. 4"... .K .i i x‘ cabs... ......q.; I 5-: . 11"". .txtt] . i.i.i§l|2 .. a a. t! 25...}. ‘5... ya 1.... tapfis»... 5.... , . 35:5. .TF. «ET Eu. , g R? $.&» . a 3. Lin. 321...} . . .2; safififiwnq .. .11qu... .8. £5,335“: :0 (uni-no.1. .pI-a'd‘ “ ibqeuahnfi... :6". a.“ ‘ l. .‘fiiw. It‘l’.’ v .1 1...: In: (no. 3355‘ . 3 ; .. a. 3 .3. . 3.... 31.13.. £33.. 1;}.1afiuso‘. ’uvtiifi.%¥.utv n ' ‘h"?’ .2. if... I .v'i‘ 1"- . ofi" §.f‘:I-:l 713? .. 13...! A of. ‘ .l 0 :"AA liznfifllltfivla Ilv-§n.14 u: v p’ 5:“; siftihflzln AK.“ 5v. (influx. If... . «9‘! . . .vO‘nu. ‘ .J h. > . ,.::;.::..;x a 5.1 Vittzluz I ERNIEC: . 181.. .a‘. (u 5-15 l v \v 0.. . . T. “I? A! II. , .1531.)- x): 1.5: A!n4~.>2..t..fi¢ .1 {3.} , ‘ f .1. flirtinkrllfi. \C ;k}!i?fi. r ‘ummmmmnnmflmmugnw 2500 3 1293 o LIBRARY Michigan State University 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 woo cJCIRCIDaloDuopfis-p.“ THE EVOL. Prc THE EVOLUTION OF PHENOTYPIC PLASTICITY IN A NATURAL WILDFLOWER POPULATION By Brian Bruce Black A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Botany and Plant Pathology Program in Ecology, Evolutionary Biology, and Behavior and W. K. Kellogg Biological Station 2000 mEEVOl llhcn me 61: mialization or ph: tam] selection. th; genetic interdtoendv~ fitness consequence affine “inter annua': mural light enxiror: Immature genetic mi Within germ: filling perform at: Natural light but correlated across “m There We re if “311311351 Matt-rm Pleated to constra: The” “'85 a; SWEéiliOndonmt ABSTRACT THE EVOLUTION OF PHENOTYPIC PLASTICITY IN A NATURAL WILDFLOWER POPULATION By Brian Bruce Black When the environment is heterogeneous, factors that influence whether genetic specialization or phenotypically plastic generalists are favored include the patterns of natural selection, the amount of genetic variation for trait plasticity, and the degree of genetic interdependence between traits. In this study, I investigated the genetics and fitness consequences of responses to variable light and litter environments in a population of the winter annual wildflower Collinsia verna. Over two generations, I measured natural light environments, and manipulated light and litter environments. I used quantitative genetic breeding designs to investigate genetic and environmental effects on traits within generations, and the effects of maternal environment and genotype on offspring performance across generations. Natural light environments were variable at a scale appropriate to favor plasticity, but correlated across years in a way that might favor the evolution of plastic maternal effects. There were maternal genotype-environment interactions for seed size and dormancy. Maternal effects sometimes improved offspring performance, but also appeared to constrain offspring performance when mothers were stressed. There was additive genetic variation in some environments for germination/dormancy, emergence date, flowering date, specific leaf area, mainstem length, mean seed 3 entironment intera Irea mainstem len entironment Speci; cflects on vegetatix taiation suggestet: mtegies. The present emergence date an I l 'V ell sun reversed 1} mm“ the size c inpatterns of seleC‘ these traits. Together, ti and plastic matern. heterogeneous and light sensitive trai' Seemingly, reprc sci-e man-0n in this tr length, mean seed mass, and reproductive investment; and strong evidence for genotype- environment interactions (genetic variation for plasticity) for flowering date, specific leaf area, mainstem length, and reproductive investment. There were no strong light environment specialists among the genotypes sampled. However, significant maternal effects on vegetative biomass, seed number, and seed mass without additive genetic variation suggested that maternal genotypes may specialize for different reproductive strategies. The presence of leaf litter reversed the direction of direct linear selection on emergence date and increased the size of direct linear selection on vegetative biomass. Full sun reversed the direction of direct linear selection on specific leaf area, and increased the size of direct linear selection on reproductive investment. These differences in patterns of selection provide evidence that leaf litter and light are selective agents on these traits. Together, the parts of this study suggest that genetic variation for emergence date and plastic maternal genetic effects on seed size and dormancy may be maintained by a heterogeneous and unpredictable leaf litter environment. In contrast, the plasticity of the light sensitive traits flowering date and specific leaf area may be at or near optimal levels. Surprisingly, reproductive investment was both heritable and under strong directional selection. Significant genotype-environment interactions appear to maintain genetic variation in this trait. Ithaflk the i JetiCOflDCL Tom ( Andi" JW and T“ eseecially than.k SL4 native to complet'fi one else during the reminders that the agenda} issue maj Any clarity that l h. I thank all t} Their friendship he} Special thanks go It throughout my time fKtii'rimental desigr manning. In pr flan my cenetic an a Char Adam 3 W aflrninistrati 33d Caroljm Ham? in E) ‘ e do". ACKNOWLEDGMENTS I thank the past and present members of my committee for support and guidance: Jeff Conner, Tom Getty, Andy J arosz, Susan Kalisz, Alan Tessier, and Steve Tonsor. Andy J arosz and Tom Getty provided critical advice and support during lonely times. I especially thank Susan Kalisz for advice and encouragement at key moments when my resolve to complete this work was wavering. Jeff Conner has been there for me like no one else during the analysis and writing stages. I especially appreciate his constant reminders that the world is drowning in verbiage, and that careful attention to every tangential issue may be scholarly but in the end will only dilute the impact of the work. Any clarity that I have achieved is to his credit. Failures are of course my own. I thank all the Kellogg Biological Station graduate students past and present. Their friendship helped make KBS a stimulating and enjoyable place to live and work. Special thanks go to Denise Thiede. Her support, guidance, and encouragement throughout my time at KBS have been invaluable. I especially thank her for advice in experimental design and analysis. I have benefitted greatly from her mastery of SAS programming. In particular, she supplied the files and information necessary to jump- start my genetic analysis, and a SAS bootstrapping macro. Char Adams, Alice Gillespie, Mike Klug, and Sally Shaw provided friendly, CXpert administrative support during my graduate career. Nina Consolatti, John Gorentz, and Carolyn Hammarsjkold were all extremely generous and helpfiil. They contributed to the development, and completion of this research. iv Sarah Blac} Olendorf. and Bret: conditions that we: Homyak. and more Dudycha Kay Gro 3 these chapters. 1 ti: promo: Financial St Science recruiting l College of Natural Tracing Group “cl faculty Direct fin. NSF Dissertation 1 from Kellogg Biol Program at MSU, ’c’tXlSU. and a Gr Finally. tc Sacrifices they ha Wk schedule W bright rump:~ 1 2 9 . It‘s “an I have f0? Sarah Black, Tara Darcy, Becky Fuller, Kari Gorentz, Christy Lynn, Rob Olendorf, and Brent Pav, contributed their time and talents in the field, often in conditions that were miserable at best. Tara Darcy, Christy Lynn, Brent Pav, Carolyn Homyak, and more volunteers than I can recall provided expert assistance in the lab. Jeff Dudycha, Kay Gross, and Kevin Kosola provided useful comments on earlier versions of these chapters. I thank the Balkema family for permission to conduct this work on their property. Financial support has come from a Michigan State University College of Natural Science recruiting fellowship, a Graduate School dissertation completion fellowship, a College of Natural Science continuing fellowship, and especially from a Research Training Group fellowship from NSF grants DIR-9113598 and DBI-9602252 to the KBS faculty. Direct financial support for research expenses came from the KBS RTG grants, NSF Dissertation Improvement Grant DEB-9701 148, a G. H. Lauff Research Award from Kellogg Biological Station, the Ecology, Evolutionary Biology, and Behavior Program at MSU, the Paul Taylor fund of the Department of Botany and Plant Pathology at MSU, and a Graduate Student Research Grant from the MSU chapter of Sigma Xi. Finally, to my family, Patricia F rueh and Henry Black, I recognize the great sacrifices they have made for me. Their tolerance of my obsessive and never ending work schedule was exemplary. I apologize for lost opportunities, but look forward to a bright future. I am most proud that we have made it through. I thank my mother, Becky Black, and sister, Sarah Black, for being there for us at critical times. I especially thank Pat and Henry for their regular reminders about what matters in life, and their patience When I have forgotten. This work is dedicated to them. usr or TABLES cm or FIGLRE :i CHAPTER 1 DTRODL'CTION . E.\\lR0.\'I\lE.\'T.—” Models: Pl Titi- Fred Models: Pl: Cost Pred The Empirical st The evoluti: Em" Mat Qua Qua Literature C CHAFTER 2 PLASTIC stem CONSEQLENCE Introductio Da- TABLE OF CONTENTS LIST OF TABLES ....................................................... x LIST OF FIGURES .................................................... xiii CHAPTER 1 INTRODUCTION: MICRO-EVOLUTIONARY CONSEQUENCES OF ENVIRONMENTAL HETEROGENEITY .................................... 1 Models: Plasticity evolution ......................................... 2 Types of models ............................................. 3 Predictions .................................................. 6 Models: Plasticity and genetic variation ................................ 8 Costs and limits of phenotypic plasticity .......................... 8 Predictable environmental variation .............................. 9 The value of models ......................................... 10 Empirical studies .................................................. 11 The evolution of phenotypic plasticity in a natural population .............. 13 Environmental heterogeneity .................................. l4 Maternal effects in heterogeneous environments ................... 14 Quantifying Genetic and Environmental Effects on Phenotypes ....... 15 Quantifying phenotypic selection ............................... 15 Literature cited ................................................... 17 CHAPTER 2 PLASTIC MATERNAL EFFECTS: GENETIC VARIATION AND FITNESS CONSEQUENCES IN A NATURAL PLANT POPULATION .................. 24 Introduction ...................................................... 24 Maternal genetic and environmental effects ....................... 24 Plastic responses to light ...................................... 26 Methods ......................................................... 28 Study system ............................................... 28 Experimental design ......................................... 29 Light treatments ...................................... 29 Breeding design ...................................... 29 Mothers (year 1, 1995-96) .............................. 33 Offspring (year 2, 1996-97) ............................. 34 Maternal effects ....................................... 34 Light measurements ................................... 35 Data analysis ............................................... 36 Environmental effects and genotype-environment interactions . . 36 Fitness consequences of maternal effects ................... 39 vi Results. . Discussit S C C Literatur CHAPTER 3 EVOLUTION C Reseoxses r POPLlATION Introduc ( I Method. ( l REsults DiSCUS Results .......................................................... 41 Light ..................................................... 41 Genetic and environmental effects on the 6 maternal traits ........... 41 Seed number (mothers) ................................. 41 Mean seed mass (mothers) .............................. 41 Proportion of offspring germinating ....................... 52 Proportion of offspring surviving ......................... 52 Offspring mean seed number ............................ 52 Offspring mean seed mass .............................. 53 Fitness consequences of maternal effects ......................... 53 Phenotypic correlations with seed mass .................... 53 Phenotypic selection ................................... 56 Discussion ....................................................... 61 Seed size plasticity and variation in dormancy ..................... 61 Genotype-environment interactions ............................. 65 Conclusion ................................................ 68 Literature cited ................................................... 70 CHAPTER 3 EVOLUTION OF REACTION NORMS: THE QUANTITATIVE GENETICS OF RESPONSES TO VARIABLE LIGHT ENVIRONMENTS IN A NATURAL PLANT POPULATION ......................................................... 78 Introduction ...................................................... 78 Genetic constraints and the evolution of reaction norms ............. 79 Plastic responses to light ...................................... 81 Methods ....... - .................................................. 82 Study system ............................................... 82 Experimental design ......................................... 83 Light treatments ...................................... 83 Breeding design ...................................... 84 Data analysis ............................................... 85 Heritability .......................................... 85 Environmental effects and interactions ..................... 86 Genetic correlations across environments ................... 88 Results .......................................................... 90 Light environment effects ..................................... 9O Additive genetic effects and narrow sense heritabilities .............. 97 Dam effects ................................................ 99 Genotype-environment interactions ............................. 99 Genetic correlations across environments ........................ 100 Discussion ...................................................... 103 Patterns of genetic variation .................................. 103 Heritability ......................................... 103 Maintenance of genetic variation ........................ 104 Environmental effects on the expression of genetic variation . . 104 vii Ma: Ge: Cor Literature t CHAPTER 4 EVOLUTION OF SELECTION IN A. Introductio: Elle Dir: Methods. . I Sillti Exp; Dal! Maternal effects ............................................ 105 Genetic variation for plasticity ................................ 107 Antagonistic pleiotropy ............................... 107 Genotype-environment interactions for fitness .............. 108 Control of plasticity via a genetic switch .................. 108 Adaptive plasticity relaxes natural selection? ............... 109 Conclusion ............................................... 109 Literature Cited .................................................. 1 11 CHAPTER 4 EVOLUTION OF REACTION NORMS: ENVIRONMENT-DEPENDENT SELECTION IN A NATURAL POPULATION OF COLLINSIA VERNA .......... 119 Introduction ..................................................... 1 19 Effects of light and leaf litter ................................. 120 Direct and indirect selection .................................. 121 Methods ........................................................ 122 Study system .............................................. 122 Experimental design ........................................ 123 Data analysis .............................................. 124 Phenotypic plasticity .................................. 124 Survival analysis ..................................... 125 Phenotypic selection .................................. 126 Path analysis ........................................ 127 Results ......................................................... 131 Phenotypic plasticity ........................................ 13 1 Survival analysis ........................................... 131 Phenotypic selection: Survival episode ......................... 139 Phenotypic selection: F ecundity episode ........................ 139 Path analysis .............................................. 147 Discussion ...................................................... 157 Traits .................................................... 159 Causes of selection ......................................... 163 When is plasticity adaptive? .................................. 165 Literature cited .................................................. 168 CHAPTER 5 CONCLUSION: ENVIRONMENTAL HETEROGENEITY, PHENOTYPIC PLASTICITY, AND THE MAINTENANCE OF GENETIC VARIATION IN A NATURAL POPULATION .............................................. 175 Environmental heterogeneity ....................................... 175 Plastic maternal effects ............................................ 176 Genetic and environmental effects ................................... 176 Phenotypic selection .............................................. 177 Future directions ................................................. 178 Ongoing analysis of this data ................................. 178 viii OL‘. Literature t Temporal variation in natural selection ................... 178 Environmental correlations and bias in selection analysis ..... 178 Genetic correlations between traits ....................... 179 Correlational selection ................................ 179 Constancy of genetic parameters ........................ 183 Costs of plasticity .................................... 187 Inbreeding depression in variable environments ............ 187 Evolutionary demography in variable environments ......... 188 Outstanding questions ....................................... 188 Heterogeneity of leaf litter and other environmental factors . . . 188 Evolution of plant architecture .......................... 189 Physiology of photosynthetic acclimation ................. 189 Functional ecology of anthocyanin pi grnentation ............ 189 Multilevel selection ................................... 190 Predicting multivariate evolution ........................ 191 Literature cited .................................................. 192 ix are not pre entironme lablel. Within—e significant and Table late 3. Pearson traits. Off planted in size. Trai germinati' OffSpring Correlatit table-Md tuP: F. a no.3 Ecofizomxm ©0320 Figure l. 31 havmhhm U—DC OH: —.:O.—" UU~UD-CU mc-w3~um>w~v=~ VKVN $2.29: 0% S VchN :ud G 30:32:58 Eon—cotéo Em: m .8333: mama wccmmto wfittsm Ndflluqfl. 68283 $03 mctamto madam 2 (plaid) Buudsrro g uotterauag 68332 $0508 W mar/35m amalgam. W m 6853 w W QEmcoufloc :26qu mo 253 Wm M. maxim one and m. z meme «2 seam om mm m have museum eaten 3538 3 m w .wctotso: 8 SE Bow 2: Soc 380:8 2.3233: vow wag Figure 2. Experimental design: Crossing and planting plan as described in text. 32 population. designated a design (TWO dams were e resulting see moist soil fr seeds to the X 8 cm grid blocks = 68; the populatic “llhin the rim naturally occ highs X S c: Moth ”3 the moth: t"I‘I‘CTgence a:- nmnbm 0f C( C'OWledOn am. COQ'IIedOH are Plants in all trl W diSP‘CTSaj d h I: at, $8 traits (er t‘fg‘ 225 to (liffEr population. In the greenhouse 50 of these field collected plants were randomly designated as sires and each crossed to three randomly chosen dams in a standard nested design (two sires with small dams were each crossed to one additional dam). Flowers of dams were emasculated in the bud stage to prevent selfing. In July 1995 I weighed the resulting seeds and planted them in 2 cm uniquely numbered plastic straws containing moist soil fiom the study site. In early August, I transferred the straws containing the seeds to the field. Three seeds fiom each firll-sib family were randomly planted into an 8 X 8 cm grid in each treatment (152 full-sib families X 3 seeds/family X 3 treatments X 5 blocks = 6840 seeds planted). This density is approximately one third of the average in the population at harvest (unpublished data). The entire study area was established within the natural Collinsia population, so that the planted seeds grew in a matrix of naturally occurring plants (Figure 1). The experimental area was enclosed with 1.2 m high 5 X 5 cm welded wire fencing. Mothers (year 1, I995-96)-Plants emerging fiom the straws in the field treatments are the mothers in this study (Figure 2, generation 2). I conducted censuses for emergence and survival weekly from September through December. In December, the numbers of cotyledons and true leaves were counted, and the diameters of the largest cotyledon and leaf were measured to the nearest 0.4 mm. Winter size was calculated as cotyledon area + leaf area. In mid-April a census for survival to flowering was done. Plants in all treatments were naturally (open) pollinated. Plants were harvested before seed dispersal in June of 1996 and scored for number of seeds and total seed mass. I call these traits (emergence date, winter size, seed number, and mean seed mass) individual traits to differentiate them from the maternal traits described below. 33 are mothers are mothers wa light mothe: fecundity, 0 size for the I additional 7. seeds were r 3 treatments distributed e generation 2 795 Offsprin: excluded the effects anal}: as described Sfptember 2. in this treatm “‘ere the Sam Consequenllj, ‘VGIer F Offspring in r! I .. Scored S Eye}. number and s;| l Offspring (year 2, 1996-97)-Seeds collected from the field grown, open pollinated mothers are the offspring in this study (Figure 2, generation 3). Seed production by mothers was quite variable, but I used equal numbers of seeds from medium and high light mothers in the subsequent planting (1325 each). Due to high mortality and low fecundity, only 58 seeds were available from low light mothers. To increase the sample size for the low light maternal environment, I included in the offspring generation an additional 737 seeds produced by naturally emerging plants in the low light plots. All seeds were randomly planted back into the 15 experimental plots (230 seeds X 5 blocks X 3 treatments = 3450) with the constraints that the grandchildren of generation 1 sires were distributed equally in each treatment and block, and one third of the seeds from each generation 2 mother were planted in each of the three environments. Because most of the 795 offspring from the low light maternal environment were of unknown relationship, I excluded these plants from the genetic analyses described below. Separate environmental effects analyses include all three maternal light environments. Planting techniques were as described above. I transferred the numbered straws containing the seeds to the field on September 24, 1996. To increase survival in the low light treatments, I raised light levels in this treatment from 10% to 18% of full sun. Census, harvest, and scoring techniques were the same as in year 1. This species has a well-documented seed bank (Kalisz 1991). Consequently, seeds that did not germinate were considered dormant. Maternal effects-To address whether mothers alter the phenotype of their offspring in response to the current environment in a way that increases maternal fitness, I scored several measures of maternal performance. Two are traditional measures: the number and size (mean mass) of the seeds produced by each mother. I measured four 34 additiona entironm sunit'ing number 0 offspring. trait value the same ‘ analyses. (analyses conservat from the t maternal 1 Produced Calculatio; genuinare fecufldity Value Of 3 [lg address an "aliathn ii “inflated? fica80m A c... ”PiOmEtEr additional traits to quantify maternal performance in terms of offspring traits in each light environment: (1) proportion of offspring germinating; (2) pr0portion of offspring surviving (both based on number of seeds per mother per light treatment); (3) mean number of seeds produced by surviving offspring; (4) mean mass of seeds produced by offspring. I calculated these later four traits for each light environment by averaging the trait values of each mother's offspring. Because these averages grouped offspring from the same treatment across all 5 blocks, I could not include block in the subsequent analyses. In separate analyses, block effects were small and mostly insignificant (analyses not shown). This approach makes the tests for maternal effects more conservative. Together, I refer to these six traits as maternal traits to differentiate them fi'om the traits of individual offspring. Finally, I calculated a cumulative measure of maternal fitness in each light environment as the product of mean number of seeds produced by offspring and proportion surviving divided by proportion germinating. This calculation estimates the mean number of grandseeds per offspring once all offspring germinate. It assumes that all dormant seeds genninate, and that their survival and fecundity would be the same as in the year of this study. Because it does not discount the value of a dormant seed, it is likely an overestimation of cumulative maternal fitness. Light measurements- I measured photosynthetically active radiation (PAR) to address three questions: Did the light manipulations effectively bracket the natural variation in the light environment? Are parent and offspring light environments correlated? Is light availability spatially variable? All light data were collected using a Decagon AccuPAR light ceptometer (Decagon Devices, Inc. Pullman, WA). The ceptometer has 20 quantum sensors arrayed along a l m wand. I measured light in the 35 experime lO sampl vegetatio samples ( samplin g Tl levels in 1 10m east to the fort afer leaf It 1997 at St POPUlatioi bCIWeen 1 DUI. Each CalCulated Values. I I experimental plots weekly at solar noon throughout both years. Each reading consisted of 10 samples taken while moving the wand horizontally across the plot 10 cm above the vegetation surface. Consequently, a reading was the average of at least 200 single point samples (20 sensors X 10 samples). Three readings were taken on each plot at each sampling time. To determine if light levels were correlated across generations, I measured light levels in the population in fifteen permanent I m2 plots at uniform intervals along three 10 m east-west transects spaced 20 m apart across a light gradient from the southern edge to the forest interior. I made these measurements at solar noon weekly for one month after leaf fall and in May of each year. Light levels for these fall and spring measurements were expressed as percent of full sun. Because ambient light levels differ between spring and fall, I calculated separate spring and fall across year correlations. To assess spatial variation in light, I made measurements 30 times in the spring of 1997 at 50 points spaced on a 20 X 20 m grid throughout the full spatial extent of the population. I took a reading fi'om a single light sensor at the end of the wand every hour between 1000 and 1500 on five days between May 7 and 24, 1997 prior to canopy leaf out. Each reading was expressed as percent of full sun at each sample interval. I calculated a light score for each point taking the mean of these 30 relative light level values. I plotted the scores for each of the 50 points in a frequency graph. Data analysis Environmental eflects and genotype-environment interactions-All traits were independently analyzed using the MIXED procedure of SAS (SAS 1992, 1997). Proc. MIXED performs mixed model analysis of variance using a restricted maximum 36 likelihood 4‘ effects. M] data (Shaw I computing tests were e The signific sided likeli‘t Fixe interaction \ nailEfill)’ CIT effects and I only mother modfil for cal emimminent and darn We Were treated afiem00n Sh While Sire, (I IandOI-n. Su- likelihood (REML) algorithm to directly compute variance components for random effects. Maximum likelihood methods are generally superior for analysis of unbalanced data (Shaw 1987, Searle et al. 1992, Littell et al. 1996). Fixed effects were tested by computing a Type III F statistic. Denominator degrees of fieedom for these fixed effects tests were estimated using the Satterthwaite approximation option (Littell et al. 1996). The significance of variance components for random effects were assessed using one- sided likelihood ratio tests (Littell et al. 1996). Fixed effects of maternal environment, offspring environment, and their interaction were examined first with the complete data set (including the offspring of naturally emerging mothers from the low light environment). Genetic and environmental effects and their interactions were then analyzed with a mixed, split plot model including only mothers of known relationship from the medium and high light environments. The model for each trait included block (except traits averaged across blocks), maternal environment, sire, darn nested in sire, offspring environment, and their interactions. Sire and dam were split plot factors, while light treatments were whole plot factors. Blocks were Heated as fixed because they captured an east-west gradient of diurnal morning to afiemoon shading. Environmental effects, and their interaction were also treated as fixed, while sire, dam nested in sire, and their interactions with fixed effects were treated as random. Sire and dam effects both reflect genetic differences among mothers. The full analysis included all main effects and two-way interactions except no darn interactions could be examined for the four maternal traits due to sample size limitations. Highly insignificant three-way and higher interactions with block were omitted from the final model for seed number and mean seed mass of the mothers. 