Inference Package) version 3.6.Distributed by the author. Department of Genome Sciences, University of Washington, Seattle. Fishman, L., A. J. Kelly, and J. H. Willis. 2002. Minor quantitative trait loci underlie floral traits associated with mating system divergence in Mimulus. Evolution 56:2138-2155. Fitzpatrick, B. M., and M. Turelli. 2006. The geography of mammalian speciation: Mixed signals from phylogenies and range maps. Evolution 60:601-615. Graham, C. H., S. F errier, F. Huettmann, C. Moritz, and A. T. Peterson. 2004a New developments in museum-based informatics and applications in biodiversity analysis. Trends in Ecology and Evolution 19:497-503. Graham, C. H., S. R. Ron, J. C. Santos, C. J. Schneider, and C. Moritz. 2004b. Integrating phylogenetics and environmental niche models to explore speciation mechanisms in dendrobatid frogs. Evolution 58: 1781-1793. Grant, A. L. 1924. A monograph of the genus Mimulus. Annals of the Missouri Botanical Garden 11:99-388. Guisan, A., and W. Thuiller. 2005. Predicting species distributions: Offering more than simple habitat models. Ecology Letters 8:993-1009. Hiesey, W. M., M. A. Nobs, and O. Bjorkman. 1971, Experimental Studies on the Nature of Species. V. Biosystematics, genetics, and physiological ecology of the Erythranthe section of Mimulus. Washington, DC, Carnegie Institution of Washington Publication 628. Jamevich, C. S., T. J. Stohlgren, D. Barnett, and J. Kartesz. 2006. Filling in the gaps: Modelling native species richness and invasions using spatially incomplete data. Diversity and Distributions 12:511-520. Jaynes, E. T. 1957. Information theory and statistical mechanics. Physical Review 106:620-630. I32 The Jepson Online Interchange for California Floristics 2004. University and Jepson Herbaria, University of California. Berkeley, CA. Jordan, D. S. 1908. The law of geminate species. American Naturalist 42:73-80. Jordan, D. S., and V. L. Kellogg. 1907, Evolution and Animal Life: An Elementary Discussion of Facts, Processes, Laws and Theories Relating to the Life and Evolution of Animals. New York, D. Appleton and Company. Kay, K. M. 2006. Reproductive isolation between two closely related hummingbird- pollinated Neotropical gingers. Evolution 60:538-552. Kiang, Y. T. 1972. Pollination study in a natural population of Mimulus guttatus. Evolution 26:308-310. Kirkpatrick, M., and V. Ravigne. 2002. Speciation by natural and sexual selection: Models and experiments. American Naturalist 159:822-S35. Kozak, K. H., C. H. Graham, and J. J. Wiens. 2008. Integrating GIS-based environmental data into evolutionary biology. Trends in Ecology & Evolution 23: 141-148. Kozak, K. H., and J. J. Wiens. 2006. Does niche conservatism promote speciation? A case study in North American salamanders. Evolution 60:2604-2621. Liu, C. R., P. M. Berry, T. P. Dawson, and R. G. Pearson. 2005. Selecting thresholds of occurrence in the prediction of species distributions. Ecography 28:385-393. Lowry, D. B., R. C. Rockwood, and J. H. Willis. 2008. Ecological reproductive isolation of coast and inland races of Mimulus guttatus. Evolution 62:2196-2214. Macnair, M. R. 1983. The genetic control of copper tolerance in the yellow monkey flower, Mimulus guttatus. Heredity 50:283-293. Martin, N. H., and .1. H. Willis. 2007. Ecological divergence associated with mating system causes nearly complete reproductive isolation between sympatric Mimulus species. Evolution 61:68-82. Mayr, E. 1942, Systematics and the origin of species. New York, Columbia University Press. —. 1947. Ecological factors in speciation. Evolution 1:263-288. —. 1963, Animal Species and Evolution. Cambridge, MA, Harvard University Press. 133 —. 1984. Species concepts and their applications, Pages 531-540 in E. Sober, ed. Conceptual Issues in Evolutionary Biology. Cambridge, MA, MIT Press. Nakazato, T., M. Bogonovich, and L. C. Moyle. 2008. Environmental factors predict adaptive phenotypic differentiation within and between two wild andean tomatoes. Evolution 62:774-792. Nakazato, T., D. L. Warren, and L. C. Moyle. 2010. Ecological and geographic modes of species divergence in wild tomatoes. American Journal of Botany 97:680-693. Nosil, P. 2008. Speciation with gene flow could be common. Molecular Ecology 17:2103-2106. Ortega-Huerta, M. A., and A. T. Peterson. 2008. Modeling ecological niches and predicting geographic distributions: a test of six presence-only methods. Revista Mexicana De Biodiversidad 79:205-216. Papes, M., and P. Gaubert. 2007. Modelling ecological niches from low numbers of occurrences: assessment of the conservation status of poorly known viverrids (Mammalia, Carnivora) across two continents. Diversity and Distributions 13:890-902. Pearson, R. G., C. J. Raxworthy, M. Nakamura, and A. T. Peterson. 2007. Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. Journal of Biogeography 34: 102-1 17. Peterson, A. T. 2003. Predicting the geography of species' invasions via ecological niche modeling. Quarterly Review of Biology 78:419-433. Peterson, A. T., J. Soberon, and V. Sénchez—Cordero. 1999. Conservatism of ecological niches in evolutionary time. Science 285:1265-1267. Phillips, S. J ., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modeling 190:231-259. Poulton, E. G. 1908. What is a species?, Pages 46-94 Essays on Evolution 1889-1907. London, Oxford at the Clarendon Press. R Development Core Team 2010. R: A Language and Environment for Satistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Ramsey, J ., H. D. Bradshaw, and D. W. Schemske. 2003. Components of reproductive isolation between the monkeyflowers Mimulus lewisii and M cardinalis (Phrymaceae). Evolution 57:1520-1534. 134 Rice, W. R., and E. E. Hostert. 1993. Laboratory experiments on speciation: What have - we learned in 40 years? Evolution 47:1637-1653. Schemske, D. W. 2000. Understanding the origin of species. Evolution 54:1069-1073. Schemske, D. W., and H. D. Bradshaw. 1999. Pollinator preference and the evolution of floral traits in monkeyflowers (Mimulus). Proceedings of the National Academy of Sciences of the United States of America 96:11910-11915. Schluter, D. 2001. Ecology and the origin of species. Trends in Ecology & Evolution 16:372-380. . Schoener, T. W. 1968. Anolis lizards of Bimini: Resource partitioning in a complex fauna. Ecology 49:704-726. Sexton, J. P., P. J. McIntyre, A. L. Angert, and K. J. Rice. 2009. Evolution and ecology of species range limits. Annual Review of Ecology Evolution and Systematics 40:415-436. Sobel, J. M., G. F. Chen, L. R. Watt, and D. W. Schemske. 2010. The biology of speciation. Evolution 64:295-315. Stebbins, G. L. 1950, Variation and Evolution in Plants. New York, Columbia University Press. Streisfeld, M. A., and J. R. Kohn. 2005. Contrasting patterns of floral and molecular variation across a cline in Mimulus aurantiacus. Evolution 59:2548-2559. Thorpe, R. S., Y. Surget-Groba, and H. Johansson. 2008. The relative importance of ecology and geographic isolation for speciation in anoles. Philosophical Transactions of the Royal Society B-Biological Sciences 363:3071-3081. Turelli, M., N. H. Barton, and J. A. Coyne. 2001. Theory and speciation. Trends in Ecology & Evolution 16:330-343. Via, S. 2001. Sympatric speciation in animals: The ugly duckling grows up. Trends in Ecology & Evolution 16:381-390. Vickery, R. K. 1959. Barriers to gene exchange within Mimulus guttatus (Scrophulariaceae). Evolution 13 2300-3 10. —. 1964. Barriers to gene exchange between members of Mimulus guttatus complex ( Scrophulariaceae ). Evolution 18:52-69. —. 1995. Speciation by aneuploidy and polyploidy in Mimulus (Scrophulariaceae). Great Basin Naturalist 55: 174-176. 135 Warren, D. L., R. E. Glor, and M. Turelli. 2008. Environmental niche equivalency versus conservatism: Quantitative approaches to niche evolution. Evolution 62:2868- 2883. Wiens, J. J. 2004. What is speciation and how should we study it? American Naturalist 163:914-923. Willis, J. H. 1993. Effects of different levels of inbreeding on fitness components in Mimulus guttatus. Evolution 47:864-876. Wisz, M. S., R. J. Hijmans, J. Li, A. T. Peterson, C. H. Graham, and A. Guisan. 2008. Effects of sample size on the performance of species distribution models. Diversity and Distributions 14:763-773. Wu, C. A., D. B. Lowry, A. M. Cooley, K. M. Wright, Y. W. Lee, and J. H. Willis. 2008. Mimulus is an emerging model system for the integration of ecological and genomic studies. Heredity 100:220-230. 136 CHAPTER 3: THE EVOLUTION OF REPRODUCTIVE ISOLATION ACROSS THE GENUS MIM UL US 137 CHAPTER 3: THE EVOLUTION OF REPRODUCTIVE ISOLATION ACROSS THE GENUS MIM UL US Abstract Identifying the forms of reproductive isolation that contribute most to total isolation at the time of speciation is a major goal of speciation studies. Two previous methods have gathered significant data on this problem, based on either 1) exhaustive treatment of multiple forms of isolation in a single pair of recently diverged species or 2) comparative study of one or two forms of isolation across multiple species pairs. These two approaches are combined in the presented study to elucidate which forms of isolation are most commonly associated with speciation in 9 pairs of species in the genus Mimulus. The barriers measured include ecogeographic isolation, temporal isolation (both in nature and in common garden), seed set isolation (potentially a combination of both garnetic and intrinsic postzygotic isolation), intrinsic postzygotic isolation based on relative hybrid fitness, and intrinsic postzygotic isolation based on relative hybrid fertility. Analyses reveal that many of the traits associated with these forms of isolation are significantly different in many species pairs, leading to large variation in how much isolation individual barriers confer. Among the strongest individual barriers are ecogeographic isolation with an average individual strength of 0.641 and intrinsic postzygotic isolation due to relative hybrid fitness with an average individual strength of 0.346. When barriers are considered within the framework of the linear sequential ordering method, ecogeographic isolation retains its individual strength as the first barrier to act throughout the life history of the organisms. However, the relative contribution of later acting 138 barriers is diminished considerably compared to their individual strength. The average relative contribution of intrinsic postzygotic isolation by fitness differences is 0.077 despite its moderate individual strength. These data confirm that early acting (prezygotic) barriers are highly important to speciation regardless of whether the linear sequential method is employed. 139 Introduction As the ultimate source of biodiversity, the process of speciation has attracted the attention of biologists from the time of Darwin (1859). Under the biological species concept (Mayr 1942), species are defined as interbreeding natural p0pulations, and it has been appreciated for many years that there are many factors that can limit gene flow in nature (Dobzhansky 193 7; Mayr 1942; Poulton 1908). Much intense debate has surrounded the issue of which of these barriers should be studied (Coyne and Orr 2004; Rice and Hostert 1993; Schemske 2000), and many would consider the ‘holy grail’ of speciation as the identification of which of these forms of isolation operates most _J commonly in nature. Forms of isolation are commonly divided into those that operate before zygote formation (prezygotic) and those that act after hybrid formation (postzygotic), and the relative importance of these two broad category of isolation has received considerable attention (e.g. Coyne and Orr 1989; Coyne and Orr 1997; Mendelson 2003). Prezygotic forms of isolation are typically based on ecological differences between species, affecting the likelihood that species will encounter each other in nature. A primary form of prezygotic isolation involves adaptation to geographically different locations, resulting in ecogeographic isolation. This form of isolation has been largely neglected in the literature (Sobel et a1. 2010). However, in the cases where it has been measured, it strongly separates many recently diverged species in nature (Sobel Chapter 2; Ramsey et al. 2003; Kay 2006). A number of other prezygotic mechanisms can also act. Temporal isolation, for example, restricts gene flow between populations or species that differ in the timing of reproduction (e.g. Dopman et al. 2010; Lowry et al. 2008; 140 Pascarella 2007 ; Yamamoto & Sota 2010). One virtually unexplored aspect of temporal isolation is that the timing of reproduction can vary both due to intrinsic and extrinsic means. Intrinsic temporal isolation is due to differences in either mating or flowering time genes. Extrinsic temporal isolation results from differences in mating or flowering times that are mediated by other factors such as habitat. For example, spatial variation in timing of precipitation may determine that flowering occurs in species A several weeks earlier than for species B (with flowering coinciding with peak soil moisture availability in each habitat). Species A and B may therefore differ in the timing of reproduction without experiencing any differences in flowering time alleles. Sexual isolation in animals and pollinator isolation in plants can also serve as a form of isolation based on mate selection. Sexual isolation is known to be strong in many cases of Drosophila, even when postzygotic barriers are absent (Coyne & Orr 1989; 1997); and pollinator isolation has been shown to generate significant isolation when plants have shifted between pollination syndromes (e.g. Cruzan and Arnold 1994; Ramsey et al. 2003; Schemske and Bradshaw 1999). Postmating, prezygotic isolation also occurs in which pollination or mating occurs, but either pollen or sperm precedence favors fertilization by conspecific gametes (Howard 1999). This form of isolation presents significant difficulties in studying, as it is often impossible to distinguish the relative impacts of gametic interactions fi‘om early zygote abortion. Postzygotic isolation occurs when hybrids are less fit than parents, either due to intrinsic or extrinsic means (Coyne and Orr 2004). Intrinsic postzygotic isolation can occur before hybrids reproduce through differences in survival and growth relative to parents, resulting in lower viability (e. g. Tang and Presgraves 2009). Hybrids may also 141 experience differences in fertility relative to parents, resulting in intrinsic postzygotic isolation based on either full or partial sterility (e. g. Phadnis and Orr 2009). Hybrids may also be unfit due to a lack of suitable ecological niche to fill. This extrinsic form of postzygotic isolation has been investigated in cases of lake populations of benthic and limnetic stickleback fish that are known to hybridize (Hatfield and Schluter 1999). One mechanism of speciation that often gets cited as a form of reproductive isolation is due to differences in mating system. For example, Martin and Willis (2007) investigated reproductive isolation between two species of Mimulus in which one member of the pair was highly selfing. While selfing was not measured directly, its impact on gene flow was inferred fi'om hybridization rates. However, selfing species are a notorious difficulty for the biological species, so including it as a legitimate form of isolation is problematic. Increases in selfing isolate individuals within a population from each other just as much as fi'om other populations or species (Coyne and Orr 2004; D. Schemske, pers. comm), so characterizing the effect it has on gene flow is not straightforward. Two approaches have most commonly been used to evaluate the relative importance of different forms of isolation. The first approach is to choose a pair of closely related species, and measure the strength of as many forms of isolation as possible to evaluate which is strongest (e.g. Ramsey et al. 2003; Chari and Wilson 2001; Husband and Sahara 2004). This method can reveal which forms of isolation are Strongest in a particular species pair, but because species pairs are generally selected non- randomly, many biases exist in which forms of isolation are likely to be encountered. An example of this approach is the work of Ramsey et al. (2003) in which Mimulus 142 cardinalis and M lewisii were found to be isolated predominantly by ecogeographic isolation and pollinator preference despite the presence of several other potential sources of isolation. An alternative approach is to measure only a limited number of isolating barriers, but to examine a large number of species pairs within a group of organisms. This approach is exemplified by Coyne & Orr’s (1989; 1997) comparative study of mating and postzygotic isolation in Drosophila. They examined the rate at which these two forms of isolation evolved by comparing the strength of each form of isolation with the time since speciation (genetic distance). This revealed that the prezygotic barrier based on mating preferences evolved faster than the postzygotic barrier based on relative hybrid fitness and fertility. This difference was primarily due to sympatric taxa, sparking a rejuvenation of interest in the process of reinforcement. While each of these approaches can help elucidate the relative importance of isolating barriers, neither are sufficient to answer the question of which forms of isolation are most important on their own. The single species pair study design may reveal the forms of isolation that are important in that pair of species, but generalizing about common trends in speciation is difficult from these limited examples. Alternatively, comparative studies of isolation in which multiple species pairs and more than one isolating barrier is measured (Coyne and Orr 1989; Coyne and Orr 1997; Mendelson 2003; Moyle et al. 2004) either rely on limited data from the literature or are so labor intensive that very few isolating barriers can be included. While these studies may reveal the relative rate at which the few included barriers arise, it is impossible to know which forms of isolation are most important when many are excluded from analysis. 