37 The: reasonable i! always natc in the temp: competitive of competitt The normality. l germination distributions each environ de'PaITure fro assumptions CaIBgon'ca] IT mp1s: Silas main Effects idemlCa] to [I PTOC. MIXED Illllllber W€re Wlien enllropmem t (is mdicfited l “:8" These analyses assume that genotypes within plots are independent, which is reasonable because the plantings were relatively low density and nearest neighbors were always naturally occurring plants. Light treatment effects may be due to light or variation in the temperature or moisture environment caused by light. They may also be due to competitive interactions with other plants when light levels affected the size and density of competitors. These factors all naturally covary with light. The germination and survival traits were trimodal and could not be transformed to normality. The offspring of more than one third of the mothers had either complete or no germination or survival, while the remaining families were normally distributed. These distributions were a consequence of several mothers with only a few offspring planted in each environment. Two approaches were used to examine the consequences of this departure from normality. First, by using data from only the largest families, normality assumptions could be met. Second, a balanced subset of the data was analyzed using categorical modeling (SAS CATMOD procedure, SAS 1989). To achieve the necessary . sample sizes for categorical modeling, a reduced model with only sire and environment main effects and their interaction was necessary. Both approaches yielded results nearly identical to the proc. MIXED analysis of the full data set with similar models, so only the proc. MIXED results are presented. Seed number (mothers) and mean offspring seed number were log transformed to achieve normality. When the full mixed model analysis indicated that the effect of offspring environment or the expression of genetic variation depended on the maternal environment (as indicated by significant interaction terms), analyses within maternal environments were done. The approach used for these analyses was as described above, with a model 38 that included only offspring environment, sire, and dam nested in sire. Reaction norm figures were constructed for each trait with significant sire or darn effects by plotting family means against environment. Genotypic differences in slope in these reaction norm plots indicate genotype-environment interactions. Statistical and graphical approaches to detecting genotype-environment interactions are complimentary and equally important because the power of statistical methods to detect genotype- environment interaction can be limited (Lewontin 1974, Via and Lande 1985, Wahlsten 1990). Fitness consequences of maternal eflects-The analyses described above revealed effects of maternal genotype and enviromnent on seed mass (see Results). This suggests that maternal effects on offspring fitness may be mediated through seed size. To understand the relationships between seed mass and other traits, phenotypic correlations (Pearson) of mean seed mass with the other five maternal traits were calculated for each environment, and correlations of mean seed mass with individual traits (emergence date, winter size, mean seed mass, and seed number) were calculated for each environment for both years. I used multivariate episodic selection analysis (Arnold and Wade 1984) to assess the fitness consequences of seed mass and phenotypically correlated traits under maternal influence. Two episodes of selection, survival to flowering and fecundity were analyzed independently for each light environment in both the mother and the offspring generations. The estimates of selection resulting from these analyses address whether a character has fitness consequences in that episode independent of selection in other episodes (Koenig et al. 1991). Traits analyzed in the survival episode were mean seed 39 mass. err included morphol: collected selection calculate mass, emergence date, and winter size. The traits analyzed in the fecundity episode included mean seed mass, emergence date, winter size, and several phenological and morphological traits of adult plants. Due to snow cover, winter size data was not collected for the offspring generation (Year 2, 1996-97). Multivariate linear selection gradients ([3,) were calculated as the partial regression coefficients of a trait on relative fitness. Linear selection gradients measure the direct selection on each trait independent of all other traits in the analysis. Relative fitness was calculated by dividing each absolute fitness measure (zero or one for the survival episode, total seeds for the fecundity episode) by the mean for that environment. All phenotypic traits were standardized within environments to a mean of zero and variance of one, and transformed as needed before analysis to improve normality. While the calculation of regression coefficients on a categorical measure of fitness like survival provides an unbiased estimate of the selection parameter, parametric significance tests are not valid (Mitchell-Olds and Shaw 1987). Significance tests of all selection parameters were performed by bootstrap resampling methods (Dixon 1993). The data set fiom each environment and year was resampled 1000 times using a SAS macro (SAS 1990). The number of observations in each resampled data set was equal to the number of observations in the original data set. Confidence intervals (95%) for all selection parameters were calculated from the bootstrapped estimates of each parameter by the shift-distribution method (Noreen 1989). Path analysis was used to examine an a-priori model describing hypothesized relationships between seed mass, emergence date, winter size, and survival. Because winter size was not measured in the second year, only the data for the first year (1995-96) 40 were USl T Within lht in: plots \ the natura The distri't distributio The interaction Table I), w mean seed lmorhers fr,- Seer; 33), and wag Insignifican genotype 0? 110,733 Sh0“~ Area efie high light mi I l were used in this analysis. RESULTS Light The light treatments effectively bracketed the average light level over a wide area within the forest over the entire growing season (Figure 3), and natural light levels in 1 m2 plots were highly correlated across years (Figure 4a). At a 20 m scale, light levels in the natural population were variable, ranging fiom 25% to 75% of full sun (Figure 4b). The distribution was not normal, with 78% of the sites falling in the lower half of the distribution. Both the mean and median were about 40% of full sun. Genetic and environmental effects on the 6 maternal traits These results draw on three separate analyses, the genotype-environment interaction analysis of the genetic data set (medium and high light mothers, Figure 5, Table 1), within-enviromnent genetic analyses of pr0portion germinating and offspring mean seed mass (Table 2), and the environmental effects analysis of the full data set (mothers from all treatments, Figure 6) Seed number (mothers)-Seed number increased significantly with light (Figure 5a), and was typical of natural plants growing in the plots. There was a marginally insignificant maternal family effect on seed number (P=0.07) suggesting that maternal genotype or environment could potentially contribute to offspring fitness. Reaction norms show that some families differed both in number of seeds and in plasticity. Mean seed mass (mothers) -The environment in which seeds were produced effected seed mass (Figure 5b). Seed mass was greater in medium light than either low or high light maternal environments, and was most variable among individuals in low light. 41 2000 Treatment ,:\ —Cl— 10% Sun -- .0711 N . + 40°/ S " 51600 - ° W ° A :8; + 100%81111 K A E“ ‘ A .3 + Forest Natural v .8 A E 1200 4 ‘ Q) ‘ S .2. A 8 .. <1 3, '8‘ .. '8 800 t‘ .c: *5 >5 8 4s ‘5 i 400 - I 0‘ I I I Sept. 1 Nov. 14 Jan. 25 Apr. 10 June 15 Figure 3. Seasonal variation in photosynthetically active radiation (PAR), Year 1 (1995- 96) in treatments and natural forest plots. Each point is the mean of 5 plots, and all data were collected at solar noon on clear days. Bars are standard errors. 42 Figure 4. Temporal stability and spatial variability of light environments. (3) Between year correlations in light levels for 15 plots spaced uniformly across a light gradient from forest to edge. (b) Variation in light availability at 50 grid points covering the spatial extent of the population. 43 (a) Across Year Correlation of Light Environments 0.9 4 0.8 - {C °.‘ 8 0.7 - 2.‘ RT 3 0.6 - Q) >-' g 0.5 - E “5 0.4 — t: .2 1: 8.0.3 — 8 g I Fall C] Spring $2 0.2 ‘ rf 311:0‘96 P<0.0001 0.1 - r sprin g=0.85 P<0.0001 0 I I I 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 Mean proportion of full sun, Year 1 (1995-96) 20 _ (b) Variation in Light Levels in the Natural Population 15 j 10-3 U . 5 7. 0 .3 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 Mean proportion of full sun, May 1997 Figure 4. Figure 5. Reaction norms for traits affected by maternal genotype and/or environment. Norms are for either paternal or maternal families depending on which effects were significant in the mixed model analysis. (a) Mean number of seeds produced by mothers, by paternal family, June 1996. (b) Mean mass of seeds produced by mothers, by maternal family, June 1996. (c) Proportion of offspring germinating, by paternal family, Fall 1996. Note that the x-axis is the maternal environment, not the germination environment, and that there were too few seeds produced in the low maternal light environment for them to be included in this analysis. ((1) Proportion of offspring surviving to flowering, by paternal family, Spring 1997. Statistics based on genetic and environmental effects mixed model REML analysis. When a variance component was estimated as zero, no P value is reported. Many families did not survive in low light, only those families with data for at least two environments are shown. The number of families varies: (a)-50 families, (b)—58 families, (c) and (d)-20 families. (13,» 45 Iment. 'le f 7 1011165. emal nine-rim hr 3 ant and 11 MS ight. oer of 12 - (a) 10 a . Effect _8 8 _. M atemal Environment g Sire (N =50) “a ‘ Dam (N=151) g 6 - Sire x Maternal Env. g . Dam X Maternal Env. 5 4 _. N =1083 é’ . 2 ‘ [— Sire Family I 0 I 10% Sun 40% Sun 100% Sun Maternal Environment 7' (b) .5 I Mean Seed Mass (mg) N L It 47/ ét/ I \\ at; \I It w t/ L_ Effect Maternal Environment esseé Sire (N =50) $3}: Dam (N =151) 77' Sire x Maternal Env. \ Darn X Maternal Env. N =949 |—— DamFamily I 1 _. O I I I 10% Sun 40% Sun 100% Sun Maternal Environment Figure 5. 46 R 0.0007 0.3148 0.0692 0.1847 0.0588 0.2987 0.0614 0.0455 (0) WWW 0.9 0.8 — _ Effect 12 0.7 - Maternal Environment 0.4622 2° 0 6 _ Offspring Environment 0.2378 g ' , Mat. Env. x Off. Env. 0.3347 g 0-5 ‘ Sire (N=38) ..... 20.4— Dam(N=111) 0.1001 g - Sire x Maternal Env. 0.0249 $03 f Sire x Offspring Env. an 0.2 — N=1002 0.1 - —— Sire Family 0 I I 40% Sun 100% Sun Maternal Environment 1 .— 00 .s ‘ a; 0-9 “ Effed 2 fl ()8 _ Maternal Environment 0.524 1,9,, 0 7 _‘ Offspring Environment 0.0001 c . I; ~ Mat. Env. X Off. Env. 0.997 5 0'6 i Sire (N =38) ...... §05r Dam(N=lll) 0.1383 0 3 Q4 — Sire X Maternal Env. ------ €33 0 3 _‘ Sire x Offspring Env. 0.0895 E . N=665 a 0.2 — “a ,5 0.1 n —— Sire Family .E - 8. 0 I f a? 18% Sun 40% Sun 100% Sun Offspring Environment Figure 5 (cont'd). 47 3398 32 AN. 5.8 hmzd coca ta> EsEmom ...... o :t: o EoE:O:>:m war—amto x 85 rrrrrr o rrrrrr O EOE—hodgcm 35032 X ohm :25 A389 63 32.0 58.8 285 3855 3:5 $8.8 85 ...... o 2a 336 8a ..... sad 8 3.0 8.0 u EC .Em magnate x .Em neon: 88.0 3..» i 22m soon 2 .2 n 22m anaconém mango 23o 8m u and 886 an i ask assassin neon: R Cobm vumvcgmv 0355mm Q Cobm 6.823me 395ng oosmtmxr Ho oesom Eocomfiov 8593 no 253% EncomEoU oocmte> co 253$ awnuz $32 comm :82 wEEmtO 5252 new as: mango 8an ho “OMEN wawm E .mcomuowuvucm EoEcobéodguocow mo 5:865 o: E 22: 3283 m oSmE E @8583 Lo: 2a was: 82:. .mmmE 38 :88 manage can #58:: 38 :88 metambo £3: 36038 of do amines 42mm BeoE c058 gumbo BEoEdESfi 98 2650 A 038. 48 Table 2. Within-environment mixed model REML analysis for maternal traits with significant interactions in the fill analyses presented in Figure 5c (germination) and Table 1 (seed mass). Maternal Light Fixed Effect or F -Value or Variance Environment (N) Source of Variance Component Est. (Std. Error) P i at' Medium (551) Offspring Environment F 2,73 = 0.07 0.9305 Sire 0 ---..... Dam(Sire) 0.0074 (0.0062) 0.1 128 Residual Variance 0.1777 (0.01 18) High (451) Offspring Environment F 2,370 = 2.66 0.0713 Sire 0.0085 (0.0081) 0.1283 Dam(Sire) 0.0156 (0.0095) 0.0161 Residual Variance 0.1372 (0.0102) WW Medium (192) Offspring Environment 1"}.I78 = 10.82 0.0001 Sire 0.1053 (0.0774) 0.0429 Dam(Sire) 0 -------- Residual Variance 1.0194 (0.1 159) High (157) Offspring Environment F1126 = 1.32 0.2705 Sire 0 ----- - Dam(Sire) 0.203 (0.1441) 0.0564 Residual Variance 1.207 (0.1735) 49 Figure 6. Germination, survival, and maternal fitness measures for the full data set including the offspring of natural mothers in the low light environment. (a) Proportion of offspring germinating, Fall 1996. (b) Proportion of offspring surviving to Spring 1997. (0) Mean seed number for offspring, June 1997. (d) Hypothetical cumulative maternal fitness if all offspring had germinated. This measure assumes no mortality prior to germination in dormant seeds, and equivalent survivorship and fecundity in dormant seeds after germination. (e) Mean mass of seeds produced by offspring. Environmental effects tested with fixed effects REML analysis. No genetic effects were included in the models. ME = maternal environment, OE = offspring environment. Bars are standard errors. 50 1 l _ 13% E (a) _ 252% E (b) ME <0.01 ME 0.89 E 0'8 ‘ OE 0.15 200.8 OE <0.01 2 _ MEXOE 0.77 E — MEXOE 0.27 E E 0.6 — 5 5 E “g- .E 0.4 — 1: o O D. g g 0.2 — n. 0 .. 18% Sun 40% Sun 100% Sun 18% Sun 40% Sun 100% Sun 1996-97 Offspring Environment 1996-97 Offspring Environment 12 12 - ram 2 (e) um 13 (d) 10 " MB 0.05 10 _ MB 0.01 <0.01 - OE <0.01 % € 0 Q) 8 g- “<95 3~MEXOE 0.11 “a 7 E “5 E 6 — *3 ’2 6- E . E E . :3 E :s z 4 — 5 z 4 ~ 5 ‘ 5 i g 2 — 2 2 — 0 — 0 _ 18% Sun 40% Sun 100% Sun 18% Sun 40% Sun 100% Sun 1996—97 Offspring Environment 1996—97 Offspring Environment MMatemalEnf ‘(Year I) ’83 g I 10% Sun g 40% Sun “D g [:l 100% Sun C 8 2 18% Sun 40% Sun 100% Sun 1996-97 Offspring Environment Figure 6. 51 The significant maternal genotype-enviromnent interaction for mean seed mass, and the considerable crossing of maternal family reaction norms indicate the presence of genetic variation for the plastic provisioning of seeds. Proportion of offspring germinating-In the genetic analysis, there were no significant main effects on offspring germination: only the sire by maternal environment interaction is significant (Figure 5c). Within maternal environments, there were significant dam effects on offspring germination when seeds were produced in high light, but not in medium light (Table 2). When the low maternal light environment was added to the analysis, the effect of maternal environment became highly significant, while the germination enviromnent continued to have no significant effect (Figure 6a). Overall, seeds produced by mothers growing in the medium and high light environments germinated at higher rates in each offspring environment, but mothers responded quite differently to these environments. Some mothers produced more dormant seeds in the medium light environment, while others did so in high light (Figure 5c). Seeds produced by mothers in low light were least likely to germinate anywhere. Proportion of offspring surviving—In both the genetic and the environmental effects analyses there were no significant effects of maternal environment on offspring survival. Proportion surviving to flowering simply increased with light (Figures 5d, 6b). In the genetic analysis the sire-offspring environment interaction term was insignificant (P=0.09), but the reaction norm figure shows that mortality patterns for some families differed dramatically across environments (Figure 5d). Ofifs'pring mean seed number-Light had highly significant effects on seed production by offspring in both analyses (Table 1, Figure 6c). Moreover, maternal 52 environment had a significant effect on offspring fecundity (P=0.02 in the genetic analysis, Table 1; P=0.05 in the environmental effects analysis, Figure 6c). The lowest mean fecundities were for the offspring of low and high light mothers in their home (maternal) environments. The offspring of medium light mothers performed as well as, or better than the offspring of other mothers for this trait (Figure 6c), and in the cumulative fitness calculation (Figure 6d). Oflspring mean seed mass-Light environment also had highly significant effects on offspring seed size in both the genetic (Table 1) and the environmental effects analyses (Figure 6e). Mean seed mass was greatest in the medium light environment, and lowest in the high light environment (Figure 6e). In the genetic analysis there was a significant maternal environment-offspring environment interaction for seed mass (P=0.048, Table 1). A within-matemal-environment genetic analysis was performed to understand this interaction (Table 2). The offspring of medium light mothers altered seed size in response to the environment (P<0.0001), but those from high light mothers did not (Pr-0.27). These results are consistent with the pattern seen in Figure 6e. Moreover, this analysis revealed genetic effects on seed provisioning expressed in offspring derived from both maternal environments (medium light P=0.04, high light P=0.056). Fitness consequences of maternal effects Phenotypic correlations with seed mass-Larger seeds were less dormant (Table 3), achieved a greater winter size (Table 4), and emerged later in the second year (Table 4). Phenotypic correlations between mean seed mass of mothers and pr0portion of offspring surviving (Table 3), and offspring seed production (Tables 3, 4), were low and insignificant when corrected for multiple tests. This result suggests that if present, 53 Table 3. Pearson product-moment correlations by light environment for the maternal traits. Offspring traits are the mean values for all the progeny of a single mother planted in each offspring environment. The number in parentheses is the sample size. Traits: Maternal mean seed number (M-Seeds), Proportion of offspring germinating (O-Germ), Proportion of offspring surviving to flower (O-Surv), Offspring mean seed mass (0- Mass), Offspring mean seed number (O-Seeds). Correlations in bold are significant afier sequential Bonferroni adjustment at a table-wide level of a=0.05 (Rice, 1989). #P<0.1, *P<0.05, **P<0.0l, ***P<0.001. M-Seeds O-Germ O-Surv O-Mass O-Seeds LQ‘ZLSJm Maternal Mean Seed Mass -0.16*** 0.21*** -0.04 0.15# 0.15# (449) (618) (324) (144) (145) 40%.5311 Maternal Mean Seed Mass -0.11# 0.26*** -0.02 0.15* 0.03 (258) (620) (338) (264) (265) LQQPALSnn Maternal Mean Seed Mass 0.25*** 0.23*** 0.09 0.10# 0.12* (180) (622) (350) (346) (346) 54 Table 4. Pearson product-moment correlations with maternal mean seed mass by light environment for traits of individual offspring. Traits: emergence date (Edate), winter size (W size), mean seed mass (Smass), seed number (Seeds). Correlations in bold are significant afier sequential Bonferroni adjustment at a table-wide level of a=0.05. #P0-0.1 —'>O'3'0'4 __._>0.1-0.2 —} 0.4-0.5 _>0.2-0.3 -—->>0-5 —> Positive " " + Negative Figure 8. 60 DISCUSSION The results of this study demonstrate that: (1) light environments in this population of Collinsia verna vary in space but are correlated across years in a way likely to favor plastic maternal effects (Figure 4); (2) there are significant maternal environmental effects on seed number (Figures 5, 6), seed size (Figure 5), and dormancy (Figure 6); (3) there are significant genotype-environment interactions in the field for seed size and dormancy (Figure 5); and (4) seed size may have important consequences for offspring fitness through its effect on size at overwintering. Below, I discuss whether these results support the hypotheses that larger seed size increases offspring establishment in reduced light environments and increased seed dormancy benefits mothers in low light environments where offspring survival is unlikely. I conclude with a discussion of what the genotype-enviromnent interactions for seed size and dormancy found in this study suggest about the factors affecting the evolution of plastic maternal effects. Seed size plasticity and variation in dormancy-In both years of this study, plants growing in the high light environments produced smaller seeds on average than plants growing in the medium light environments (Figures 5b, 6e). In low light in both years mean seed mass was lowest and most variable. Reduced seed mass in the low light environments was probably a consequence of absolute resource limitation. The mean number of seeds produced by mothers in low light (10% sun) was only slightly more than one (Figure 5a), while offspring in the low light environment (18% sun) produced about three seeds (Figure 6e). The significant negative correlation between seed size and number in the low light environment but not in medium or high light (Table 3), suggests 61 a seed size-number tradeoff in low light. Increases in seed mass as light resources decline from high to medium levels would be beneficial if offspring experience light environments similar to their mothers (as shown here), and if offspring from large seeds had higher establishment, survival, or fecundity in medium but not high light environments. However, the other results of this study provide limited evidence that larger seeds may be indirectly beneficial to offspring survival and fecundity at all light levels. Although the offspring of medium and high light mothers survived at similar rates across all environments (Figure 6b), and there was no direct selection on seed size in the survival episode (Figure 7), the path analysis suggested that increased seed size benefits offspring indirectly in all environments by contributing to larger overwinter size, and higher survival to flowering (Figure 8). Seed mass was only very weakly correlated with subsequent seed production (Table 4), and was under no direct selection in the fecundity episode (results not shown), but the heavier offspring of medium light mothers produced as many or more seeds than offspring of high or low light mothers by both the direct and cumulative measures (Table 1, Figures 6c-d). Other studies have consistently shown direct and indirect effects of increased seed size on dormancy, emergence time, early size, early growth rates, and many traits later in the life history (recent reviews in Westoby et al. 1997, Rees 1997). In competitive situations, differences in seedling size can persist throughout the life cycle, and lead to differences in fitness (Gross 1984, Stanton 1984, Fenner 1985, Morse and Schmitt 1985, Stratton 1989, Gross and Smith 1991). Although Thiede's (1996) two-year study of selection on early life history traits in this population of Collinsia verna found no direct 62 selection on seed mass, she did find direct selection for later emergence and larger winter size. In addition, both emergence time and winter size had positive genetic correlations with seed size. Other univariate (Kalisz 1986, Winn 1988, Biere 1991b), and multivariate (Mitchell-Olds and Bergelson 1990, Stratton 1992, Bennington and McGraw 1995) selection studies also have detected relationships between juvenile traits (seed mass, emergence date, and juvenile size) and survival or fecundity. Donohue and Schmitt (1998) showed that greater seed mass in Plantago lanceolata increases individual fitness; however, in contrast to this study, they found that maternal fitness was enhanced in some environments when mothers reduced seed mass and produced more seeds. The reduced performance of the offspring of low and high light grown mothers (Figures 6c-e) may have many causes. Many of these offspring started from smaller seeds. Resource limitation in the low light treatment likely accounts for the lack of seed set in more than 60% of the plants that survived to harvest. In addition, plants in low and high light environments flowered out of synchrony with the rest of the population (Chapter 4). Consequently, these plants may have received poor pollinator service, and their seeds may have been produced through self pollination (Kalisz et al. 1999). There is modest inbreeding depression in this species (Kalisz 1989), and population (Kalisz et al. unpublished data). As a result, the possible positive maternal effects on survival seen in the home environments of offspring of low and high light mothers may be offset later in the life history by the expression of inbreeding depression. Other selective factors not addressed in this study may favor small seeds in high light environments. Seed predation has been shown to be higher on forest edges (e.g. 63 Jules and Rathcke 1999, Manson et al. 1999), and some evidence suggests that seed foraging rodents favor species with larger seeds (e. g. Kelrick et al. 1986, Reader 1993, Hulme 1998). But there is little direct evidence that seed predators select larger seeds within species, and there are many examples of non-size-selective seed predation.(e. g. Kerley and Erasmus 1991, Meiners and Stiles 1997). It has been suggested that seed predation on species with a soil seed bank should select for smaller seeds (Thompson 1987) The significant maternal environment-offspring environment interaction effect on the size of seeds produced by offspring is intriguing (Table 1). Essentially, the size of grandchildren (and possibly their fitness) is in part determined by the environment of their grandmothers. Grandparent environment has been shown to effect grandprogeny phenotype in the lab in both plants (e. g. Case et al. 1996 and references therein) and animals (Fox and Savalli 1998 and references therein), but studies showing such effects in the field are rare. In the present study, the offspring of medium light mothers differentially provisioned seeds depending on their own light environment. In contrast, the offspring of high light mothers were insensitive to their own environment, and always made large seeds. Because in the second year plants in high light produced no more seeds than plants growing in medium light (Figure 6c), the results presented here suggest that high light could also be a stressfirl environment. The higher dormancy in offspring of low light mothers (Figure 6a) may allow these mothers to disperse their offspring further in space or time to better environments, or may be beneficial if pre-reproductive mortality in reduced light is variable from generation to generation. However, a strong test of these hypotheses would be very 64 difficult because individual dormant seeds would have to be followed for many years. Moreover, these more dormant seeds may simply be smaller and less viable because of resource limitation in their mothers, or they may be the less viable products of self fertilization. One other aspect of the offspring of low light mothers deserves comment. The insignificant trend toward lower mortality in low light and higher mortality in high light for the progeny of low light parents (Figure 6b), may be the result of a beneficial maternal effect, but the result suggests that this effect is mediated through some aspect of seed quality not measured by seed size. First, as mentioned above, each year low light plants on average produced smaller seeds (Figures 5b, 6e). Moreover, the higher rate of dormancy in the offspring of low light mothers suggests that they may have thicker seed coats that would reduce the proportion of total seed mass allocated to resource storage. Differences in nutrient concentration (nitrogen, etc.) in seeds, or the use of more concentrated energy stores (lipids instead of carbohydrates) might affect seed quality but not seed size (Westoby et al. 1992). Genotype-environment interactions-The most striking results here are the genotype-environment interactions for dormancy and seed size (Figures 5b-c). Nearly all studies that have investigated maternal environment effects on seed mass have found them (reviewed by Roach and Wulff 1987, Platenkamp and Shaw 1993, but see Weiner et al. 1997). More recent studies have demonstrated significant genetic variation in seed mass, germination date, and seedling size (e.g. Shaw et al. 1995, Helenurm and Schaal 1996, Sultan 1996, Byers et al. 1997, Husband and Gurney 1998, Thiede 1998). But studies that show genetic variation for plastic maternal effects are rare (Donohue and 65 Schmitt 1998). Previous studies in Collinsia verna have shown that seed size and seed dormancy are variable (Kalisz 1989, Thiede 1996), and under both additive genetic and matemal genetic control (Thiede 1998, unpublished data). The results of the present study suggest that any main effects of either genotype or environment on these traits should be interpreted with caution. The genotype-environment interactions for seed size and dormancy can be interpreted as genetic variation for plasticity in these traits. Artificial selection on the plasticity of dormancy and seed size would almost certainly succeed in altering reaction norms. However, the alternate question is why does this genetic variation persist? Other factors may be limiting plasticity evolution and contributing to the maintenance of genetic variation in plasticity of seed size and dormancy (Mitchell-Olds 1992, Schmitt 1995). Future evolution in depends on the patterns of natural selection on these traits, gene flow between environments, and on the genetic correlations among traits. In particular, variable selection can maintain genetic variation in natural populations if there are genotype-environment interactions for components of fitness. The basic quantitative genetic equation for multivariate evolution, A2 = GB, shows that trait evolution depends on both the selection gradients, B, and the genetic covariance matrix, G. Two hypotheses stand out as explanations for the persistence of genetic variation in the plasticity of dormancy and seed size. First, direct selection (B) on these traits may be variable within light environments so that optima vary and fluctuate. Second, correlated responses to selection may maintain some traits away from their univariate optima. In both cases, genetic variation would be maintained. 66 Course grained, between- generation variation in survival to reproduction can select for increased dormancy (reviews in Brown and Venable 1986, Evans and Cabin 1995, Rees 1997). If this selection differs across environments, mothers should produce seeds that are more dormant in some environments than others. However, if variation in survival to reproduction is fine grained (occurring within environments or generations), and there are no good cues to predict offspring survival, then genetic variation in dormancy may be maintained. In a companion study to this one, I have found a twofold difference in survival to reproduction within light environments as a consequence of the presence or absence of leaf litter (Chapter 4). The distribution of leaf litter in deciduous forests during the fall is unpredictably variable at the single leaf scale (Frankland et al. 1963, Sydes and Grime 1981, Facelli and Carson 1991, Molofsky and Augspurger 1992, personal observation). In this population, leaf litter is quite transient, with substantial decomposition occurring before the onset of winter (personal observation). Consequently, leaf litter may be an unpredictable cause of fine grained variable natural selection on dormancy independent of light environment, and may lead to the persistence of genetic variation in dormancy and its plasticity. There is strong evidence that selection on seed size in this population is indirect through emergence date and winter size (this study, Thiede 1996). Other studies in Collinsia verna have found significant positive genetic correlations between seed size and emergence date (Kalisz 1989, Thiede 1998) and seed size and winter size (Thiede 1998); and significant negative genetic correlations between emergence date and winter size (Thiede 1998, Chapter 5). Consequently, selection for larger winter size as is consistently seen produces correlated responses for larger seeds and earlier emergence. However, 67 since selection for early emergence produces correlated responses for smaller seeds, the overall effects on seed size are unclear. Further, my study of natural selection in this population (Chapter 4) found that the direction of direct selection on emergence date (and consequently selection on seed size) changes depending on the presence or absence of leaf litter across all light environments. Leaf litter selects for later emergence and larger seeds; the absence of leaf litter selects for earlier emergence and smaller seeds. The path analysis (Figure 8) suggests that early emergence and larger seed size are alternate strategies that both allow seedlings to achieve a size sufficient to survive the winter. Larger seeds may be further advantaged if they emerge later and thereby avoid the negative effects of leaf litter in early fall. More interesting still, seed size appears to be largely controlled by maternal genes (Chapter 3, Thiede 1998), while timing of emergence is an additive genetic trait (Chapter 3, Thiede 1998). Consequently, the evolution of seed size will have very complicated and unpredictable dynamics where selection in parents and offspring is likely to conflict (Westoby et al. 1992). Conclusion-The spatial and temporal variation in light in this population favors the evolution of plastic maternal effects, and seed size and dormancy are plastic. But genetic variation in plasticity of seed size and dormancy may be maintained by variable direct selection or correlated responses. Environmental factors like leaf litter may contribute to this unpredictable, heterogeneous selective environment resulting in the maintenance of a diversity of specialized genotypes with different seed provisioning and dormancy strategies. Offspring of mothers from the two extreme light environments produced fewer, and often smaller seeds when grown in their home environments. This result suggests that some plasticity of seed size seen in this study may be a maladaptive 68 consequence of stressful maternal environments. This study demonstrates the value of performing environmental manipulations in the field to study micro-evolutionary processes. By quantifying maternal effects, plasticity, genetic variation, and fitness components in an ecologically and evolutionarily relevant context, it was possible to assess both the adaptive value of plasticity and the factors that might affect the future evolution of the population. Wider application of this approach will help to answer fundamental questions about maternal effects, plasticity evolution, and the factors that maintain genetic variation in populations. 69 LITERATURE CITED Arnold, S. J., and M. J. Wade. 1984. On the measurement of natural and sexual selection: theory. Evolution 38:709—719. Baskin, J. M., and C. C. Baskin. 1983. Germination ecology of Collinsia verna, a winter annual of rich deciduous woodlands. Bulletin of the Torrey Botanical Club 110:311-315. Bell, G. and M. J. Lechowicz. 1991. The ecology and genetics of fitness in forest plants. 1. environmental heterogeneity measured by explant trials. Journal of Ecology 79:663-685. Bennington, C. C., and J. B. McGraw. 1995. Natural selection and ecotypic differentiation in Impatiens pallida. Ecological Monographs 65:303-323. Biere, A.l99la. Parental effects in Lychnisflos-cuculi. 1. Seed size, germination and seedling performance in a controlled environment. Journal of Evolutionary Biology 3:447-465. Biere, A.l99lb. Parental effects in Lychnisflos-cuculi. H. Selection on the time of emergence and seedling performance in the field. Journal of Evolutionary Biology 3:467-486. Brown, J. S. and D. L. Venable. 1986. Evolutionary ecology of seed-bank annuals in temporally varying environments. American Naturalist 127231-47. Byers, D. L., G. A. J. Platenkamp, and R. G. Shaw.1997. Variation in seed characters in Nemophila menziesii: evidence of a genetic basis for maternal effect. Evolution 51:1445-1456. Carriere, Y. 1994. Evolution of phenotypic variance: non-Mendelian parental influences on phenotypic and genotypic components of life-history traits in a generalist herbivore. Heredity 72:420-430. Case, A. L., E. P. Lacey, and R. G. Hopkins. 1996. Parental effects in Plantago lanceolata L. 2. Manipulation of grandparental temperature and parental flowering time. Heredity 76:287-295. Dixon, P. M. 1993. The bootstrap and jackni fe: describing the precision of ecological indices. Pages 290-319 in S. M. Scheiner and J. Gurevitch, editors. Design and analysis of ecological experiments. Chapman and Hall, New York, New York, USA. 70 Donohue, K,. and J. Schmitt. 1998. Maternal environmental effects in plants: adaptive plasticity? Pages 137-158 in T. A. Mousseau and C. W. Fox, editors. Maternal Effects as Adaptations. Oxford University Press, New York, New York, USA. Dudley, S. A., and J. Schmitt. 1996. Testing the adaptive hypothesis: Density-dependent selection on manipulated stem length in Impatiens capensis. American Naturalist 147:445-465. Endler, J. A. 1993. The color of light in forests and its implications. Ecological Monographs 63:1-27. Evans, A. S., and R. J. Cabin. 1995. Can dormancy affect the evolution of post- germination traits? The case of Lesquerella fendleri. Ecology 76:344-356. Facelli, J. M., and W. P. Carson. 1991. Heterogeneity of plant litter accumulation in successional communities. Bulletin of the Torrey Botanical Club 118:62-66. Fenner, M. 1985. Seed Ecology. Chapman and Hall, London, UK. Forbes, L. S. 1991. Optimal size and number of offspring in a variable environment. Journal of Theoretical Biology 150:299-304. Fox, C. W. and U. M. Savalli. 1998. Inheritance of environmental variation in body size: superparasitism of seeds affects progeny and grandprogeny body size via a nongenetic maternal effect. Evolution 52:172-182. F rankland, J. C., J. D. Ovington, and C. Macrae. 1963. Spatial and seasonal variations in soil, litter, and ground vaegetation in some Lake District woodlands. Journal of Ecology 51:97-112. Goldberg, D. E. 1990. Components of resource competition in plant communities. Pages 27-50 in J. Grace and D. Tilrnan, editors. Perspectives on plant competition. Academic Press, San Diego, California, USA. Gross, K. L. 1984. Effects of seed size and growth form on seedling establishment of six monocarpic perennial plants. Journal of Ecology 72:369-387. Gross, K. L., and A. D. Smith. 1991. Seed mass and emergence time effects on performance of Panicum dichotomiflorum Michx. across environments. Oecologia 87:270-278. Haig, D. And M. Westoby. 1988. Inclusive fitness, seed resources, and maternal care. Pages 60-79 in J. and L. Lovett-Doust, editors. Plant Reproductive Ecology. Oxford University Press, Oxford, UK. 71 Helenurm, K., and B. A. Schaal. 1996. Genetic and maternal effects on offspring fitness in Lupinus texensis (F abaceae). American Journal of Botany 83:1596-1608. Hulme, P. E. 1998. Post-dispersal seed predation and seed bank persistence. Seed Science Research 8:513-519. Husband, B. C., and J. E. Gurney. 1998. Offspring fitness and parental effects as a function of inbreeding in Epilobium angustifoliurn (Onagraceae). Heredity 80:173-179. Janzen, F. J., and H. S. Stern. 1998. Logistic regression for empirical studies of multivariate selection. Evolution 52:1564-1571. Jules, E. S., and B. J. Rathcke. 1999. Mechanisms of reduced Trillium recruitment along edges of old-grth forest fi'agments. Conservation Biology 13:784-793. Kalisz, S. 1986. Variable selection on the timing of emergence in Collinsia verna (Scrophulariaceae). Evolution 40:479-491. Kalisz, S. 1989. Fitness consequences of mating system, seed weight, and emergence date in a winter annual, Collinsia verna. Evolution 43:1263-1272. Kalisz, S. 1991. Experimental determination of seed bank age structure in the winter annual Collinsia verna. Ecology 72:575-585. Kalisz, S. and M. A. McPeek. 1992. Demography of an age-structured annual: projection matrices, elasticity analyses and seed banks. Ecology 73:1082-1093. Kalisz, S. and M. A. McPeek. 1993. Extinction dynamics, population growth and seed banks. Oecologia 95:314-320. Kalisz, S., L. Horth, and M. A. McPeek.l997. Fragmentation and the role of seedbanks in promoting persistence in isolated populations of Collinsia verna. Pages 286-312 in M. W. Schwartz, editor. Conservation in highly fragmented landscapes. Chapman and Hall, New York, New York, USA. Kalisz, S., D. Vogler, B. Fails, M. Finer, E. Sheppard, T. Herman, and R. Gonzales. 1999. The mechanism of delayed selfing in Collinsia verna (Scrophulariaceae). American Journal of Botany 86:1239-1247. Kelrick, M. I., J. A. MacMahon, P. R. Parmenter, and J. A. Sisson. 1986. Native seed preferences of shrub-steppe rodents, birds, and ants: the relationships of seed attributes and seed use. Oecologia 68:327-337. Kerley, G. I. H. and T. Erasmus. 1991. What do mice select for in seeds? Oecologia 72 86:261-267. Koenig, W. D., S. S. Albano, and J. L. Dickinson. 1991. A comparison of methods to partition selection acting via components of fitness: Do larger male bullfrogs have greater hatching success? Journal of Evolutionary Biology 4:309-320. Lacey, E. 1991. Parental effects on life-history traits in plants. Pages 735-744 in E. C. Dudley, editor. The Unity of Evolutionary Biology. Volume II. International Congress of Systematic and Evolutionary Biology IV. Dioscorides, Portland, Oregon, USA. Lacey, E. P., S. Smith, and A. L. Case. 1997. Parental effects on seed mass: Seed coat but not embryo/endosperrn effects. American Journal of Botany 84:1617-1620. Lewontin, R. C. 1974. The analysis of variance and the analysis of causes. American Journal of Human Genetics 26:400-411. Littell, R. C., G. A. Milliken, W. W. Stroup, and R. D. Wolfinger. 1996. SAS system for mixed models. SAS Institute Inc., Cary, North Carolina, USA. Lynch, M., and B. Walsh. 1998. Genetics and Analysis of Quantitative Traits. Sinauer Associates Inc., Sunderland, Massachusetts, USA. Manson, R. H., R. S. Ostfeld, and C. D. Canham. 1999. Responses of a small mammal community to heterogeneity along forest-old-field edges. Landscape Ecology 14:355-367. McGinley, M. A., D. H. Temme, and M. A. Geber. 1987. Parental investment in offspring in variable environments: Theoretical and empirical considerations. American Naturalist 1302370-398. Meiners, S. J. and E. W. Stiles. 1997. Selective predation on the seeds of woody plants. Journal of the Torrey Botanical Society 124367-70. Mitchell-Olds, T. 1992. Does environmental variation maintain genetic variation? A question of scale. Trends in Ecology and Evolution 72397-398. Mitchell-Olds, T., and J. Bergelson. 1990. Statistical genetics of an annual plant, Impatiens capensis. II. Natural selection. Genetics 124:417-421. Mitchell-Olds, T., and R. Shaw. 1987. Regression analysis of natural selection: statistical inference and biological interpretation. Evolution 41:1149-1161. Molofsky, J ., and C. K. Augspurger. 1992. The effect of leaf litter on early seedling establishment in a tropical forest. Ecology 73:68-77. 73 Morse, D. H., and J. Schmitt. 1985. Propagule size, dispersal ability, and seedling performance in Asclepias syriaca. Oecologia 67:372-3 79. Mousseau, T. A., and C. W. Fox, editors. 1998a. Maternal effects as adaptations. Oxford University Press, New York, New York, USA. Mousseau, T. A., and C. W. Fox. 1998b. The adaptive significance of maternal effects. Trends in Ecology and Evolution 13:403-407. Noreen, E. W. 1989. Computer-intensive methods for testing hypotheses: an introduction. John Wiley and Sons, New York, New York, USA. Philippi, T. 1993. Bet-hedging germination of desert annuals: Variation among populations and maternal effects in Lepidium lasiocarpum. American Naturalist 1421488-507. Pigliucci, M. 1996. How organisms respond to environmental changes: from phenotypes to molecules (and vice versa). Trends in Ecology and Evolution 11:168-173. Platenkamp, G. A. J ., and R. G. Shaw. 1993. Environmental and genetic maternal effects on seed characters in Nem0phila menziesii. Evolution 47: 540-555. Reader, R. J. 1993. Control of seedling emergence by ground cover and seed predation in relation to seed size for some old-field species. Journal of Ecology 81:169-175. Rees, M. 1997. Evolutionary ecology of seed dormancy and seed size. Pages 121-142 in J. Silvertown, M. Franco, and J. L. Harper, editors. Plant life histories: ecology, phylogeny, and evolution. Cambridge University Press, Cambridge, UK. Roach, D. A., and R. D. Wulff. 1987. Maternal effects in plants. Annual Review of Ecology and Systematics 18:209-235. Rossiter, M. C. 1996. Incidence and consequences of inherited environmental effects. Annual Review of Ecology and Systematics 27:451-476. SAS Institute Inc. 1989. SAS/STAT Users Guide, Version 6, Fourth Edition. SAS Institute, Cary, North Carolina, USA. SAS Institute Inc. 1990. SAS Guide to Macro Processing, Version 6, Second Edition. SAS Institute, Cary, North Carolina, USA. SAS Institute Inc. 1992. SAS/STAT software: changes and enhancements, release 6.07. SAS Technical Report P-229. SAS Institute, Cary, North Carolina, USA. SAS Institute Inc. 1997. SAS/STAT software: changes and enhancements through release 74 6.12. SAS Institute, Cary, North Carolina, USA. Schaal, B. A. 1984. Life-history variation, natural selection, and maternal effects in plant populations. Pages 188-206 in R. Dirzo and J. Sarukan, editors. Perspectives on plant population ecology. Sinauer, Sunderland, Massachusetts, USA. Schlichting, C. D. 1986. The evolution of phenotypic plasticity in plants. Annual Review of Ecology and Systematics 172667-693. Schmitt, J. 1995. Genotype-enviromnent interaction, parental effects, and the evolution of plant reproductive traits. in P. C. Hoch and A. G. Stephenson, editors. Experimental and molecular approaches to plant biosystematics. Monographs in Systematic Botany fiom the Missouri Botanical Garden. 53: 199-214. Schmitt, J ., J. Niles, and R. D. Wulff. 1992. Norms of reaction of seed traits to maternal environments in Plantago lanceolata. American Naturalist 1391451 -466. Schmitt, J ., A. C. McCormac, and H. Smith. 1995. A test of the adaptive plasticity hypothesis using transgenic and mutant plants disabled in phytochrome-mediated elongation responses to neighbors. American Naturalist 146:937-953. Searle, S. R., G. Casella, and C. E. McCulloch. 1992. Variance components. John Wiley and Sons, Inc., New York, New York, USA. Shaanker, R. U., K. N. Ganeshaiah, and K. S. Bawa. 1988. Parent-offspring conflict, sibling rivalry, and brood size patterns in plants. Annual Review of Ecology and Systematics 19: 177-205. Shaw, R. G. 1987. Maximum -likelihood approaches to quantitative genetics of natural populations. Evolution 41:812-826. Shaw, R. G., G. A. J. Platenkamp, F. H. Shaw, and R. H. Podolsky. 1995. Quantitative genetics of response to competitors in Nemophila menziessi: A field experiment. Genetics 1392397-406. Stanton, M. L. 1984. Seed variation in wild radish: effect of seed size on components of seedling and adult fitness. Ecology 65:1105-1112. Stratton, D. J. 1989. Competition prolongs the expression of maternal effects in seedlings of Erigeron annuus (Asteraceae). American Journal of Botany 76:1646—1653. Stratton, D. J. 1992. Life-cycle components of selection in Erigeron annuus: I. Phenotypic selection. Evolution 46292-106. Stratton, D. J. and C. C. Bennington. 1998. F ine-grained spatial and temporal variation 75 does not maintain genetic variation in Erigeron annuus. Evolution 52:678-691. Sultan, S. E. 1987. Evolutionary implications of phenotypic plasticity in plants. Evolutionary Biology 21:127-176. Sultan, S. E. 1995. Phenotypic plasticity and plant adaptation. Acta Botanica Neerlandica 44:363-383. Sultan, S. E. 1996. Phenotypic plasticity for offspring traits in Polygonum persicaria. Ecology 77:1791-1807. Sultan S. E., and F. A. Bazzaz. 1993. Phenotypic plasticity in Polygonum persicaria. I. Diversity and uniformity in genotypic response to light. Evolution 47:1009-1031. Sydes, C., and J. P. Grime. 1981. Effects of tree litter on herbaceous vegetation in deciduous woodland. Journal of Ecology 69:237-248. Thiede, D. A. 1996. The impact of maternal effects on adaptive evolution: combining quantitative genetics and phenotypic selection in a natural plant population. Ph.D. Dissertation. Michigan State University. East Lansing, Michigan, USA. Thiede, D. A. 1998. Maternal inheritance and its effect on adaptive evolution: A quantitative genetic analysis of maternal effects in a natural plant population. Evolution 522998-1015. Thompson, K. 1987. Seeds and seed banks. New Phytologist 106223-34. van Tienderen, P. H. 1991. Evolution of generalists and specialists in spatially heterogeneous environments. Evolution 45: 13 17-133 1. Via, S., and R. Lande. 1985. Genotype-environment interaction and the evolution of phenotypic plasticity. Evolution 39:505-522. Wahlsten, D. 1990. Insensitivity of the analysis of variance to heredity-environment interaction. Behavioral and Brain Sciences 13:109-161. Weiner, J ., S. Martinez, H. Muller-Scharer, P. Stoll, and B. Schmid. 1997. How important are environmental maternal effects in plants? A study with Centaurea maculosa. Journal of Ecology 85:133-142. Westoby, M., E. Jurado, and M. Leishman. 1992. Comparative evolutionary ecology of seed size. Trends in Ecology and Evolution 72368-372. Westoby, M., M. Leishman, and J. Lord. 1997. Comparative ecology of seed size and dispersal. Pages 143-162 in J. Silvertown, M. Franco, and J. L. Harper, editors. 76 Plant life histories: ecology, phylogeny, and evolution. Cambridge University Press, Cambridge, UK. Winn, A. A. 1988. Ecological and evolutionary consequences of seed size in Prunella vulgaris. Ecology 69: 1537-1544. Wolf, J. 13., E. D. Brodie 111, J. M. Cheverud, A. J. Moore, and M. J. Wade. 1998. Evolutionary consequences of indirect genetic effects. Trends in Ecology and Evolution 13:64-69. Wulff, R. D. 1995. Environmental maternal effects on seed quality and germination. Pages 491-505 in J. Kigel, and G. Galili, editors. Seed development and germination. Marcel Dekker Publication, New York, New York, USA. Wulff, R. D., A. Caceres, and J. Schmitt. 1994. Seed and seedling responses to maternal and offspring environments in Plantago lanceolata. Functional Ecology 8:763- 769. 77 Chapter 3 EVOLUTION OF REACTION NORMS: THE QUANTITATIVE GENETICS OF RESPONSES TO VARIABLE LIGHT ENVIRONMENTS IN A NATURAL PLANT POPULATION INTRODUCTION The consequences of environmental heterogeneity are an important focus of theoretical and empirical studies in evolutionary ecology. Quantitative traits in all organisms generally exhibit plastic responses to changes in the biotic or abiotic environment (Travis 1994, Roff 1997). Recent studies demonstrate that environmental heterogeneity can cause spatial and temporal variation in natural selection (e. g. Kalisz 1986, Kelly 1992, Stratton 1992, 1995). Depending on patterns of gene flow, the magnitude of fitness differences, and stochastic factors, variable selection can both maintain genetic variation within populations and lead to adaptive differentiation between populations or subpopulations (e. g. Gillespie and Turelli 1989, Mitchell-Olds 1992; see also Hedrick et al. 1976, Hedrick 1986). Other theory predicts that with gene flow between environments, variable selection may lead to phenotypically-plastic generalist genotypes that perform well across a broad range of environments (e. g. Via and Lande 1985, de Jong 1995, van Tienderen 1997, Scheiner 1998; reviews in Scheiner 1993, Roff 1997) However, many factors may constrain reaction norm evolution and result in the persistence of specialist genotypes and the maintenance of genetic variation. These factors include the scale of environmental heterogeneity (course vs. fine; Lynch and Gabriel 1987, van Tienderen 1991, Gabriel and Lynch 1992), a lack of predictability 78 about future environmental states (Bradshaw 1965, Moran 1992, Getty 1996), developmental limitations (van Tienderen 1990), physiological costs of plasticity (Via and Lande 1985, van Tienderen 1991, DeWitt et al. 1998), or a variety of genetic constraints (discussed below, Falconer 1952, Via and Lande 1985). Data from natural populations regarding the genetics of responses to variable environments is scarce (recent reviews in Roff 1997, Schlichting and Pigliucci 1998). Genetic constraints and the evolution of reaction norms Bradshaw (1965) recognized that the characteristics of plastic responses are specific to particular environmental factors and are under genetic control. In this context, the terms generalist and specialist describe the relative ability to maintain fitness in variable environments through coordinated, plastic responses in underlying traits. Generalists have adaptive plasticity in physiological and/or morphological traits that allows them to maintain performance when resources are scarce, and to boost performance when resources are abundant (de Jong 1990, Thompson 1991). Whether generalists can be as fit in particular environments as specialists on those environments remains controversial, and can affect whether variable selection favors plasticity or the maintenance of genetic variation (Gillespie and Turelli 1989). Specialists for favorable environments lack the underlying plasticity in physiological, morphological, and behavioral traits necessary to maintain high relative fitness in less favorable environments. Specialists for stressful environments may be more fit than generalists in those environments, but may lack the ability to respond when resources are abundant, and so have low, flat reaction norms for fitness. Because different combinations of plasticity and canalization in underlying 79 developmental, physiological, or morphological traits can trade off to produce equally fit genotypes, there is no simple relationship between fitness across variable environments and plasticity in a trait (Via 1987, Schlichting and Pigliucci 1995). For example, in a series of greenhouse experiments, Sultan and Bazzaz (1993a, b, c) found that diverse patterns of physiological and morphological plasticity to light, water, and nutrients can produce convergent reaction norms for reproductive performance. Thus, it is necessary to consider how groups of traits are integrated to produce generalists or specialists. Genetic constraints on the evolution of plasticity include an absolute lack of genetic variation and genetic interdependence between the expression of single traits in different environments. When the same genes determine a trait in different environments, the cross-environment genetic correlation approaches one or negative one. Selection in one environment will change the expression of the trait in all other environments. Values of the cross-environment genetic correlation significantly different from one or negative one indicate the presence of genotype-environment interaction, which is a measure of genetic variation for phenotypic plasticity (Via 1987). As the cross-environment genetic correlation for a trait approaches zero, genetic variation for plasticity of the trait increases. Depending on the relationship between a trait and fitness in different environments, significant negative or positive cross-environment genetic correlations can either slow or accelerate reaction norm evolution. For example, Shaw and coworkers (1995) found a negative genetic correlation for dry mass across manipulated competitive environments in a study of Nemophila menziesii. This negative relationship is expected to maintain specialist genotypes in the population and slow the evolution of a competitive generalist. Similarly, genetic correlations between different traits within and across 80 environments may also affect plasticity evolution. Plastic responses to light Light is an important resource for photosynthetic plants, and light availability is a strong selective agent in plant populations (reviews in Goldberg 1990, Sultan and Bazzaz 1993a). Moreover, because light availability varies in space and time, selection should favor generalist genotypes possessing plastic traits that would both allow tolerance of low light levels and the conversion of high light availability into increased growth and reproduction. Genotypic differences in the plastic response of many physiological, morphological, and life history traits to light quantity or quality have been characterized many times in greenhouse or common garden environments (e. g. Clough et al. 1980, Sultan and Bazzaz 1993a, Schmitt 1993, Andersson and Shaw 1994, Pigliucci et al. 1995, Dudley and Schmitt 1995). However, it is uncertain if the genotypic differences seen in these studies would be expressed in the field, or if expressed how they might affect fitness or the dynamics of plasticity evolution under natural conditions. A better understanding of the evolution of phenotypic plasticity requires the careful manipulation of selectively relevant environmental factors and the characterization of additive genetic variation for phenotypic plasticity within natural populations of plants (Schmitt 1995, Sultan 1995, Pigliucci 1996). In plants, maternal effects are known to be an important source of resemblance between relatives (Roach and Wulff 1987, Donohue and Schmitt 1998). Moreover, it has been shown that maternal effects can obscure genetically based tradeoffs in performance across environments (Shaw et al. 1995). Relatively few studies have characterized additive genetic variance and covariance independent of maternal effects within natural plant populations (e. g. 81 Mitchell-Olds 1986, Mitchell-Olds and Bergelson 1990, Schwaegerle and Levin 1991, Montalvo and Shaw 1994, Schoen et al. 1994, Campbell 1996, 1997a, b). Fewer still have estimated these parameters across a range of natural field environments (Shaw et al. 1995, Bennington and McGraw 1996, Wulff 1998). Recent empirical comparisons suggest that heritability estimates may be similar across different environments (Roff 1997). However, it is the genetic correlations that are keys to plasticity evolution, and recent comparisons of genetic correlations across environments and between populations have found no consistent patterns (reviews in Roff 1997). In this study, two generations of half-sib Collinsia verna plants were grown in the field in three manipulated and two natural light environments. Here I address the following questions concerning the evolution of phenotypic plasticity: (1) Are there additive genetic and/or maternal effects on traits with plastic responses to light environment? (2) Are there genotype-environment interactions indicative of genetic variation in plasticity to light? (3) Are there fitness differences among genotypes that would suggest specialization for particular light resource environments? By addressing these questions under natural field conditions, this study avoids many of the interpretive limitations of lab, greenhouse or common garden experiments. METHODS Study system Collinsia verna is a winter annual wildflower native to moist woods and floodplain forests in eastern North America. In many Collinsia habitats, light is so limiting during the summer months that the forest floor is nearly barren. All growth of C. verna occurs during the seasonal light-gap between forest canOpy senescence in the fall 82 and leafout in the spring. The one hectare study population resides along the south facing edge of a woodlot adjacent to an agricultural field in Kalamazoo County, southwest Michigan. Seed germination in the fall is cued by diurnal temperature fluctuations (Baskin and Baskin 1983). Seedlings emerge between September and December, and overwinter with only cotyledons or a single pair of leaves. Flowering begins in late April and ceases as the forest canOpy closes in May. By mid June fruits ripen, seeds are passively dispersed, and plants die. Fecundity depends on the micro-environment, ranging fiom five to fifty seeds. Experimental design Light treatments-Light in this population is spatially and temporally variable. Peak irradiance in full sun at solar noon varies between 500 and 2000 umoles m'2 sec‘l over the October to May growing season. Depending on adjacent woody vegetation and proximity to the southern edge, small patches within the forest receive from 25% to 75% of this maximum (Figure 4 of Chapter 2). To establish different light levels in the field, in July 1995 I cleared all trees and shrubs from a 15 X 20 m area along the southern edge of the woodlot in the densest area of the Collinsia population. I then established 15 1.2 m2 plots in a 5 X 3 grid with 1 m spacing between plots (5 blocks of 3 plots each). I randomly assigned one of three light treatments to the plots of each block: 100% sun (high), 40% sun (medium), and 10% sun (low) (Figure l of Chaper 2). Light treatments were constructed using a wood lattice that continued to allow sun-flecks to reach the plots (Chapter 2). Mortality in low light during the first year (1995-96) was greater than 95%. To increase survival for the second year (1996-97), light levels in this treatment were increased from 10% to 18% of full sun in September 83 1996. For the second year, 10 additional natural plots were established. These plots were chosen to represent the range of natural light experienced by the population. Five plots were located in the forest interior (45% Sun) and five were located along the southern edge of the woodlot (70% Sun). Breeding design-I used two independent breeding designs to generate the families used in this study. Plants were collected prior to flowering at 5 to 20 m intervals fiom the full area of the population (204 in year 1, 50 in year 2). Crosses were performed in the greenhouse at Kellogg Biological Station. Sires were randomly chosen (year 1: 50; year 2: 12). Each sire was mated to three unique dams in standard nested half-sib designs (North Carolina Design I, Lynch and Walsh 1998, Chapter 18). In the first year, two sires with small initial mates were each mated to one additional dam. Flowers of dams used in crosses were emasculated in the bud stage to prevent self pollination. The crosses produced 9338 and 17 76 seeds in each year respectively. In year one, an average of 45 seeds from each full-sib family were randomly planted into the field plots (6840 seeds total: 152 full-sib families X 3 seeds/family X 3 light treatments X 5 blocks). In year two, an average of 50 seeds from each full sib family were randomly planted into each field plot (1776 seeds total: 36 full-sib families X 2 seeds/family X 5 light treatments X 5 replicates). Seeds were individually planted in 2 cm long uniquely numbered plastic straws inserted in the ground. In each year emergence and survival censuses were conducted weekly from September through December. When emergence ceased the number of cotyledons and true leaves on each seedling were counted, and the diameter of the largest cotyledon and leaf were measured. These data were used to calculate winter size (cotyledon area + leaf 84 area). I conducted a census for survival to flowering in mid-April. Starting in late April, the date of first flowering was recorded daily. After a plant flowered, a uniform sized piece of leaf tissue was collected with a hole punch from the youngest firlly expanded leaf. Samples were dried and weighed to determine specific leaf area. This trait quantifies the morphological changes made in leaf characters to optimize photosynthetic ability in different light environments. Plants surviving to the end of the experiment were harvested prior to seed dispersal in June of each year and scored for mainstem length, above ground vegetative biomass, number of seeds, and total seed mass. Reproductive investment was calculated as the proportion of total above ground biomass allocated to seeds. Because plants were dead or dying, this trait measures the efficiency with which they were able to convert vegetative biomass to seed mass. The light manipulations used in this study may simultaneously change many aspects of the biotic and abiotic environment. To better understand these affects, I measured light (Chapter 2) and several other parameters in the study plots. I counted number of conspecifics at overwintering and harvest; I measured soil temperature throughout the season using Onset Hobo data loggers; and I measured volumetric water content of the soil weekly in the fall using time-domain reflectometry. Data analysis Heritability-Within environment heritabilities (hz) were estimated using MTDFREML (Multiple Trait Derivative Free