143 A commonly overlooked issues in speciation research is the need to study the forms of isolation that reduce gene flow the most in nature (Rice and Hostert 1993; Schemske 2000; Sobel et al. 2010). Coyne & Orr (1989) first recognized that in calculating the strength of total reproductive isolation, postzygotic isolation can only account for any isolation that was unaccounted for by prezygotic isolation. Ramsey et al. (2003) provided an important insight that many forms of isolation can be treated sequentially because they operate at distinct stages in the life history of organisms. In this approach, late acting barriers are discounted by any barriers that affect gene flow at an earlier stage. By relating each barrier’s individual strength to the amount of gene flow that it actually has the potential to interrupt, this method allows a direct calculation of the relative contribution of each barrier to total reproductive isolation. In the presented work, the comparative approach of Coyne & Orr (1989) was combined with the single species pair approach of Ramsey et al. (2003) to ask the following: 1) Which forms of isolation have the highest individual strengths among recently diverged species pairs? 2) What are the relative contributions of different forms of isolation to total reproductive isolation? and Are there general trends in which forms of isolation are strongest? 3) Are early acting barriers more likely to interrupt gene flow than late acting barriers? I44 Materials & Methods Studv svstem The genus Mimulus consists of approximately 120 species of wildflowers, approximately 75% of which are found in western North America (Grant 1924). The genus exhibits a broad diversity in traits interesting to ecologists and evolutionary biologists including: growth form (Lowry et al. 2008), mating systems (Fishman et al. ‘ L... ..4 .F..il"flk'l1 2002; Willis 1993), ploidy levels (Vickery 1995; Beardsley et al. 2004), pollination syndromes (Bradshaw and Schemske 2003; Bradshaw et a1. 1995; Schemske and Bradshaw 1999; Streisfeld and Kohn 2005), edaphic specialization (Macnair 1983), and . I climatic tolerances (Angert and Schemske 2005). Previous phylogenetic work has elaborated relationships at broad taxonomic levels (Beardsley and Olmstead 2002), and a species-level phylogeny is available with nearly complete coverage of the North American species (Beardsley et al. 2004). A rich history of ecological and genetic work (Hiesey et al. 1971; Kiang 1972; Vickery 1959; Vickery 1964), combined with more recent advances in developing genomic resources (Wu et al. 2008) has elevated the genus to a model system in ecology and evolutionary biology, especially in studies of adaptation and speciation (F ishman and Willis 2001; 2006; Martin and Willis 2007; Ramsey et al. 2003; Sweigart et al. 2007). Selection of species pm Species pairs were selected across the genus using the phylogenetic hypothesis of Beardsley et al. (2004) (Table 3.1). The decision to include species pairs involved a combination of factors including high bootstrap support of the relationship, practical 145 ability to collect and propagate specimens, and desired broad coverage of the traditional sectional delineations (Grant 1924). In practice, closely related species pairs were identified in the phylogeny, and field collections were attempted for one or both members of the pair. Because many Mimulus species are rare, a combination of efforts for locating species was needed, including referencing herbarium records, utilizing the expertise of local forest and park service botanists, and extensive field surveys. When one member of a pair was located, increased efforts were made to locate the alternate member of a pair. The result is selection of species pairs that is not truly random across the genus, but was somewhat haphazard, and more importantly, species were selected with no prior knowledge of reproductive barriers that were in operation. Each species pair represents very recent divergence in order to ensure that estimates of reproductive isolation obtained are as close as possible to the isolation that was present at the moment of speciation. Therefore, in most cases, species pairs represent each other’s closest extant relative. However, due to phylogenetic uncertainty or inability to collect specimens in the field, some species pairs consist of two very close relatives in which a third may be more closely related to either of the two selected. For example, Mimulus kelloggii, M douglasii, and M congdonii are a trio of closely related species in section Oenoe. The phylogeny of Beardsley et al (2004) suggests that M douglasii and M congdonii are sister species. However, M congdonii proved elusive in the field, and therefore, M kelloggii and M douglasii were instead used as a species pair. In order to maintain phylogenetic independence, and avoid the need for a phylogenetic correction, it is only necessary for each of the included species pairs to be more closely related to each other than to any other species in the study (F elsenstein 1985; Harvey and Pagel 1991), 146 and this method of species selection achieves this goal (see Sobel, Chapter 2- Figure 2.1). Nine species pairs in the genus Mimulus were sampled in the field between 2005 and 2008 (Figure 3.1). Populations were located in the spring and early summer, and GPS coordinates were recorded (Table 3.1). Collections of seeds were carried out in late summer or early fall. Overview In the following sections, methods employed for measuring each form of isolation are described. The order of isolating barriers presented corresponds with the sequence in which they act in nature, progressing from barriers that act early in the life stage of Mimulus (prezygotic barriers), and concluding with barriers that act at later stages (postzygotic barriers). Finally, details are provided for how the relative importance of each barrier was measured, both by differences in absolute strengths and in their relative contribution to total isolation. Termgoral isolation Two types of data were collected to measure the strength of temporal isolation in pairs of Mimulus species. The first consisted of collection dates for specimens within the Jepson Online Interchange for California Floristics (UC-JEPS2004). Dates were collected from all specimens in the database (both georeferenced and non-georeferenced), and converted to the Julian date system. This represents the distribution of dates that each species flowers in nature, giving an indication of the total isolation experienced by species pairs based on flowering time. This measure of isolation can include both 147 intrinsic biological differences in flowering time genes, and/or differences in flowering time mediated by habitat (extrinsic temporal isolation). Because collection records do not provide sufficient information to estimate isolation using approaches that take relative abundance into consideration, isolation was estimated using the simplified R17 equation presented in Chapter 1: R I . 1_ ( Shared sp “165A Shared + Unshared A . In this equation, “Shared” refers to the Julian dates of co-flowering for species A and B, while “Unshared/1,” is the duration of the dates that species A flowered while species B did not. Gaps in Julian dates were assumed to be due to patchiness of the collection record rather than actual biology, so timing of flowering was essentially treated as a block from earliest to latest date of specimen collection. Temporal isolation was calculated separately for each species of a pair and then averaged together for a composite estimate of reproductive isolation. Distributions in flowering time were ofien non-normally distributed, so the non-parametric Mann-Whitney U test implemented in JMP version 8.0 was used to test for significant differences between species. Plants were also grown in common garden conditions in the Michigan State greenhouses to evaluate the amount of temporal isolation due to intrinsic genetic differences between species. Assays were conducted to establish the optimal gibberellic acid concentration for each species (Sobel, unpublished). Seeds were treated with appropriate gibberellic acid concentration overnight to break dormancy, rinsed thoroughly, sterilized with 25% bleach solution, and sown on Phytoblend agar (Caisson Laboratories) plates with Gamborg BS plus sucrose nutrient medium (BIOPLUS). Plates 148 were randomly assigned positions within growth chambers using settings optimized for each species pair. A minimum of 200 seeds from 6 maternal lines were sown, and each seed was given a unique identification number and followed to record day of germination and day of flowering. Upon germination, seedlings were transplanted into a mixture of Baccto High Porosity soil mix with supplemental perlite and vermiculite. After transplanting, seedlings were placed under misters for 3-4 days to acclimate to greenhouse conditions, and then were randomly assigned positions on benches. Seeds and seedlings were monitored for germination and flowering respectively three days a week until germination rates dropped precipitously. This aspect of temporal isolation was estimated using equation R13 from Chapter Ai Bi RI =l-2 , x . A 21(At0tal Ai+Bi) Many of these species can be maintained in the greenhouse indefinitely, and do not naturally senesce without being pollinated. Therefore, an arbitrary cutoff date of three weeks past the last opened flower was used as the flowering duration of the entire population. This convention for estimating temporal isolation results in a symmetrical distribution of flowering times, and therefore a single isolation estimate was calculated for each species as a pair. IsoLation due to differences in seed set Experimental hybridizations were performed on plants from the temporal isolation growouts described above. For each species pair (Table l), intraspecific crosses 149 and interspecific crosses in both directions were done. In previous experiments (Sobel, unpublished), it was determined whether each species would set seeds autogamously. For those species that do set autogamous seeds, dissections were performed to remove anthers previous to stigma receptivity. Flowers were marked as emerging buds, and only flowers that were 2-3 days past opening were used in crosses. On any given day of crossing, flowering individuals were counted to insure that a roughly equal numbers of intraspecific and interspecific crosses could be performed. Non-pollinated controls were also conducted on each day of crossing. Two independent anthers were collected from the paternal parent of each cross, and pollen was applied to the receptive stigma by directly rubbing the anther across the stigmatic surface. Because Mimulus has reactive bilabiate stigmas that close when perturbed, pollinations had to be performed quickly to ensure a saturating amount of pollen had been applied before the stigma closes. Therefore, for each cross, an additional identical pollination was performed on each flower after the stigma re-opened (10 minutes to 2 hours later). For intraspecific crosses, individuals were chosen from independent maternal lines to minimize issues related to inbreeding. A minimum of 20 interspecific, and 20 intraspecific crosses were performed for each species as maternal parent. Seeds were collected prior to dehiscence and counted. Two separate analyses of seed set were conducted. First, within each class of cross, success or failure of a pollination (either resulting in some offspring or zero offspring) was treated as a categorical variable and analyzed using contingency table analysis in JMP version 8.0. Among crosses that did produce offspring, seed number was analyzed as a continuous variable in a Mann Whitney U test in which the number of seeds produced in 150 interspecific crosses was compared to intraspecific seed set. This comparison was made within each species such that the intraspecific seed set of species A was compared to the interspecific cross between species A and B in which species A was the maternal parent. Individual values of reproductive isolation were calculated for both the failure analysis and the seed set analysis using equation R15 from Chapter 1: R1 =1-2( ” ) C+H In the failure analysis, the success rate of interspecific cross was used as the H term and the success rate of intraspecific crosses was the C term. Average seed set from interspecific (H) and intraspecific (C) crosses were used directly in this equation. The two RI values were combined to produce a composite seed set isolation estimate. Because differences in seed set between cross types could result from either gametic interactions or from early zygote abortion, these data can not unambiguously be classified as either pre- or postzygotic. Therefore this form of isolation is treated separately from other measures of intrinsic postzygotic isolation. Intrinsic postggotic isolation - Fitness Intrinsic postzygotic isolation was measured by examining hybrid fitness relative to each parent species. In the growout described above in the temporal isolation section, seeds were monitored for germination, and an overall rate of germination of hybrids versus parents was compared. For a given species pair, both parents and interspecific offspring from both directions of cross were sown on agar and placed randomly in a growth chamber. Most species were represented by at least 200 seeds per genotypic class 151 (mean of 210), but due to seed limitation, a few classes were represented by as little as 64 seeds. Because seed germination was highly variable among lines, germination data were pooled into each genotypic class (Pl, F 1(P1), F 1(P2), P2) for analysis. Fitness was set relative to 1.0 within each comparison in order to facilitate analyses across groups and life history stages. Of the seeds from the above growouts that germinated, each seedling was transplanted and monitored for flowering (described in temporal isolation section). Within each species pair, survival rate was calculated for each genotypic class by comparing the number of individuals that flowered to the total number of seeds that germinated. As before, fitness was set relative to 1.0. Of the individuals that survived until flowering, the total number of flowers produced throughout the lifetime of each individual was recorded. This measure of relative hybrid fitness was also relativized to 1.0. A combined metric of relative hybrid fitness was obtained by multiplying across each of these three life history stages (Ramsey et al. 2003). This combined hybrid fitness estimate was used in equation R15 to obtain estimates of reproductive isolation based on the relative fitness of hybrids to parents. As before, all comparisons were made between a parent species, and the fitness of hybrids produced when that species was the maternal parent in the interspecific crosses. Intrinsic postzygotic isolation - Fertility Intrinsic postzygotic isolation was also estimated through its effect on male fertility with pollen viability assays. For each genotypic class (P1, F 1(P1), F 1(P2), P2), a minimum of 8 individuals from independent maternal lines were assayed. Each individual 152 was measured 4 times, and these repeated measures were averaged to increase confidence in individual estimates. Flowers were marked in the bud stage, and anthers were collected on the first day of dehiscence. Pollen fertility was measured using a modified standard viability assay (Kearns and Inouye 1993). Anthers were placed in a microcentrifuge tube containing a solution with an optimized sucrose concentration specific to each species (ranging from 10-25% sucrose; Sobel, unpublished). In addition, the solution contained lOOmg/L boric acid, 300mg/L calcium nitrate, 200mg/L magnesium sulfate heptahydrate, and 100mg/L potassium nitrate. The tube was vortexed briefly to dislodge dehiscent pollen from the anthers and the solution was incubated at room temperature for approximately 2 hours. Two hundred pollen grains were counted per sample, and the number of pollen grains that had germinated was used as an indication of pollen viability. As with previous analyses of intrinsic postzygotic isolation, pollen viability was compared between a parent of a cross and the interspecific offspring from crosses where that parent was maternal. Mann-Whitney U tests were performed for all parent hybrid combination, and the average number of viable pollen grains from hybrids was used as the H term while the average number of viable pollen grains from parents was used as the C term in equation R15 from Chapter 1. @ubgtion of total reproductive isolation The linear sequential method of estimating total reproductive isolation that was introduced in Coyne and Orr (1989) and expanded in Ramsey et al. (2003) was employed to calculate both the total reproductive isolation experienced by each pair of species and the relative contribution of each barrier to total reproductive isolation. By this method, 153 the strength of each individual barrier is essentially discounted by the gene flow that has already been interrupted by any barriers that act previously within the life history of the organism. The relative contribution of any barrier is therefore the amount of gene flow that is actually prevented by the action of that barrier. For the first barrier that acts (ecogeographic isolation), the individual strength is equivalent to its relative contribution. For the second barrier to act (temporal isolation), the relative contribution is the individual strength of that barrier multiplied by the remaining potential gene flow after the first barrier has acted (1-R11)* R12. This process is repeated through all barriers in the analysis in the following order: 1) ecogeographic isolation, 2) temporal isolation (from collection records), 3) seed set isolation, 4) intrinsic postzygotic isolation — fitness, 5) intrinsic postzygotic isolation — fertility. Relationshifletween genetic distance and strmgth of isolation Genetic distance was calculated from molecular sequence data used to construct the phylogeny in Beardsley et al. (2004). Loci included in the analysis were trnL/F, ITS, and ETS, and the distance matrix was extracted from the maximum likelihood phylogenetic analysis conducted in PHYLIP (F elsenstein 2005) presented in Chapter 2 (Figure 2.1). Genetic distance was used as a proxy for time to evaluate the rate at which various forms of isolation and total reproductive isolation evolved. While there are too few data points to provide significant relationships by linear or logistic regression, examining the few data available may still be informative to patterns of reproductive isolation evolution. 