Restricted Maximum Likelihood, Boldman et al.1995). MTDFREML is a set of programs deveIOped to apply the animal model (description in Lynch and Walsh 1998, pp. 755-758) to the estimation of genetic variance components and the prediction of breeding values. Unlike ANOVA methods, the animal 85 model uses all available pedigree information to produce restricted maximum likelihood (REML) estimates of genetic variance components. From these variance components the BLUP (best linear unbiased prediction) breeding values were calculated. Likelihood ratio Chi-square tests were used to assess the significance of the additive genetic variance components. Vegetative biomass and seed number were log transformed to improve normality prior to all analyses. All assumptions of the nested design for estimation of quantitative genetic parameters were met in this study (Mitchell-Olds 1986, Mitchell-Olds and Rutledge 1986). Although Collinsia verna is self-compatible, outcrossing rates in this population were above 0.9 in each year of a three-year study (Kalisz et al. unpublished data), so upward bias in heritabilities due to inbreeding should be negligible. The high pre- flowering mortality in some environments in each year could bias the estimation of genetic parameters for subsequent traits if mortality was associated with some sibships. However, there were no significant effects of sire or sire-environment interaction on survival, indicating that no measurable selection on half-sib families occurred at this stage (see Results). Environmental eflects and interactions-Main effects of environment and genotype, and genotype-environment interactions were analyzed statistically using maximum likelihood methods and graphically by examining reaction norm plots for genotypic differences in slope. The approaches are complimentary and equally valuable because the power of statistical methods to detect genotype-environment interaction can be limited (Lewontin 1974, Via and Lande 1985, Wahlsten 1990). Reaction norm figures were constructed for each year by plotting paternal half-sib family trait means 86 against light enviromnents. Genetic variation for trait plasticity is suggested when trait values for families change rank across environments. Changes in rank for fitness suggest that negative cross-environment genetic correlations may retard the evolution of generalists (Via 1984). Mixed model REML analysis was used to examine the main and interactive effects of light environment, block, sire, and dam on each trait in each year. A split-plot model was used, with light environment as the whole plot factor, and sire and dam as subplot factors. Maximum likelihood methods are generally superior for analysis of unbalanced data (Searle et al. 1992, Littell et al. 1996). All analyses were completed using the MIXED procedure of SAS (SAS 1992, 1997). Blocks were treated as fixed because they were designed to capture an east-west gradient of diurnal morning-afternoon shading. Environmental effects were also treated as fixed, while sire, dam nested within sire, and their interactions with fixed effects were treated as random. Because the new natural plots in the second year could not be included in blocks with the existing experimental treatments, no block effect was included in the full analysis this year. Block effects in the first year and in an analysis of the three treatments in the second year were all insignificant. The final analysis included all main effects and two way interactions, except for the darn by block interactions. When examined in reduced models, the darn by block interactions and all three way interactions were highly insignificant. The significance of random effects were assessed using likelihood ratio tests, while fixed effects were tested by computing a Type III F statistic using the Satterthwaite option to estimate denominator degrees of freedom (Littell et al. 1996). No sequential Bonferroni 87 corrections (Rice 1989) have been applied to the results of these analyses because correction for multiple tests is not apprOpriate when correlations are expected between traits (Manly 1991, Simons and Roff 1996, Lynch and Walsh 1998 page 641). It is the general patterns that are of interest here, rather than specific comparisons among traits or environments. Because germination and survival are discrete variables they were analyzed using maximum likelihood logistic regression (SAS LOGISTIC procedure, SAS 1989). This is the most appropriate analysis for data of this type, but the procedure cannot accommodate the nested structure of the data (dams nested within sires). The fit of models with sire effects were compared to the fit of models with dam effects using likelihood ratio chi- square tests. Models with darn effects generally provided a poorer fit to the data, and never significantly improved the likelihood of sire based models. Consequently, the results presented are for models including terms for sire, light environment, and sire by light environment interaction. Genetic correlations across environments-The genetic correlations (r8) between the same trait expressed in different environments indicate the degree to which traits are free to evolve independently in those environments. They are directly interpretable in terms of evolutionary quantitative genetic theory and they provide a dimensionless measure for comparisons between traits and across environments (Via 1984, 1987, Via and Lande 1985). The best way to estimate genetic correlations across environments is the subject of much discussion and research (Via 1984, Dutilleul and Potvin 1995, Windig 1997, Dutilleul and Carriere 1998). The half-sib breeding designs used in this study allow the calculation of the narrow sense additive genetic correlations (unbiased by 88 any dominance or maternal effects) by two complementary methods. First, the cross environment additive genetic correlations were calculated as the Pearson product-moment correlations of the within environment predicted sire breeding values fiom the MTDFREML analyses described above. This method is similar to the family means approach (e. g. Via 1984, Simons and Roff 1996) but it eliminates bias due to unbalanced data and fractional contributions of dominance and environmental effects (Shaw et al. 1995, Lynch and Walsh 1998). Limitations of this approach are that genetic correlations cannot be calculated when additive genetic variance components are estimated as zero, and as in the case of family mean correlations, sampling error may cause the calculated genetic correlations to underestimate the true values (Cameron 1993, Notter and Diaz 1993, Mathur and Horst, 1994). This bias means that traits may not be as genetically independent across environments as it appears based on the correlations among breeding values. For comparison the cross environment genetic correlation was also estimated fi'om the variance components of mixed-model REML analyses (SAS proc. MIXED) according to the formula: 2 r ___ 6 sire g 2 2 ) '\/(C sireEl X0 sireE 2 where aim is estimated by the sire variance component from a two-way analysis, and of mg, and offing] are estimated by the sire variance components from separate one-way analyses in each environment (Yarnada 1962, Fry 1992, Windig 1997). As with estimating the genetic correlation from breeding values, the genetic correlation estimated by this technique is undefined whenever oz,,,,_.£, or o’mfl are zero. When 02,,,,E, or (Jig-m5; 89 are near zero, sampling error can produce correlations that are outside of theoretical bounds, sometimes resulting in estimates of rg that can be much in excess of one (Fry 1992) The advantage of calculating product-moment correlations from predicted breeding values is that they produce estimates that are within the theoretical bounds for correlations and the usual t test of the null hypothesis r, = 0 is easily applied. However, the use of Fisher's z transformation (Sokal and Rohlf 1981) to test the alternative null hypothesis for genetic data (rg= 1) has been shown to produce biased (Roff and Preziosi 1994) and highly unreliable results (Windig 1997). Fortunately, for the cross environment genetic correlations calculated from mixed model REML analyses, the test of the sire-environment interaction term is a test of the null hypothesis rg = 1, and the test of the sire effect is a test of the null hypothesis r1, = 0 (Fry 1992). An insignificant sire- environment interaction component suggests that the genetic basis of a trait is the same in each environment. Other recent studies have applied jackknife or bootstrap resampling methods to directly calculate standard errors of genetic correlations (e. g. Roff and Preziosi 1994, Windi g 1994, Reusch and Blanckenhom 1998, Phillips 1998). However, the appropriate resampling level in complex designs such as this one is unclear (Shaw 1992). The available tests are sufficient to allow full interpretation of the results. RESULTS Light environment effects All traits were significantly plastic in at least one year (germination, emergence date, winter size, survival to flowering, flowering date, specific leaf area, mainstem length, vegetative biomass, seed number, mean seed mass, and reproductive investment; 90 Figure 9a-v). The primary difference between years occurred in traits expressed early in the life cycle. The autumn of the first year was warm and dry. As a result, germination rates were more variable, and emergence was delayed. This resulted in less variation in emergence date and winter size. A more favorable germination environment in the second year resulted in a higher fraction germinating, and earlier, more variable emergence dates. Earlier emergence gave plants more time to respond to variation in resources and resulted in greater size variation in December. The light manipulations affected other aspects of the environment. There was a higher incidence of grazing on C. verna by small herbivores in the low and medium light treatments. Under winter snow cover, the temperature of the top 1 cm of soil remained constant near freezing in the natural and full sun treatments, but the lattice covers prevented snow accumulation, and temperatures in these treatments plunged to as much as -15 degrees C with wide diurnal fluctuations. In spring, the t0p 1 cm of soil in the high light treatments was warmer and had more variable temperatures. The daily maximum temperature in these plots in May 1997 often exceeded 35 degrees C compared to less than 20 degrees C in shaded plots. The soil was drier in the high light treatments. The mean weekly volumetric water content of top 5 cm of soil during October 1996 in high light was 0.12 m3 water / m3 soil compared to 0.15 m3 water / m3 soil in each of the other treatments (0.01 s.e. in all cases). In contrast to these pattems, there were no clear relationships between the light treatments and the density of naturally occurring conspecifics at harvest in either year. The range was broad (from 75 plants/m2 in high light in year two, to 879 plants/m2 in high light in year one), but most plots in each year had between 100 and 200 plants/m2 at 91 Figure 9. Paternal half-sib family reaction norms for Years 1 (50 sires and 3 light treatments), and 2 (12 sires and 5 light treatments (3 manipulated and 2 natural)). The significance values for all traits except germination and survival are based on univariate mixed mode] REML analysis. Year 1 traits were analyzed with a model that included sire, dam(sire), light, block, and all two way interactions, except the dam-by-block interaction. The model for the second year omitted all block terms. Germination and survival were analyzed by logistic analysis with a simplified model including only sire, light, and their interaction. See text for details. The bars on the right side of the figures represent one average standard error for all sires in the most variable environment. No error bars are presented on the vegetative biomass and seed number graphs because this data is plotted on a logarithmic scale. The same symbol is used for the same sire throughout all figures. The light level in the forest interior plots averaged 45% of full sun and on the edge it averaged 70% of fill] sun. 92 a. Yearl on t: c: to E E 305— 0 o c: g t‘.‘ o o. o I— O— Sire 0.01 Light <0.01 ‘ Sire x Light 0.79 . Sire 0.01 _ Light 0.36 Sire x Light 0.95 0 r l l I l l r c Year 1 (I. Year 2 11/19 Sire <0.01 Dam(Sire) 0.08 ‘ Lig1t<0.01 SireXLight 0.35 3 Dam X Light 0.29 a 11/9- 8 = - I ED 0 . E Sire 0.22 m W30“ Dam(Sire) 0.06 _ Light 0.39 Sire x Light >0.5 Dam x Light >0.5 10/20 r I e. Year 1 f. Year 2 150 - Sire 0.34 - Dam(Sire) <0.01 2 “Light <0.2 N51 0 jSireX Light >0.5 V 0 ,Damx Light >05 .8 - l (0 4 , .423 _; Sire 0.2 ’35 5° . Dam(Sire) <0.01 . Light <0.01 . SireX Light 0.34 ‘ DamX Ligrt 0.07 0 l r 1 1 T r 1 9)“ so“ 39“ so“ '0‘ 0%“ 59“ \00/0 A0 [o \QQ \%0lo A0010 cg'X‘xxc“ Ya \Qoolo ‘3‘ Light Environment Figure 9. 93 g. Yearl h. Year2 Sire 0.98 Sire X Ligrt 0.79 Sire 0.82 ‘ Light <0.01 8 Light <0.01 E” i Sire X Light 0.44 E - g -t m 0.5 4 I l S: .9. - ‘5 1 D. E . e. 0 W T T I I I I I 1. Year 1 j. Year 2 5/31 . Sire <0.01 Dam(Sire) <0.01 Ligrt <0.01 .g 5/21_ Sire X Light >05 90 Darn X Light 0.13 e l I “3’ ”" Dam(Sire) 0.22 \ ‘ \ Light <0.01 Sire X Light 0.05 Dam X Light >0.5 5/1 I I k. Year 1 1 Year 2 800 ._ H Sire . 0.34 Sire 0.19 0? Dam(Sire) 0.01 Dam(Sire) 0.15 5700 — 2\ Light . <0.01 Light <0.01 ‘5 , (K \\ SrreX Lrght 0.02 SireX Light >0.5 1.600 _ .t\\\ Darn x L1ght>0.5 Dam x Light 0.3 < -:°&\\ ‘11-. \ \\ \ \: \ I I 8 00 - . “‘ W 300 r 200 I l l I I I I I 03°“ . s s09 59“ so“ {19‘ 9%“ 9‘“ W", [to I“ \oB°/° \‘3/0 110°]; eta-“9‘6 e \09°’° Light Environment Figure 9 (cont'd). 94 111. Year 1 25 Sire 0.11 Dam(Sire) >05 A Light <0.01 5 20 ~ SireX Light 0.16 s: Dam x Light 045 g 15 — . I M 10 - O 5 2° 5‘ . 3 Ser 0.37 Dam(Sire) <0.01 Light 0.15 Sire x Light >05 Dam x Light >05 0 V I I I I I I I 0. Year 1 p. Year 2 50° . Sire 0.34 Dam(Sire) >05 . Ligit <0.01 E” _ SireXLight 0.17 g 100 -: .9. - m I .“>_’ - E _ 80 —J . . . g SIre 0.42 Dam(Sire) 0.04 Light 0.19 Sire X Light >05 Dam x Light 0.22 10 r I I I I q. Year 1 r. Year 2 4° - Sire 0.31 Dam(Sire) 0.06 ,1, — Light <0.01 ’ SireX Light 0-18 ;s:¢s(l/lé-\ ‘ ‘— ' . /">-va"“é-./.’/l-2S=——To 10 : DamX L1ght>0.5 j... #8., ,3”? ~_.- ‘- : ,ZOWf-E: ». w ‘ «fix-2:22;? ‘5; - ,/¢/;r’/,/ /-- V3 ‘ f/fr'j't'éz?!’ // ’ l 1 ’ I Sire 0.32 Dam(Sire) 0.01 Light <0.04 ‘ Sire x Light >05 Darn x Light 0.13 0-2 I I I I I fl I I 0“ ho he 0“ 05‘ or or,“ 9“ \0°°$ It,o°/°s @0905 \%°/°$ 1.85% ‘esk\“‘e{\ «a (000% 0 Light Environment Figure 9 (cont'd). 95 5. Year 1 t. Year 2 7 6 .— GB :5. 5 - g; 4_ go: 2 “'0' l I B 3 - g o 0.3 g 2 -I 0.06 g Light 0.08 1 - Sire X Light 0.35 Sire >0.5 Dam(Sire) <0.01 Light <0.01 Darn X Light 0.05 Sire X Light >0.5 Dam X Ligrt 0.43 0 I I I I I I I r A u. Year 1 v. Year 2 E; 0'4 Sire >05 Dam(Sire) 0.0; :0 Isighii L gh 3(0); ‘ Vt ire i t . I “ 30-3‘DamXLi t022 ,. ’ §n i, g" nos‘Le-e/ 4%- r: /’7»'—'-—-‘§7’ O.5 Darn X Light >05 "8’ 0 I I I I I E I I T 0. 0° 0 0° 0“ - i 0 00 g \00/0 S A00/0 $® 000/0 8“ \%0/0 S AQOIO S ‘\‘\‘et\0 gag 000/0 S \ ‘65 \ $0 Light Environment Figure 9 (cont'd). 96 harvest. Densities were reduced in the second year most likely because all plants in the first year were harvested prior to seed dispersal. Consequently, most of the plants in year two were recruited from the seed bank. Additive genetic effects and narrow sense heritabilities Overall, there were significant narrow sense heritabilities in at least one environment for six of nine traits (Table 5). Surprisingly, the reduced number of sires in year two does not appear to have limited the ability of these analyses to detect significant genotypic effects and heritabilities. Paternal families differed significantly in the proportion of seeds that germinated in both years (Figures 9a-b). There were also significant sire effects on emergence date, flowering date, and reproductive investment in at least one year (Figures 9c-d, i-j, u-v), and these traits were frequently heritable within environments (Table 5). Although there were no significant sire effects on specific leaf area, mainstem length, and mean seed mass (Figures 9k-n, s-t), all were significantly heritable in at least one environment (Table 5). Significant within environment heritabilities in the absence of significant overall sire effects are evidence for genotype- environment interactions (Feldman and Lewontin 1975). Winter size, survival to flowering, vegetative biomass, and seed number had no significant sire effects (Figures 9e-h, o-r), and were not significantly heritable in any environment (Table 5). This result is expected of traits closely linked to fitness because they are likely under strong selection. There were more non-zero and significant heritability estimates in both years in the higher light environments (full sun and natural edge, Table 5). In particular, flowering date, specific leaf area, and mean seed mass were most heritable in high light. 97 98 SONG eNNe no: Ed aNNv SN: 83 Rd :3 o N see 26 53 ..NN.o 39 N8 : 255355.28: :ONV 3.0 EN: 0 aNNv :8 8e 9 :3 o N 8...: :.o 53 o as o _ 3:2 B8 :82 35 mod ENC 9 SN: o AS: o fie: o N so: 2 3 see 88.0 an o _ 888 we: CNN: 0 GE o ANeNv o No: o as o N $3 :3 2:: 86 AN: o _ 82:55 :3 CNN: .33 ENG 82.0 ANN: o :8: o as o N SN: 5o 8: 83 68 o _ 583 e882 8: o ANN: 0 an: 83 E: o an o N sec tNNd 55 :9: :8 o L 82:8: 206on EN: 98 Go: #3 5N: have 83 o 2.8 S N €N$ :35 not 83 at eNd _ one: 85332: ANN: NS 8va o 2va mod 8N: 3o :38 2.0 N 58 o 32 : mod 38: o _ 98 L253 ES .283 8N: :3 :3 o :8 *3 65 38 N 6:: o 82: 0 $2: 0 _ one: ooeowLoEm owe: 82:238.: 58 .82: gm .8... 58 .52 so.» :5 Ever—832m A88 2:8me €398: 8:813:th .AmnofioE 08v 8:250:35 888 98 $3: 5220: 8:28—0:00 9 0:: 82:58:23 8: 0.8 88 032:8 :8 292 8:85:38 mo 8:289:00 25:09:50 85:? ocofim 03:28 2.: .«o 82 0:8 :02on: :o :83 2: 82m.» 8:85:3m 8:8: on :8 88 Q A368: m 30> 8:8: m2 :8 88 cm ”693:: g 80> .888 2:88 03 885:2“: E 838:: 2a. .8833: 62583:» 555 6852:: :_ 9:2 .3 :o 58:39 AEEmQPZ 36: 88880 3:08:88 8:88 28:8 02.68 on. :o :83 8:8: 95:58:: we 8:23:23: 8:8 3052 .n 2:5. Aodvmi. .mcdvm... .movmn .EdVKw Emergence date was more heritable in the second year when this trait was more phenotypically variable. Only reproductive investment was consistently heritable across all environments in each year. Dam effects Significant dam effects on a trait in the absence of significant additive genetic variation (sire effects) may be attributable to dominance, maternal genotype, or maternal environment. There were highly significant dam effects on winter size in both years (Figures 9e-t). Interestingly, in the second year there were significant dam effects on all late life-cycle traits except specific leaf area (flowering date, mainstem length, vegetative biomass, seed number, mean seed mass, and reproductive investment, Figures 9i-v). Genotype-en vironment interactions Statistical tests provide evidence for significant genotype-environment interactions in the first year for flowering date (Figure 9i), specific leaf area (Figure 9k) and reproductive investment (Figure 9n). As mentioned in the previous section, the contrasting results of significant within enviromnent heritability estimates (Table 5) but insignificant sire effects for specific leaf area (Figures 9k-l), mainstem length (Figures 9m-n), and mean seed mass (Figures 9s-t) also suggest genotype-environment interactions. In the second year, no statistical tests for additive genetic variation in reaction norms approached significance. The reduced number of sires in this year limits the power of these tests. The reaction norm plots show considerable diversity in reaction norm shape among sires, indicative of genotype-environment interactions and genetic variation for plasticity (Figure 9). Reaction norms cross between environments for all traits, but often 99 it is just a few genotypes that are responsible for most of the diversity in their shape. This diversity is particularly important for evolution, but is not likely to be detected in statistical tests of genotype-environment interaction (Lewontin 1974). Although mortality in some enviromnents was quite high, there was no evidence of differential mortality or cross environment tradeoffs among paternal families in either year (Figures 9g-h). Genetic correlations across environments The cross-environment genetic correlations are generally consistent with the genotype-environment interaction results. In year 1, very few cross-environment genetic correlations were significantly different from zero by either method, indicating considerable cross-environment independence of traits (Table 6). Correlations for flowering date and specific leaf area across the medium and high light environments were significantly greater than zero by the breeding value method and significantly less than one by the variance component method. In year 2 most genetic correlations were significantly greater than zero by both methods, and none were significantly less than one by the variance component method (Table 7). All cross-environment correlations in year 2 were larger in magnitude than their comparable values in year 1. These results suggest that in contrast to year 1, the genetic basis of traits in year 2 was very similar across environments. There were no significant negative genetic correlations in either year that would indicate the presence of cross environment tradeoffs or genetic constraints on plasticity evolution. 100 Table 6. Cross environment additive genetic correlations. Data from 50 sires from Year 1 (1995-96). Methods: bv-from predicted breeding values. o’,,,,-from variance components. Empty cells occur when there was no additive genetic variance for the trait in at least one of the environments. Significance testing for the bv method done with individual t tests. Significance testing for the variance component method is based on the results of mixed model REML analyses. Because estimated variance components were not greater than zero in at least two environments, cross-environment genetic correlations could not be estimated for emergence date, winter size, and mean seed mass in Year 1, and specific leaf area, vegetative biomass, and seed number in Year 2. Significance values: bv method- Ho: rg=Oz $Pm~ mmwz UDOW 2...... nembd N5 888.32 is... to... 1.8.. ton. ionN IQNN 2...... :2... .30... imp... .13.... .9... m... a. 2.... a. n. NN.. .N... a... a... N... 2...... *3... 3m... .6... 3.8... :E... 33...... a. samba; :Nm... .18.. :5... is... :2... 3N2 2...... *8... *8... is... 3.... am... :2... a. 23 .m swam cm... 538.). a... a... 6282 33 gm... aw... 5282 85...: :2... -620... -owcm -omem -owcm -320". -fifiom -380... -8232 -33 -33 o 2an E 2. 2.9.5 A863: N 30> 88.. 8.3 N. 8.. 8...... So... mecca—oboe ozocow 02:2... #882835 .380 .N. 038 WEE 2.25.235 3-83 .03. So.» 102 DISCUSSION This study of a natural population of Collinsia verna has found a surprising diversity of genetic effects expressed in diverse field environments. There was additive genetic variation in at least some environments for germination, timing of emergence, flowering date, specific leaf area, mainstem length, mean seed mass, and reproductive investment. There was strong evidence for genotype-environment interactions (genetic variation for plasticity) for flowering date, specific leaf area, mainstem length, and reproductive investment. The lack of evidence for genotype-environment interactions in the fitness components survival, vegetative biomass, and seed number suggest that there were no strong light environment specialists among the genotypes sampled. However, significant maternal effects on vegetative biomass, seed number, and seed mass in the absence of additive genetic variation suggest that maternal genotypes specialize for different reproductive strategies. Patterns of genetic variation Heritability-The magnitudes of the heritabilities estimated in this study are low but compare favorably with other narrow sense, field-based heritability estimates for non- floral traits (e. g. Mitchell-Olds 1986, Bennington and McGraw 1996, Campbell 1997a, Thiede 1998). As in numerous other studies, traits closely linked to fitness (vegetative biomass, seed number) display little heritability. These traits are under strong selection (see Chapter 4) Two traits in this study, emergence date and winter size, were previously studied in this population (Thiede 1998). The significant heritability estimates for emergence date found in all but the high light environment (Table 5) are similar to previous estimates for this trait in the greenhouse (I12 = 0.14) and field (112 = 0.25) (Thiede 103 1998). Winter size was most strongly influenced by maternal effects in this study and that of Thiede (1998), but she also found a significant narrow sense heritability in the greenhouse (h2 = 0.27). In this study the narrow sense heritability of winter size was non- zero in all but the forest interior in at least one year, but was never significant. Maintenance of genetic variation-In Chapter 4, I found that patterns of survival and selection on emergence date depend on the presence or absence of leaf litter. Survival to flowering is generally higher in the absence of leaf litter. Since germinating seedlings are unable to predict if they will be trapped under a fallen leaf, this variable selection could maintain the genetic variation for dormancy found in this study. The absence of leaf litter also selects for early emergence, while the presence of litter selects :for late emergence. The unpredictability of these litter effects again may maintain genetic variation in emergence date. Environmental ejfects on the expression of genetic variation-There are many ideas about how the expression of genetic variation might change along resource gradients or in novel or stressful environments. For example, novel or stressful environments have been predicted to increase genetic variance (Holloway et al. 1990). Alternatively, if stressful conditions are common and result in strong selection, genetic variation would be lost faster. Similarly, relaxed selection in benign and high resource environments may allow more genetic variation to persist. Another possibility is that genetic variation may be lowest in the most common natural environment where selection occurs most fi'equently. All these patterns have been found in nature. Novel or stressful environments have been shown to increase heritability for many animal traits, but no patterns are seen in data from plants (review in Hoffrnann and Parsons 1991). More recent work in plants 104 has found a decrease in genetic variance under stressful conditions (Sultan and Bazzaz 1993b, Bemrington and McGraw 1996, J. Conner unpublished data). Increased genetic variance under conditions of resource abundance has been demonstrated for other plant species across resource gradients (Clough et al. 1980, Schwaegerle and Bazzaz 1987), but the patterns are not always consistent, even within a species. For example, Polygonum genotypes were found to display increased genetic variance along a soil moisture gradient (Sultan and Bazzaz 1993b), but not with respect to a light gradient (Sultan and Bazzaz, 1993a). In this study both the full sun and the low light environments are novel, and consequently may be stressful. Natural variation in light availability in this population ranges from 25% to 7 5% of full sun (Chapter 2). The results show that heritability estimates along a light gradient vary, depending on the trait. Heritability estimates for flowering date, specific leaf area, and mean seed mass were high or highest in full sun, while estimates for emergence date and reproductive investment were highest in low light. Winter size and mainstem length were most heritable in the intermediate or natural light environments. Overall, trait heritability was lowest in the low resource environments (Table 5: 10% sun, 40% sun, and forest interior). It is likely that genetic variation is absent in low light simply because all genotypes performed poorly. Maternal effects In a half-sib design, the maternal variance component a fraction of the additive genetic variance plus a portion of dominance variation and maternal environmental and genetic effects (Falconer and Mackay 1996). In this study, winter size and mean seed mass in both years, and length of mainstem, vegetative biomass, and seed number in the 105 second year had no additive genetic variance, but substantial among darn variance. It is quite possible that these effects indicate genetic differences among mothers in their effects on offspring phenotype (rather than dominance or maternal environment effects). First, the relatively uniform greenhouse conditions under which the dams produced seeds should minimize maternal environment effects. Second, other studies of this population have found substantial maternal genetic effects on seed mass and winter size (Chapter 2, Thiede 1998). Finally, there is no evidence for dominance variation in seed mass or winter size in this population (Thiede 1998). The persistence of maternal effects in the absence of additive genetic variation late in the life cycle in the second year is a striking result. The significant differences among maternal families for seed number in the second year are clear evidence that selection differentiated among maternal families. Further, the marginal dam-by-light interaction term (Figure 9r) raises the possibility that different maternal genotypes may be favored in different environments, and suggests that maternal plants specialize for different seed provisioning strategies. A greenhouse study of this Collinsia population also found that maternal effects persisted up until flowering for two size related traits (Thiede 1998). In contrast, most other studies have found that maternal effects decline through the life cycle (e. g. Biere 1991, Montalvo and Shaw 1994, Schmid and Dolt 1 994). The between year differences in dam effects late in the life cycle may be due to differences between years in the timing of emergence. Although there are strong darn effects on winter size in both years, late emergence in the first year resulted in little Phenotypic variation in winter size (Figures 9e-f ). Without a strongly established size 106 hierarchy in the fall, over winter mortality and spring growth could not differentiate among maternal families. Consequently, maternal effects could not persist into later life- history stages. Earlier emergence in the second year resulted in much more variation in overwinter size (Figures 9e-f), and possibly more variation among maternal families in survival and spring growth (Figures 9m-v). Several other studies have found that the expression of maternal effects may depend on the offspring environment (Stratton 1989, Schmitt et a1. 1992, Schmid and Dolt 1994, Thiede 1998). Genetic variation for plasticity The integration of plastic traits in coordinated responses to environmental heterogeneity remains one of the most complex and poorly understood aspects of the general phenomenon of phenotypic plasticity (Schlichting and Pigliucci 1998). The genetic and environmental effects on the suite of traits studied here were diverse and complex. Consequently, generalizations about the effects of light gradients on the expression of genetic variation, or plasticity evolution would be foolhardy. However, the evidence found here for genetic variation for plasticity in several traits sheds light on several issues. Antagonistic pleiotropy-Mainstem length and reproductive investment are under consistent directional selection (Chapter 4). Moreover, estimates of genetic correlations found strong and significant positive genetic correlations between mainstem length and vegetative biomass (Chapter 5, Table 15), and vegetative biomass is also under very strong directional selection (Chapter 4). Under these circumstances, little genetic variation would be expected to remain for these traits or their plasticity. Yet there is strong evidence for genetic variation for plasticity for reproductive investment, and 107 modest evidence for mainstem length. A possible explanation for this pattern is that there are fundamental genetic tradeoffs between the ability to efficiently utilize resources and convert them to seeds (the reproductive investment trait), and the ability to rapidly acquire resources and convert them to biomass (the mainstem length and vegetative biomass traits). If there are negative genetic correlations between reproductive investment and the size traits, selection for greater Opportunistic growth ability and biomass accumulation may also select for reduced efficiency in reproductive investment and vice versa. This antagonistic pleiotropy may maintain genetic variation. Genetic correlations of reproductive investment with mainstem length and vegetative biomass were indeed negative, but they were insignificant (Chapter 5, Table 15). Genotype-environment interactions for fitness-There is a simpler explanation for genetic variation in plasticity of reproductive investment. The strong selection on this trait across all environments simply favors different genotypes in different environments. Consequently, genotype-environment interactions for fitness maintain genetic variation for plasticity in this trait. Control of plasticity via a genetic switch-Some have suggested that plants should be generalists for resource utilization regardless of the pattern of genetic correlations between underlying traits (Chapin 1991, Chapin et al. 1993). Ideally, plants would be able to both tolerate stressful low resource conditions and grow rapidly when resources are abundant. Interspecific comparisons show that compared to sun plants, shade tolerant plants have low relative growth rates, low photosynthetic rates, low transpiration and stomatal conductance, low leaf turnover, and high ability to use sun flecks (e. g. Grime 1979, Chapin 1980, Chapin et al. 1993). Interestingly, when high light plants are grown 108 under shady conditions, they show many of these shade plant characteristics, which may represent an adaptive "stress resistance syndrome" (Chapin 1991, Chapin et a1. 1993). Chapin argues that conversion of a high resource genotype to a stress tolerant one may involve a simple genetic switch. Hormonal regulators of plant growth, development, and stress responses are known which meet the criteria of a genetic switch (V oesenek and Blom 1996, Schlichting and Pigliucci 1998). If plastic responses are under single gene control, quantitative genetics is an inappropriate model for understanding the evolution of plasticity in these traits. It should be relatively easy for adaptively plastic responses to resource limitation to evolve regardless of the genetic correlations among traits, and consequently, we would expect most genotypes to be generalists with respect to light. It is notable that there was no evidence in this study for differences among paternal families in survival or fecundity. At this level, all genotypes appeared to be generalists. Adaptive plasticity relaxes natural selection ?-F lowering date and specific leaf area also show strong evidence for genetic variation for plasticity. But in contrast to most other traits, they are under little direct selection (Chapter 4). In this case, the absence of selection may allow genetic variation to persist. Intriguingly, adaptive plasticity may also be responsible for the absence of selection: By being plastic, all genotypes produce phenotypes appropriate for their environment. Conclusion The results of this study of Collinsia verna suggest that the patterns of genetic and enviromnental effects on each trait are unique. Genotypes appear to specialize for particular patterns of germination, emergence, and reproductive investment, while at the same time they may be adaptively plastic for flowering date and specific leaf area. 109 Persistent maternal effects suggest that maternal genotypes may specialize for different reproductive strategies. 110 LITERATURE CITED Andersson, S., and R. G. Shaw. 1994. Phenotypic plasticity in Crepis tectorum (Asteraceae): genetic correlations across light regimens. Heredity 72:113-125. Baskin, J. M., and C. C. Baskin. 1983. Germination ecology of Collinsia verna, a winter annual of rich deciduous woodlands. Bulletin of the Torrey Botanical Club 110:311-315. Bennington, C. C., and J. B. McGraw. 1996. Enviromnent-dependence of quantitative genetic parameters in Impatiens pallida. Evolution 50: 1083-1 097. Biere, A. 1991. Parental effects in Lychnisflos-cuculi. I. Seed size, germination and seedling performance in a controlled environment. Journal of Evolutionary Biology 32447-465. Boldman, K. G., L. A. Kriese, L. D. Van Vleck, C. P. Van Tassel] and S. D. Kachman. 1995. A manual for use of MTDFREML. a set of programs to obtain estimates of variances and covariances [DRAFT]. U. S. Department of Agriculture, Agricultural Research Service. Software available at http://chuck.agsci.colostate.edu/cvantass/mtdfreml.html. Bradshaw, AD. 1965. Evolutionary significance of phenotypic plasticity in plants. Advances in Genetics 13:1 15-155. Cameron, N. D. 1993. Methodologies for estimation of genotype with environment interaction. Livestock Production Science 35:237-249. Campbell, D. R. 1996. Evolution of floral traits in a hennaphroditic plant: Field measurements of heritabilities and genetic correlations. Evolution 50:1442-1453. Campbell, D. R. 1997a. Genetic and environmental variation in life-history traits for a monocarpic perennial: A decade-long field experiment. Evolution 51:373-382. Campbell, D. R. 1997b. Genetic correlation between biomass allocation to male and female functions in a natural population of Ipomopsis aggregata. Heredity 79:606-614. Chapin, F. 8., HI. 1980. The mineral nutrition of wild plants. Annual Review of Ecology and Systematics 11:233-260. Chapin, F. S., III. 1991. Integrated responses of plants to stress. BioScience 41 :29-36. 111 Chapin, F. S., 111, K. Autumn, and F. Pugnaire. 1993. Evolution of suites of traits in response to environmental stress. American Naturalist 142 S78-S92. Clough, J. M., R. S. Alberte, and J. A. Teeri. 1980. Photosynthetic adaptation of Solarium dulcamara L. to sun and shade environments. III. Characterization of genotypes with differing photosynthetic performance. Oecologia 44: 221-225. de J ong, G. 1990. Quantitative genetics of reaction norms. Journal of Evolutionary Biology 32447-468. de Jong, G. 1995. Phenotypic plasticity as a product of selection in a variable environment. American Naturalist 145:493-512. DeWitt, T. J., A. Sih, and D. S. Wilson. 1998. Costs and limits of phenotypic plasticity. Trends rn Ecology and Evolution 13: 77- 81. Donohue, K,. and J. Schmitt. 1998. Matemal environmental effects in plants: adaptive plasticity? Pages 137-158 in T. A. Mousseau and C. W. Fox, editors. Maternal Effects as Adaptations. Oxford University Press, New York, New York, USA. Dudley, S. A., and J. Schmitt. 1995. Genetic differentiation in morphological responses to simulated foliage shade between populations of Impatiens capensis from open and woodland sites. Functional Ecology 9:655-666. Dutilleul, P., and Y. Carriere. 1998. Among-environment heteroscedasticity and the estimation and testing of genetic correlation. Heredity 80:403-413. Dutilleul, P., and C Potvin. 1995. Among-environment heteroscedasticity and genetic autocorrelation: Implications for the study of phenotypic plasticity. Genetics 139:1815-1829. Falconer, D. S. 1952. The problem of environment and selection. American Naturalist 86:293-298. Falconer, D. S., and T. F. C. MacKay.1996. Introduction to quantitative genetics. Fourth edition. Addison Wesley Longman Limited, Essex, UK. F eldman, M. W., and R. C. Lewontin. 1975. The heritability hang-up. Science 19021163- 1 168. Fry, J. D. 1992. The mixed-model analysis of variance applied to quantitative genetics: Biological meaning of the parameters. Evolution 46:540-550. Gabriel, W., and M. J. Lynch. 1992. The selective advantage of reaction norms for environmental tolerance. Journal of Evolutionary Biology 5:41-59. 112 Getty, T. 1996. The maintainence of phenotypic plasticity as a signal detection problem. American Naturalist 148:378-385. Gillespie, J. H., and M. Turelli. 1989. Genotype-environment interactions and the maintenance of polygenic variation. Genetics. 121:129-138. Goldberg, D. E. 1990. Components of resource competition in plant communities. Pages 27-50 in J. Grace and D. Tilrnan, editors. Perspectives on plant competition. Academic Press, San Diego, CA. Grime, J. P. 1979. Plant strategies and vegetation processes. John Wiley, New York. Hedrick, P. W. 1986. Genetic polymorphism in heterogeneous environments: A decade later. Annual Review of Ecology and Systematics 172535-566. Hedrick, P. W., M. E. Ginevan, and E. P. Ewing. 1976. Genetic polymorphism in heterogeneous environments. Annual Review of Ecology and Systematics 7:1-32. Hoffrnann, A. A., and P. A. Parsons. 1991. Evolutionary genetics and enviromnental stress. Oxford University Press, Oxford. Holloway, G. J ., S. R. Povey, and R. M. Sibly. 1990. The effect of new environment on adapted genetic architecture. Heredity 64:323—330. Kalisz, S. 1986. Variable selection on the timing of emergence in Collinsia verna (Scrophulariaceae). Evolution 40:479-491. Kelly, C. A. 1992. Spatial and temporal variation in selection on correlated life-history traits and plant size in Chamaecristafasiculata. Evolution 46: 1658-1673. Lewontin, R. C. 1974. The analysis of variance and the analysis of causes. American Jom'nal of Human Genetics 26:400-411. Littell, R. C., G. A. Milliken, W. W. Stroup, and R. D. Wolfinger. 1996. SAS system for mixed models. SAS Institute Inc., Cary, NC. Lynch, M., and W. Gabriel. 1987. Environmental tolerance. American Naturalist 129:283-303. Lynch, M., and B. Walsh. 1998. Genetics and Analysis of Quantitative Traits. Sinauer Associates Inc., Sunderland, MA. Manly, B. F. J. 1991. Randomization and monte carlo methods in biology. Chapman and Hall, London. 113 Mathur, P. K., and P. Horst. 1994. Methods for evaluating genotype-environment interactions illustrated by laying hens. Journal of Animal Breeding and Genetics 111:265-288. Mitchell-Olds, T. 1986. Quantitative genetics of survival and growth in Impatiens capensis. Evolution 40: 107-1 16. Mitchell-Olds, T. 1992. Does environmental variation maintain genetic variation? A question of scale. Trends in Ecology and Evolution 7:397-398. Mitchell-Olds, T., and J. Bergelson. 1990. Statistical genetics of an annual plant, Impatiens capensis. I. Genetic basis of quantitative variation. Genetics 124:407- 415. Mitchell-Olds, T., and J. J. Rutledge. 1986. Quantitative genetics in natural plant populations: a review of the theory. American Naturalist 127:379-402. Montalvo, A. M., and R. G. Shaw. 1994. Quantitative genetics of sequential life-history and juvenile traits in the partially selfing perennial, Aquilegia caerulea. Evolution 48:828-841. Moran, N. A. 1992. The evolutionary maintenance of alternate phenotypes. American Naturalist 139:971-989. Notter, D. R., and C. Diaz. 1993. Use of covariances between predicted breeding values to assess the genetic correlation between expressions of a trait in 2 environments. Genetics Selection Evolution 25:353-372. Phillips, P. C. 1998. Genetic constraints at the metamorphic boundary: Morphological development in the wood frog, Rana sylvatica. Journal of Evolutionary Biology 11:453-463. Pigliucci, M. 1996. How organisms respond to environmental changes: from phenotypes to molecules (and vice versa). Trends in Ecology and Evolution 112168-173. Pigliucci, M., J. Whitton, and C. D. Schlichting. 1995. Reaction norms of Arabidopsis. I. Plasticity of characters and correlations across water, nutrient and light gradients. Journal of Evolutionary Biology 82421-438. Reusch, T. and W. U. Blanckenhom. 1998. Quantitative genetics of the dung fly Sepsis cynipsea: Cheverud's conjecture revisited. Heredity 81:111-119. Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution 43:223-225. Roach, D. A., and R. D. Wulff. 1987. Maternal effects in plants. Annual Review of 114 Ecology and Systematics 18:209-235. Roff, D. A. 1997. Evolutionary quantitative genetics. Chapman and Hall, New York. Roff, D. A., and R. Preziosi. 1994. The estimation of the genetic correlation: the use of the jackkrrife. Heredity 73:544-548. SAS Institute Inc. 1997. SAS/STAT software: changes and enhancements through release 6.12. SAS Institute, Cary, NC. SAS Institute Inc. 1992. SAS/STAT software: changes and enhancements, release 6.07. SAS Technical Report P-229. SAS Institute, Cary, NC. SAS Institute Inc. 1989. SAS/STAT User’s guide, Version 6, Fourth Edition, Volume 2, SAS Institute Inc., Cary, NC. Scheiner, S. M. 1993. Genetics and evolution of phenotypic plasticity. Annual Review of Ecology and Systematics 24:3 5-68. Scheiner, S. M. 1998. The genetics of phenotypic plasticity. VII. Evolution in a spatially- structured environment. Journal of Evolutionary Biology 11:303-320. Schlichting, C. D., and M. Pigliucci.l995. Gene regulation, quantitative genetics and the evolution of reaction norms. Evolutionary Ecology 9: 154-168. Schlichting, C. D., and M. Pigliucci. l 998. Phenotypic evolution: a reaction norm perspective. Sinauer Associates, Inc., Sunderland, MA. Schmid, B., and C. Dolt. 1994. Effects of maternal and paternal enviromnent and genotype on offspring phenotype in Solidago altissima L. Evolution 48: 1525- 1 549. Schmitt, J. 1993. Reaction norms of morphological and life-history traits to light availability in Impatiens capensis. Evolution 47: 1654-1668 Schmitt, J. 1995. Genotype-environment interaction, parental effects, and the evolution of plant reproductive traits. in P. C. Hoch and A. G. Stephenson, editors. Experimental and molecular approaches to plant biosystematics. Monographs in Systematic Botany from the Missouri Botanical Garden. 53:199-214. Schmitt, J ., J. Niles, and R. D. Wulff. 1992. Norms of reaction of seed traits to maternal environments in Plantago lanceolata. American Naturalist 139:451-466. Schoen, D. J ., G. Bell, and M. J. Lechowicz. 1994. The ecology and genetics of fitness in forest plants 4. Quantitative genetics of fitness components in Impatiens pallida 115 (Balsarninaceae). American Journal of Botany 81:232-239. Schwaegerle, K. E., and F. A. Bazzaz. 1987. Differentiation among nine populations of Phlox: response to environmental gradients. Ecology 68:54-64. Schwaegerle, K. E., and D. A. Levin. 1991. Quantitative genetics of fitness traits in a wild population of phlox. Evolution 45:169-177. Searle, S. R., G. Casella, and C. E. McCulloch. 1992. Variance components. John Wiley and Sons, Inc., New York, New York, USA. Shaw, R. G. 1992. Comparison of quantitative genetic parameters: reply to Cowley and Atchley. Evolution 46: 1967-1969. Shaw, R. G., G. A. J. Platenkamp, F. H. Shaw, and R. H. Podolsky. 1995. Quantitative genetics of response to competitors in Nemophila menziessi: A field experiment. Genetics 139:397-406. Simons, A. M., and D. A. Roff. 1996. The effect of a variable environment on the genetic correlation structure in a field cricket. Evolution 50:267-275. Sokal, R. R., and F. J. Rohlf. 1981. Biometry, second edition. W. H. Freeman and Company, New York. Stratton, D. J. 1989. Competition prolongs the expression of maternal effects in seedlings of Erigeron annuus (Asteraceae). American Journal of Botany 76:1646-1653. Stratton, D. J. 1992. Life-cycle components of selection in Erigeron annuus: I. Phenotypic selection. Evolution 46292-106. Stratton, D. J. 1995. Spatial scale of variation in fitness of Erigeron annuus. American Naturalist 146:608-624. Sultan, S. E. 1995. Phenotypic plasticity and plant adaptation. Acta Botanica Neerlandica 44:363-383. Sultan, S. E., and F. A. Bazzaz. 1993a. Phenotypic plasticity in Polygonum persicaria. I. Diversity and uniformity in genotypic response to light. Evolution 47:1009-1031. Sultan, S. E., and F. A. Bazzaz. 1993b. Phenotypic plasticity in Polygonum persicaria. II. Norms of reaction to soil moisture and the maintenance of genetic diversity. Evolution. 47: 1032-1049. Sultan S. E., and F. A. Bazzaz. 1993c. Phenotypic plasticity in Polygonum persicaria. III. The evolution of ecological breadth for nutrient environment. Evolution 4711009- 116 1031. Thiede, D. A. 1998. Maternal inheritance and its effect on adaptive evolution: A quantitative genetic analysis of maternal effects in a natural plant population. Evolution 52:998-1015. Thompson, J. D. 1991. Phenotypic plasticity as a component of evolutionary change. Trends in Ecology and Evolution 62246-249. Travis, J. 1994. Evaluating the adaptive role of morphological plasticity. Pages 99-122 in P. C. Wainwright and S. Reilly, editors. Ecomorphology: integrative organismal biology. University of Chicago Press, Chicago. van Tienderen, P. H. 1990. Morphological variation in Plantago lanceolata: Limits of plasticity. Evolutionary Trends in Plants. 4:35-43. van Tienderen, P. H. 1991. Evolution of generalists and specialists in spatially heterogeneous enviromnents. Evolution 452 13 17-1331 . van Tienderen, P. H. 1997. Generalists, specialists and the evolution of phenotypic plasticity in sympatric populations of distinct species. Evolution. 51 :1372-1380. Via, S. 1984. The quantitative genetics of polyphagy in an insect herbivore. II. genetic correlations in larval performance within and among host plants. Evolution , 38:896-905. Via, S. 1987. Genetic constraints on the evolution of phenotypic plasticity. pp. 47-71 In: V. Loeschcke ed. Genetic constraints on adaptive evolution. Springer-Verlag, Berlin. Via, S., and R. Lande. 1985. Genotype-environment interaction and the evolution of phenotypic plasticity. Evolution 39:505-522. Voesenek, L. A. C. J ., and C. W. P. M. Blom. 1996. Plants and hormones: an ecophysiological view on timing and plasticity. Journal of Ecology 84:111-119. Wahlsten, D. 1990. Insensitivity of the analysis of variance to heredity-environment interaction. Behavioral and Brain Sciences 13:109-161. Windig, J. J. 1994. Genetic correlations and reaction norms in wing pattern of the tropical butterfly Bicyclus anynana. Heredity 73:459-470. Windig, J. J. 1997. The calculation and significance testing of genetic correlations across environments. Journal of Evolutionary Biology 10:853-874. 117 Wulff, R. D. 1998. Intraspecific variability in the response to light quality in Crotalaria incana and Impatiens sultanii. Canadian Journal of Botany 76:699-703. Yamada, Y. 1962. Genotype by environment interaction and genetic correlation of the same trait under different environments. Japanese Journal of Genetics 37:498-509. 118 Chapter 4 EVOLUTION OF REACTION NORMS: ENVIRONMENT-DEPENDENT SELECTION IN A NATURAL POPULATION OF COLLINSIA VERNA INTRODUCTION Many recent studies have documented spatial and temporal variation in natural selection within natural plant populations (e. g. Kalisz 1986, Kelly 1992, Stratton 1992, 1995, Gross et al. 1998). Depending on a variety of ecological and genetic factors, theory suggests that the variable selection seen in these studies may lead to phenotypically plastic generalist genotypes or to the persistence of specialist genotypes and the maintenance of genetic variation (reviews in Hedrick 1986, Scheiner 1993, Roff 1997, Schlichting and Pigliucci 1998). To date, we have little ability to predict which of these alternatives might prevail in particular populations. Moreover, a particular selective regime may simultaneously result in plasticity in one suite of traits, and the maintenance of genetic variation in another. A better understanding of the evolutionary consequences of variable selection requires detailed studies of how variation in environmental factors affect phenotypic traits, plant fitness, and the expression of genetic variation in natural populations. In particular, manipulation of potential selective agents is necessary to identify which environmental factors are the causes of natural selection (Mitchell-Olds and Shaw 1987, Wade and Kalisz 1990, Dudley and Schmitt 1996). Once a cause is identified, information about the magnitude and predictability of variation in the selective agent is necessary to predict whether this environmental variation would favor the evolution of 119 adaptive plasticity or the maintenance of genetic variation in particular traits (e. g. Bradshaw 1965, Via and Lande 1985). Effects of light and leaf litter Two factors that can be expected to function as strong selective agents on plants are light and leaf litter. Light is a primary plant resource, and has been shown to have a strong effect on female fitness in many plant populations (reviews in Goldberg 1990, Sultan and Bazzaz 1993). Light availability for understory plants in forests can be highly variable in time and space (Chapter 2), but in very predictable spatial patterns, and diurnal and seasonal cycles. Thus it is likely that this variation in light availability should favor plastic traits that would allow plants to tolerate low light levels and convert high light availability into increased fitness. If this selective regime has allowed the plasticity of light responsive traits to evolve close to their Optima, we would expect to see relatively little direct selection on these traits in all but the most extreme light environments. Leaf litter also can have a major impact on plant fitness, particularly through reductions in the establishment and survival of seedlings (e. g. Goldberg and Werner 1983, Bergelson 1990, Carson and Peterson 1990, Foster and Gross 1997). Litter inhibits establishment through physical interference with the growth of emerging seedlings and the reduction of light quantity and quality below the compensation point (F acelli and Pickett 1991b, Foster and Gross 1998). Indirect negative effects of leaf litter on plant fitness may include promotion of fungal pathogens, and creation of habitat for litter dwelling seed predators and herbivores (Facelli 1994). However, in some habitats leaf litter may facilitate seedling establishment through the amelioration of abiotic stresses (e.g. desiccation, Fowler 1986, Willrns et al. 1986, Hamrick and Lee 1987). An 120 important abiotic stress affecting establishment of winter annuals in temperate deciduous forests may be the diurnal freeze-thaw cycle that occurs at the soil surface during winter months. Leaf litter may reduce the amplitude and frequency of these thermal cycles (Facelli and Pickett 1991a). Thus, the effects of litter may be negative at germination, but positive at later stages of growth. Importantly, the incidence, quantity, and persistence of leaf litter at particular locations in forests are unpredictable (F rankland et al. 1963, Sydes and Grime 1981, Facelli and Carson 1991, Molofsky and Augspurger 1992). Direct and in direct selection Since the development of multiple regression methods for the study of phenotypic selection (Lande and Arnold 1983, Arnold and Wade 1984a, b), selection gradients have received the most attention in the evolutionary ecology literature because they measure only direct selection on a trait independent of variation in other traits, and they are easily interpretable in terms of quantitative genetic equations for multivariate evolution (Brodie et al. 1995). Selection differentials are problematic because although they measure total selection on a trait. The indirect selection that is part of the differential is evolutionarily important only if the phenotypic correlations among traits are representative of the genetic correlations. Ofien, information about genetic correlations is unavailable or would be prohibitively difficult to obtain. However, if traits are genetically correlated, then correlated responses to selection can be a primary cause of evolutionary change in a trait, and failure to consider correlated responses may lead to erroneous conclusions. A great challenge for ecological genetic studies is to assess when indirect selection may have important evolutionary consequences, especially in cases where selection gradients and selection differentials suggest very different relationships between a trait and fitness 121 (Brodie et al. 1995). Path analysis (structural equation modeling) is an underutilized tool that can aid in understanding the potential sources of indirect selection on traits (Kingsolver and Schemske 1991, Mitchell 1992, Conner 1996, Conner et al 1996). In this study, I manipulated light and leaf litter within a natural population of the forest winter annual Collinsia verna to quantify their effects on six phenotypic traits: emergence date, date of first flowering, specific leaf area, plant height, above ground vegetative biomass, and reproductive investment. The relationship of these traits to survival and three multiplicative components of lifetime fitness in each environment was estimated. I ask: Are traits plastic in response to these factors? Does the magnitude or direction of selection on traits change across light and/or leaf litter environments? If selection changes between particular environments, can light or leaf litter be characterized as causes of natural selection? Do patterns of indirect selection differ from those of direct selection in a way suggestive of tradeoffs between traits? If traits are plastic, is the direction of phenotypic change consistent with the observed pattern of selection in such a way that the plasticity could be characterized as adaptive? METHODS Study system Collinsia verna is a winter annual wildflower of moist woods in eastern North America. The winter annual life history allows plants to grow in a window of light availability between canopy leaf senescence in the fall and leafout in the spring. Seed germination is cued by diurnal temperature fluctuations in the fall (Baskin and Baskin 1983). Flowering in this mostly out-crossing species begins in late April and ceases as the canopy closes in May. Fruits ripen, seeds fall passively to the ground and plants die 122 by mid-June. Biotic and abiotic conditions during the growing season are inherently unpredictable, and can result in variable emergence, survival, and seed set (Chapter 2 and this Chapter). Reproductive assurance in this mostly outcrossing species is achieved through backup selfing (Kalisz et a1. 1999). Long term persistence is ensured through the production of a significant fraction of dormant seeds (Kalisz 1991). Experimental design The study population (described in Thiede 1998), occurs along the south-facing edge of a woodlot adjacent to an agricultural field in Kalamazoo County, southwest Michigan. Eight different light and leaf litter environments were created at the field site. Each light and litter combination was replicated five times for a total of 40 1.2 m2 plots. To control light levels, the entire woody canopy along the southern edge of the population was cleared, and three light treatments were assigned randomly to the 15 plots in this area: full sun, 40% of full sun, and 18% of full sun (Figure l of Chapter 2). Reduced light was achieved by placing a wood lattice (3.5 cm wood strips in a 15 cm or 7 cm grid) over the plots that allowed sunflecks to reach the plants throughout the day. Because light was being manipulated in these plots, leaf litter was not allowed to accumulate (for details see Chapter 2). To understand how light and litter manipulations relate to the conditions the plants naturally experience, 25 additional plots were established. In 15 of these plots leaf litter was lefi intact. These plots were placed along the natural light gradient with five in the forest interior, five along the southern edge, and five in the full sun. The remaining ten plots, five along the southern edge and five in the forest interior, had all leaf litter removed on a twice weekly basis. Light was more variable in the plots in the forest and along the edge than in the cleared area where light 123 levels were experimentally manipulated. The mean light availability in the forest interior was about 45% of full sun, while along the edge it was about 70% of full sun. Overall, there were five light levels in the study: low (18%), medium (40%), forest interior (45%), edge (70%), and full sun (100%). The low and medium light levels had all leaf litter removed, while the three highest light levels had both litter and no litter treatments. In the fall of 1996, I conducted weekly censuses of all naturally emerging Collinsia verna seedlings. Plants were tagged with color coded washers. Mortality censuses were done weekly through December, again just prior to flowering in April, and at harvest in June. In April and May of 1997, survivors were tagged for date of first flowering and sub-sampled for specific leaf area. A circular piece of leaf tissue from the youngest fully expanded leaf was collected with a hole punch, air dried, and weighed to determine specific leaf area for approximately 250 randomly chosen individuals in each treatment, 50 from each plot. Plants were harvested prior to seed dispersal in June. After harvest, plants were air dried and scored for length of the mainstem, aboveground vegetative biomass, number of flowers, number of fruits, number of seeds, and total mass of seeds. Multiplicative fitness components (fruits per flower and seeds per fiuit) were calculated from these measures for use in path analysis. Reproductive investment was calculated as the proportion of aboveground biomass that was seeds. Because all plants were dying at this point, this trait measures the efficiency with which they were able to convert vegetative biomass to seeds, not simply reproductive allocation. Data analysis Phenotypic plasticity-The first goal of the analysis was to determine if the traits were plastic in response to the environmental manipulations. For this analysis, each of 124 the eight environments was considered a unique treatment. Fixed effects of treatment on emergence date, flowering date, specific leaf area, mainstem length, vegetative biomass, reproductive investment, flowers, fi'uits/flower, seeds/fi'uit, and seeds were tested with a one-way MANOVA (SAS GLM procedure, SAS 1989). This approach is conservative, correcting for any correlations among traits. Although specific contrasts are of interest in the selection parameters (below), these phenotypic traits and fitness components were compared across all pairs of environments using univariate post-hoe tests. Traits were log-transformed as necessary to improve normality. Survival analysis-The effects of leaf litter, light level, and emergence date on survival to flowering were analyzed with proportional hazards regression (SAS PHREG procedure, SAS 1997). Litter was coded as present or absent (0 or 1), and light levels were ordered from low (1) to full sun (5). The lifespan of each individual was determined in days fiom the date the first seedling emerged. Because seeds that emerged late were not exposed to the same conditions as those which emerged early, individuals that emerged late were treated as missing values in the risk set until they emerged (Allison 1995). Data for plants that survived to flowering was treated as uncensored. There was very little mortality between flowering and harvest, when all plants died. Treating survivors as censored data increased the