154 Results Temmral isolation Differences in flowering time as calculated by differences in collection dates in the Jepson collection indicate that flowering times typically differ moderately between species of Mimulus (Figure 3.2). Only 2 of the 9 species pairs did not show significant differences by Mann-Whitney U test- M cusickii / M nanus (Figure 3.2 E; p = 0.074) and M gracilipes / M palmeri (Figure 3.2 H; p = 0.188). In both of these species pairs, one member is extremely rare, and therefore, the analysis was conducted on a relatively small sample size. The magnitudes of these differences in flowering time results in moderately strong reproductive isolation across the species pairs, with an average of 0.246 (Table 3.2). Common garden experiments reveal that much of the temporal isolation experienced by species pairs does not arise due to intrinsic genetic differences between species. Differences in flowering time between species are presented in Figure 3.3. Several species pairs exhibit significantly different average days to flowering (M brevipes/johnstonii, Figure 3.3C; M constrictus/whitneyi; Figure 3.3D; M cusickii/nanus, Figure3.3E; M floribundus/norrisii, Figure 3.3G; M gracilipes/palmeri, Figure 3.3H; and M jungermanm'oides/washingtonensis, Figure 3.31). While these differences were statistically significant, their biological significance is minimal. Even when a member of a species pair initiated flowering earlier than the alternate species, the two populations were co-flowering in the greenhouse for the majority of their lifespan. This resulted in very modest estimates of reproductive isolation for each of these species pairs, ranging from 0.004 to 0.092, with a mean of 0.052 (Table 3.2). 155 Several species pairs showed reversed patterns of flowering times in the greenhouse and collection data. For example, in the Jepson collection records, M constrictus flowers significantly earlier (average of June 8‘“) than M whitneyi (average of July 17) (Figure 3.2D). This difference results in a composite strength of isolation between the two species of 0.37. However, in the greenhouse, M whitneyi flowers significantly earlier (mean days to flower 46.5 :1: 12.0) than M constrictus (63.6 i 11.0) F- (Figure 3.3D). This pattern was also seen in M brevipes/johnstonii. Postmating (seed SQ) Patterns of seed set differences between experimental intra- and interspecific crosses varied highly across species pairs (Figure 3.4). Failure rates (instances of pollinations resulting in zero offspring) did not differ significantly in most crosses (Figure 3.4). However, in the species pairs M constrictus/whitneyi and M cusickii/nanus, interspecific crosses were more likely to fail than intraspecific crosses. This was consistent in both directions of the cross for each species pair. M constrictus had a failure rate of 56.5% in interspecific crosses, while only 20.6% of intraspecific pollinations failed (Figure 3.4A; N=80, x2=10.4, p=0.0012). Similarly, in the reciprocal cross, M whitneyi showed significantly higher failure rates in interspecific crosses (79.8%) versus intraspecific crosses (37.8%) (N=194, X2=35.6, p=<0.0001). This results in moderately strong reproductive isolation for this species pair with a composite metric of 0.40. In the other species pair with significant differences in failure rates, M cusickii sets fails in interspecific crosses with M nanus 61.0% of the time, while intraspecific crosses fail only 12.2% of the time (Figure 3.4 E; N=82, x2=21.0, p=<0.0001). In the 156 reciprocal cross, M nanus interspecific crosses fail 88.2% of the time, while they fail 10.3% of the time in intraspecific crosses (N=90, x2=54.2, p=<0.0001). Quantitative differences among fruits that bore seeds were also highly variable across species pairs (Figure 3.4). For every species pair at least one member of the pair exhibited significant differences in seed set between inter- and intraspecific crosses. 6 of the 18 species showed no differences between the cross types: M angustatus (Figure 3.4A, M filicaulis (Figure 3.4B), M johnstonii (Figure 3.4C), M nanus (Figure 3.4B), M gracilipes (Figure 3.4H), and M washingtonensis (Figure 3.41). M nanus should be viewed with caution however, as the high failure rate of interspecific crosses led to very few fruits with viable seeds to count. The remaining species showed significant differences in seed set between the cross types falling into two categories: A) interspecific crosses yield fewer seeds than intraspecific (isolation) and B) interspecific crosses yielding more seeds than intraspecific (heterosis). Species falling into the isolation category include: M angustatus, M bicolor, M brevipes, M constrictus, M whitneyi, M cusickii, M douglasii, M kelloggii, M floribundus, M norrisii, M palmeri, and M jungermannioides (Figure 3.4). Only M pulchellus exhibited heterosis (Figure 3.4A); however, M filicaulis and M washingtonensis also trended in this direction. Reproductive isolation estimated from the combined effects of pollination failure and quantitative differences in seed set are presented in Table 3.2. While many species pairs exhibited differences in seed set, overall isolation was relatively low, with an average RI value of 0. 1 81 across all pairs. Two species pairs exhibited strong isolation by seed set differences: M constrictus/whimeyi (RI composite = 0.664) and M cusickii/nanus (RI composite = 0.708). 157 Intrinsic postzygotic isolation — Viability Three components of relative hybrid fitness were measured to assess intrinsic postzygotic isolation due to viability. Table 3.3 gives provides germination rates, survival rates, and numbers of flowers produced set relative to l for each direction of a cross. In most cases, offspring of intraspecific crosses germinated at higher rates than offspring of interspecific crosses. The exceptions included M angustatus, in which intraspecific offspring germinated at a rate of 23.9% of hybrids and M floribundus in which intraspecific offspring germinated at a rate of 23.6% relative to hybrid offspring. Several instances of moderately strong reproductive isolation by germination rate differences exist, including M pulchellus, M bicolor, M constrictus, M cusickii, M norrisii, M palmeri, and M jungermannioides (Table 3.3). This form of isolation also gives several instances of complete reproductive isolation, as some classes of seeds did not germinate at all. These include hybrid progeny of: M brevipes, M johnstonii, M whitneyi, and M nanus. For these 4 species, further measures of reproductive isolation are therefore absent because no hybrids were produced to measure. Survival and flowering rates also differed considerably among the remaining classes of offspring (Table 3.3). In particular, despite reasonably high germination rates, no hybrid offspring M cusickii maternal crosses survived to adulthood. M norrisii and M washingtonensis hybrids also showed significant decreases in F1 survival to flowering despite reasonable germination rates. Each of these three fitness components were multiplied together (as in Ramsey et al. 2003) to produce a composite metric of relative hybrid fitness, and reproductive isolation was calculated from this value. A broad range 158 al.-n. 4.51: .. 4 of reproductive isolation values were obtained, ranging from the above-mentioned instances of complete isolation to instances of strong heterosis. In an example of the latter, M angustatus hybrids consistently outperformed conspecific offspring in all three categories, resulting in an isolation coefficient of -0.826. M floribundus also showed evidence of heterosis, with a total combined metric of -0.708 for isolation. Both of these values indicate that the relative fitness of hybrids would actually facilitate gene flow I“- rather than limit it. Intrinsic postzygotic isolation - Fertility Relative fertility of pollen reveals that several species experience reduced F 1 fertility relative to parent fertility (Figure 3.5). M angustatus, M bicolor, M constrictus, M floribundus, and M jungermannioides all show evidence of significant declines in the fertility of their interspecific offspring. In contrast, only one species, M pulchellus, exhibits a strong significant increase in pollen viability in hybrids. One species pair, M gracilipes / M palmeri shows marginally significant differences with M gracilipes showing a decline in hybrid fertility, and M palmeri showing an increase. The isolation that results from these differences in fertility mostly results in insignificant amounts of reproductive isolation, with an average strength of 0.082 (Table 3.2). Total reproductive isolation Ecogeographic isolation data presented in Chapter 2 were added to these measures of individual barriers to compare the strengths of individual barriers. Due to a lack of collection records available, ecogeographic isolation and temporal isolation were 159 not calculated for M jungermannioides / M washingtonensis. Therefore, only the later acting barriers can be included in comparisons for this species pair. Figure 3.6 shows the distribution of estimates for each form of reproductive isolation. One-way ANOVA and posthoc comparisons reveal that ecogeographic isolation is significantly stronger than either form of temporal isolation, seed set isolation, and intrinsic postZygotic isolation by fertility. It is indistinguishable from gametic isolation (due to small sample size of pollen D competition experiments) and intrinsic postzygotic isolation due to relative fitness. No other barriers are significantly different from each other by this analysis. Calculations of total reproductive isolation are given in Table 3.2. Three species show particularly low values of isolation: M kelloggii has an isolation value of 0.539 from M douglasii, M filicaulis has an isolation of 0.248 from M bicolor, and M gracilipes has isolation of 0.274 from M palmeri. The remaining have an isolation value of at least 0.8 (overall mean is 0847:1226), suggesting that the majority of isolation has been accounted for in most species pairs. Figure 3.7 demonstrates the difference between the measured individual barrier strength and the relative contribution of each barrier to total isolation. Due to the sequential nature of isolating barriers, early acting barriers have an increased potential to affect gene flow between species. Figure 3.7A shows the individual barrier strengths broken down by species pair. The individual isolation values for each species has been averaged together to create a single value of isolation per pair. There is considerable variation in which barriers are strongest in particular pairs (as also shown in Table 3.2); however, ecogeographic isolation and intrinsic postzygotic isolation due to relative hybrid fitness commonly have the highest values. 160 In calculating the relative contributions of each, the sequential nature of barriers leads to a reduction in the relative strength of late acting barriers across all species (Table 3.4). In all species pairs where ecogeographic isolation was measured (all but M jungermannioides/washingtonensis), it is the strongest relative contributor to isolation. Even cases where individual barriers were complete, such as intrinsic postzygotic isolation due to fitness in M brevipes/johnstonii or M cusickii/nanus, relatively little P gene flow ever reaches those barriers to give them the opportunity to act (Figure 3.7B) Relationship between isolation and genetic distance - While the species pairs presented in this study all represent recent cases of w speciation, sufficient variation in genetic distance occurs to examine the relationship between genetic distance and values of reproductive isolation. Figure 3.8 shows the relationship between sequence divergence and total reproductive isolation, total prezygotic, and total postzygotic isolation. While only suggestive, it appears that prezygotic isolation mediated through differences in habitat (both ecogeographic isolation and temporal isolation) are strongest in early stages of divergence, while postzygotic isolation evolves after more time has passed. Discussion Uncovering the relative importance of different forms of isolation is a major goal of speciation research (Coyne and Orr 2004); however, relatively few data are available to speculate on this important topic. The presented study offers a unique approach to answer the question of which forms of isolation are responsible for isolating species by 161 combining the comparative approach of Coyne & Orr (1989; 1997) with a detailed study of isolation between species pairs known to be recently speciated (e. g. Kay 2006; Matsubayashi and Katakura 2009; Ramsey et al. 2003; Chari and Wilson 2001). By selecting recently diverged species pairs randomly from across a phylogenetic hypothesis, much of the bias that plagues studies of reproductive isolation was avoided (Sobel and Randle 2009). In addition, choosing recently diverged species pairs helps F7 assure that the strength of barriers measured today are as similar as possible to the strength at the moment of speciation. This sampling method also avoids the need for i phylogenetic corrections on data (Felsenstein 1985), which can obscure patterns of E ; isolation evolution. By broadly sampling across the genus and placing these results into the linear sequential framework of Coyne and Cm (1989; 1997) and Ramsey et al. (2003), a general picture of the evolution of reproductive isolation in the genus Mimulus emerges. Geographic isolation clearly plays a prominent role in speciation (Barraclough and Vogler 2000; Fitzpatrick and Turelli 2006), and its sequence within life history (first to act) virtually assures that geographic differences will play a large role in limiting gene flow. However, by considering only the fraction of geographic isolation that is based on intrinsic genetic differences (Sobel et al. 2010), this work confirms that closely related species are separated not only by simple geography (Jordan 1908), but are also commonly separated by ecological differences (Stebbins 1950). This confirms that geographic differences observed in these species are truly ecogeographic differences that can legitimately be included as a barrier within the biological species concept. 162 Previous work in the genus compares favorably with the findings presented here. For example, Ramsey et al. (2003) studied the closely related species pair Mimulus cardinalis and M. lewisii. This species pair was used in Chapter 2 in order to apply the ecological niche modeling technique described there to the problem of estimating the strength of ecogeographic isolation. This verified that ecogeographic isolation was indeed very strong between these two species, and similarly to many species pairs in this study, this was by far the biggest contributor to total reproductive isolation despite some low levels of intrinsic postzygotic isolation that were detected. This study also revealed that this species pair experiences substantial reproductive isolation by differences in pollinators, resulting in isolation of 0.976 in sympatry and 0.403 relative contribution to total isolation. While substantial pollinator data for this study were not collected due to unfortunately timed drought events in the Sierra, anecdotal evidence suggests that many of the species pairs in the present study share primary pollinators (Sobel, unpublished). Therefore, despite abundant work on the genetics and ecology of pollinator isolation in this species pair (Bradshaw and Schemske 2003; Bradshaw et al. 1995; Ramsey et al. 2003; Schemske and Bradshaw 1999), this form of isolation is most likely only mildly important in the majority of speciation events in the genus. Unfortunately, very few other studies of reproductive isolation include an ecogeographic isolation aspect, limiting the potential to compare these findings with other measures. For other forms of isolation, previous studies have also shown similar findings to those presented here. The work of Martin & Willis (2007) provides an example of reproductive isolation in a different pair of Mimulus species, M. guttatus and M. nasutus. In this study, among other factors, these two species are moderately isolated 163 by differences in flowering time, which the authors speculate is driven mainly by habitat differences. Similarly, Lowry et al. (2008) found that inland and coastal races of Mimulus guttatus are strongly isolated by a habitat mediated temporal isolation. The temporal isolation measured in this study was almost completely mediated by differences in habitat. For example, Mimulus constrictus flowers significantly earlier in the field than M whitneyi, leading to moderately strong reproductive isolation. Interestingly, the pattern of flowering is reversed in the common garden conditions of the greenhouse, with M whitneyi flowering earlier than M constrictus (though the isolation that results from this difference is modest). This pattern of flowering time is consistent with the biology of species that occupy distinct altitudes. M. constrictus is found at mid- range elevations in the southern Sierra between approximately 800-2000m (2600—6500ft), and M whitneyi grows between 1500-3000m (4900-9800fi). The difference in flowering times observed in the collection records (Figure 3.2D) occurs because suitable habitat for the high elevation M whitneyi is often still under snow cover when M constrictus starts to germinate and flower. However, at high altitudes, growing seasons are constricted, and many species are expected to be under selection to reproduce earlier. Therefore, it is perhaps unsurprising that the later flowering M whitneyi has a shorter duration between germination and flowering than M constrictus when grown in the greenhouse (Figure 3.3D). In other systems, Kay (2006) found that the highest contributions to total isolation in two species of Neotropical spiral ginger came from geographic isolation and mechanical floral isolation. While mechanical isolation was not measured in this study, the similarity of floral shape and size between closely related species suggests this too is 164 I J L .T v!- unimportant in Mimulus. In a study of the isolation between diploid and tetraploid fireweeds in the genus Chamerion, Husband & Sabara (2004) found that geography also plays a significant role despite the strong postzygotic isolation involved in the ploidy shift. Previous comparative work is especially difficult to relate to the presented work because none of the studies published thus far have included an estimate of how ecogeographic isolation affects speciation, and very few include more than a single form of isolation. Comparative studies that include at least one prezygotic barrier and one postzygotic barrier include studies of Drosophila by Coyne & Orr (1989; 1997), Etheostoma fish by Mendelson (2003), and three genera of plants by Moyle et al. (2004). Both Coyne & Orr (1989; 1997) and Mendelson (2003) found that prezygotic isolation evolves faster than postzygotic, while Moyle et al. (2004) showed no significant difference between the two types of isolation. In addition, the genetic distances covered by these three studies extends far beyond recent speciation events, limiting their applicability to the presented work. Given that ecogeographic isolation has been very strong in nearly every study that has ever included it, adding some estimate of this form of isolation to these comparative studies would greatly enhance our understanding of speciation. Definitions of importance While in the presented work, the working definition of the most important barriers to speciation has been the forms that prevent the most gene flow at the time of speciation, other definitions of importance could clearly be used. For example, many would argue 165 that intrinsic postzygotic isolation deserves its historical prominence (Coyne and Orr 2004) because it has the potential to be more permanent than prezygotic forms of reproductive isolation such as ecogeographic. This may be true once isolation is complete; once an isolation of l is achieved by intrinsic postzygotic means, it is unlikely that this form of isolation will decay and be reduced in the future. On the other hand, ecogeographic isolation may go to completion, but then be reduced as species ranges ”7 change through climate changes and adaptation. Therefore, complete isolation by these two forms differ in their permanence, but incomplete isolation may have the opposite .- impact. Consider an example where two allopatric populations evolve some incomplete E : amount of intrinsic postzygotic isolation without any adaptive divergence due to habitat. If the two populations come back into contact and do not differ in ecological niche, selection may favor the elimination of segregating alleles involved in postzygotic isolation in order to increase the number of available mates rather than promoting enhancement of prezygotic isolation through reinforcement. Under this scenario, postzygotic isolation could be just as ephemeral as ecogeographic isolation if not more so. Another definition of importance of isolating barriers may hinge on considering only those barriers that completely prevent gene flow on an individual basis. For example, in the presented work, intrinsic postzygotic isolation based on relative hybrid fitness showed complete reproductive isolation for two species pairs, M cusickii/nanus and M brevipes/johnstonii. One somewhat counterintuitive result of using the linear sequential combination approach to estimating the strength of total isolation is that complete reproductive isolation (RI=1) can only fully be achieved if a single individual 166 barrier is equal to 1. Therefore, while the relative contribution of postzygotic isolation to total isolation in M. cusickii/nanus and M. brevipes/johnstonii is relatively low due to previously acting barriers, only these two species pairs achieved complete total isolation (Table 3.2) due to the action of intrinsic postzygotic isolation in preventing all gene flow. Unmeasured forms of isolation and error in estim_ation While most of the isolation potentially experienced by species pairs was captured by this study (mean total R1 of 0.85i0.26, Table 3.2), several important barriers were not measured, and could account for the remaining isolation. For example, microspatial habitat isolation can be play an important role in limiting gene flow (e. g. Lynch 1978; Howard et al. 1997; Rand and Harrison 1989). The ecogeographic isolation estimates presented in Chapter 2 measure habitat based isolation on a broad geographic scale, but will miss substantial local differences in habitat that vary at spatial scales below the grid size of the analysis. In the field, it is extremely uncommon to find two closely related species of Mimulus at the same site, unless one member of the pair is highly selfing (Sobel, personal observation). Despite extensive searching in the field, none of the species in this study were ever found at a sympatric site despite some broad overlap in geographic ranges in the ecogeographic analysis. Therefore, it is quite possible that this form of isolation both acts in nature and has a large magnitude. Another important form of isolation related to habitat that was not measured is extrinsic postzygotic isolation. Extrinsic postzygotic isolation is thought to be important in stickleback fish (Hatfield and Schluter 1999), and could play a significant role in limiting gene flow between species that can and do hybridize. The same traits that confer 167 J '- ‘1‘”... ‘L‘ WV “:3 k; .- ecogeographic isolation could also lead to extrinsic postzygotic isolation if/when hybrids are formed. In the course of performing this study, several thousand field collected seeds were grown, and no instances of obvious F 1 hybrids were ever recorded. Given that these seeds were grown under the benign common garden environment of the greenhouse, the lack of hybrids suggests that extrinsic postzygotic isolation is probably not commonly important, and lends support to the idea that there is some portion of prezygotic isolation that has not been measured. One possible explanation is that the ecogeographic isolation estimated by ecological niche modeling is underestimating the true strength. Species that might be susceptible to underestimation would be those whose taxonomy is easily confused. For example, specimens of Mimulus whitneyi are commonly misidentified as its sister species M. constrictus (Sobel. pers. obs.). While efforts were made to correct for this problem, it is possible that the some of the herbaria records used to construct the niche models created more overlap than species actual experience because of these misidentifications. This limitation of herbarium/museum records can hinder efforts to reconstruct species ranges, and points to the need for competent systematic work in maintaining collections (Graham et al. 2004). Another source of error in estimates of total reproductive isolation may arise from difficulty in obtaining accurate estimates of isolation. One form of isolation could conceivably have significant error is the germination rate component of relative hybrid fitness. Many species of Mimulus, especially those in sections Oenoe and Eunanus, have very strong seed dormancy, which requires treatment with gibberellic acid to break. For each species pair, an appropriate concentration was identified such that both parents and 168 'J‘ ...! both classes of hybrids could be germinated at a common concentration of gibberellic acid. However, in some species, especially in the Eunanus section (M brevipes/ johnstonii, M. cusickii/nanus, and M. constrictus/whitneyi) seed germination rates were chronically low and often failed unexpectedly. Germination is also a difficult measure of hybrid fitness because high germination rates in hybrids (such as in M. angustatus hybrids; Table 3.3) may be an indication of short-circuiting an adaptive dormancy trait rather than indicative of higher fitness. Three species showed markedly low total postzygotic isolation compared to the remaining species. Interestingly, biological aspects may be involved in at least 2 of the 3. M. kelloggii was estimated as being isolated from M. douglasii at a strength of 0.539. Isolation in the opposite direction (M douglasii in reference to M kelloggii) is also not complete at 0.879. One possible explanation for this species pair’s low value of isolation despite very clear morphological differentiation is that M douglasii is highly selfing, producing both autonomous seeds when unmanipulated and producing fully cleistogamous flowers under certain conditions. While selfing is often treated as a form of isolation (e. g. Martin and Willis 2007), there are considerable difficulties in using this as a form of isolation under the biological species concept because selfing individuals can be just as isolated from other members of their own population as they are from any other population (Coyne and Orr 2004). Therefore, estimating an isolation coefficient that has any meaning for interruption of gene flow is problematic. While this was probably only a problem in this species pair, many small-flowered and presumably highly selfing Mimulus species exist, and methods for dealing with these species is necessary to gain a complete picture of speciation in the genus. 169 In another example of low isolation, M gracilipes showed relatively low isolation from its closest relative, M palmeri. One possible explanation is that temporal isolation may occur between these two species on a different time scale than what was measured here. M gracilipes is a very rare inhabitant of mid-elevations of the Sierra foothills, and is extremely sensitive to variation in precipitation. The population that was used in this study was located in 2005, an El Nifio year. In 2005 it bloomed densely, forming a dense carpet of flowers (Sobel, personal observation). However, the same site was visited every year between 2006 and 2009, and not a single flowering individual was found in any of those years. If this species is restricted to only surviving to flowering in years of extremely high precipitation, temporal isolation may actually be much higher than estimated. An alternative to the possibility of unmeasured barriers limiting gene flow is that some species pairs may not actually be biological species. For example, Mimulus filicaulis appears to have very limited isolation from M bicolor despite strikingly different coloration of the corolla. The habitat of M filicaulis appears to be reasonably suitable to M bicolor according to the ecogeographic isolation estimate (Chapter 2, Figure 2.5C), and the two species are highly interfertile. While it is possible that aspects of habitat could have been missed (such as micro-edaphic differences) by the ecological niche modeling technique, it is also possible that these two species are separated due to completely historical processes. Geographic ranges are the result of both ecological and historical processes (Endler I982; Sobel et al. 2010), and the ecogeographic isolation estimated here only measures the portion of geographic isolation that is based on ecological differences between species. Therefore, these two species may occupy 170 g ’ ' a..-"-n.'t. different geographic ranges not due to biological reasons, but mainly due to stochastic processes of history. If true, these two taxonomic species may collapse if dispersal of the more common M bicolor occurs into the range of M filicaulis. These results clearly have ramifications for appropriate conservation strategies to ensure that the M filicaulis taxonomic species is not lost to gene flow. Summary The divergence of species often requires the formation of multiple forms of reproductive isolation, and revealing which forms of isolation contribute most to cessation of gene flow is a major goal of speciation research. The presented study examined reproductive isolation in recently diverged species across the wildflower genus Mimulus. The source of isolation with the highest relative contribution to total isolation was ecogeographic isolation, owing mainly to the sequential nature of isolating barriers and that it is the first barrier to act on gene flow. Ecogeographic isolation was also commonly strong on an individual barrier basis, suggesting that regardless of whether one ascribes to the sequential nature of barriers or not, this barrier deserves to attract further attention from speciation biologists. Other forms of isolation were strong on an individual basis, especially intrinsic postzygotic isolation due to These findings suggest that, consistent with Darwin’s (l 859) views, understanding adaptive divergence and its direct effects on reproductive isolation is the key to understanding the origin of species. 171 352m 0: 600.3 30 nwmdm ofl ._ 3.30 we 03.? w: .333 on Gm.mm a: .mem hm 300.0 w: .253 0m 83% w: .03.: 9:. 03.00 om_ .wvmdm mm 30.0w NE .80.? mm 502m 2N2 .85mm 00 €32 03 .30de :0 003% w: .03.: cm 03.? w: .338 2m wvofiv C— .3 Sum 0m 2.3.2 a: .0093. 0m 53.8 a: .mc—dm mm 3 bwmd: .2 30.2“ 30.0 02 .206 mm 3 med: .2 wmosm 303200000 mmO MO 5:000 0:000 .320 .2 3:23 2 000200 MO 5:000 E250 .0022 20:00—00 <0 5:000 033,—. .230 3:232 20000m <0 5:000 0:32“— .320 m 3:232 202m <0 5:000 03—00. .230 3:232 20000m <0 5:000 0:305 .320 m 3:232 20000m <0 .0550 0380 _m .0880 080.02 .0380 _0 <0 5:000 0002 620302 >230» =0:m <0 5:000 :33; .320 0 3:23 2 :03004 <0 5:000 00022 .320 ”2 3:23 2 02002 <0 5:000 0522 .3222 3:232 20000m <0 5:000 220,—. 2.03 3:23 2 20000m <0 5:000 m0_0w:< 30.— .3200 3:232 m0_0w:< <0 5:000 23:0> .320 m 3:232 3200; 304 <0 5:000 0:80—03. 320m 3:23 2 m00_mESm <0 £00000 0:32.02 .320 m 3:232 202m <0 50000 0:50.000. .320 m 3:232 0022:00m <0 5:000 082m .320 ”— 30232 202m :20200000 300:0w :2302 :230:00 023—02 02300202 00 .803 32330800 2 0000 230 3020000 0035.03 Ad 030,—. 30502330 005000030 0020:0003; 00:00 30:00:00 0000000 30:00:00 00550200 00:00 0.0000200w23003 .3 0020320500M§N .2 30.20% .3 0020000000 .2 20.200: .3 030052.83 .3 0000000 .2 030.80 .2 00:0: .30 0.20.0000 .3 3053.: .3. 05200200 .3 0:20:02 .3. 0020000 .3. 0380.0 .2 0200.5 .3. 0:-0&0~0& .2 05000038 .3 :230m 3030 2 00w0203 3200mm 172 C :30 2:0 2 2223 2:0me20 02030022 0: 000 2020200022000: .22 \ 0020.5:02:0M§.~ .22 00202 :0: 3000 3:0: :002 ...... N 02920 20.2 __. 173 @000 $000 8000 2000.00 8000 . . . . . . . :30 300 0:20 3:0 3:0 00 00 30 0 :3 00 03 0 03 00 :0 0 02.0 .03 so: ~20- 2:0- 03. $2 0000- 0:00 :27: S7: 00.220228... .2 .200 8:0 80.0 .22 2:0 0:00 32 $2 80.02.55.022 2 000.0 02.0- ~30 $2 $20 :000 0.00.0 300 0.2.000 2 030 3:0 030- .22 2:0 :000 000.0 000.0 800.020 2 000.: $2 000.: <2 0:0 000 000.0 02.0 00.0.. 2 000.0 300 000.0- .22 02.0 00.0 300 000.0 0000:0022 2 0.3.0 300. 080 S2 0.2.0 :00 0: :0 000.0 000.22 2 050 000.0 08.0 <2 09.0 :00 020 300 0800.50 .2 000.: $2 000.: S7: 000.0 R00 000 20.0 00.0.. 2 000.: $2 000.: .22 000.0 20.0 000.0 :20 5.0.0.5 2 000.: $7: 000.: $2 020 20.0 :20 030 .0052... .2 200 3:0 3:0 32 :000 20.0 030 2.00 30:00.8 2 000.: S2 000.: $2 00:0 000 2:0 030 0:20.02 2 000.: $2 000.: S7: :00 000.0 :00 0:00 30.32.: .2 020 8:0- :20 0~:.0 080- 30.0 000.0 0:00 038.0% 2 :000 2:0 300.0 :20 20.0 000.0 300 0000 8:82 2 020 E0- 300 080 200- 000.0 :00 0000 3:20:02 2 20.0 03.0 030- 020 E0 0000 3:0 :000 30:03.8 2 00:29:03 Gwen—flab chow-am 9:380:03 «nomad—0mm o Gown—0mm coma—0mm nomad—0mm c3332 GOEEOQV €050 0mg 0 gnaw mofioom 3: H 20002300 00035300 0:00:30 :03 000m :23_0m: . _ . E . m . . 