magnitude of parameter estimates, but signs and significance levels were identical. Plants growing in the medium and low light treatments experienced high mortality due to grazing by an unknown herbivore in early spring. The shade lattice appeared to provide a refuge for these grazers, who removed the cotyledons and leaves of many plants, killing them (personal observation). Because this mortality is likely an 125 artifact of the lattice not directly related to the light and leaf litter manipulations of interest, these treatments were excluded from the full analysis. Phenotypic selection-Natural selection on emergence date, flowering date, specific leaf area, mainstem length, vegetative biomass, and reproductive investment, was analyzed using multivariate episodic selection analysis (Arnold and Wade 1984a). For this analysis, all traits were standardized to a mean of zero and unit variance. Linear and nonlinear selection differentials (S, C) and gradients ([3, 7) were calculated with simple (S, C) and multiple regression (B and y) for two episodes of selection, survival to flowering and fecundity (SAS REG procedure, SAS 1989). Variance inflation factors (VlFs) were less than ten for all terms in all analyses, indicating that there were no problems with multicollinearity (Neter et al. 1985). Although the selection parameters calculated for each episode are not additive (Lynch and Arnold 1988), they do address the fitness consequences of traits independent of selection in other episodes (Koerri g et al. 1991). Moreover, because only one trait, emergence date, is common to both selection episodes little would be gained by transforming the selection parameters to make them additive. Selection gradients are a measure of direct selection on a trait, while the selection differentials measure total selection on a trait including any indirect selection through phenotypically correlated traits. The survival episode was a univariate analysis of emergence date because no other traits were measured in this interval. The fecundity episode included all six traits. Relative fitness was calculated for each episode by dividing a plant's survival (0 = died or 1 = survived) or seed number by the mean survivorship or seed number for all plants within the same environment. Confidence 126 intervals for selection parameters were calculated using bootstrap resampling methods (Noreen 1989, Dixon 1993). The data sets for each of the eight environments were resampled 1000 times using a SAS macro (SAS 1990). The number of observations in each resampled data set was equal to the number of observations in the original data set. Bootstrapping and parametric significance tests produced nearly identical results. Calculation of selection parameters for the survival episode using logistic regression (SAS LOGISTIC procedure, SAS 1997), and back transforming the results (Janzen and Stern 1998) also produced similar results (not reported). Overall differences across environments in selection parameters were analyzed with heterogeneous slope tests (ANCOVA in SAS GLM procedure, SAS 1989). For traits where the linear selection gradients were significantly different, planned contrasts were constructed using the contrast statement in proc GLM. Six contrasts were evaluated for each trait: the three unmarripulated light environments (forest interior, edge, and full sun) contrasted with each other, and presence versus absence of leaf litter within each light environment. Path analysis-Structural equation modeling was used to investigate the relationships among traits in order to understand the sources of indirect selection in each environment (SAS CALIS procedure, SAS 1997). The multiple regression technique used in selection gradient analysis is analogous to a path model where all traits are directly linked to fitness (Figure 10a). This model assumes no causal relationships among traits and may be misleading when causal relationships among traits are expected. In this study, it is likely that emergence date, flowering date, specific leaf area, and mainstem length are causally related to the performance traits total vegetative biomass 127 Figure 10. Path models representing possible relationships between traits and fitness. Straight lines with single headed arrows represent hypothesized causal relationships, while curved lines with double headed arrows represent correlations. (a) Multiple regression of all traits on fitness. This path diagram represents the multivariate directional gradient analysis. (b) Path model 1 with vegetative biomass and reproductive investment as intermediate traits, and three multiplicative fitness components. (0) Path model 2 with only vegetative biomass as an intermediate trait. Traits: emergence date (Edate); flowering date (F date); specific leaf area (SLA); length of mainstem (Mainstem); reproductive investment (RI); vegetative biomass (Mass); residual unexplained variation (U). 128 Multivariate Selection Model (b) . , Model 2 (C) Edate ‘— U / Mass \ U 1 Flowers Fruits! D SLA ._ Flower D U Main Seeds/ Stem Fruit 1 U Figure 10. RI flflflfl and reproductive investment. It is also possible that these relationships may change depending on the light environment. Two a-priori path models were analyzed (Figure 10b-c). In model one, vegetative biomass and reproductive investment were treated as intermediate traits linking emergence date, flowering date, specific leaf area and mainstem length to fitness components. In model two, only vegetative biomass was treated as an intermediate trait. Model two may be more appropriate in high resource conditions, where abundant resources make efficient conversion of resources to seeds less important. Model one may be more appropriate in low resource conditions where the ability to efficiently use scarce resources is critical. In both path models, seed number was partitioned into three multiplicative components: flowers, fruits/flower, and seeds/finit. Correlations among all traits, and correlations of multiplicative fitness components with seed number are provided to aid interpretation, but the analysis was run on the covariance matrices, as this produces more reliable standard errors (Hatcher 1994). Emergence date, flowering date, specific leaf area, vegetative biomass, and number of flowers were log transformed based on the results of the diagnostics for multivariate normality reported by the CALIS procedure. In structural equation modeling, outliers do not have a strong effect on path coefficients, but can cause highly inaccurate standard errors (Hatcher 1994). Path coefficients were estimated from the full data set using the maximum likelihood method. For significance testing, outliers were identified and excluded using the multivariate kurtosis diagnostics in the CALIS procedure. The significance of individual paths was detenrrined fi'om t tests calculated by dividing each path coefficient by its standard error. Because models one and two are not 130 nested, there is no way to compare them statistically. The fit of each model to the data was compared to that of a null model (no correlations among traits) with Chi-square tests and three goodness-of—fit measures. RESULTS Phenotypic plasticity All traits and fitness components were highly plastic in response to light (MANOVA, P<0.0001 for each, Figures 11, 12). The effects of leaf litter were variable. In the forest interior, litter had no effect on early life history traits, but had a dramatic effect on vegetative biomass, reproductive investment, flowers, and seeds per fruit. On the edge, leaf litter affected traits that were expressed throughout the life cycle: emergence date, specific leaf area, mainstem length, and number of seeds. In the firll sun, leaf litter only affected emergence date and specific leaf area. Traits varied in response to light following several different patterns. There were maximums at intermediate light levels for traits closely linked to fitness (vegetative biomass, reproductive investment, number of flowers, and number of seeds), suggesting that the extreme environments were stressful. Mainstem length also had maximum values in intermediate light environments. Emergence date was earliest in the forest interior and edge environments. Flowering date and specific leaf area declined with increasing light. These are characteristic light responses in most plants. Fruits per flower showed a modest increase with light. Survival analysis The goal of this analysis was to understand the combined effects of leaf litter, light environment, and emergence date on patterns of mortality. Because mortality in the 131 Figure 11. Means and standard errors of phenotypic traits. Fixed effects of the eight environments were analyzed in a single one-way MANOVA that included all phenotypic traits and all fitness components (Figure 12) except survival as dependent variables (GLM procedure, SAS 1989). All traits differ across environments at P<0.0001. All pairwise means comparisons were made using the tukey option of the means statement of proc. GLM. Columns sharing a letter are not significantly different from each other at P<0.05. 132 2 Q mutuaoi c6 a e m m .. saw :3. .m .. . I L u C -_ owpm M m E D 8:8... . .88”. b Em s? W . a w, m an s... A _ m m m m m m m o 4. 3 .. 1 o 5 5 4 4 AS... 88.2.82 mo 59.3 0 Q 0. 0.. Amcuom mm 3.2.85 a... 48>... diam cam =3”. ownm 1.2.8... . .88". .Sm .xbv @ m + a We, 55 s... _ _ _ _ _ _ _ _ A B B % M 0 0 O 0 0 0 5. 4. 3 2 J O U H W W w w m N w 0 0 0 O 0 8mm venomous". Erma... no.5. meet. on. comm Am. 3385 o>_§owo> Figure 11. 133 Figure 12. Means and standard errors of fitness components. All pairwise means comparisons were made as in Figure 11. Survival is not a continuous variable, so it could not be analyzed this way. Environmental effects on survival were analyzed separately using proportional hazards regression (Tables 8 and 9). 134 .. new :3. b .. w m , . .wmé. 0 R. m er... .3»... .1. L m c 8.8.... .m o .808”. R N new $3 I D b c 56 Ax... 0 8.8.... ,. . .88.. 25m gov ) ) (at (er d cam fin: _ _ _ q _ _ 1 8 6 4 2 O 1 0 0 O. 0. 0 0. . 2 .33.... o. w...>.>...m 8.22.0... 832%..2”. mpoom 135 Figure 12. medium and low light manipulations was caused by herbivory that was unrelated to light or litter, these Mo treatments were excluded from the full analysis. Overall, the presence of leaf litter more than doubled the probability that plants would die before flowering (Table 8), but this result was accompanied by a highly significant three-way interaction between leaf litter, light environment, and emergence date. The risk ratio describes the change in hazard of death as a consequence of a one unit change in the variable. Risk ratios greater than one indicate an increasing hazard, less than one indicate a decreasing hazard. The risk ratio of 0.998 for emergence date in Table 8 indicates that the hazard of death before flowering declined 0.2% for every day that emergence was delayed or about 6% per month. The risk ratio of 1.005 for the significant two-way interaction of emergence date and light indicates higher risk of death with greater combinations of these two (late emergence in high light), while the risk ratio of 0.994 for the three-way interaction suggests this relationship did not hold in the presence of litter as seen in Figure 12a. This result is consistent with the hypothesis that leaf litter might reduce the number of diurnal freeze-thaw cycles. To better understand the three-way litter by light by emergence date interaction, separate analyses of the effect of emergence date on survival were then run within each litter and light level combination, including the low and medium light manipulations (Table 9). When leaf litter was intact, there were no significant effects of emergence date on the hazard function in any light environment. In the absence of leaf litter, the risk of death increased with delayed emergence by 33-3 6% per month in the forest interior and full sun, but by 63% per month along the edge. In the low light treatment (excluded fiom the full analysis), the risk ratio of 0.995 indicates a 15% decline in the risk of death for 136 Table 8. Proportional hazards regression survival analysis for effects of emergence date, leaf litter, and light for forest interior, edge, and full sun environments (N=9836). Variable P Risk Ratio Emergence Date 0.6564 0.998 Leaf Litter 0.0138 2.099 Light 0.6124 0.969 Emergence Date X Leaf Litter 0.3831 1.007 Emergence Date X Light 0.0003 1.005 Leaf Litter X Light 0.5289 1.051 Emergence X Litter X Light 0.002 0.994 137 Table 9. Proportional hazards regression survival analysis for effects of emergence date on survival of Collinsia verna within environments. Environment N P Risk Ratio EulLAnébLsis Forest, With Leaf Litter 1402 0.5037 0.999 Forest, No Leaf Litter 1582 0.0001 1.011 Edge, With Leaf Litter 2100 0.657 0.999 Edge, No Leaf Litter 2165 0.0001 1.021 Full Sun, With Leaf Litter 1466 0.6368 0.999 Full Sun, No Leaf Litter 1121 0.0001 1.012 iv ct 18% Sun, No Leaf Litter 2982 0.0001 0.995 40% Sun, No Leaf Litter 2156 0.49 0.999 138 every month that emergence is delayed. The herbivores preferred earlier emerging seedlings. Phenotypic selection: Survival episode The direction of selection on emergence date changed in the unmarripulated light environments depending on the presence or absence of leaf litter (Figure 13a). Moreover, there was significant upward curvature in the fitness surface in all environments (positive nonlinear gradients, Figure 13b). This means that there were dramatic nonlinear increases in survival with late emergence in the presence of leaf litter, but with early emergence in the absence of leaf litter. Overall, this result implies strong disruptive selection on emergence date. In the low and medium light manipulations where herbivory was the primary source of mortality, positive directional selection on emergence date indicates that late emerging plants were less likely to be eaten. These results are generally consistent with the survival analysis, but selection for late emergence by leaf litter appears much stronger here than suggested in the previous analysis. Phenotypic selection: F ecundity episode The most striking result of this analysis was that indirect selection differed in sign from direct selection for nearly all traits and environments (Figure 14). In contrast to the survival episode, there were no changes in the direction of direct linear selection except for specific leaf area (Figure 14). Direct selection favored higher specific leaf area (a shade phenotype) in the forest interior and on the edge, but lower specific leaf area in the undisturbed full sun environment (Figure 14c). Total linear selection (open bars, Figure 14) differed significantly in magnitude across enviromnents for all traits except emergence date, and direct selection (shaded bars, Figure 14) differed for specific leaf 139 é 0‘3 (a) P=0.0001 302-1 '8 630.1— —} . ~ g 0 R k u Egg-0.1— 1 I T Eo-().2— § IeafLrtter l E DNoLeafLitter -0.3 g 51.58! & l E ”3 Vi $5 .3." :23 0'6 (b) __ P=00001 s: 0.5- C204—T I § - é . l Foe—11' \]§T$ e l\ V s 02- \ §l§- 8 _ \ 5 Oc:k~.l\o kc —. <1- Figure 13. Survival episode standardized linear (B) and nonlinear (7) selection gradients for emergence date. Because emergence date is the only measured trait that was expressed during this episode, selection gradients and differentials were very similar. Consequently, only the gradients are presented. Bars are 95% confidence intervals based on 1000 bootstrap resampled data sets. P-value is the result of an overall ANCOVA testing whether selection gradients differ across environments. The Y-axis indicates the proportion by which relative fitness would change with a change of one standard deviation in the trait. (a) Linear gradients. (b) Nonlinear gradients. 140 Figure 14. F ecundity episode standardized linear selection differentials (S) and gradients ([3). Bars are 95% confidence intervals based on 1000 bootstrap resampled data sets. P- values are the results of AN COVAs testing wether differentials or gradients differ across environments. The Y-axis indicates the proportion by which relative fitness would increase with an increase of one standard deviation in the trait. Note that the scale of the Y-axis differs between figures. (3) Emergence date. (b) Flowering date. (c) Specific leaf area. ((1) Mainstem length. (e) Vegetative biomass. (1) Reproductive investment. 141 xS\<€<1<~\\‘<< 11.“: “Q. (R. 2. .. .._.:5m a: ~“1NKIVW“1‘“‘\‘11111.1§ V ‘11" 11V| ‘V“V14 00 6 6 0. 0 k \ .nnwxxxxxsx m 1 1 0 m 0. m. . “unease 0. 0 m r... n... k ( P w .2. mil . . . _ . _ _ . _ _ _ . 2 0 2. 4. 6. 6 4. 2. 0 2 4. 6. 4 2. 0 80A. 9:530... :95... E8282 H.5..Em..v>=. 02.030803. 1 00 0 o. 0 \m 0 k. \e/ ( ( ( 4. _ ._ .11 _._. _._._._._ 2 0 2. 4. 1 3. 5. 4. 2 1 8 6. 4 2. 0 800 005925. _ m2< .00.. 0:6on Linear Gradient - No Leaf Litter Linear Gradient — Leaf Litter [:1 Linear Differential - No Leaf Litter Linear Differential — Leaf Litter Figure 14. 142 area, vegetative biomass and reproductive investment. Planned contrasts for these three traits showed different patterns of change in the linear selection gradients across environments (Table 10). Selection on specific leaf area and reproductive investment differed between the full sun and the other unmarripulated environments (Figures 14c, f). In contrast, selection on vegetative biomass was similar across light environments, but increased in the presence of leaf litter in the forest interior and on the edge (Figure 14e). There was little direct nonlinear selection (nonlinear gradients, 7,.) on any trait in any environment, except vegetative biomass (Figure 15, shaded bars). There was strong and variable upward curvature in the fitness function for vegetative biomass. This increasing slope suggests that fitness accelerated as plants got bigger across all environments. Total nonlinear selection was generally negative for most traits (Figure 15, Open bars). When viewed in conjunction with the negative linear terms, these results suggest that plants with particularly late emergence or flowering, or high specific leaf area had very low fitness. With one exception, correlational selection gradients (yij) were not significantly different from zero (results not shown). There was significant correlational selection to increase the covariance between vegetative biomass and reproductive investment across all environments (range: yij = 0.2 in the forest interior to yij = 0.4 in the full sun). This result simply suggests that plants that were able to allocate resources to seeds in proportion to their size had higher fitness. 143 s.e.. £80 $2.0 $3.0 gm :3. 280 30.0 682 8B wtwd Coed 33.0 8.08:. 580m 0340.212 480.0 $3.0 $8.0 5m =3. .3 85 200.0 0250 .380 cum :5. .9 SEE. neon. mom... 96.3 m . mm... "swam. .m> 3.08:. 820.”. fig “Cour—30>... o>$oguohaom wmQEOmm 0>390wo> «02 .30... 2.20on .meCOU 200808.25 .8... .3. .05 Em: 8.. 300.3% 00500.3 800.. .0 380000 0058.0 00.. a. 002? 0058..in .9. 03m... 144 Figure 15. F ecundity episode standardized nonlinear selection differentials (C) and gradients (7). See Figure 14 for details. 145 P=0.2588 P=0.0022 \\\\\ P=0.0001 00m :0”. we 40:80. .800... 03m :9. ma .0: 00m :0". we 40:80. .800... dam :9. 5m 32 800 3:030... .308. £82.02. — d '4' 2 0 0 . 0020330. 03.00.0053. +1. ll _. 2 0 P=0.9349 P u \I C l.\ fi P=0.0001 n 03. m. In. 3341...: V w b} ~I \ V5 ~sz. 4“! -P P=0.0001 A. .0». .IL. m... 40:30. Nonlinear Gradient - No Leaf Litter Nonlinear Gradient - Leaf Litter 4. 0 00w :0“. ~08 00:20. .800... 0.0 $9. E5 3.: 5m =3. .08 3.20.". dam $3. Em sh: _ A O 0 1 -0.3 - -O 5 . . 2 6. 8 3 mu . 0. n. O . 005 800908.". 02< .00.. 0:6on .._. 33.3197. 111100 0.5 {100.0001 T qu— o.3 { _ 1|” 1 o 6 30.005 0>:80wo> 3. 0 Cl Nonlinear Differential - No Leaf Litter § Nonlinear Differential - Leaf Litter Figure 15. 146 Path analysis The path analyses revealed sources of indirect selection on traits in the fecundity episode, and also showed how different traits contribute to the multiplicative fitness components in different ways. As suggested by many reviews of the use of structural equation models (e. g. Petraitis et al. 1996, Shipley 1997, Grace and Pugesek 1998), Table 11 presents the raw data (means, standard deviations, and correlations) used in these analyses. Correlations with seed number are also presented in this table to aid in interpreting paths leading to multiplicative fitness components. Goodness of fit indices for both path models (Figures lOb-c) in all environments are presented in Table 12. The Chi-square statistic reported here is a measure of how well the specified model fits the data, with higher P values indicating a better fit. Values of the normed fit index (NFI), non-normed fit index (NNFI), and the comparative fit index (CFI) greater than 0.9 indicate an acceptable fit between model and data (Hatcher 1994). In all cases model one provided a better fit to the data, so only the results of the analysis of this model are presented. Results for model two were essentially identical to those for model one. Consistent with the results of the fecundity selection episode, the relationships among traits in the path analysis were quite similar across all environments (Tables 13 and 14). The path diagram for the natural forest environment is representative of these results (Figure 16). Earlier emergence was predictive of greater vegetative biomass. Earlier flowering was predictive of greater vegetative biomass and reproductive investment, but later flowering increased number of seeds per fi'uit. Selection favored higher specific leaf area (shade phenotype) directly through its relationship with flower number and indirectly through reproductive investment. But this selection was balanced 147 00.0 00.. 00.0 00.0 00.0 0.0 00.0 00.0 0.0- 0.0- 0.0- 00.0 00.0 00.000 00000 00.0 00.0 0.0 00.0 0 00.0 0.0 00.0 00.0 00- 00- 00.0 00.0 .000 0300080 000 0.0 0.0 3.0 0.0- 00.0 0.0 00- 0.0- 00- 00.0 00.0 00.0 E. 5320050 00.0 00.. 0.0 000 0 _ .0 00.0 0 _ .0 00.0 00- 0.0- 0.0- 00.0 00.0 00300 02.5: 00.0 00.0 00.0 00 00.0 00.0 000 E0 0.0- 0.0- 0.0- 00.0 00 08>. 0.0820 203000> 00.0 00.0 00.0 00.0 00.0 00.0 00- 00.0 00.0 0.0- 00- 00.0 .00 000 002582. 2.080280 00.0 :0 00.0 000 000 00.0 00.0 0.0 0.0- 00- 00- 00... 0. .020 .0003 820502 00 00.0 0.0- 00.0 0.0- 0.0- 3.0 0.0- 0.0- 00.0 00 00 00.0 00.500 8208-. 00.800 00.0 00.. 0.0- 0.0- 0.0- 0.0- 0.0- 0.0- 0.0- 00.0 00.0 00.0 00.. 00.0-.0 200 005300 .00 0.1 0.0- 0.0- 00- 0.0- 0.0- 0.0- 00- 00.0 00.0 0.0 .00 00.000 200 850500 .......->... - .4 .0. ......-) ....0...,.>>..-.. .. .4 ....:..3. 000 :82 00 00 0... 00 m> .0 0: <0 00 00 000 502 25,2 / 025020 3200 awn-C. .GuEdo-._>cm 6000.00.00.00 w0. 003 0.00.03 w». ..000w0..0 00.. 30.00. .000 000.0 0000 00000000300 00000....5 0.00.00 .00.. 0.: 0. 00000.00. 00 000 000.00.000.02 0.0.0.000 000.. 0. 00.0.0.0? 000.50.. 000000000300 0.5.3 000.00.00.60 .000 A95. 0000:0300 000.0080 £000.). .2 0.0.0... 148 .00 0.... 00.0 00.0 00.0 .00 00.0 00.0 ..0- 0.0- 0.0- 00.0 00.. 00.000 0080 00.0 0.0 .00 0 00.0 00.0 00.0 00.0 :0 ..0 .0- 00.0 .00 .00. 03-00080 0. .0 00.0 00.0 .0- .0- 00.0 00.0 00.0 .0- 0.0- .0- 0.0 00.0 0.0. 0030.00.30 00.0 00.. 00.0 0. .0 ..0- .00 .0- 0.00 .0- 0.0- 0.0- 00.0 3.. 003-... 0.030.“. 00.0 00.0 00.0 00.0 00.0 0.0 0.0- 00.0 0.0- 0.0- 0.0- 00.0 00.0 020;. 00.2.5.0 o>00..-00> 00.0 00.0 00.0 00.0 00.0 0 0.0- .0- 0.0 0.0- 0.0- 00.0 00.0 50 0.2503... 02.802000 .00 0.00 00.0 0 0 0.00 00.0 .0- 0.0- .0- .0- 00.0 0.0 .02. 505-. 0.22.02 00.0 .00 0.0- 0 .0- 0.0- 0.0- 00.0 0.0- 0.0 .0- 00.0 00.0 00.500 8.503 00.800 ..0 :.. 0.0- ..0- 0.0- 0.0- 0.0- 0.0- 0 00.0 00.0 00.0 0... 00.000 0.0.0. 00530.”. 00.0 00.. .0- 00.0 0.0- .0- ..0- .0- .0 00.0 00.0 00.0 00.. 00.00.. 200 850.200. 0....v.¢.>o... - .1 a. 0.0.0..- .-0 o.--- .0. 000 082 00 ”.0 0... 00 m> .0 02 <00 00 00 000 :82 26.2 / 025005 30.00 0.00-.- 0000000030-0. .0008. .. 0.000 149 .00 3.. 3.0 00.0 00.0 00.0 00.0 0.00 ..0- .00- 0.0- 00.0 00.. 00.000 0080 00.0 .00 .00 0 0.0 0.0 00.0 00.0 00.0 0 .0- 00.0 0.00 .00. 020.0030 0.0 00.0 00.0 00.0 .0- .00 .00 0 0 0 ..0 ..0 0.0 .0... 030.0005". .00 0 . .. 0.0 .0- 0.0 00.0 .0.0 00.0 0 0- .00- 00- .0.0 .0. 003-... 0.030.“. 0.0 00.0 00.0 0 00.0 00.0 0.0- .00 0 0- .00- ...0- 00.0 00.0 00.050 00050.0 02.0.00; 00.0 .0.0 00.0 00.0 00.0 .0- 0.0- 0.0- 00.0 0 0 00.0 .00 000 .5802... 00.80203. 00... 00. 00.0 0.0 0.0 00.0 3.0 0 0 0 .0- 0.0 0.00 .02. 5000-. 8200...... 00.0 00.0 0.0- 00.0 .0- 0.0- 0.0- .00 0.0- ...0 00.0 00.0 00.0 0020000258. 2002.0 0.0 0.0 0.0- .0 .0- 0.0- 0.0- .0- . 0- 0.0 .00 .00 .00 00.0.0. 0.00 000030.“. 00.0 00.. 0.0 00.0 ..0- .00- 00- .0- . 0- 00.0 00.0 00.0 00.. 00.00.. 200 850.050 .....;... 3..-; .. .1 0...- 000 :82 00 00 00 00 0> .0 02 <00 00 00. 000 :82 26.2 / .2500... 30.00. 0.00-.- 00080900.". .0008. .. 0.00-0 150 00.0 00.. 0.00 00.0 00.0 .00 00.0 .00 0.0- 0.0- 0.0- 00.0 00. . 00.00. 00000 .00 00.0 0.00 ..0- 00.0 00.0 00.0 00.0 .0- .0- .0- 0.0 0.0 .00. 030.080 0.0 .00 00.0 0.0 0.0 00.0 0.0 0.0 .0- ..0- 0 0.0 00.0 .00. 030.0003”. .00 00.. 00.0 00.0 .00 .00 00.0 00.0 0.0- .00- 0.0- 00.0 00.. 00.00. 0030.0 00.0 0.0 .00 .00 00.0 00.0 0.0- 00.0 0.0- 0.0- 0.0- 00.0 00.0 00.05. 0088.0. 0>§000> 00.0 00.0 .00 00.0 00.0 .0- 0.0- 0.0- 00.0 0.0- 0 00.0 00.0 00. 005.02.. 02.00.0280 0.0 .0. .00 00.0 .00 00.0 0.0 .0- .0- .0- .0- 00... 0.0. .02. 0.0.00 0.0.2.02 .00 0.0 .0- 00.0 .0.0 0.0- 0.0- 0.0 .0- 00.0 00.0 ...0 0.0.0 00200. 002.000 00.800 0.0 0.0 0.0- ..0- ..0- .00- .00- 00- 0.0- 00.0 00.0 0.0 00.0 00.0.0. 0.00 000030.”. 0.0 00.. 0.0- 00.0 0 0.0- 0.0- .00 0.0- 00.0 0.0.0 0.0 0.... 00.00.. 0.00 850.200 00030200000001.0003 00.0 9.0.). 00 00 00 00 00> .0 00>. <00 00 00. 000 :82 0>o0< . .2000... 30.00. “mm-C.- EOE-09.3.0.1”. 00.0.8. .. 0.0.... 151 Table 12. Goodness of fit indices for path models. df= degrees of freedom; NFI = normed fit index; NNFI = non-normed fit index; CFI = comparative fit index. Model Chi-square df P NFI NNFI CFI MW n=214 Null Model 941.6317 36 <0.000 0 Model 1 25.2778 6 0.0003 0.973 0.872 0.979 Model 2 27.2053 7 0.0003 0.971 0.885 0.978 ' e itt n=205 Null Model 1099.373 36 <0.000 0 Model 1 13.1631 6 0.0405 0.988 0.960 0.993 Model 2 24.4493 7 0.0009 0.978 0.916 0.984 Bones-1.1111009102102011 n=190 Null Model 761.2273 36 <0.000 0 Model 1 16.0655 6 0.0134 0.979 0.917 0.986 Model 2 30.5346 7 0.0001 0.960 0.833 0.968 W n=223 Null Model 847.4237 36 <0.000 0 Model 1 9.3355 6 0.1556 0.989 0.975 0.996 Model 2 20.3378 7 0.0049 0.976 0.916 0.984 W 11:224 Null Model 893.9886 36 <0.000 0 Model 1 5.3995 6 0.4937 0.994 1.004 1 Model 2 8.1612 7 0.3186 0.991 0.993 0.997 W n=231 Null Model 865.8145 36 <0.000 0 Model 1 14.0028 6 0.0296 0.984 0.942 0.990 Model 2 33.954 7 0.0001 0.961 0.833 0.968 W n=316 Null Model 14072935 36 <0.000 0 Model 1 8.002 6 0.238 0.994 0.991 0.999 Model 2 17.9727 7 0.0121 0.987 0.983 0.992 W n=272 Null Model 1335.4301 36 <0.000 0 Model 1 11.304 6 0.0794 0.992 0.976 0.996 Model 2 20.1626 7 0.0052 0.985 0.948 0.990 152 Table 13. Path model one standardized path coefficients for all causal paths. *P<0.05, **P<0.01, ***P<0.001. Environments: 1. Low light no leaf litter (18% Sun), 2. Medium light no leaf litter (40% Sun), 3. Forest Interior natural, 4. Forest Interior no leaf litter, 5. Edge natural, 6. Edge no leaf litter, 7. Full sun natural, 8. Full sun no leaf litter. Traits: Emergence Date (ED), Vegetative Biomass (VB), Flowering Date (FD), Flowers (FL), Fruits/F lower (FF), Seeds/Fruit (SF), Reproductive Investment (RI), Specific Leaf Area (SLA), Mainstem Length (MS). Environment Path 1 2 3 4 5 6 7 8 ED>VB -0.03 0 -O.16** -0.05 -0.l9*** -0.2*** -O.1 1*" -0.08 ED>VB -0.29*** 037*" -0.22*** -0.21*** -0.26*** 01* -o.19*** -0.15*** FD>FL -0.08** -0.03 -0.02 -0.01 O -0.09*** -0.06*** -0.02 FD>FF 0.1 0.06 -O.11 -O.18*** -0.02 0.01 0.03* 0.09 FD>SF 0.12* 0.1 0.2** 0.08 0.09 0.18”“ 0.13”“ 0.1 1* FD>RI -0.2*** -0.23** -O.23*** -0.18* -0.07* ~O.14 -0.22*** -0.25*** SLA>VB -0.2*** -0.11* -0.19** -0.18** -O.26*** -O.35*** -0.28*** -0.23*** SLA>FL 0 004* 0.09*** 004*" 0.05** o.13*** 0.04 0.03 SLA>RI 0.14** 0.22** 0.34*** 0.11 0.29 0.36*** 0 0.34m MS>VB 0.69*** 056*“ 0.47*** 052*" 047*" 032*" 064*” 063*" MS>FL 029*" 0.03 0.03 -0.03* 0 -0.08*** -0.04** 0.02 MS>FF —O.33** -0.07 -0.02 O -0.03 0 0.01 -O.17* MS>SF -0.06 -0.03 -0.03 -0.07 0.01 0.13** 0.06 0.02 MS>RI 0.07 O -0.01 -O.l 1* -0.15** 0.05 -0.18* -0.11"' VB>FL 056*“ 0.93*** 093*” 097*“ 0.97*** 1*" 097*" 096*“ VB>FF o.33** 036* 0.14 0.02 0.03 0.37*** 0.27*** 034*" VB>SF 0.15 0.29" 0.25" 027*“ 0.28*“ 0.12" 0.15* 0.21" RI>FL 0.04 O.l4*** 0.06 02*" 021*“ 024*" 0.2*** O.l3*** RI>FF 043*" 0.33*** 0.33*** 0.21 0.04 0.35 024* 0.47*** RI>SF O.62*** O.68*** 062*” 0.63*** 06*“ 057*M 0.65*** 066*" 153 Table 14. Path model one unanalyzed correlations. Details as in Table 13. Environment Path 1 2 3 4 5 6 7 8 ED<->FD o.39*** o.39*** 0.29*** 0.26*** 0.34*** 0.33*** 0.35*** o.42*** ED<->SLA 0.1 0.08 -005 0.06 0.06 0.08 0.06 0.06 ED<->Ms -0.13* -011 -0.06 01* -013 -0.06 -0.07* -015 FD<->SLA 0.27*** 0.32*** 0.19 022* 011* 0.2* 0.06 0.23*** FD<->MS 013* -0.33*** -o.07* 0.02 -002 -007 -0.06* -o.21*** 3LA<->Ms -005 016* -0.16*** -o.17** -003 -0.29*** 014* -0.11* VB<->RI -0.08 -0.26*** -o.29*** -o.21*** -0.12* -0.29*** -o.17** -o.2*** FL<->FF -0.54*** -0.56*** -0.56*** -o.4*** -o.37*** -o.33*** -0.29*** -0.36*** FL<->SF -0.27** -o.21** -0.17** -0.26*** -0.44*** -o.49*** -0.35*** -o.24*** FF<->SF -0.29*** -0.15* -O.23*** -0.3*** -0.04 --0.1 -O.31*** -O.16* 154 Figure 16. Representative path diagram showing typical relationships among traits. This diagram is for the natural forest interior plots. The size of the arrow indicates the magnitude of the correlation or path coefficient. Solid lines are positive, dotted lines are negative. "U" represents residual unexplained variance. All values are from structural equation models based on the correlation matrices in Table 11. For other environments see Tables 13 and 14. H 00 . ova? * * D ' . * * * ,***w .. mo A w 4 ..... HQOVanw * .13." D , 325. >5 E. 84 . *daom . ........... 803582 4 8.9m... ‘ . ...... A... _ . ...... ***... DZEWDZA. . . v .. :95 is.” .... 035,51 3% .85.. A. .. ‘ . . £6on . 3.32 4 ”.... u 3.0-346 ....** . 4*... - . ... *2. 8mm mmoumm 0 ‘II ... *** 3.0-2.0 Al .. ...... 2 6-8.0 Al 3.... D *Mmmeem. .. . . 