3:09:0H -00w00m GEE-5:: DEE-an: 3.89th- .<\2 02>» 003202 0.3 000:: 00 00: 0300 0202200002 2223 :00 32025 .0202: 2 32003 .20 230 2 220.30 02800022 3002202 .20 £223 20 .20 003230 .N.m 030k. Table 3.3. Relative parent and F1 viability in Mimulus species pairs. For each stage in life history (germination, survival to flowering, and number of lifetime flowers produced), fitness values were set relative to l in each direction of a cross between species. Total relative viability was calculated as the product of the independent stages. Instances where earlier stages showed zero hybrid fitness prevented further estimates and N/A is listed. Species Germination Survival Flowers Total rate rate produced viability M. angustatus 0.239 0.751 0.531 0.096 F1 (angu) 1.000 1.000 1.000 1.000 M. pulchellus 1.000 1.000 1.000 1.000 F1 (pulch) 0.556 0.888 0.533 0.263 M. bicolor 1.000 1.000 0.754 0.754 F1 (bico) 0.672 0.985 1.000 0.662 M. filicaulis 1.000 0.936 1.000 0.936 F1 (fili) 0.809 1.000 0.699 0.565 M. brevipes 1.000 1.000 N/A 1.000 F] (brev) 0.000 0.000 N/A 0.000 M. johnstonii 1.000 1.000 N/A 1.000 F1 (john) 0.000 0.000 N/A 0.000 M. constrictus 1.000 0.694 1.000 0.694 F1 (const) 0.522 1.000 0.891 0.465 M. whitneyi 1.000 1.000 1.000 1.000 F1 (whit) 0.000 0.000 0.000 0.000 M cusickii 1.000 1.000 N/A 1.000 F1 (cusi) 0.542 0.000 N/A 0.000 M. nanus 1.000 N/A N/A 1.000 F1 (nanus) 0.000 N/A N/A 0.000 M. douglasii 1.000 1.000 1.000 1.000 F1 (doug) 0.898 0.866 0.772 0.600 M. kelloggii 0.911 1.000 1.000 0.911 F1 (kell) 1.000 0.967 0.906 0.876 M. floribundus 0.236 1.000 0.702 0.166 F1 (flor) 1.000 0.972 1.000 0.972 M norrisii 1.000 1.000 1.000 1.000 F1 (nor) 0.195 0.302 0.000 0.000 M. gracilipes 0.907 0.819 0.612 0.454 F1 (grac) 1.000 1.000 1.000 1.000 M. palmeri 1.000 1.000 1.000 1.000 F1 @alm) 0.325 1.000 0.840 0.273 M. jungermanm'oides 1.000 1.000 0.870 0.870 F1 (iunger) 0.460 0.797 1.000 0.367 M. washingtonensis 1.000 0.493 1.000 0.493 F1 (wash) 0.971 1.000 0.763 0.741 174 D. Ly. 5E3 Sod shod owed wwod God we 8:35:00 033—2 owfio>< 5:.8 o :86 39o $3 $3 $9 :2. =3?me Mme; 90:38.3 cons—0mm @855 coca—0mm coca—0mm couflofl ozowmfimom ocowmfimom 8m 38 gas—om :ouflom— .8388. ofiqfiwoowoom 653me .«o 88on Bwbfiv £258. 5 sous—8m o>tosvoao~ _82 8 33:55:00 2528 98 5mg? BEE 1263265 mo :859200 in 03.5. 175 .335 £5 5 now: 33:52 macaw o5 E 360% go mtg newbie onoom .—.m unsur— mmncmceucEmEs .2 822::ogmuci .2 .5223 .2 83:6er .2 Etc: .2 magneto: .2 2.69:3 .2 5.83:8 .2 3 to: .2 .5633 .2 33:23 .2 33.538 .2 £382.: .2 328.5 .2 3:233 .2 “28396 .2 ......5; _"3...._...:.,..%. a . 3:": .32. 176 A 320« B 320: 280- 280< E 240‘ ‘5 240: a 200: I. ‘2 200: E 160‘ __. 32 160- 2 120- I 2 120- 804 80 401 . 40 . . . M. angustatus M. pulcheI/us M. bICO/Of M. fillcauhs N 51 54 N 176 34 mean (31. dev.) 133.5 (25.3) 160.6 (30.7) mean (st. dev.) 155.3 (23.4) 170.4 (238) MW U 2015.5 MW U 4544.5 p <0.0001 p 0.0032 C 3201 D 320- 280- 280- __._. u - m - 1'6 240- E 240- D I — O I c 200_ F— : 200_ g 160- g 160- 3-. 120: 3 120: 804 30- 40 , 1 . _, 40 . . . . M. brewpes M. johnstonii M. constnctus M. Whitney: N 450 93 N 126 91 mean (st. dev.) 139.4 (30.1) 178.8 (30.1) mean (st. dev.) 159.3 (20.8) 198.0 (31.0) MW U 39200 MW U 13928 p <0.0001 p <0.0001 Figure 3.2A-D. Differences in flowering time based on herbarium collection for species pairs Mimulus angustatus /pulchellus, M. bicolor /filicaulz's, M. brevipes /j0hnst0m'i, M constrictus / whitneyi. Specimen collection dates from the Jepson Online Interchange (UC-JEPS2004) were compiled, and differences between species pairs were analyzed by Mann Whitney U test. Boxes represent the 25th , 50th , and 75th percentiles, while the stems show the 10th and 90th percentiles. Means are shown with the connected line between species in a pair. All species pairs show significant differences in average flowering time except M. gracilipes / M. palmeri (H). 177 E 320- F 3201 u 280- 280- ?B’ 240; g 2404 o - :1: ; c 200‘ a c zooq .2 160‘ t: g 160- 3 ‘ :1 ‘ —. 120- —~ 120- 80‘ 80: 40 . .. , 40 .. .. M. cus:cku M. nanus M. douglasu M. kelloggii N 8 105 N 110 158 mean (st. dev.) 192.8 (21.6) 180.07 (22.0) mean (st. dev.) 95.7 (24.4) 119.2 (26.9) MW U 616 MW U 10487 p 0.074 p <0.0001 G 320'“ H 320: 280‘ a 280: ‘6' 240- E 240- 200~ . ‘2 200- _— N 160~ .2 160- ’ — — I": 2 120— .2 120: %—’ f 80~ 804 40“ I 40.1 . . I - M. flon’bundus M. norrisii M. gram/mes M. palmen N 471 20 N 14 156 mean (st. dev.) 170.8 (48.6) 109.8 (23.5) mean (st. dev.) 123.6 (13.9) 133.1 (27.9) MW U 1386.5 MW U 963 p <0.0001 p 0.188 Figure 3.2E-H. Differences in flowering time based on herbarium collection in the genus Mimulus. Specimen collection dates from the Jepson Online Interchange (UC-JEP82004) were compiled, and differences between species pairs were analyzed by Mann Whitney U test. Boxes represent the 25th , 50th , and 75th percentiles, while the stems show the 10th and 90th percentiles. Means are shown with the connected line between species in a pair. All species pairs show significant differences in average flowering time except M. gracilipes / M. palmeri (H). 178 or—n — ——— ————-—————7--- O) '- 1 A -— B d" c- C 3'0”! Eat?! 3:”, W H,_, .' 0 ° 1 :3" .. M. angusratus. i. M bicolor . ' m g m g . ~M. pulcheuus ~ -M. filicaulis a: o'j 5," RI=0004 " . RI='-v0.005fl .- o. . ...... . e ' i a“ -‘. ‘30 a°1 ' 7.7» -. -. _ 1 _, . 0° go M. brewpes g g . _ ' "M. johnstoniig 8 . _ .... ° ° RI = 0.098 N; g iii-mu ' ' . 3".“- I v I . ‘ ‘7 . 9° ; 0 1o 20 30 40 50 50 700 10 20 30 40 50 60 0 20 4o .60 days to flowering days to flowering days to flowering 0,3175 "H"? p—r- " E””‘""__"}—fi—"“ 7.8 “"7 " "‘ 1;“ . c - I . .0. 8. .7 .5 g ‘4 .J I . 4.3- -‘ 3 t J. ’0' (‘6‘ F o ‘0 ‘ - o o "A s a: o . f g .-" s .- ‘ r ‘ o n o D- a d .0 1; s .‘ i 3 N. - g; M. oonstn'ctus ' -' '° M 0090“” " _- a M. douglasg’fiiiii E o 1 .' V 1' M. whitneyi _ , -' M. nanus 1 . M. “”099” a o J I ' RI _-. 0,073 . ... : ; RI 2 0.073 ' :d— 1 R] =2__0r1_7fl!§ o 20 4o 60 80. 100 1200 20 40 60 80. 100 1200 20 40 60 80_ 100 120 days to flowering days to flowenng days to flowering O r. ______ r“ ,_' «1 «IE—_- 1 "-21- ’ 2’ 1 G .- ..e H h l I 9' - .5 2 J :- .~' m‘n'"g 1 6:9 - g i :0 mg _n" - ‘ a. .0 O I 0 —, I . n a: g 1 - y, . o. 1 " '5‘ . o ‘- 1 O . ——-— ‘9 *0 a 0' 4: .' - M. flonbundus, . . . - 3 : .° x, 332:; j I o M. gracilipes . , _-' x M . E 24 " .f“ . 2:42-“- . ° _ °M. palmeri -' { M11533? 8 °j .' .— 3 we ,. L_fl.‘-;9;QPJ_J i =3 : LR’ = 0013 § _ o - ‘ _ o k‘“* * M i o vm—r'*'* y-—~- . t r .. . "w f—‘w- —v-——-*’ 7‘ ‘**"v r y r v 7': o 20 4o 60 _80 100 o 20 40 60. 80 0 20 40 60 .80 100 days to flowenng days to flowering days to flowenng Figure 3.3. Comparisons of flowering time differences between Mimulus species under common garden conditions in the greenhouse. For each pair of species in the study, the cumulative proportion of plants flowering is plotted against days from germination to flowering. Statistical comparisons by Mann Whitney U reveals that many of these comparisons are statistically different (p<0.05; C, D, E, G, H, I); however, reproductive isolation that results from these differences is relatively modest. 179 Seed set Seed set 1.0 : -0.S interspecific (201 intraspecific (27) M. angustatus, U=404, p=0.104 interspecific (23) ' intraspeclfic (20) M. pulchellus, U=338, p=0.0134 1.0 6%. L05 intersfecific (46) ' intrafpecific (61) M. bicolor, U=2180, p=0.0557 215 r A interspecific (4 S) intraspecific (58) M. fi/icau/is, U=2634, p=0.2067 1.0 400- 3504 300- 250- 200- 150- 100- 50‘ <> 1 - <> C * 0 intersfecific (18) ' intrasfecific (26) M. brevipes, U=260, p=0.0006 intersficific (17) ' intraspecific (13) M. johnstonii, U=202, p=0.999 Figure 3.4A-C. Postmating reproductive isolation in the species pairs Mimulus angustatus /pulchellus, M bicolor /filicaulis, and M brevipes /j0hnst0m'i. Wide horizontal lines represent the average seed set from crosses resulting in progeny for both inter- and intraspecific crosses (left vertical axis). Sample sizes are in parentheses. Narrow horizontal lines indicate standard deviation, and the vertices of diamonds represent 95% confidence intervals. Species names in bold indicate instances of significant differences between inter- and intraspecific seed set, and Mann-Whitney U statistic and p—value are given next to the species name. Failure rate (proportion of crosses resulting in zero progeny) is shown by a star for each cross category (right vertical axis). 180 9121 amuej 3121 6Jn|!E_-j 3191 aJn|IEj 180 1.0 160— D 140- * _n g 120- g _ * c .0 100 -0.5 76 g 80- — x a W 60‘ 8 40- e e“ e - O interspecific (20) ' intrasFecific (27) interspecific (21) ' intraspecific (57) 0 M. constrictus, U=356.5, p=0.0081 M. whitneyi, U=586, p=0.0062 110 1.0 100~ * 901 E ; 80- _ "n . 3 70- 9. ' m 60 * — - E "U 7 -0.5 ‘ °’ <> 3 0’ e w W 40" _ H 30- (D 20 a- - _ 10-1 _ * * interspecific (16? intraspecific (36) intersficific (6) f intraspecific (3S) 0 M. cusickii, U=246, p=0.0004 M. nanus, U=99, p=0.328 90 1.0 80- F 70- ._ 11 a so- 2.: g 50- 2k — - 0 s S 3 30- _ 5* 2"' 7‘ 10~ interspecific (10) ' intraspecific (12) interspecific (19) ' Intraspecific (26) M. douglasii, U=73, p=0.0062 M. kelloggii, U=273, p=0.0002 Figure 3.4D-F. Postmating reproductive isolation in the species pairs Mimulus constrictus / whitneyi, M cusickii / nanus, and M douglasii / kelloggii. Wide horizontal lines represent the average seed set from crosses resulting in progeny for both inter- and intraspecific crosses (left vertical axis). Sample sizes are in parentheses. Narrow horizontal lines indicate standard deviation, and the vertices of diamonds represent 95% confidence intervals. Species names in bold indicate instances of significant differences between inter- and intraspecific seed set, and Mann-Whitney U statistic and p-value are given next to the species name. Failure rate (proportion of crosses resulting in zero progeny) is shown by a star for each cross category (right vertical axis). 181 160 1.0 140« G — ‘0'; 120* _ 3‘ m 100— _ _ E '0 80- <>. -0 5 $3 8 e B U) 60‘- e 8 40- * * — 20- 2k 0 intergecific (50) ' intrasEcific (69) interspecific (S4) ' intraspeciftc (70) 0 M. floribundus, U=2574, p=0.0221 M. norrisii, U=2966, p=0.0395 350 1.0 300-1 H — 250- 'n 3'5 8.- W 200'4 — C '8 _ —0.5 3 <0 150- _ a W - F6 10°“..- 50- _ _ X ... .1. ’1‘ interspecific (17) ' intrasKecific (18) intersB'ecific (17) r intraspecific (21) M. gracilipes, U=269.5. p=0.2345 M. palmeri, U=249, p=0.0161 160 1.0 140- I - '- ... 120- _ — 35" 31°“ <> = , C B 80- r—O.5 F6 u q (A 60-<> 2*. (D 40- — 204 _ at: :1: O * _- interspecific (39) ' intraspecific (39) interspecific (33) ' intraspecific (26) M. jungermannioides, U=1761, p=0.0279 M. washingtonensis, U=675, p=0.1106 Figure 3.4G-I. Postmating reproductive isolation in the species pairs Mimulus floribundus / norrisii, M gracilipes /palmeri, and M jungermannioides / washingtonensis. Wide horizontal lines represent the average seed set from crosses resulting in progeny for both inter- and intraspecific crosses (lefi vertical axis). Sample sizes are in parentheses. Narrow horizontal lines indicate standard deviation, and the vertices of diamonds represent 95% confidence intervals. Species names in bold indicate instances of significant differences between inter- and intraspecific seed set, and Mann- Whitney U statistic and p-value are given next to the species name. Failure rate (proportion of crosses resulting in zero progeny) is shown by a star for each cross category (right vertical axis). 182 3.5 A B 3 >. f‘é‘ 2.5 t .12 c: 2 ‘ .22 g m 1.5 .2 ‘65 is 1 * m 0.5 01 w F1 (angu) F1 (pa/ch) F1 (bico) F1 (fill) M. angustatus M. pulchellus M. bicolor M. filicaulis N (Inter):8 N (inter):8 N (inter):9 N (Inter): 10 N (lntra): 13 N (intra): 14 N (intra):9 N (intra):9 MWhit U: 139 MWhlt U:41 MWhlt U:53 MWhit U: 103 p = 0.0003 p = 0.0006 p = 0.0047 p = 0.3068 Figure 3.5A-B. Relative pollen fertility measures in species pairs Mimulus angustatus / pulchellus and M bicolor /filicaulis. Each box contains data for a given species pair. Black bars indicate parent values while gray bars show hybrids. Estimates of pollen fertility were relativized to the value of each parent; therefore, the amount a hybrid is under or over 1 gives an estimate of the standardized change in pollen fertility. Species names in bold represent comparisons in which the intraspecific offspring differed significantly from interspecific offspring. 183 1 4 C D E 1.24 Q 'E' 1 1 .92 C :3 0.8 o a. g 0.61 .5 £2 (“f 0.4. 0.2 0 F1 (const) F1 (doug) F1 (kell) F1 (flor) M. oonstn‘ctus M. douglasii M. kelloggii 81- florlbundus N (inter): 10 N (lnter):9 N (inter): 9 N (Inter): 8 N (lntra):9 N (intra):9 N (lntra):8 N (Intra): 10 MWhit u: 71 MWhit U:68 MWhit U:82 MWhlt U=116 p = 0.1307 p = 0.133 p = 0.361 p = 00004 Figure 3.5C-E. Relative pollen fertility measures in species pairs Mimulus douglasii / kelloggii and individual species M constrictus and M floribundus. Each box contains data for a given species pair. Black bars indicate parent values while gray bars show hybrids. Estimates of pollen fertility were relativized to the value of each parent; therefore, the amount a hybrid is under or over 1 gives an estimate of the standardized change in pollen fertility. Species names in bold represent comparisons in which the intraspecific offspring differed significantly from interspecific offspring. 184 1.8 F G 1.6 ,1? 1.4 E 1.2 .93 C Q 1 2 (D 0.8 .2 E 0.6 a; 0.4 0.2 0 . F1 (grac) F1 (pa/m) F1 aunger) F1 (wash) M. gracilipes M. palmeri M. jungermannloldes M. washingtonensis N (inter):9 N (Inter): IO N (Inter): 12 N (inter): 12 N (Intra):10 N (intra):9 N (Intra):13 N (Intra):11 MWhIt U: 114 MWhIt U: 1 I4 MWhit U: 205.5 MWhit U: 150 p = 0.0546 p = 0.0550 p = 0.0077 p = 0.282 Figure 3.5F-G. Relative pollen fertility measures in species pairs Mimulus gracilipes / palmeri and M jungermannioides / washingtonensis. Each box contains data for a given species pair. Black bars indicate parent values while gray bars show hybrids. Estimates of pollen fertility were relativized to the value of each parent; therefore, the amount a hybrid is under or over 1 gives an estimate of the standardized change in pollen fertility. Species names in bold represent comparisons in which the intraspecific offspring differed significantly from interspecific offspring. 185 A BC BC BC AB C g 14 I O '5 1 g "g . 3 O . .9 0.5‘ Q . . 9 - 3 ' g a '8 o : . I i D ‘-1 ““““““““““““““““““““““““““ .— ------------- ‘- ‘1 '0 o o . :2. * . . r1 11.)--05q :, 0: . g ' o _1 O l A I A I .... l l E =' PL 8 E .3 .v 3 :8” 8 '8 E 1:: L1 8’ E E’ 8 5 “o ' " 9 8 E .. 175 D m 8 g 3. a 3 “J l— E .9 .% o g .g E 2 s ..— Figure 3.6. Individual estimates of reproductive isolation across the genus Mimulus. Each dot represents the barrier strength for a species at the given component of isolation. The barriers pictures are ecogeographic isolation, temporal isolation based on collection records (both intrinsic and extrinsic), temporal based on common garden (intrinsic only), seed set (both gametic and intrinsic postzygotic isolation), intrinsic postzygotic isolation based on viability, and intrinsic postzygotic based on fertility. Barriers differed significantly by one-way ANOVA (P: 7.686, df=5, 92, p<0.0001). Posthoc analysis of pairwise barrier comparisons using Tukey-Kramer technique provides separation of means indicated with A’s, B’s, and C’s. 186 Figure 3.7A. Individual reproductive barrier strengths in species pairs of Mimulus. Reproductive isolation data for each species was averaged to create a composite estimate for each pair. Relative seed set is treated as the first postmating form of isolation due to the inability to separate gametic from early intrinsic postzygotic isolation within this type of data. Intrinsic postzygotic isolation is separated into a viability component (relative germination rates, survival rates, and numbers of flowers produced) and a fertility component (pollen fertility assays). 187 1:] Intrinsic postzygotic (viability) E] Intr1n51c postzygotic (fertility) I Ecogeographic isolation I Relative seed set isolation I Temporal isolation 1 Ln U} in o ". o “I o 0 01609113 1911190 IenpwpuI -0.25 _ 188 Figure 3.78. Relative contribution of reproductive barriers to total reproductive isolation in species pairs of Mimulus. Reproductive isolation data for each species was averaged to create a composite estimate for each pair. Relative seed set is treated as the first postmating form of isolation due to the inability to separate gametic from early intrinsic postzygotic isolation within this type of data. Intrinsic postzygotic isolation is separated into a viability component (relative germination rates, survival rates, and numbers of flowers produced) and a fertility component (pollen fertility assays). Relative barrier contributions were calculated using the method of Coyne & Orr (1989) and Ramsey et al. (2003) in which sequentially acting barriers are discounted by the amount of isolation acting in previous barriers. 189 A .1: E g ‘8 t 7 156305 c '5 ">' 19.’ 300‘. g g v v _ 0‘10 095‘s '2 c .111 .3 £3. :51 1°“910‘°“ n 8 H o o ' s)“ ' é: -- a a > > ... U 10 . 9\ E 3 'c B U 9 $19 a) U) U) ‘_ . . 6 0- .9 0 0 E22: 0‘“? 1‘3 — 8 D- Q- 918 \\ O1 ‘9 u u ‘1. «\5 O \- 3.171171 5 “0 0’ 8. “a 1: c 605' O) E 2 T". t . ()0 . 8 a) w a a - 011° “J l— 1 H H L fifi J “0 "\ HI «3' 3‘1 3 as) 00¢» j t“ , '— U ; a“ '6‘“ ,. J’ 005‘ L— . . 7 ‘5. “(1064‘ f' N f . 0 3 51(60“ 1‘ 001‘ J w G“ ._ Y‘ 5‘0 1° 1- 099 l 019 # w r- \\5 7 011m, 1» 0‘ 5 0° (\0 J“ “\Gx‘e V 06' 9 0959‘ 099 A “0 LO Ln 0 L0 L63 cs 01 01 o 0 CI) uonelos! 1910101 uounqgnuoo 911112193 190 - Total isolation 1 ‘ ' ‘o ‘ l:l Prezygotic isolation 0.9 - ‘ I: A Postzygotic isolation : O 8" A fl _ .7- n o g 0.6- 'n D .5 0.5- I g 0.44 ma 5 0.3“ A ‘ a g 0.2- A .3 0.1d 1; O l l l T h F r l a 0 0.005 0.015 0.025 0.035 0.045 ; _' Sequence divergence (avg #subst/site) L-J Figure 3.8. Relationship between total reproductive isolation of species pairs in Mimulus and genetic distance. Linear and logistic regression show no significant relationship between the variables. However, these species pairs represent recent speciation events, so very little variation in genetic distance is available. 191 Literature Cited Angert, A. L., and D. W. Schemske. 2005. The evolution of species' distributions: Reciprocal transplants across the elevation ranges of Mimulus cardinalis and M. lewisii. Evolution 59: 1671-1684. Barraclough, T. G., and A. P. Vogler. 2000. Detecting the geographical pattern of speciation from species-level phylogenies. American Naturalist 155:419-434. Beardsley, P. M., and R. G. Olmstead. 2002. Redefining Phrymaceae: The placement of Mimulus, tribe Mimuleae and Phryma. American Journal of Botany 89:1093- 1102. Beardsley, P. M., S. E. Schoenig, J. B. Whittall, and R. G. Olmstead. 2004. Patterns of evolution in Western North American Mimulus (Phrymaceae). American Journal of Botany 91:474-489. Bradshaw, H. D., and D. W. Schemske. 2003. Allele substitution at a flower colour locus produces a pollinator shift in monkeyflowers. Nature 426:176-178. Bradshaw, H. D., S. M. Wilbert, K. G. Otto, and D. W. Schemske. 1995. Genetic mapping of floral traits associated with reproductive isolation in monkeyflowers (Mimulus). Nature 376:762-765. Chari, J ., and P. Wilson. 2001. Factors limiting hybridization between Penstemon spectabilis and Penstemon centranthifolius. Canadian Journal of Botany 79: 1439- 1448. ' Coyne, J. A., and H. A. Orr. 1989. Patterns of speciation in Drosophila. Evolution 43:362-381. Coyne, J. A., and H. A. Orr. 1997. "Patterns of speciation in Drosophila" revisited. Evolution 51:295-303. Coyne, J. A., and H. A. Orr. 2004. Speciation. Sinauer Associates, Sunderland, MA. Cruzan, M. B., and M. L. Arnold. 1994. Assortative mating and natural selection in an iris hybrid zone. Evolution 48:]946-1958. Darwin, C. R. 1859. On the Origin of Species. John Murray, London. Dobzhansky, T. 193 7. Genetics and the Origin of Species. Columbia University Press, New York. Endler, J. A. 1982. Problems in distinguishing historical from ecological factors in biogeography. American Zoologist 22:441-452. 192 Felsenstein, J. 1985. Phylogenies and the comparative method. American Naturalist 125:1-15. F elsenstein, J. 2005. PHYLIP (Phylogeny Inference Package) version 3.6. Distributed by the author. Department of Genome Sciences, University of Washington, Seattle. F ishman, L., and J. H. Willis. 2006. A cytonuclear incompatibility causes anther sterility in Mimulus hybrids. Evolution 60:1372-1381. —. 2001. Evidence for Dobzhansky-Muller incompatibilities contributing to the sterility r] of hybrids between Mimulus guttatus and M. nasutus. Evolution 55:1932-1942. .. Fishman, L., A. J. Kelly, and J. H. Willis. 2002. Minor quantitative trait loci underlie floral traits associated with mating system divergence in Mimulus. Evolution . 56:2138-2155. h .3" Fitzpatrick, B. M., and M. Turelli. 2006. The geography of mammalian speciation: Mixed signals from phylogenies and range maps. Evolution 60:601-615. Graham, C. H., S. F errier, F. Huettmann, C. Moritz, and A. T. Peterson. 2004. New developments in museum-based informatics and applications in biodiversity analysis. Trends in Ecology and Evolution 19:497—503. Grant, A. L. 1924. A monograph of the genus Mimulus. Annals of the Missouri Botanical Garden 1 1 :99-3 88. Harvey, P. H., and M. D. Pagel. I991. The Comparative Method in Evolutionary Biology. Oxford University Press, New York. Hatfield, T., and D. Schluter. 1999. Ecological speciation in sticklebacks: Environment- dependent hybrid fitness. Evolution 53:866-873. Hiesey, W. M., M. A. Nobs, and O. Bjorkman. 1971. Experimental Studies on the Nature of Species. V. Biosystematics, genetics, and physiological ecology of the Erythranthe section of Mimulus. Carnegie Institution of Washington Publication 628, Washington, DC. Howard, D. J. 1999. Conspecific sperm and pollen precedence and speciation. Annual Review of Ecology and Systematics 30:109-132. Howard, D. J., R. W. Preszler, J. Williams, S. Fenchel, and W. J. Boecklen. 1997. How discrete are oak species? Insights from a hybrid zone between Quercus grisea and Quercus gambelii. Evolution 51:747-755. ' 193 Husband, B. C., and H. A. Sabara. 2004. Reproductive isolation between autotetraploids and their diploid progenitors in fireweed, Chamerion angustifolium. New Phytologist 161:713-713. Jepson Online Interchange for California F loristics 2004. University and Jepson Herbaria, University of California. Berkeley, CA. Jordan, D. S. 1908. The law of geminate species. American Naturalist 42:73-80. Kay, K. M. 2006. Reproductive isolation between two closely related hummingbird- pollinated Neotropical gingers. Evolution 60:538-552. Keams, C. A., and D. W. Inouye. 1993. Techniques for Pollination Biologists. University Press of Colorado, Niwot, CO. ' I‘ ‘4 'II-u'YAlLI 1’ l‘ ' Kiang, Y. T. 1972. Pollination study in a natural population of Mimulus guttatus. Evolution 26:308-310. Lowry, D. B., R. C. Rockwood, and J. H. Willis. 2008. Ecological reproductive isolation L“; of coast and inland races of Mimulus guttatus. Evolution 62:2196-2214. Lynch, J. D. 1978. Distribution of leopard frogs (Rana blairi and Rana pipiens) (Amphibia, Anura, Ranidae) in Nebraska. Journal of Herpetology 12:157-162. Macnair, M. R. 1983. The genetic control of copper tolerance in the yellow monkey flower, Mimulus guttatus. Heredity 50:283-293. Martin, N. H., and J. H. Willis. 2007. Ecological divergence associated with mating system causes nearly complete reproductive isolation between sympatric Mimulus species. Evolution 61 :68-82. Matsubayashi, K. W., and H. Katakura. 2009. Contribution of multiple isolating barriers to reproductive isolation between a pair of phytophagous ladybird beetles. Evolution 63 :2563-2580. Mayr, E. 1942. Systematics and the origin of species. Columbia University Press, New York. Mendelson, T. C. 2003. Sexual isolation evolves faster than hybrid inviability in a diverse and sexually dimorphic genus of fish (Percidae: Etheostoma). Evolution 57:317-327. Moyle, L. C., M. S. Olson, and P. Tiffin. 2004. Patterns of reproductive isolation in three angiosperm genera. Evolution 58:1195-1208. 194 Pascarella, J. B. 2007. Mechanisms of prezygotic reproductive isolation between two sympatric species, Gelsemium rankinii and G. sempervirens (Gelsemiaceae), in the southeastern United States. American Journal of Botany 94:468-476. Phadnis, N., and H. A. Orr. 2009. A single gene causes both male sterility and segregation distortion in Drosophila hybrids. Science 323:3 76-379. Poulton, E. G. 1908. What is a species? Pp. 46-94. Essays on Evolution 1889-1907. Oxford at the Clarendon Press, London. Ramsey, J., H. D. Bradshaw, and D. W. Schemske. 2003. Components of reproductive .- isolation between the monkeyflowers Mimulus lewisii and M. cardinalis n (Phrymaceae). Evolution 57:1520-1534. 5 " . Rand, D. M., and R. G. Harrison. 1989. Ecological genetics of a mosaic hybrid zone: Mitochondrial, nuclear, and reproductive differentiation of crickets by soil type. . . Evolution 43:432-449. j Rice, W. R., and E. E. Hostert. 1993. Laboratory experiments on speciation: What have we learned in 40 years? Evolution 47:1637-1653. Schemske, D. W. 2000. Understanding the origin of species. Evolution 54:1069-1073. Schemske, D. W., and H. D. Bradshaw. I999. Pollinator preference and the evolution of floral traits in monkeyflowers (Mimulus). Proceedings of the National Academy of Sciences of the United States of America 96:11910-11915. Sobel, J. M., G. F. Chen, L. R. Watt, and D. W. Schemske. 2010. The biology of speciation. Evolution 642295-315. Sobel, J. M., and A. M. Randle. 2009. Comparative approaches to the evolution of reproductive isolation: A comment on Scopece et al. 2007. Evolution 63:2201- 2204. Stebbins, G. L. 1950, Variation and Evolution in Plants. New York, Columbia University Press. ' Streisfeld, M. A., and J. R. Kohn. 2005. Contrasting patterns of floral and molecular variation across a cline in Mimulus aurantiacus. Evolution 59:2548-2559. Sweigart, A. L., A. R. Mason, and J. H. Willis. 2007. Natural variation for a hybrid incompatibility between two species of Mimulus. Evolution 61 :141-151. Tang, S. W., and D. C. Presgraves. 2009. Evolution of the Drosophila nuclear pore complex results in multiple hybrid incompatibilities. Science 323:779-782. 195 Vickery, R. K. 1959. Barriers to gene exchange within Mimulus guttatus (Scrophulariaceae). Evolution 13:300-3 10. Vickery, R. K. 1964. Barriers to gene exchange between members of Mimulus guttatus complex ( Scrophulariaceae ). Evolution 18:52-69. Vickery, R. K. 1995. Speciation by aneuploidy and polyploidy in Mimulus (Scrophulariaceae). Great Basin Naturalist 55:174-176. Willis, J. H. 1993. Effects of different levels of inbreeding on fitness components in Mimulus guttatus. Evolution 47:864-876. r1 Wu, C. A., D. B. Lowry, A. M. Cooley, K. M. Wright, Y. W. Lee, and J. H. Willis. 2008. Mimulus is an emerging model system for the integration of ecological and genomic studies. Heredity 100:220-230. Yamamoto, S., and T. Sota. 2009. Incipient allochronic speciation by climatic disruption . of the reproductive period. Proceedings of the Royal Society B-Biological I J Sciences 276:2711-2719. ‘: I96 CHAPTER 4: CONTRASTING PATTERNS OF INTROGRESSION IN TWO PAIRS OF MIMUL US SPECIES JJ 197 CHAPTER 4: CONTRASTING PATTERNS OF INTROGRESSION IN TWO PAIRS OF MIM UL US SPECIES Abstract While estimates of reproductive isolation obtained through laboratory and field work provide valuable information regarding the origin of species, few studies have corroborated that the strength of isolation obtained corresponds to the amount of gene flow experienced by natural populations of closely related species. Two species pairs were selected representing a species pair with strong total reproductive isolation, Mimulus constrictus and M. whitneyi and a species pair with incomplete reproductive J isolation M. bicolor and M. filicaulis. Sequencing of neutral genetic markers was performed on multiple populations of each species, from populations that were both nearly sympatric and clearly allopatric with regards to the alternate species. Basic polymorphism analysis revealed that the completely isolated species pair (M. constrictus/M. whitneyi) has at least a single fixed difference between the two species and multiple cases of exclusive polymorphisms. Alternatively, the incompletely isolated pair (M. bicolor/M. filicaulis) shows a general pattern of many shared polymorphisms, fewer exclusive polymorphisms, and no fixed differences between species. Pairwise F ST values confirm that the loci of M. constrictus/M. whitneyi are highly differentiated, while the FST values between M bicolor/M filicaulis are much lower, comparable with differentiation between populations rather than species. Taken together, these data suggest that the low value of reproductive isolation for M bicolor/M. filicaulis obtained previously is probably accurate, and that rampant gene exchange is currently occurring (or has recently occurred. The high values of isolation obtained for M. constrictus/M. 198 whitneyi are also substantiated by the genetic analysis, suggesting that the high values of isolation are indeed resulting in a nearly complete restriction of gene flow. ‘5‘. . .. .. ' O . I. 31' I .- 0:3". ...J 199 Introduction A long-standing debate in speciation research is how often the process occurs in the face of gene flow (Bolnick 2004; Bolnick and Fitzpatrick 2007; Coyne and Orr 2004; F eder et al. 2005; Mallet 2008; Nosil 2008; Via 2001). Sympatric and parapatric models of speciation assume that some amount of gene flow occurs throughout divergence, and determining how often speciation occurs while experiencing this constraint is a subject receiving considerable attention (Bolnick and Fitzpatrick 2007). While under most circumstances speciation should be slowed or halted by introgression, some believe that introgression may even act to facilitate speciation through the formation of novel trait combinations capable of assisting a population in traversing an adaptive valley (Arnold and Hodges 1995; Hodges et al. 1996; Rieseberg et al. 2003; Rieseberg et al. 1999). These factors have resulted in a focus on studying speciation in populations that are incompletely isolated so that the effects of gene flow on divergence can be approached (e.g. Egan et al. 2008; Nosil et al. 2008; Turner et al. 2005). At the extreme, some advocate the abandonment of organismal species concepts altogether in favor of a ‘genic’ view of speciation (Lexer and Widmer 2008). While these approaches have indeed improved our understanding of how intraspecific divergence can occur in the face of introgression, an alternative point of view is that not all diverging populations will become species (Magurran 1998). Under this framework, it is important to appreciate that it may only be possible to understand the nature of speciation by studying pairs of taxa that have recently become completely isolated. For example, if the goal is to determine which forms of isolation are most important at the time of speciation, only taxa that have 200 completed speciation will be fully informative, and it is important to be able to identify the point at which species have ceased to exchange genes. Estimates of total reproductive isolation using the linear sequential ordering of barriers estimated by laboratory or field experiments of multiple isolating barriers (e. g. Ramsey et al. 2003; Sobel, Chapter 3) can offer an excellent means of identifying the endpoint of speciation. If all of the barriers likely to be important to a group are 33 measured, this approach can be thought of as a test of species status under the Biological if i Species Concept (Mayr 1942). If pairs of taxa are found to have complete (or nearly so) isolation by this method, they are diagnosed as species, and if they fall short of complete be”! isolation, they are not yet full biological species (and there is no guarantee they will ever become so). Unfortunately, high estimates of isolation measured in the lab can be misleading. For example, Llopart et al. (2005) found an example of a hybrid zone in two species of Drosophila in which a class of hybrid never produced under laboratory conditions was the primary hybrid present in nature. Similarly, Yatabe et al. (2007) found evidence for introgression in two species of Helianthus despite strong reproductive barriers. These examples indicate that in addition to estimates of reproductive isolation, some genetic corroboration of reproductive isolation is necessary to fully assess whether speciation is complete. Until recently, diagnosing speciation completion using molecular tools was difficult. Shared polymorphisms between populations were often detected in nature, but it is difficult to separate shared ancestral polymorphisms from those due to introgression. However, recent advances using coalescent simulations provides an opportunity to identify patterns of polymorphisms that either depart from or are consistent with neutral 201 expectations (Wakeley and Hey 1997; Wang et al. 1997). While these methods do not necessarily detect which loci have experienced gene flow, they can help determine if the overall pattern of polymorphism suggests that gene flow occurs at all. These methods are being increasingly utilized to corroborate measures of reproductive isolation and to estimate other population parameters important at the time of speciation (e. g. Counterman and Noor 2006; Kliman et al. 2000; Strasburg and Rieseberg 2008). F“ Mimulus bicolor/filicaulis and M. cortstrictus/whimeyi are two excellent species pairs in which measured strength of reproductive isolation can be compared with patterns of introgression through coalescent approaches. M. bicolor is a relatively widespread species throughout the Sierra Nevada, while M. filicaulis is restricted to a very small region outside of Yosemite National Park. While both have similar grth form, the flowers of the two species are distinct. The lower half of the corolla in M. bicolor is yellow and white on the upper (though some populations are dominated by an all yellow morph). In contrast, M. filicaulis is strikingly marked with purple markings and yellow nectar guides in the throat of the corolla. Despite the morphological differentiation, reproductive isolation seems to be relatively low between this species compared to other closely related species of Mimulus (Sobel, Chapter 3). There are no reproductive barriers that completely isolate the two species, and it appears that historical factors may be the key limit to species maintenance. Ecogeographic isolation is only moderately strong (average 0.56; Chapter 3), with indications that the habitat of M filicaulis would be hospitable to M. bicolor. Other forms of isolation are almost completely missing, and a total composite value for isolation between the two species is found to be 0.60 (Sobel, Chapter 3). 