23 1 . ogflowuyx . oocomhofim . mo 0V Tl ***. , ._' . ....... .23.. D A..:::::.: ...... A m. 38% 8:584 fig» 88:: seem m. 156 by indirect selection for lower specific leaf area (sun phenotype) through vegetative biomass. Lower specific leaf area, and longer mainstems were consistently predictive of greater biomass and consequently more flowers and more seeds per fruit. Earlier flowering and higher specific leaf area were predictive of higher reproductive investment and consequently more fruits per flower and more seeds per fruit. Vegetative biomass increased fitness primarily through the production of more flowers, while reproductive investment functioned through increases in fruits per flower and seeds per fruit. Seed production was consistently more highly correlated with flower number than with fruits per flower or seeds per fruit (Table 11). There were consistent negative correlations between biomass and reproductive investment and between the three multiplicative fitness components (Table 14). These negative correlations suggest tradeoffs among these various fitness components. In only two cases were there clear environmentally-dependent changes in the signs of path coefficients. In each case the magnitudes of the path coefficients were small but significant (FD>F F and MS>F L, Table 13). Early flowering plants resulted in more fruits per flower in the forest, but the opposite was true in the sun. In each of these environments plants flowering at these times were more likely to be flowering synchronously with the bulk of the population. Longer mainstems were predictive of more flowers in the low light manipulation, but fewer flowers in three other environments. DISCUSSION The results of this study show that the presence of leaf litter caused a change in the direction of direct linear selection on emergence date. Full sun resulted in a change in 157 the direction of direct linear selection on specific leaf area. These differences in selection provide direct evidence that leaf litter and light were causes of selection in this experiment (Wade and Kalisz 1990). The strength of selection on vegetative biomass and reproductive investment was also environment-dependent in some cases (Table 10). In other cases (flowering date and mainstem length), direct linear selection was surprisingly consistent across all environments, and quite small in magnitude. That flowering date, specific leaf area, and mainstem length are highly plastic, but subject to little direct selection in any environment suggests that the plasticity in these traits is near optimal. Selection always favored greater vegetative biomass and higher reproductive investment. However, there were increasing fitness returns (nonlinear gradients) for greater biomass, but not for greater reproductive investment. Instead, the nonlinear differentials suggest that fitness returns plateau for reproductive investment. Indirect linear selection was almost always in an opposing direction to direct selection. Consequently, total selection on traits often differed in sign from direct selection. Indirect selection of a greater magnitude and opposite direction from direct selection as seen in this study is a common finding in multivariate selection studies of many species (e. g. Impatiens pallida Mitchell-Olds and Bergelson 1990, Bennington and McGraw 1995, Gross et al. 1998, Labelia Johnston 1992, Diodz'a teres Jordan 1991, Chamaecristafasiculata Kelly 1992). In this study, the path analyses revealed consistent relationships among traits across all environments that may lead to the observed patterns of indirect selection. 158 Traits Several univariate studies have detected selection for earlier emergence (Kalisz 1986, Miller 1987, van der Toom and Pons 1988, Biere 1991), and some multivariate studies have found direct influences of emergence time on early survival (Kelly 1992, Stratton 1992). However, most multivariate studies show that selection on emergence date is indirect and occurs through phenotypic correlations with size-related traits expressed later in the life-history (Mitchell-Olds and Bergelson 1990, Kelly 1992, Stratton 1992, Bennington and McGraw 1995, Thiede 1996). In Collinsia, the size that seedlings achieve by the onset of winter is an important determinant of overwinter survival (Chapter 2, see also Thiede 1996). Large size at overwintering is also correlated with greater fecundity (Thiede 1996), but at a relatively low level compared with other traits expressed later in the life-history (Chapter 2). Interestingly, there is some evidence that sufficient size at overwintering can be achieved either through early emergence or large seed size (Figure 8 of Chapter 2). Within generations, the phenotypic correlations between these traits are negative (late emerging plants produce small seeds, unpublished data, see also Thiede 1996). However, across generations the phenotypic correlations are positive (large seeds emerge later, Chapter 2, see also Kalisz 1989, Thiede 1996), and the traits have a significant positive genetic correlation (Kalisz 1989, Thiede 1998). Three episodes of selection on emergence date in an Illinois population of Collinsia verna were studied over two generations by Kalisz (1986): survival to spring, survival from spring to fruiting, and fecundity. There was significant direct selection for early emergence in the survival to spring episode of the first year (B=-0.06), and in both fecundity episodes (year 1: [310.2, year 2 [3=-0.08). In the first year an autumn flood 159 removed most litter, creating conditions similar to those in the no litter environments in this study, where selection also favored early emergence (Figure 13a). In the fecundity episode, Kalisz found selection for early emergence (range B=-0.05 to B=—0.33). Because Kalisz's study was a univariate analysis, these results are best compared to the selection differentials in this study (Figure 14a). Again, there are striking similarities, both overall, and in her by transect analysis. Kalisz found that total lifetime selection on emergence date at the quadrat scale was highly variable (year 1 range B=-0.55 to 0.2, P<0.01; year 2 range B=-0.96 to 0.68, P<0.89). Considering the small sample sizes in these calculations, it is possible that the presence or absence of leaf litter accounts for these results. There was no direct selection on flowering date in any environment (Figure 14b). In contrast, many other studies in annual plants have shown that early flowering may have significant fitness benefits (e. g. Mazer 1987, Lechowicz and Blias 1988, Brassard and Schoen 1989, Lotz 1990, Bennington and McGraw 1995, Petit and Thompson 1998, but see Ollerton and Lack 1992). This difference may be due to the size related traits in my analysis. Benefits of early flowering here are accrued indirectly through the greater biomass and higher reproductive investment achieved by early flowering plants. However, some studies still found positive direct benefits of early flowering even when size related traits were included in the analysis (height: Bennington and McGraw 1995, stern height: Petit and Thompson 1998). Selection on flowering date may depend on variable weather conditions. Spring storms in some years could eliminate the benefits of early flowering. The path analysis suggests that plants flowering later in the spring benefit directly by producing more seeds per fi'uit (Table 13). It is possible that pollinator service is better later in the season when 160 conditions are more benign. Specific leaf area is a trait that integrates over all the physiological and morphological changes that plants make to optimize photosynthetic performance in different light environments. Consequently, it is expected to be highly plastic. The selection differentials and gradients for specific leaf area in this study (Figure 14c) are remarkably similar to the pattern of selection on this trait found in studies of Iris pumila (Tucic et al. 1998) and Diodz'a teres (Jordan 1991). Dudley (1996) found similar changes in selection on two other photosynthetic traits, water-use efficiency and leaf area between two moisture environments. The consistency of these results is noteworthy, suggesting physiological and morphological changes in the photosynthetic machinery are critical for success when light and moisture environments are variable. Indirect selection on specific leaf area favored a denser, sun leaf morphology, but the path analysis (Figure 16, Tables 13 and 14) suggested a persistent tradeoff across all environments. Plants with thick leaves produced more biomass, but plants with thin leaves were more efficient at converting vegetative biomass to seeds. Resources invested in thick, dense leaves may be less labile. Not surprisingly, positive direct selection for size related traits is a common finding in nearly all selection studies (e. g. Impatiens pallida, Mitchell-Olds and Bergelson 1990, Bennington and McGraw 1995, Gross et al. 1998; Erigeron annuus, Stratton 1992). Interestingly, although overall selection favored taller, heavier plants, direct selection in this study favored shorter plants (Figure 14d). Most multivariate selection studies that include a plant height trait have not found significant direct selection for smaller plants (e. g. Kelly 1992, Bennington and McGraw 1995, Petit and 161 Thompson 1998, but see Gross et al. 1998). I could find no previous selection studies that have included both measures of plant height and biomass, so it is difficult to assess the generality of this finding. Selection for shorter plants is likely only when an analysis includes other size traits more strongly correlated with fitness. A-priori, selection would be expected to favor taller plants when inter and intra- specific competition for light is greatest. Indeed, just this sort of adaptive plasticity in stem elongation has been demonstrated in Impatiens capensis (selection gradients for height at 25 plants/m2: [3 = -0.12, at 1111 plants/m2 B = 0.25, Dudley and Schmitt 1996), and other species (Schmitt et al. 1995, Pigliucci and Schmitt 1999). The plots in the current study spanned a range of densities, yet direct selection consistently favored shorter plants. At harvest, densities ranged &0m 75 plants/m2 in the full sun no litter environment to 244 plants/m2 in the edge no litter environment. It is possible that direct selection would favor longer stems at densities greater than the range encountered in this study. In contrast to other traits, total selection on vegetative biomass and reproductive investment was less than direct selection (Figures 14e-f). The results of the path analysis provide an explanation for this pattern: there is a tradeoff between these traits, and they contribute to different fitness components (Figure 16, Table 13). The relationship between vegetative biomass and flower number is obvious: larger plants have more flowers. The relationship between reproductive investment and the number of seeds per fruit suggests that reproductive investment may be a measure of the ability to self pollinate (Kalisz et al. 1999) and/or of various display characters that affect the rate of pollinator service (flower size, color, scent, rewards). However, it is difficult to see how 162 investment in display characters or the ability to self pollinate could result in the negative correlation between vegetative biomass and reproductive investment. Given this negative relationship, it is likely that reproductive investment is in part a measure of the ability to reallocate resources to seeds. The tradeoff reflects a basic energetic constraint: large plants must invest pr0portionally more resources in structural tissues, which are then unavailable to provision seeds. Causes of selection The observation that there is spatial and temporal variation in the magnitude of selection within populations is a common finding, but rarely can the cause of this variation be identified (e.g. Kalisz 1986, Stewart and Schoen 1987, Kelly 1992, Stratton 1992, Gross et al. 1998). Reciprocal transplants can demonstrate environment-dependent selection and can suggest possible causal agents of population differentiation (e.g. Jordan 1991, Bennington and McGraw 1995, Petit and Thompson 1998). However, actual environmental manipulations are necessary to identify specific causal agents of selection (Mitchell-Olds and Shaw 1987, Wade and Kalisz 1990). Surprisingly few studies in plants have used this powerful approach (Dudley 1996, Mauricio and Rausher 1997, Tucic et al. 1998, Winn 1999). Recently, researchers have manipulated both environment and phenotype to gain a more mechanistic understanding of environmental effects on plant phenotype and fitness (reviews in Schmitt 1999, Schmitt et a1. 1999, Schmitt 1997). By inducing the production of the "wrong" phenotype in relevant environments, this approach allows both the identification of causes of selection and the testing of adaptive plasticity hypotheses. By manipulating plant height using mutants (Schmitt et al. 1995, Callahan et al. 1999, 163 Pigliucci and Schmitt 1999) and light and hormonal cues (Dudley and Schmitt 1996, Cipollini and Schultz 1999), it has been shown that increased plant density selects for shade-avoidance responses, and that plasticity in these traits is adaptive. Other studies of similar design have demonstrated the potential adaptive value of hormonally-induced chemical defenses and the effectiveness of herbivores as selective agents (Baldwin 1998, Agrawal et al. 1999). The results of this study show that in this population, leaf litter was a selective force on emergence date (Figure 133). Further, leaf litter and light interact in their effects. In the presence of leaf litter, selection for late emergence was constant across different light levels. Other studies have shown that leaf litter can be a source of mortality, both directly through physical interference and shading, and indirectly through its affect on pathogens and herbivores abundance (F acelli 1994). In this population, ' seedlings become etiolated and fragile when they emerge beneath litter. The environment under the litter also provides an ideal habitat for foraging slugs which occur in abundance in some autumns (Thiede 1996). In the absence of leaf litter, the strength of selection for early emergence increased with light availability. Here, the diurnal freeze/thaw cycle in the upper soil layer over the winter may cause mortality, and this cycle may be stronger in high light. Late emerging, shallowly rooted plants are vulnerable to being heaved from the soil (personal observation). That leaf litter selects for later emergence suggests that the litter layer buffers these temperature cycles, and that late emerging seedlings are less likely to become trapped under litter. Results also showed that light environment was a selective force on specific leaf 164 area (Table 10, Figure 14c). Low light favored a shade leaf morphology, while full sun favored a sun leaf morphology. Environment-dependence in the pattern of selection on vegetative biomass and reproductive investment was also evident (Table 10), but the causal effects of environment are more difficult to characterize because these results were not consistent across all environments (Figures 14e-f). When is plasticity adaptive? Because the plants in this study were not induced to produce the wrong phenotype in each environment, this study cannot answer this question definitively. However, the results for flowering date, specific leaf area, and mainstem length are consistent with the hypothesis that plasticity is adaptive. The lack of strong direct selection on these traits within environments in spite of significant plasticity across environments suggests that plants are producing the appropriate phenotypes in each environment. The apparent conflict between direct and indirect selection for each of these traits may mean that tradeoffs among traits prevent a closer match between phenotype and environment. In contrast, there is strong evidence that plasticity in timing of emergence is not adaptive. Emergence date shows considerable plasticity in response to both leaf litter and light (Figure 11a), but plasticity is nearly always opposite to the direction favored by selection. Except on the edge, plants emerged earlier in the presence of leaf litter, but selection in these environments favored later emergence. In the absence of litter, plants in high light emerged late, but selection favored early emergence. Positive directional selection on vegetative biomass and reproductive investment across all environments (Figure 14) combined with the intermediate maxima seen for these traits (Figure 11) suggests that plasticity in these traits is not adaptive. Because both the low light and full 165 sun environments increase mortality and reduce fecundity, these environments can be characterized as stressful for this population (Hoffman and Parsons 1991, Bennington and McGraw 1995). A weakness of all the approaches used in this study of natural selection is that they are based upon correlations among traits (Mitchell-Olds and Shaw 1987, Wade and Kalisz 1990, Brodie et al. 1995). These relationships may be due to causal relationships among traits and/or between traits and fitness. They may also be caused by selection on phenotypically correlated, but unmeasured traits, or by environmentally induced covariance between traits and fitness (Rausher 1992). Selection on specific leaf area is a good example. Although specific leaf area reflects changes in leaves that occur in response to light availability, the actual traits under selection differ in different light environments. The large thin leaves produced in shady environments where photons are scarce and moisture more abundant probably maximize light harvesting ability (Sultan and Bazzaz 1993). In contrast, smaller thicker leaves produced in more illuminated, drier environments probably maximize water use efficiency (Dudley 1996). F inc-scale variation in water availability in shady environments could result in the positive selection on specific leaf area if plants in moist patches produced larger, thinner leaves and bigger, more fecund plants. Similarly, fine-scale patchy distribution of light due to variation in the density of competitors could be responsible for negative selection on specific leaf area in the fill] sun if plants growing in more illuminated patches produced thicker leaves and had higher fitness. These limitations are one reason why experimental manipulations of phenotypes to address questions regarding causal agents of selection and adaptive responses are so 166 exciting. Still, there are limitations to phenotypic manipulations. Non-lethal mutations and hormonal manipulations can have far reaching pleiotropic effects on the phenotype unrelated to the focal traits (Ketterson and Nolan 1999, Preziosi et al. 1999, Purrington and Bergelson 1999, Tatar 1999). Additionally, they can produce phenotypes outside the natural range, raising questions about their relevance to evolutionary processes in natural populations. Consequently, more studies of the multivariate phenotype in the context of environmental manipulations within natural populations will also be very valuable (Schmitt 1999). 167 LITERATURE CITED Agrawal, A. A., S. Y. Strauss, and M. J. Stout. 1999. Costs of induced responses and tolerance to herbivory in male and female fitness components of wild radish. Evolution 53:1093-1104. Allison, P. D. 1995. Survival analysis using the SAS system: a practical guide. SAS Institute Inc., Cary, NC. Arnold, S. J ., and M. J. Wade. 1984a. On the measurement of natural and sexual selection: theory. Evolution 38:709-719. Arnold, S. J ., and M. J. Wade. 1984b. On the measurement of natural and sexual selection: applications. Evolution 38:720-734. Baldwin, 1. T. 1998. J asmonate-induced responses are costly but benefit plants under attack in native p0pulations. Proceedings of the National Academy of Sciences of the USA 95:8113-8118. Baskin, J. M., and C. C. Baskin. 1983. Germination ecology of Collinsia verna, a winter annual of rich deciduous woodlands. Bulletin of the Torrey Botanical Club 110:311-315. Bennington, C. C., and J. B. McGraw. 1995. Natural selection and ecotypic differentiation in Impatiens pallida. Ecological Monographs 65:303-323. Bergelson, J. 1990. Life after death: site pre-emption by the remains of Poa annua. Ecology 71 :2157-2165. Biere, A.l991. Parental effects in Lychnisflos-cuculi. II. Selection on the time of emergence and seedling performance in the field. Journal of Evolutionary Biology 31467-486. Bradshaw, A. D. 1965. Evolutionary significance of phenotypic plasticity in plants. Advances in Genetics 13 :1 15-155. Brassard, J. T. and D. J. Schoen. 1990. Analysis of phenotypic selection among location in Impatiens pallida and Impatiens capensis. Canadian Journal of Botany 68:1098-1105. Brodie III, E. D., A. J. Moore, and F. J. Janzen.1995. Visualizing and quantifying natural selection. Trends in Ecology and Evolution 5:313-318. 168 Callahan, H. S., C. L. Wells, and M. Pi gliucci. 1999. Li ght-sensitive plasticity genes in Arabidopsis thaliana: Mutant analysis and ecological genetics. Evolutionary Ecology Research 1:731-751. Carson, W. P., and C. J. Peterson. 1990. The role of litter in an old—field community - impact of litter quality in differetn seasons on plant-species richness and abundance. Oecologia 85:8-13. Cipollini, D. E., and J. C. Schultz. 1999. Exploring cost constraints on stem elongation in plants using phenotypic manipulation. American Naturalist 153:236-242. Conner, J. K. 1996. Understanding natural selection: an approach integrating selection gradients, multiplicative fitness components, and path analysis. Ethology, Ecology and Evolution 82387-397. Conner, J. K., S. Rush, and P. Jennetten. 1996. Measurrnents of natural selection on floral traits in wild radish (Raphanus raphanistrum). I. Selection through lifetime female fitness. Evolution 50:1127-1 136. Dixon, P. M. 1993. The bootstrap and jacknife: describing the precision of ecological indices. Pages 290-319 in S. M. Scheiner and J. Gurevitch, editors. Design and analysis of ecological experiments. Chapman and Hall, New York, NY. Dudley, S. A. 1996. Differing selection on plant physiological traits in response to environmental water availability: A test of adaptive hypotheses. Evolution 50:92- 102. Dudley, S. A., and J. Schmitt. 1996. Testing the adaptive hypothesis: Density-dependent selection on manipulated stem length in Impatiens capensis. American Naturalist 147:445-465. Dudley, S. A., and J. Schmitt. 1995. Genetic differentiation in morphological responses to simulated foliage shade between p0pulations of Impatiens capensis from open and woodland sites. Emotional Ecology 92655-666. Facelli, J. M. 1994. Multiple indirect effects of plant litter affect the establishment of woody seedlings in old fields. Ecology 75:1727-1735. Facelli, J. M., and W. P. Carson. 1991. Heterogeneity of plant litter accumulation in successional communities. Bulletin of the Torrey Botanical Club 118:62-66. Facelli, J. M., and S. T. A. Pickett. 1991a. Plant litter - Its dynamics and effects on plant community structure. Botanical Review 57:1-32. Facelli, J. M., and S. T. A. Pickett. 1991b. Plant litter - Light interception and effects on 169 an old-field plant community. Ecology 72: 1024-103 1. Foster, B. L., and K. L. Gross. 1998. Species richness in a successional grassland: Affects of nitrocen enrichment and plant litter. Ecology 79:2593-2602. Foster, B. L., and K. L. Gross. 1997. Partitioning the effects of plant biomass and litter on Andropogon gerardi in old-field vegetation. Ecology 78:2091-2104. Fowler, N. L. 1986. Microsite requirements for germination and establishment of three grass species. American Midland Naturalist 115:131-145. Frankland, J. C., J. D. Ovington, and C. Macrae. 1963. Spatial and seasonal variations in soil, litter, and ground vaegetation in some Lake District woodlands. Journal of Ecology 51 :97-1 12. Goldberg, D. E. 1990. Components of resource competition in plant communities. Pages 27-50 in J. Grace and D. Tilman, editors. Perspectives on plant competition. Academic Press, San Diego, CA. Goldberg, D. E., and P. A. Werner. 1983. The effects of size of opening in vegetation and litter cover on seedling establishment of golderrods (Solidago spp). Oecologia 60: 149-155. Grace, J. B. and B. H. Pugesek. 1998. On the use of path analysis and related procedures for the investigation of ecological problems. American Naturalist 152:151-159. Gross, J ., B.C. Husband, and S. C. Stewart. 1998. Phenotypic selection in a natural population of Impatiens pallida Nutt. (Balsaminaceae). Journal of Evolutionary Biology 11:589-609. Hamrick, J. L., and J. M. Lee. 1987. Effect of soil surface-topography and litter cover on the germination, survival, and growth of musk thistle (Carduus nutans). American Journal of Botany 74:451-457. Hatcher, L. A. 1994. A step-by-step approach to using the SAS system for factor analysis and structural equation modeling. SAS Institute Inc.,Cary, NC. Hedrick, P. W. 1986. Genetic polymorphism in heterogeneous environments: A decade later. Annual Review of Ecology and Systematics 172535-566. Hoffrnann, A. A., and P. A. Parsons. 1991. Evolutionary genetics and environmental stress. Oxford University Press, Oxford. J anzen, F. J ., and H. S. Stern. 1998. Logistic regression for empirical studies of multivariate selection. Evolution 52:1564-1571. 170 Johnston, M. O. 1992. Effects of cross and self-fertilization on progeny fitness in LobeIia cardinalis and L. siphilitica. Evolution 46:688-702. Jordan, N. 1991. Multivariate analysis of selection in experimental populations derived fi'om hybridization of two ecotypes of the annual plant Diodia teres W. (Rubiaceae). Evolution 45: 1760-1772. Kalisz, S. 1991. Experimental determination of seed bank age structure in the winter annual Collinsia verna. Ecology 72:575-585. Kalisz, S. 1989. Fitness consequences of mating system, seed weight, and emergence date in a winter annual, Collinsia verna. Evolution 43:1263-1272. Kalisz, S. 1986. Variable selection on the timing of emergence in Collinsia verna (Scrophulariaceae). Evolution 402479-491. Kalsiz, S., D. Vogler, B. Fails, M. Finer, E. Sheppard, T. Herman, and R. Gonzales. 1999. The mechanism of delayed selfing in Collinsia verna (Scrophulariaceae). American Journal of Botany 86: 1239-1247. Kelly, C. A. 1992. Spatial and temporal variation in selection on correlated life-history traits and plant size in Chamaecristafasiculata. Evolution 46:1658-1673. Ketterson, E. D., and V. Nolan. 1999. Adaptation, exaptation, and constraint: a hormonal perspective. American Naturalist 154:S4-S25. Kingsolver, J. G., and D. W. Schemske. 1991. Path analyses of selection. Trends in Ecology and Evolution 6:276-280. Koenig, W. D., S. S. Albano, and J. L. Dickinson. 1991. A comparison of methods to partition selection acting via components of fitness: do larger male bullfrogs have greater hatching success? Journal of Evolutionary Biology 4:309-320. Lande, R., and S. J. Arnold. 1983. The measurement of selection on correlated characters. Evolution 37:1210—1226. Lechowicz, M. J ., and P. A. Blais. 1988. Assessing the contributions of multiple interacting traits to plant reproductive success: environmental dependence. Journal of Evolutionary Biology 1:25 5-273. Lotz, L. A. P. 1990. The relation between age and size at first flowering of Plantago major in various habitats. Journal of Ecology 78:757-771. Lynch, M., and S. J. Arnold. 1988. The measurement of selection on size and growth. Pages 47-59 in B. Ebenman and L. Persson, editors. Size-Structured Populations. 171 Springer-Verlag, Berlin. Mauricio, R., and M. D. Rausher. 1997. Experimental manipulation of putative selective agents provides evidence for the role of natural enemies in the evolution of plant defense. Evolution 51 21435-1444. Mazer, S. J. 1987 . The quantitative genetics of life history and fitness components in Raphanus raphanistrum L. (Brassicaceae): ecological and evolutionary consequences of seed-weight variation. American Naturalist 1302891-914. Miller, T. E. 1987. Effects of emergence time on survival and growth in an early old-field plant community. Oecologia 72:272-278. Mitchell, R. J. 1992. Testing evolutionary and ecological hypotheses using path analysis and structural equation modeling. Functional Ecology 62123-129. Mitchell-Olds, T., and J. Bergelson. 1990. Statistical genetics of an annual plant, Impatiens capensis. II. Natural selection. Genetics 124:417-421. Mitchell-Olds, T., and R. Shaw. 1987. Regression analysis of natural selection: statistical inference and biological interpretation. Evolution 41:1149-1161. Molofsky, J ., and C. K. Augspurger. 1992. The effect of leaf litter on early seedling establishment in a tropical forest. Ecology 73268-77. ' Neter, J ., W. Wasserrnan, and M. H. Kutner. 1985. Applied Linear Statistical Models, Second Edition. Irwin, Homewood, IL. Noreen, E. W. 1989. Computer-intensive methods for testing hypotheses: an introduction. John Wiley and Sons, New York, NY. Ollerton, J ., and A. J. Lack. 1992. Flowering phenology: an example of relaxation of natural selection? Trends in Ecology and Evolution 72274-276. Petit, C., and J. D. Thompson. 1998. Phenotypic selection and population differentiation in relation to habitat heterogeneity in Arrhenatherum elatius (Poaceae). Journal of Ecology 86:829-840. Petraitis, P. S., A. E. Dunham, and P. H. Niewiarowski. 1996. Inferring multiple causality: the limitations of path analysis. Functional Ecology 10:421-431. Pigliucci, M., and J. Schmitt. 1999. Genes affecting phenotypic plasticity in Arabidopsis: pleiotropic effects and reproductive fitness of photomorphogenic mutants. Journal of Evolutionary Biology 122551-562. 172 Preziosi, R. F ., W. E. Snyder, C. P. Grill, and A. J. Moore. 1999. The fitness of manipulating phenotypes: implications for studies of fluctuating asymmetry and multivariate selection. Evolution 53:1312-1318. Purrington, C. B., and J. Bergelson. 1999. Exploring the physiological basis of costs of herbicide resistance in Arabidopsis thaliana. American Naturalist 1542S82-S91 . Rausher, M. D. 1992. The measurement of selection on quantitative traits: biases due to environmental covariances between traits and fitness. Evolution 46:616-626. Roff, D. A. 1997. Evolutionary quantitative genetics. Chapman and Hall, New York, NY. SAS Institute Inc. 1997. SAS/STAT sofiware: changes and enhancements through release 6.12. SAS Institute Inc., Cary, NC. SAS Institute Inc. 1990. SAS Guide to Macro Processing, Version 6, Second Edition. SAS Institute Inc., Cary, NC. SAS Institute Inc. 1989. SAS/STAT User's guide, Version 6, F outh Edition. SAS Institute Inc., Cary, NC. Scheiner, S. M. 1993. Genetics and evolution of phenotypic plasticity. Annual Review of Ecology and Systematics 24235-68. Schlichting, C. D., and M. Pigliucci.l998. Phenotypic evolution: a reaction norm perSpective. Sinauer Associates, Inc., Sunderland, MA. Schmitt, J. 1999. Introduction: experimental approaches to testing adaptation. American Naturalist 154:S1-S3. Schmitt, J. 1997. Is photomorphogenic shade avoidance adaptive? Perspectives from population biology. Plant Cell and Environment 20:826-830. Schmitt, J ., S. A. Dudley, and M. Pi gliucci. 1999. Manipulative approaches to testing adaptive plasticity: phytochrome-mediated shade-avoidance responses in plants. American Naturalist 154:S43-854. Schmitt, J ., A. C. McCormac, and H. Smith. 1995. A test of the adaptive plasticity hypothesis using transgenic and mutant plants disabled in phytochrome-mediated elongation responses to neighbors. American Naturalist 146:937-953. Shipley, B. 1997. Exploratory path analysis with applications in ecology and evolution. American Naturalist 14921113-1138. Stewart, S. C., and D. J. Schoen. 1987 . Pattern of phenotypic viability and fecundity 173 selection in a natural population of Impatiens pallida. Evolution 41 : 1290-1301. Stratton, D. J. 1995. Spatial scale of variation in fitness of Erigeron annuus. American Naturalist 146:608-624. Stratton, D. J. 1992. Life-cycle components of selection in Erigeron annuus: I. Phenotypic selection. Evolution 46:92-106. Sultan S. E., and F. A. Bazzaz. 1993. Phenotypic plasticity in Polygonum persicaria. I. Diversity and uniformity in genotypic response to light. Evolution 47:1009-1031. Sydes, C., and J. P. Grime. 1981. Effects of tree litter on herbaceous vegetation in deciduous woodland. Journal of Ecology 69:237-248. Tatar, M. 1999. Transgenes in the analysis of life span and fitness. American Naturalist 1542867-881. Thiede, D. 1998. Maternal inheritance and its effect on adaptive evolution: A quantitative genetic analysis of maternal effects in a natural plant population. Evolution 52:998-1015. Thiede, D. 1996. The impact of maternal effects on adaptive evolution: combining quantitative genetics and phenotypic selection in a natural plant population. Ph.D. Dissertation. Michigan State University. East Lansing, MI. Tucic, B., V. Tomic, S. Avramov, and D. Pemac. 