202 In contrast, Mimulus constrictus and M. whitneyi are much closer to complete isolation. The two species live in very different habitats, with M constrictus residing at mid elevations of the Coastal and Sierra ranges, and M. whitneyi typically growing at higher altitudes. Therefore, the two species experience limits to gene flow based on moderate to strong ecogeographic isolation (0.62), moderate habitat-mediated temporal isolation (0.37), and moderately strong isolation due to relative interspecific seed set p; (0.66) and hybrid viability (0.60) (Sobel, Chapter 3). A total composite index of approximately 0.97 makes this species pair much more completely isolated than M bicolor/filicaulis. These two species pairs present an opportunity to corroborate or falsify the estimates of isolation obtained in the lab and greenhouse (Sobel, Chapter 2 and 3). My hypothesis is that, while distinct in nature, the lower values of isolation suggest that M bicolor/filicaulis are much more likely to experience gene flow with each other than the alternate species pair. Given their high value of reproductive isolation M constrictus/ whitneyi should exhibit much lower levels of introgression. I here provide a multilocus test of introgression for both of these species pairs to assess whether estimates of isolation provided in Chapter 3 are supported by genetic evidence. Materials & Methods Collections & population sampling Two species pairs were selected for this study, Mimulus bicolor/filicaulis from section Paradanthus, and M. constrictus/whitneyi from section Eunanus. Each species in a pair represents each others’ closest extant relative, with limited time since speciation 203 (Sobel, Chapter 2). Collection of plant material for DNA isolation was conducted in 2005 and 2006. For the first species pair, seven populations of Mimulus bicolor was collected from across a majority of its natural distribution, spanning latitudes of approximately 37.0°N to 40.2°N (populations Bl-B7; Figure 4.1A). M filicaulis occurs over a much more limited extent than M bicolor, occurring in only a few populations near the western border of Yosemite National Park. Attempts were made to collect as many isolated populations of this species as possible, but given the close proximity among sites, they may all behave like a single population. Five populations of M filicaulis were collected (Figure 4.1A). For the purposes of this analysis, the two M bicolor populations flanking the range of M filicaulis were considered the most likely to exchange genes and therefore were classified as ‘sympatric’ (B4 & B5; Figure 4.1A), while the remaining populations were classified as ‘allopatric’ (Bl-3; B6-7; Figure 4.1A). This classification may be somewhat misleading, as these two species never co-occur at a single site (Sobel, pers. obs.), and may instead be thought of as ‘near’ and ‘far’ populations. In the second species pair, Mimulus constrictus was collected from three sites across its range, nearly spanning the northern and southern limits of the species (approx. 34.8°N to 36.5°N). Similarly, M whitneyi was collected from three populations spanning its rather limited geographic extent in the southern Sierra (approx. 36°N to 37°N) (Figure 1.4B). One population of M constrictus (C3; Figure 4.1B) occurs in the transition between the Coastal and Transverse ranges in southwestern California in an area where M whitneyi is not found. This population was classified as ‘allopatric,’ while the other two populations of M constrictus occurring in mountain ranges with M whitneyi were classified as ‘sympatric’ for purposes of analysis. As above, these classifications are more 204 related to geographic proximity than actual sympatry or allopatry because populations of these two species do not co-occur in nature either (J. Sobel, pers. obs). In efforts to maximize the effectiveness of parameter estimation in coalescent simulations, a minimum of six individuals per population were collected from each population (Felsenstein 2006). Leaf material from each plant was clipped and immediately placed in silica gel to desiccate the sample. Within three weeks of collection, DNA was isolated from the dried leaf material using MP Biomedicals a ‘1‘.- “.."" ‘. FastDNA kits and F astPrep tissue homogenizer. DNA was suspended in water and stored at -20C. J Selection of loci for sequencing Molecular markers were obtained through a collaboration with J. Willis and colleagues at Duke University. They have produced an EST (expressed sequence tag) library for one species of Mimulus, M gutattus. EST sequences have been blasted against the Arabidopsis genome in search of homologues of the Mimulus guttatus ESTs, and EPIC (exon primed, intron crossing) primers are developed that amplify introns within genes that the two species have in common. Primers that successfully amplify a product in M guttatus have been made available on the public website, www.mimulusevolution.org (Wu et al. 2008). Because these primers amplify putative introns, the resulting sequences are assumed to be non-coding, and are treated as neutral nuclear genetic markers. Fifty primers were selected from among those that have successfully amplified products in M guttatus, and have also been shown to amplify in Mimulus cardinalis and 205 M lewisii (T. Bradshaw, pers. comm). Marker development in M bicolor/filicaulis and M constrictus/whitneyi was carried out to identify PCR conditions that successfully amplify these four included species. Eleven markers were selected for Mimulus bicolor/filicaulis (Table 4.1A), and seven were selected for M constrictus/whitneyi (Table 4.1B). Markers were selected that both amplified easily and spanned multiple linkage groups in a map of M cardinalis/lewisii (Table 4.2) (T. Bradshaw, unpublished). While linkage maps are not available for either M bicolor/filicaulis or M constrictus/whitneyi, selecting markers that occur on multiple linkage groups in M cardinalis/lewisii helps assure that the genomes of these species pairs will be effectively represented. PCR and sequencing DNA from field-collected tissue was amplified using standard PCR conditions with slight modifications to anneal temperatures as determined in optimization. Primer sequences used are provided in Table 4.2, and more details about each marker can be obtained through www.mimulusevolution.org. Amplified PCR products were run on ~2% agarose gels stained with SYBRSafe® DNA gel stain (www.invitrogen.com). PCR products were visualized by exciting gels with ~300nm blue light and observing through an orange emission filter. Agarose gels were poured with an extra row of wells, such that DNA was loaded into one set of wells and run toward the other set. The desired PCR product was monitored visually until it entered the second set of wells, upon which time the sample was recovered. Alternate lanes were used in loading to assure no cross contamination occurred between samples. DNA from recovered products was 206 precipitated by treatment with 3M sodium acetate pH 5.2, 100% ethanol, and incubation on ice. DNA was pelleted by centrifugation and resuspended in water after drying. 9uL of DNA sample was combined with 30pmol (in 3ul) of the appropriate forward primer, and sequencing was performed on an ABI 3730 Genetic Analyzer by the Michigan State University genomics center (www.rtsf.msu.edu). Sequence analysis Raw chromatograms were checked for quality manually using geneious 5.0 bioinformatics software (Drummond et al. 2010). Ambiguous base calls were corrected by hand, and alignments were performed in Clustal W version 1.82 (Thompson et al. 1994) using default parameters. Processed alignments were extracted into PHYLLIP format (Felsenstein 2005) for further analysis. Basic polymorphism analyses were conducted with the software SITES (Hey and Wakeley 1997; distributed by J. Hey; http://genfaculty.rutgers.edu/hey/software#SlTES). SITES provides data for each marker on the total number of polymorphisms, and further divides polymorphisms into exclusive, shared, and fixed differences between subgroups. Fixed polymorphisms are those that are polymorphic with respect to the entire group being analyzed, but are not polymorphic within groups and are fixed for different base pairs. Shared polymorphisms are those that are segregating within and between groups. Exclusive polymorphisms are those that are polymorphic in one subgroup, but fixed in another subgroup. For these analyses, each species pair contained one species that had both ‘allopatric’ and ‘sympatric’ sites with respect to the other species in the pair. Polymorphism analysis was conducted separately to compare allopatric and sympatric 207 populations to the alternate pair, such that the following comparisons were made (species A is the more widespread of the pair): 1) allopatric populations of species A vs. species B, 2) sympatric populations of species A vs. species B, and 3) all combined populations of species A vs. species B. Distributions of polymorphisms were used to estimate per locus mutation rates and recombination parameters, and pairwise FST values were calculated for allopatric, sympatric, and total populations. 3 Tests of neutrality were conducted using HKA software (distributed by J. Hey; http://genfaculty.rutgers.edu/hey/software#HKA), which uses coalescent simulations to implement the Hudson, Kreitman, and Aguidae (HKA) statistical test for selection on loci U (Hudson et al. 1987). Actual values of Tajima’s D (1989) obtained from the polymorphism analysis were compared to simulated distributions in HKA from 10,000 coalescent simulations to test for significant departure from neutrality. Significant positive values of Tajima’s D are consistent with balancing selection, estimates that do not differ from zero are assumed to be neutral, and significant negative values are consistent with excess rare variants. Because the markers employed in this study are introns, significant departures from expectations under neutrality are likely to arise by introgression, and can be taken as evidence for gene flow between species. Values estimated from observed polymorphisms in SITES analysis were used to test for significant departure from an isolation model of speciation implemented in WH software (distributed by J. Hey; http://genfaculty.rutgers.edu/hey/software#Wl-I) (Kliman et al. 2000; Wakeley and Hey 1997; Wang et al. 1997). Mutation and recombination rates generated in polymorphism analysis were used to parameterize 1000 coalescent Simulations, and both x and WH test statistics were examined to test for Significant 208 departures from randomness among shared polymorphisms. The simple speciation model fit is one where zero migration occurs, and significantly higher amounts of shared polymorphisms is consistent with introgression after speciation. M Within the Mimulus bicolor/filicaulis species pair, patterns of shared and h exclusive polymorphisms do not indicate a qualitative difference between ‘allopatric’ and ‘ ‘sympatric’ comparisons of M bicolor with M filicaulis (Table 4.3A). If gene flow between the two species was higher in ‘sympatric’ populations, the expectation is that shared polymorphisms would be higher and exclusive polymorphisms lower in sympatric populations. However, there is no evidence of differences among the distribution of these polymorphism types (Table 4.3A). There are no fixed differences between M bicolor and M filicaulis in any subgrouping suggesting either very recent divergence or high levels of gene flow. FST values confirm the lack of differentiation between subgroups of M bicolor and M filicaulis (Table 4.4A), with no significant differences among groups (one-way ANOVA; p=0.756). Included in this comparison is M bicolor in allopatry to other M bicolor in the region of sympatry, suggesting that M filicaulis is no more differentiated at the sequences analyzed than M bicolor is from other populations. Several loci show evidence of departure from neutrality as revealed by HKA analysis (Hudson et al. 1987). Actual values of Tajima’s D calculated from polymorphism analysis show significant departure from null distributions provided by coalescent simulations. In Mimulus bicolor 7 of the 11 loci depart from a random expectation and 3 of the 11 depart in M filicaulis (p<0.05; Table 4.5A). The extremely 209 low levels of differentiation made fitting an WH isolation model difficult because the model does not perform well when expectations of fixed differences approach zero. Of 1000 coalescent simulations attempted with mutation and recombination parameters from polymorphism analysis only 46 yielded estimates of expected distributions of polymorphisms across all types (shared, exclusive, and fixed differences). The other 954 simulations ended with no estimates due to null values in the expected fixed differences . . 2 . . category. While comparing x and WH test statistics generated from the actual data to only 46 simulated values is not an appropriate statistical test, both actual values were higher than any of the simulated values, suggesting that the isolation model can be rejected in this species pair. Patterns within Mimulus constrictus and M whitneyi differ markedly from M bicolor/filicaulis. While polymorphism analysis reveals little in the way of fixed differences (only 1 fixed difference in total analysis; Table 4.3B), exclusive polymorphisms outnumber shared polymorphisms by more than 2.5 to 1, providing much higher levels of differentiation between subgroups within this species pair. While exclusive polymorphisms and shared polymorphisms do not differ significantly between allopatric and sympatric pairings of M constrictus with M whitneyi (p=0.482 and p=0.996 respectively), introgression is suggested by the fact that there are 7 fixed differences when comparing allopatric M constrictus to M whitneyi, while there are only 2 fixed differences between sympatric M constrictus and M whitneyi. FST values indicate that there is no difference in average F ST between allopatric M constrictus and M whitneyi (mean Fs1~=0.512 (0.21)) and sympatric M constrictus and M whitneyi 210 (mean F 31:0.466 (0.056)) (Table 4.4B; p=0.59). However, individual loci show interesting patterns of differentiation. For example, MgS T S 558 has an FST value of - 0.008 when comparing allopatric populations of M constrictus to sympatric populations of the same species, but when comparing to M whitneyi, it has much higher levels of F ST (0.255 in allopatric and 0.377 in sympatric comparisons). MgSTS79 shows a similar h pattern (Table 4.4.B). Departure from neutrality as revealed by HKA tests shows that a handful of loci have more negative values of Tajima’s D than expected from null simulations (Table 4.5B). These include M constrictus copies of MgSTS 46, MgST S 79, and MgST S 595. While significant negative D values indicate an overabundance of rare alleles, the fact that these three loci do not exhibit relatively low FST values indicates that gene flow is most likely not responsible for the pattern. Due to the higher levels of differentiation, simulations performed under WH isolation model performed much better in this species pair. In the total analysis (both allopatric and sympatric M constrictus compared to M whitneyi), of the 1000 simulations, 322 provided estimates to compare actual values of x2 and WH test statistics. In both cases, calculated test statistics for the actual data did not differ significantly from the simulated data, indicating that an isolation can not be rejected (x2=39.718, p=0.506; WH=12.00, p=0.640). The WH isolation model was also tested for only sympatric M constrictus vs. M whitneyi. In this simulation, 382 informative estimates were made from the 1000 runs, and similarly, an isolation could not be rejected (x2=42.146, p=0.644; WH=10.00,p=O.471). 211 Discussion Estimates of reproductive isolation suggest that Mimulus bicolor and M filicaulis are incompletely isolated (Sobel Chapter 3). When two species show incomplete isolation in the lab, two alternative exist: 1) forms of isolation that were not measured are maintaining species and/or 2) gene exchange is occurring at a rate related to estimates of isolation. In the recently diverged species pair M bicolor and M filicaulis it appears that the second alternative is at play. Polymorphism analysis and coalescent simulations provided evidence that introgression is occurring between these two species, which corroborates the relatively low value of reproductive isolation obtained for this species pair. Indeed, along with evidence from isolation estimates, additional phylogenetic work on this species pair may reveal that there is insufficient genetic differentiation to categorize these two taxa as distinct biological species. Given the rarity of M filicaulis, these results, along with evidence for a near total lack of crossing barriers (Sobel, Chapter 3), suggests that conservation strategies are necessary to insure that this taxonomic species is not lost due to hybridization with M bicolor. While the total analysis suggested rampant gene exchange in Mimulus bicolor and M filicaulis, individual sequences showed distinct patterns. For example, locus MgSTS 133, showed significantly higher values of F ST for between species comparisons than within M bicolor, suggesting that this locus may be resisting introgression due to selection. In addition, locus MgSTS 46 showed extremely low levels of polymorphism relative to other sequences in this study, suggesting a history of purifying selection at a nearby locus. However, polymorphism was low in both species of this pair, so selection near this locus preceded divergence. 212 Estimates of reproductive isolation between Mimulus constrictus and M whitneyi suggest that this pair of species is closer to complete isolation than M bicolor and M filicaulis. However, the estimated strength of 0.97 leaves the possibility that the remaining isolation could either be accounted for by unmeasured barriers, or could allow some amount of introgression to occur between the species. The overall analysis provided evidence that all loci included in the study are substantially differentiated between these species. Coalescent simulations failed to find evidence for a pattern of polymorphism that departs from neutrality, but the pattern of fixed differences in allopatry and sympatry suggests that these two species may sometimes hybridize. The allopatric population of M U i constrictus had seven fixed differences relative to M whitneyi, and these differences were ‘1 reduced to just a single fixed difference at MgST S 130 when comparing all M constrict‘us sequences to M whitneyi. This means that the fixed differences in allopatry must be exclusive polymorphisms in sympatric populations of M constrictus, and in sympatry the exclusive polymorphism consists of the two base pairs that are fixed in the alternate allopatric and M whitneyi samples. This suggests that sample size and power may be responsible for failing to find significant departures from neutrality for this species pair, and additional sequencing may be necessary to fully investigate introgression in this species pair. However, the overall result is consistent with the measured isolation. In previous studies, it has been common to find discordance between measures of isolation and estimates of gene flow. For example, in a study of hydrenid beetles, Urbanelli (2002) found that isolated populations showed significantly higher gene flow than expected based on the isolation estimated by geography, and that adjacent populations showed lower levels than expected. This example highlights the importance 213 of using geographic isolation carefully as a measure of isolation. While complete allopatric separation may prevent the flow of genes, only geographic differences based on genetic differences are bound to maintain any degree of permanence over time (Sobel et al. 2010). Therefore, these ecologically similar but geographically distinct groups are bound to exchange more genes than ecologically differentiated populations that are closer in proximity. More commonly, studies of gene flow find less evidence of introgression than expected based on laboratory measurements of isolation (e. g. Counterman and Noor 2006; Llopart et al. 2005). In these circumstances, laboratory measurements suggest that mating preferences are incomplete and intrinsic postzygotic isolation is missing or weak, leading to the assumption that gene flow will be quite high in nature. The finding of lower introgression rates than expected may again be related to estimates of isolation based upon habitat preferences. Very few estimates of isolation have been made for either ecogeographic isolation (but see Ramsey et al. 20003) or microhabitat isolation (but see Kay 2006). These forms of ecological isolation precede mating isolation and intrinsic postzygotic isolation in the sequence of barriers experienced by organisms in nature. Therefore, they have the potential to greatly impact gene flow before later acting barriers have a chance to do so. The close correspondence between measures of isolation and gene flow in the presented study suggests that including measures of ecogeographic isolation in estimates of total isolation will aid in harmonizing measures of isolation and introgression. 214 Summag Estimates of introgression can help corroborate measures of reproductive isolation obtained in the laboratory. In this study, introgression was investigated for two species pairs of Mimulus that differ in their estimated degree of reproductive isolation. Mimulus bicolor/filicaulis had been previously estimated to have medium strength isolation, far from an amount necessary to complete speciation. Estimates of introgression based on coalescent simulations of polymorphism data corroborate this value, suggesting that these two taxonomic species share genes frequently (or have done so in the very recent past). Through two lines of evidence, this taxonomic species pair have been shown to not be biological species. Mimulus constrictus/whimeyi exhibit higher values of reproductive isolation in the laboratory by ecogeographic isolation, relative seed set differences, and hybrid inviability. The high level of total reproductive isolation inferred for this species pair was corroborated by gene flow analysis. No evidence of excess shared polymorphism was detected, though comparisons of allopatric and sympatric groups suggest small amount of gene flow may occur. Taken as a whole, contrasting patterns of introgression between these two species pairs suggests that measurements of reproductive isolation presented in chapter 3 are reasonably accurate. This accuracy was achieved in large part through incorporation of ecogeographic isolation into estimates of total reproductive isolation. 215 Table 4.1A. Molecular markers sequenced for analysis of introgression within Mimulus bicolor / M filicaulis species pair. Populations of M bicolor that flank the range of M filicaulis were considered sympatric (sym), while others were considered allopatric (allo) (see Figure 4.1A). Number of individuals sequenced Marker Aligned M bicolor M bicolor M bicolor M filicaulis length (bp) (allo) (sym) (total) MgST S 28 220 31 17 48 18 MgSTS 46 270 34 13 47 20 MgSTS 50 283 35 18 53 21 MgSTS 51 68 33 24 57 24 MgSTS 55 123 31 20 51 17 MgSTS 122 170 24 17 41 19 MgSTS 133 165 34 19 53 18 MgSTS 219 122 32 19 51 27 MgSTS 273 75 36 26 62 26 MgSTS 283 153 34 22 56 15 MgSTS 345 148 27 25 52 2] 216 Table 4.18. Molecular markers sequenced for analysis of introgression within Mimulus constrictus / M whitneyi species pair. Populations of M constrictus that occur within same mountain range as populations of M whitneyi were classified as sympatric (sym), while others were considered allopatric (allo) (see Figure 4. l B). Number of individuals sequenced Marker giggled M constrictus M constrictus M constrictus M w hi tneyi (bp) (allo) (sym) (total) MgSTS 46 189 7 10 17 22 MgSTS 79 151 10 ll 21 19 MgSTS 104 284 5 3 8 14 MgSTS130 362 7 5 12 12 MgSTS 464 85 10 11 21 21 MgST S 558 83 9 6 15 14 MgSTS 595 145 10 ll 21 19 217 OU<mm:_5xm 55.8% 58.5 5>_m:_oxm 555% “55me 535.28%. 558% 55352 .305 3 554.84% 355.23.: .3 5.553528 .3 5:85:83 55:5 55 mo .35 Set .3555: .5585. 55 3:55:53 .3 555235—55 on :55 1553.: .3 5 585290820: 535285 mo 5QE:: 55 .2855: .3553... .3 8 55:55.55: 5 5:2: 5:: 5.553528 .3 80¢ 5:85:88 =< 3553.: .3 5:5 883555 555 55255: 85:55.5.35 555.: no 52:5: 5:: 8 5.555: .555». .559»: 5555 .3 £555 we 5852:8833 52505—53: mo 59:5: 55 9 5.5: .39. 55.. 535288. :55 .5552? .3553; .3 5:5 5.535.:8 5.33533 :8 A33 5.5535 5:5 53 AmE~m5553¢05\>5§v5.m:5w58.33553535at: S5: ._. .3 555355.68 5538 mmbm mEm: max—5:5 85390830: mo £85m .mmé 535,—. 220 Table 4.4A. Pairwise FST values from molecular polymorphism analysis performed in SITES (distributed by J. Hey; http://genfaculty.rutgers.edu/hey/software#SITES) (Hey and Wakeley 1997) for Mimulus bicolor and M. filicaulis. ‘Sym’ refers to populations of M. bicolor that flank M filicaulis, while ‘allo’ refers to all other populations (See Figure 4. 1 A). Marker M. bicolor M bicolor (allo) M bicolor (sym) M bicolor v. (allo v. sym) v. M filicaulis v. M filicaulis M filicaulis (overall) MgSTS 28 0.041 0.015 -0.02 -0.006 MgSTS 46 0.039 -0. 174 0.065 -0.061 MgSTS 50 0.032 0.049 0.083 0.056 MgSTS 51 0.004 0.123 0.121 0.121 MgSTS 55 0.01 -0.052 -0.013 -0.039 MgSTS 122 0.023 0.1 11 0.07 0.091 MgSTS133 0.126 0.313 0.372 0.315 MgSTS 219 0.022 0.056 0.038 0.043 MgSTS 2 73 0.166 0.375 0.178 0.245 MgSTS 283 0.028 0.085 0.096 0.082 MgSTS 345 0.07 0.181 0.037 0.103 Mean (st. dev.) 0.051 (0.05) 0.098 (0.15) 0.093 (0.11) 0.086 (0.11) 221 Table 4.4B. Pairwise FST values from molecular polymorphiSm analysis performed in SITES (distributed by J. Hey; http://genfaculty.rutgers.edu/hey/software#SITES) (Hey and Wakeley 1997) for Mimulus constrictus and M whitneyi. ‘Sym’ refers to populations of M constrictus that occur in mountain ranges also containing M whitneyi, while ‘allo’ refers to all other populations (See Figure 4.1B). . M. constrictus M constrictus M constrictus v. M consmctus . . Marker (allo v sym) (allo) v. (sym) v. M. whitneyi ' M whitneyi M whitneyi (overall) MgSTS 46 0.318 0.581 0.556 0.52 MgSTS 79 0.091 0.465 0.452 0.456 MgSTS 104 0.438 0.319 0.459 0.281 MgSTS 130 0.172 0.606 0.508 0.553 MgSTS 464 0.325 0.887 0.441 0.647 MgSTS 558 -0.008 0.255 0.377 0.305 MgSTS 595 0.138 0.471 0.468 0.467 222 223 62: 5.. 7 93.8 83 34.8 8m. _- 5.8 $3 9% E 502 ES 8%. mad 2 ..o $3- $3 2% £32 SS 83. £3 83. $2- a a .o EN. 2%: ES :4. 7 23 $3. EV. _- 2.3. mm mam: 3...: m3. 7 2 3 ES 30.. _- mm _ .o q: gas 33 3V. _- 48d 83. :3- 32. 3 E32 255 ~85 ME; 8:... 83. $3 m2 E3: 3...... 0%. _- $2 83 2:. mad 2 $3: $3 33- :3 as... «:5- Ed R Emu: 5... 3. _- S _ .o 2...: 3w. 7 e2 .9 a £3: £3. M32- :55 :5... Sod- wood 2. HQ: Rod «5.. 7 83 5... Se. _- :2 mm mam: QESIH D Pmfirg c nigh D thmEH a c852 azesi .2 L383 .3 .28 a as 8%; E635 .Ewfi .3 c 5.0.33: 850%? 5 38 93:13 owflswxx Ea .cwgmouvm £82.: 05 9558295 £5: cows—25m 08.2 80¢ mos—g coamcocow $8258 mo 8:35va 2: 9 @8388 08>» Q mhfiwnmh mo 838, 333.. A33 >283? EB .65 A§E#033¢8\>o§co.mcowé.baoflaowtuaus Sam A 3 BSaEflE 2.3/mew gm mafia $532386 “gnome—moo mam: 383:8 we? AmEmEEoEbom cams—88 can .3866 .8058b% coxm dd mEmEmcoEm—oa mo maognwfimn com €3.98: 89a agave EmoEGwmm .86on Ezuofix. we 98 .8393 @335“: 5 ha 83 can 32$ m “80:83“. 88 538:8 05 98 A933 0 Each m3. .«o moumfiumm .-m Q PmEmmfl. a 8:87: Q @586; a 8x82 382.5% .2 83.588 .3 .209 5 8a 823.: EdomEEm A33 .3 8 883.: oofiouEwmm :8 58 $25 0839.. v8 .58on 883$ 05 wEEoEo—QEM £5: 88386 83: 80¢ 82? 868:8 x8858 .«o 8:358: 05 3 @8388 88> Q $98.me mo 82? EBo< A32 >283? :8 >05 Agmfioaatoflxofizvo8owé.b_:ofl:ow\\”§: Sum ... 3 8:58:83 caning gm wfim: 828388 E8838 wEm: 888:8 83 AmEmEncoEbo: aims—88 c8 €89? $0888.86 682.: .06 mEmE&oEboa mo 8225886 :8 38:5: 88.: 83.88: 885:me .86on “©2885: .2 :8 8.5.9388 832:: E :3 83 8: 323 m 8888:: 88 8:82: 05 98 8%: Q 98:.an mo 8883mm .Mmé 93.; Figure 4.1A. Distribution of populations sampled for molecular analysis in Mimulus bicolor and M. filicaulis. A projected ecological niche map is used as background (Chapter 2, Figure 2.5C) with black areas corresponding to unsuitable habitat, dark gray suitable to M. bicolor alone, white suitable to M. filicaulis alone, and light gray suitable to both species. Populations of M bicolor that flank the distribution of M. filicaulis were classified as sympatric (B4-BS). and all other populations were classified as allopatric (Bl-B3; Bo-B7). Five populations of M. “filicaulis reside within a very limited geographic extent in the cluster marked Fl-FS. 225 Figure 4.1 B. Distribution of populations sampled for molecular analysis in Mimulus constrictus and M. whitneyi. A projected ecological niche map is used as background (Chapter 2, Figure 2.5G) with black areas corresponding to unsuitable habitat, dark gray suitable to M. constrictus alone, white suitable to M. n-‘hitneyi alone. and light gray suitable to both species. Two populations ofM constrictus occur in middle altitudes of the Sierra, where M. whitneyi occurs at high elevation. These two populations were classified as sympatric with M whitneyi for purposes of analysis (C 1, C2). The other M. constrictus populations (C3) occurs in transition area between the Coastal and Transverse ranges in an area without M. whitneyi. This population was considered allopatric for purposes of analysis. 226 Literature Cited Arnold, M. L., and S. A. Hodges. 1995. Are natural hybrids fit or unfit relative to their parents. Trends in Ecology & Evolution 10:67-71. Bolnick, D. I. 2004. Waiting for sympatric speciation. Evolution 58:895-899. Bolnick, D. I., and B. M. Fitzpatrick. 2007. Sympatric speciation: Models and empirical evidence. Annual Review of Ecology Evolution and Systematics 38:459-487. Counterinan, B. A., and M. A. F. Noor. 2006. Multilocus test for introgression between the cactophilic species Drosophila mojavensis and Drosophila arizonae. American Naturalist 168:682-696. Coyne, J. A., and H. A. Orr. 2004, Speciation. Sunderland, MA, Sinauer Associates. Drummond, A. J., B. Ashton, S. Buxton, M. Cheung, A. Cooper, J. Heled, M. Kearse et al. 2010.Geneious v5.0, Available from http://www.geneious.com. Egan, S. P., P. Nosil, and D. J. Funk. 2008. Selection and genomic differentiation during ecological speciation: Isolating the contributions of host association via a comparative genome scan of Neochlamisus bebbianae leaf beetles. Evolution 62:1162-1181. Feder, J. L., X. F. Xie, J. Rull, S. Velez, A. Forbes, B. Leung, H. Dambroski et al. 2005. Mayr, Dobzhansky, and Bush and the complexities of sympatric speciation in Rhagoletis. Proceedings of the National Academy of Sciences of the United States of America 102:6573-6580. F elsenstein, J. 2005.PHYLIP (Phylogeny Inference Package) version 3.6.Distributed by the author. Department of Genome Sciences, University of Washington, Seattle. —. 2006. Accuracy of coalescent likelihood estimates: Do we need more sites, more sequences, or more loci? Molecular Biology and Evolution 23:691-700. Hey, J .,.and J. Wakeley. 1997. A coalescent estimator of the population recombination rate. Genetics 145:833-846. Hodges, S. A., J. M. Burke, and M. L. Arnold. 1996. Natural formation of iris hybrids: Experimental evidence on the establishment of hybrid zones. Evolution 50:2504- 2509. Hudson, R. R., M. Kreitman, and M. Aguade. 1987. A test of neutral molecular evolution based on nucleotide data. Genetics 116:153-159. 227 Kay, K. M. 2006. Reproductive isolation between two closely related hummingbird- pollinated Neotropical gingers. Evolution 60:53 8-552. Kliman, R. M., P. Andolfatto, J. A. Coyne, F. Depaulis, M. Kreitrnan, A. J. Berry, J. McCarter et al. 2000. The population genetics of the origin and divergence of the Drosophila simulans complex species. Genetics 156:1913-1931. Lexer, C., and A. Widmer. 2008. The genie view of plant speciation: recent progress and emerging questions. Philosophical Transactions of the Royal Society B-Biological Sciences 363:3023-3036. Llopart, A., D. Lachaise, and J. A. Coyne. 2005. An anomalous hybrid zone in Drosophila. Evolution 59:2602-2607. Magurran, A. E. 1998. Population differentiation without speciation. Philosophical Transactions of the Royal Society of London Series B-Biological Sciences 353:275-286. Mallet, J. 2008. Hybridization, ecological races and the nature of species: Empirical evidence for the ease of speciation. Philosophical Transactions of the Royal Society B-Biological Sciences 363:2971-2986. Mayr, E. 1942, Systematics and the origin of species. New York, Columbia University Press. Nosil, P. 2008. Speciation with gene flow could be common. Molecular Ecology 17:2103-2106. Nosil, P., S. R. Egan, and D. J. Funk. 2008. Heterogeneous genomic differentiation between walking-stick ecotypes: "Isolation by adaptation" and multiple roles for divergent selection. Evolution 62:316-336. Ramsey, J ., H. D. Bradshaw, and D. W. Schemske. 2003. Components of reproductive isolation between the monkeyflowers Mimulus lewisii and M cardinalis (Phrymaceae). Evolution 57: 1520-1534. Rieseberg, L. H., 0. Raymond, D. M. Rosenthal, Z. Lai, K. Livingstone, T. Nakazato, J. L. Durphy et al. 2003. Major ecological transitions in wild sunflowers facilitated by hybridization. Science 301 :121 1-1216. Rieseberg, L. H., J. Whitton, and K. Gardner. 1999. Hybrid zones and the genetic architecture of a barrier to gene flow between two sunflower species. Genetics 152:713-727. Sobel, J. M., G. F. Chen, L. R. Watt, and D. W. Schemske. 2010. The biology of speciation. Evolution 64:295-315. 228 Strasburg, J. L., and L. H. Rieseberg. 2008. Molecular demographic history of the annual sunflowers Helianthus annuus and H. petiolaris - large effective population sizes and rates of long-term gene flow. Evolution 62: 1936-1950. Tajima, F. 1989. Statistical method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics 123:585-595. Thompson, J. D., D. G. Higgins, and T. J. Gibson. 1994. Clustal-W: Improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Research 22:4673-4680. Turner, T. L., M. W. Hahn, and S. V. Nuzhdin. 2005. Genomic islands of speciation in Anopheles gambiae. PLoS Biology 3:1572-1578. Urbanelli, S. 2002. Genetic divergence and reproductive isolation in the Ochthebius (Calobius) complex (Coleoptera : Hydraenidae). Heredity 88:333-341. Via, S. 2001. Sympatric speciation in animals: The ugly duckling grows up. Trends in Ecology & Evolution 16:381-390. Wakeley, J., and J. Hey. 1997. Estimating ancestral population parameters. Genetics 145:847-855. Wang, R. L., J. Wakeley, and J. Hey. 1997. Gene flow and natural selection in the origin of Drosophila pseudoobscura and close relatives. Genetics 147:1091-1106. Wu, C. A., D. B. Lowry, A. M. Cooley, K. M. Wright, Y. W. Lee, and J. H. Willis. 2008. Mimulus is an emerging model system for the integration of ecological and genomic studies. Heredity 100:220-230. Yatabe, Y., N. C. Kane, C. Scotti-Saintagne, and L. H. Rieseberg. 2007. Rampant gene exchange across a strong reproductive barrier between the annual sunflowers, Helianthus annuus and H-petiolaris. Genetics 175:1883-1893. 229 l l LIBRARIES 20 8930 l l 3 1293 032 l V. H s R E V N U E T A T S N A m H m M