1998. Testing the adaptive plasticity of Iris Pumila leaf traits to natural light conditions using phenotypic selection analysis. Acta Oecologica 192473-481. van der Toom, J ., and T. L. Pons. 1988. Establishment of Plantago lanceolata L. and Plantago major L. among grass 11. Shade tolerance of seedlings and selection on time of germination. Oecologia 76:341-347. Via, S., and R. Lande. 1985. Genotype-environment interaction and the evolution of phenotypic plasticity. Evolution 392505-522. Wade, M. J ., and S. Kalisz. 1990. The causes of natural selection. Evolution 4421947- 1955. Willms, W. D., S. Smoliak, and A. W. Baily. 1986. Herbage production following litter removal on Alberta native grasslands. Journal of Range Management 39:536-540. Winn, A. A. 1999. Is seasonal variation in leaf traits adaptive for the annual plant Dicerandra linearzfolia? Journal of Evolutionary Biology 12:306-313. 174 Chapter 5 CONCLUSION: ENVIRONMENTAL HETEROGENEITY, PHENOTYPIC PLASTICITY, AND THE MAINTENANCE OF GENETIC VARIATION IN A NATURAL POPULATION Understanding the evolutionary consequences of spatially and temporally variable environments remains a central goal of plant evolutionary ecology. Presently, there is an abundance of theory, but comparatively little empirical data to bear on the outstanding questions. To meet this goal, we must determine the scale, pattern, and predictability of environmental heterogeneity within populations. We must investigate whether environmental heterogeneity results in variable patterns of natural selection. And we must determine the biotic and abiotic causes of natural selection. Variable selection is most interesting when there is genetic variation that can fuel a response to selection. Therefore, information is needed regarding the genetics of responses to variable environments and the genetic interdependence between traits expressed in different environments. Finally, because maternal effects on seed and offspring traits can be environment-dependant and can have a substantial impact on offspring fitness, more study is needed about the importance of these cross-generation genetic and environmental interactions. In this study I have attempted to answer some of these questions in a natural population of Collinsia verna. ENVIRONMENTAL HETEROGENEITY Light availability ranged fiom 25% to 75% of full sun within the population. Natural light environments were correlated across years at a scale appropriate to favor the evolution of plastic maternal effects. 175 PLASTIC MATERNAL EFFECTS There were important individual fitness consequences of traits influenced by maternal genotype and environment, and there were genotype-environment interactions for seed size and dormancy. The surviving offspring of intermediate light mothers consistently produced as many or more seeds than offspring of low and high light mothers in all environments. The results suggest that maternal effects in plants can improve offspring performance in variable environments, but also may constrain offspring performance when mothers are stressed. Genetic variation for plastic maternal effects can be maintained by a heterogeneous and unpredictable selective environment. GENETIC AND ENVIRONMENTAL EFFECTS There was additive genetic variation in at least some environments for germination, emergence date, flowering date, specific leaf area, mainstem length, mean seed mass, and reproductive investment. There was strong evidence for genotype- environment interactions (genetic variation for plasticity) for flowering date, specific leaf area, mainstem length, and reproductive investment. Finally, there were no genotype- environment interactions for survival, vegetative biomass, or seed number suggesting that there were no strong light environment specialists among the genotypes sampled. However, significant maternal effects on vegetative biomass, seed number, and seed mass without additive genetic variation suggested that maternal genotypes may specialize for different reproductive strategies. The results suggest that genotypes may specialize for particular patterns of germination, emergence, and reproductive investment, while they may be adaptively plastic generalists for flowering date and specific leaf area. 176 PHENOTYPIC SELECTION All traits were highly plastic, and selection varied across environments for emergence date, specific leaf area, vegetative biomass, and reproductive investment. The presence of leaf litter reversed the direction of selection on emergence date and increased selection on vegetative biomass. Full sun reversed the direction of selection on specific leaf area, and increased selection on reproductive investment. These differences in patterns of selection provide direct evidence that leaf litter and light availability are selective agents on these traits. Indirect selection on emergence date, flowering date, specific leaf area, and mainstem length was larger in magnitude and in an opposing direction to direct selection. Together, the parts of this study have simultaneously addressed the relationships between patterns of environmental variation, patterns of variation in phenotypic selection, and patterns of genetic variation and phenotypic plasticity within a natural plant population. Genetic variation for emergence date and plastic maternal genetic effects on seed size and dormancy may be maintained by a heterogeneous and unpredictable leaf litter enviromnent. The presence or absence of leaf litter can directly or indirectly alter the direction of selection on these juvenile and maternal traits. In contrast, the plasticity of the light sensitive traits flowering date and specific leaf area appears to be at or near optimal levels. The absence of substantial genotype-environment interactions for either of these traits supports this argument. Reproductive investment was both heritable and under strong directional selection. Significant genotype-environment interactions appear to maintain genetic variation in this trait. 177 FUTURE DIRECTIONS This research has produced several exceptionally rich data sets. Beyond the contents of this dissertation, I have several analyses underway. Moreover, observations during this study and the results above suggest several areas for future study. Ongoing analysis of this data Temporal variation in natural selection-Chapter 4 presents results of phenotypic selection for a single growing season, 1996-97. Temperature and soil moisture data I collected during this field season, and anecdotal observations suggest that this was a benign year for all plants compared to the 1995-96 season. There was ample fall moisture leading to earlier emergence and larger over winter size than the previous year. Spring was warm and sunny compared to the previous year. In addition, the density of plants in my study plots was lower in this second year. Together, these factors resulted in plants that were much larger and more fecund in 1996-97 than those in 1995-96. Plants from 1995-96 remain to be processed so that I can compare selection in the two years. Environmental correlations and bias in selection analysis—The Lande-Arnold multiple regression approach to measuring phenotypic selection (1983) used here in Chapters 2 and 4 can give biased results if environmental factors also contribute to the covariances between traits and fitness (review in Mauricio and Mojonnier 1997). Phenotypic selection analysis in plants may be particularly prone to bias for two reasons. First, plant size varies with local resource conditions, and traits that vary with size or resources may be subject to environmental covariances with fitness. Second, spatial heterogeneity in enviromnental conditions is expected to result in stronger enviromnental covariances in sessile organisms like plants than in more mobile organisms (Mauricio and 178 Mojonnier 1997). Rausher (1992) has developed an approach to selection analysis that can eliminate this bias. Rausher's method is identical to the Lande-Arnold approach, except it uses estimates of breeding values instead of phenotypic values. To be successful, the method requires genetic data for many sires, and the traits of interest must be genetically variable. My 1995-96 data set with 50 sires should be ideal for applying this method. The results can then be compared to the phenotypic selection analysis for natural plants in this year. Genetic correlations between traits-The next major task that needs to be completed is a multivariate analysis of the genetic data, giving unbiased between trait additive genetic correlations. These correlations are of interest because they would suggest additional genetic constraints on the independent evolution of traits. Using the univariate analysis of the genetic data presented in Chapter 3 I calculated additive genetic correlations from the correlations of breeding values. This approach biases the correlation estimates toward zero. However, all traits had significant genetic associations with at least one trait in at least one year, excepting mean seed mass for which only one year of correlations was available due to a lack of genetic variation (Table 15). Flowering date in year one and specific leaf area in year two were independent of other traits. In contrast to the cross environment genetic correlations, strong negative correlations between traits were common, especially in the second year. These negative relationships are expected, generally occurring between timing traits (emergence and flowering) and size related traits. Correlational selection-If traits are genetically correlated, then correlated responses to selection may be a very important force shaping the phenotype. 179 Table 15. Between trait phenotypic and additive genetic correlations. Phenotypic correlations include all data, sample sizes are as indicated in Table 5. Genetic correlations were calculated from BLUP breeding values with environment treated as a fixed effect. Year 1: data from 50 sires from 1995-96. Year 2: data from 12 sires from 1996-97. The sire variance component for mean seed mass in Year 2 was estimated as zero, so no genetic correlations could be calculated. Correlations in bold are significant after a sequential Bonferroni adjustment within years and type of correlation (0t=0.05). #P<0.1, *P<0.05, **P<0.01, ***P<0.001. 180 181 enn~v0 fimm0 nee0~0 «xiv—.0 0 5532.0- ...«enN0 .5— —0- N «reum0 «3,534.0 fee—n.0- eermm0u are0~0 eiemu0n v0.0- 3.00.0 ~ .502: 650$ «nnNNO «ac-90 nntmn0 «¢N~.0 enav~0u necmN0 ere—N0- N *0N0 m00- «en—N0 «*«mN0 ~00- 02200.0 #00 ...w00- — 332 000m ......550 «enmw0 fat-n60 «influ0 rtevm0- £13320 rerbficn N ere—$0 ~00 .:.«wm0 3:590 «enva0- «nevn0- anew—.0 000.0 _ mmoom mo-H 00.0 *m00 3:300 ernmN0 etc-V01 renwv0 .ffam0n N “KNO- v00 ......wm0 22300.0 s.e.-500- «nemN0- enn0N0 .3000- ~ 332 “cm—m 0590- v0.0- *0w0 nee—“M0 flew—.0- ?«rwm0 s.e.-N0... N w ~ .0- 500- **wm0 «£4050 f3m~0- retNNO- 3:00 «23,000- _ 508532 000- :0- 0 ~ 0- m00- 0.22.4320 ...000- 00.0 N neeNm0 w~.0- .Lm0 .3320- _*wN.0- 03380.0 rib—0- _0.0 _ mou< .0qu .mm «nemw0- .3000- 5m0- mv0 50.0 reemm0- teeN~0 N m ~ .0- thN0- :0- NNO- w — .0- 00.0 seeVNO- m00- fl 8mm— .530; .100 ,2. ~00 *mm0 N0- N00- *xmh0- nae-000.. N #mN0- VHO ~0 *mm0 *mm0 0—0- :0- «*«mm0- _ 0N5 .5053 ...-N00- :50- N; .0- *2 m0 N— .0- 322.0 3.050.. N NNO ~00 0~0 —.0- fimN0- w00 00.0 nrn000- ~ Baa .vmbam .502: 822 mvoom 822 Ewen-H m8< EBA 2mm 35 2mm 50> 23,—. Baum team :82 we: :83 805522 2.28% 3303 c353 owaoEm . . .. .- .2 2an ”v. 11522. 1 .‘ 511621.»- l‘a‘fi l 3.11 3w- \ v I ...-r‘" H \ 11:1 ' a | l In the current study, preliminary genetic correlation data (Table 15) suggests that selection for greater biomass will also select for earlier emergence, earlier flowering, lower specific leaf area, and longer mainstems. These results are in accord with the results of the path analysis (Figure 16). Selection for greater reproductive investment may also select for earlier emergence, earlier flowering, and lower specific leaf area, but shorter mainstem length (opposing indirect selection through biomass). Given that many genetic correlations are small, it seems possible that strong indirect selection in one direction may be balanced by weaker direct selection in the opposite direction resulting in relative stasis in the traits. Vegetative biomass and reproductive investment are unique among the traits in that total selection is smaller than direct selection (Chapter 4 Figures 14e-l). This may be due to negative indirect selection through each other (Chapter 4, Figure 16). However, vegetative biomass and reproductive investment appear to have little direct genetic relationship (Table 15), so the tradeoff between these traits seen in the path analysis may not be evolutionarily important. The genetic correlations between emergence date and other traits are most interesting. Emergence date was subject to variable selection, it was the most consistently heritable trait in the second year, and the plasticity of this trait appeared to be maladaptive. These observations may be causally related: unpredictable selection on emergence date by leaf liter could maintain genetic variation in the trait. Moreover, variable selection on emergence date may indirectly maintain genetic variation in genetically correlated traits (seed mass, winter size, flowering date, and reproductive investment; Table 15). Interestingly, the three traits with the strongest genetic correlations with emergence date (flowering date, mainstem length, and reproductive 182 investment) have more genetic variation than the two traits with low genetic correlations with emergence date (specific leaf area and vegetative biomass). Consequently, variable selection on emergence date may be maintaining genetic variation in other traits through correlated responses. Constancy of genetic parameters-Changes in quantitative genetic parameters across environments may affect predictions about the evolution of phenotypic plasticity and the maintenance of genetic variation (e. g. Via and Lande 1985, 1987, Mitchell-Olds 1992). Moreover, a basic assumption of quantitative genetic models for predicting evolutionary change is that the additive genetic variance-covariance matrix (G), is constant (Lande 1979). Several recent studies of natural plant populations demonstrate environment-dependence in quantitative genetic parameters (Mazer and Schick 1991, Shaw and Platenkamp 1993, Anderson and Shaw 1994, Shaw et al. 1995, Bennington and McGraw 1996). Phillips and Arnold (1999) pr0pose the use of common principle component (CPC) analysis (Flury 1988) for comparing the structure of genetic covariance matrices. The technique represents a powerful new approach to this question, allowing the testing of a hierarchy of hypotheses from unrelatedness to shared principle components, to proportionality, to matrix equality (Arnold and Phillips 1999). I used CPC software (Phillips 1998) to compare pairs of additive genetic covariance matrices across environments and across the two years of this study. Additive genetic covariance matrices were calculated for each environment and each year as the covariance of the BLUP breeding values. As with the calculation of genetic correlations by this method, sampling error can cause covariance components to be significantly underestimated. 183 Because sampling error may change across environments, this could contribute to differences between matrices. However, the traits that differ most between matrices are emergence date and winter size. These traits have the largest sample sizes in the data sets, and consequently are least subject to underestimation due to sampling error. The number of traits that could be included in the covariance matrix varied from three to seven depending on the number of nonzero additive genetic variance component estimates for the pair of environments or years. I omitted specific leaf area from all matrices because the extreme size of its (co)variances created problems for statistical comparisons between common principle component models. Based on these analyses, the G matrices for the two years in this study and for most environment pairs are unrelated, lacking even a single shared principle component (Table 16). The one exception was the comparison between forest and edge environments in year two, where one test suggested unrelated structure, while the other suggested equality. The low sample size in this year (12 sires) provides little power to compare different models of matrix relatedness. Together, these results suggest that the genetic relationships between traits may change dramatically across different resource environments or between years. A labile G matrix would greatly complicate efforts to make evolutionary predictions based on environment dependent selection. An additional question of interest to quantitative geneticists is whether phenotypic correlations are reasonable estimates of genetic correlations (Cheverud 1988, Rolf 1995, 1996, 1997, Waitt and Levin 1998). Estimates of phenotypic correlations are significantly easier to obtain and are much more precise than genetic correlations. The patterns in this data set match those of these previously published reviews (Table 15). 184 Table 16. Comparison of genetic covariance matrices using common principle components analysis. Only matrices with at least three traits in common were compared. Traits: emergence date (ed), winter size (ws), flowering date (fd), mainstem length (ms), vegetative biomass (vb), seeds (sd), mean seed mass (sm), reproductive investment (1i). G Matrix 1 G Matrix 2 Year Environment Year Environment Traits in Matrices Best Model 1 --- 2 ---- ed, ws, fd, ms, vb, sd, ri Unrelated 1 Medium Light 1 High Light fd, ms, vb, sd, ri Unrelated 2 Low Light 2 Edge ed, ws, fd Unrelated 2 Medium Light 2 Edge ed, ws, ri Unrelated 2 High Light 2 Edge ws, fd, sm, ri Unrelated 2 Forest 2 Edge ed, fd, ms, ri Equal/Unrelated 1 Medium Light 2 High Light ws, fd, ri Unrelated 1 Medium Light 2 Forest fd, ms, ri Unrelated 1 Medium Light 2 Edge ws, fd, ms, sd, ri Unrelated 1 High Light 2 High Light fd, sm, ri Unrelated 1 High Light 2 Edge fd, ms, sd, sm, ri Unrelated 185 Phenotypic and genetic correlations are generally of the same sign, while genetic correlations are often larger in magnitude (in spite of their tendency to be underestimated by the methods used here). Phenotypic and genetic correlations are significantly correlated in each year (Pearson correlations: r=0.77, P<0.0001 in year 1; r=0.67, P<0.0001 in year 2). Reviews have found more congruence between phenotypic and genetic correlations for morphological traits than for life-history traits (Roff 1995, Simons and Roff 1996). There was no evidence for this pattern in this data set. To address this question, log plant mass and log seeds were selected as major fitness components. The relationships between phenotypic and genetic correlations involving at least one of these traits were little different fiom the whole data set (Pearson correlations: r=0.81, P=0.0003 in year 1; r=0.82, P=0.0006 in year 2). Although these similarities between phenotypic and genetic correlations are striking, there are convincing reasons why phenotypic correlations should not be used to predict phenotypic evolution (Willis et al. 1991). Moreover, although the phenotypic correlations were very similar across years (r=0.81, P<0.0001, n=36), the results of the CPC analysis show a significant change in the genetic architecture of these traits across years. It has also been argued that due to antagonistic pleiotropy the genetic correlations between major fitness components and other traits will be negative more often than genetic correlations between non fitness components (Roff 1996). This pattern is not supported in this study (Table 15). In the two years, 27% and 54% of the genetic correlations including log plant mass or log seeds were negative, while 57% and 67% of the genetic correlations between other traits were. 186 Costs of plasticity-The ability to respond adaptively to environmental variation may be costly (review in DeWitt et al. 1998). Van Tienderen (1991) proposed a method of measuring these costs that combines genetic data and phenotypic selection analysis in a way that is similar to Rausher's (1992) technique for reducing bias in the measurement of selection. Recent applications of this technique in snails (DeWitt 1998) and Daphnia (Scheiner and Berrigan 1998) have found little evidence of costs. However, a study in Iris (Tucic et al. 1998) found evidence for a fitness cost of producing plastic change in leaf length. My genetic data sets are ideal for the application of this method. Inbreeding depression in variable environments-There is considerable interest in plant mating system evolution. Models focusing on the role of inbreeding depression suggest mixed mating should be rare (reviews: Lande and Schemske 1985, Charlesworth and Charlesworth 1987, Uyenoyama et al. 1993), but other models suggest that mixed mating is an evolutionary stable strategy when pollinator service is unpredictable (Lloyd 1979, Schoen and Brown 1991, Sakai 1995). Species like Collinsia verna with mixed mating systems are ideal for tests of the theory (Kalisz et al.1999). The frequency of self fertilization and the expression of inbreeding depression could both be environment- dependent. The results of chapter one are consistent with this idea: delayed flowering in low light mothers may have resulted in more inbred offspring. These offspring performed as well as offspring of other mothers in the intermediate light environment, but their seed production was reduced in the extreme environments (Chapter 2, Figure 6c). In 1996-97 I planted all the self fertilized offspring of each sire. These offspring germinated at the same rate as their outcrossed half-sibs in all environments. Selfed and outcrossed progeny did not differ in timing of emergence, survival, or specific leaf area in 187 any environment. However, in each environment, selfed offspring were smaller at overwintering, flowered later, were smaller at maturity, and produced fewer, smaller seeds. The coefficient of inbreeding depression (5 = 1 - wsclfed / Woutcrosscd) was modest in the natural forest (0.28) and edge (0.23) environments, but was more substantial in the manipulated environments (low = 0.45, medium = 0.51, high = 0.41). These higher levels of inbreeding depression in extreme environments are unlikely exert selection against self-fertilization if outcrossing is rare in these environments. Evolutionary demography in variable environments-As part of this research I have collected detailed demographic data for each of the light and leaf litter environments. I hope to apply population dynamic models and explore the impact of variation in these environmental factors on demographic processes. Surprisingly, results fiom Chapter 4 suggest that even in the low light environment many plants can produce enough seeds to guarantee persistence for a few more generations (Chapter 4 Figure 12a, e). Outstanding questions Together, these chapters and ongoing analyses address only some of the factors that will be important in the future evolution of this population. Like much scientific research, field observations during this study and the results suggest at least as many questions as are answered. There are several other areas of potentially fruitfirl research in this population. Heterogeneity of leaf litter and other environmental factors-Leaf litter is clearly an important factor in this population, but the frequency and predictability of different leaf litter environments are unknown. Also unexarnined is variation in selection or genetic parameters associated with other variable aspects of the environment such as 188 moisture, nutrient supply, and interspecific competition. Evolution of plant architecture-Data from this study suggests that there is a simple genetic basis to a major change in plant architecture. Typical plants have paired cotyledons, leaves, and branches, but two mothers had these organs in threes and produced trifoliate offspring. Approximately 0.01% of the plants in the population have the trifoliate phenotype. This phenotype could be advantageous in high light environments where greater leaf area could increase competitive ability and/or additional branches could increase seed production. Physiology of photosynthetic acclimation-My research has used specific leaf area as a measure of all the physiological and morphological changes that plants make to maximize photosynthesis in different light environments. At a physiological level it is well known that acclimation to high or low light alters the light compensation and saturation points, and water use efficiency. It is less well known if there is genetic variation and/or genotype-environment interaction for light compensation and saturation points, and water use efficiency. Further, the degree to which acclimation is a physiological phenomena and thus highly labile, as opposed to a consequence of developmentally fixed morphological changes is little studied in an ecological genetic context. Functional ecology of anthocyanin pigmentation-Anthocyanin pigmentation in Collinsia verna leaves varies between families, between seasons, and across light levels. Studies in other species have shown that anthocyanin production is cued by low temperatures and high light levels. Besides their importance as a pigment in flowers and fruits, anthocyanins have no proven function in plants. There are several hypotheses in 189 the literature: 1. A screen against ultra-violet light damage. 2. A mechanism for elevating leaf temperature. 3. A mechanism confening cold hardiness or freeze tolerance. 4. Defense against herbivory via secondary compounds or camouflage. 5. Aposematic coloration. 6. Part of some physiological mechanism like photosynthesis. 7. An artifact of another physiological process that performs one of these functions, or some other unknown function. These questions should be easy to investigate because the pigments are easily extracted and quantified, and a great deal is known about their biosynthesis. Multilevel selection-There is no active seed dispersal in Collinsia, setting up conditions under which kin selection processes could operate (Thiede 1996). Moreover, in areas of North America uncovered after the last glaciation, a novel mechanism has been introduced that may be intensifying the potential for kin selection. This region has no native earthworms. Alien, midden building earthworms were introduced by European immigrants. Earthworms are long lived (10 years), and their burrows can persist for hundreds of years (Edwards and Bohlen 1996). Earthworms collect leaf litter, twigs, and living plant material for their middens. My research plots each contained dozens of stable earthworm burrows and middens. In the fall of 1995, I observed extraordinarily high densities of seedlings in these middens (3-5/cm2 or 50,000/m2). In May 1996 I observed dozens of dying plants being pulled into these middens. At harvest, some middens contained the flower tags, seeds, and decomposing remains of 10-15 plants collected fiom an area within about 10 cm of the burrows. As a result, worms may be concentrating genetically related seeds in a very small area. Upon germination, the seedlings are likely to compete very intensively with each other. Experiments to investigate if kin selection has reduced the intensity of competition between genetically 190 related individuals would be simple to design. Predicting multivariate evolution-Finally, the synthesis of all of the factors affecting the multivariate evolution of traits and their plasticity in a single evolutionary model has never been attempted. Indeed, it may never be a very rewarding or productive venture given that predictions based on quantitative genetic parameters have only short- terrn, local relevance. However, it should be possible to develop matrix models that incorporate multiple environments, the fi'equencies of those environments, genetic variances, covariances, and selection gradients within each environment, and genetic covariances across environments. Studies like the present one could provide the data necessary to parameterize such models. 191 LITERATURE CITED Andersson, S., and R. G. Shaw. 1994. Phenotypic plasticity in Crepis tectorum (Asteraceae): genetic correlations across light regimens. Heredity 7 221 13-125. Arnold, S. J., and P. C. Phillips. 1999. Comparison of genetic covariance matrices. H. Coastal-inland divergence in the garter snake Thamnophis elegans. Evolution 53:1516-1527. Bennington, C. C., and J. B. McGraw. 1996. Environment-dependence of quantitative genetic parameters in Impatiens pallida. Evolution 50:1083-1097. Charlesworth, B., and D. Charlesworth. 1987. Inbreeding depression and its evolutionary consequences. Annual Review of Ecology and Systematics. 18:237-268. Cheverud, J. M. 1988. A comparison of genetic and phenotypic correlations. Evolution 42:958-968. DeWitt, T. J. 1998. Costs and limits of phenotypic plasticity: Tests with predator-induced morphology and life history in a freshwater snail. Journal of Evolutionary Biology 11:465-480. DeWitt, T. J ., A. Sih, and D. S. Wilson. 1998. Costs and limits of phenotypic plasticity. Trends in Ecology and Evolution 13:77-81. Edwards, C. A. and P. J. Bohlen. 1996. Biology and ecology of earthworms. Chapman and Hall, New York. F lury, B. 1988. Common Principle Components and Related Multivariate Models. Wiley, New York. Lande, R. S. 1979. Quantitative genetic analysis of multivariate evolution applied to brain: body size allometry. Evolution 33:402-416. Lande, R. S., and D. W. Schemske. 1985. The evolution of self fertilization and inbreeding depression in plants. 1. Genetic models. Evolution 39:24-40. Kalisz, S., D. Vogler, B. Fails, M. Finer, E. Sheppard, T. Herman, and R. Gonzales. 1999. The mechanism of delayed selfing in Collinsia verna (Scrophulariaceae). American Journal of Botany 86:1239—1247. Lande, R., and S. J. Arnold. 1983. The measurement of selection on correlated characters. Evolution 37:1210—1226. 192 Lloyd, D. G. 1979. Some reproductive factors affecting the selection of self-fertilization in plants. American Naturalist 113267-79. Mauricio, R. and L. E. Mojonnier. 1997. Reducing bias in the measurement of selection. Trends in Ecology and Evolution 12:433-436. Mazer, S. J ., and C. T. Schick. 1991. Constancy of population parameters for life-history and floral traits in Raphanus sativus L. 2. Effects of planting density on phenotype and heritability estimates. Evolution 4521888-1907. Mitchell-Olds, T. 1992. Does environmental variation maintain genetic variation? A question of scale. Trends in Ecology and Evolution 72397-398. Phillips, P. C. 1998. CPC: common principle components analysis. University of Texas at Arlington. Software available at www.uta.edu/biology/phillips/software. Phillips, P. C., and S. J. Arnold. 1999. Hierarchical comparison of genetic variance- covariance matrices. 1. Using the Flury hierarchy. Evolution 53:1506-1515.. Rausher, M. D. 1992. The measurement of selection on quantitative traits: biases due to environmental covariances between traits and fitness. Evolution 462616-626. Roff, D. A. 1995. The estimation of genetic correlations from phenotypic correlations: a test of Cheverud's conjecture. Heredity 74:481-490. Roff, D. A. 1996. The evolution of genetic correlations: An analysis of patterns. Evolution 50:1392-1403. Roff, D. A. 1997. Evolutionary quantitative genetics. Chapman and Hall, New York. Sakai, S. 1995. Evolutionary stable selfing rates of hermaphroditic plants with competing and delayed selfing modes with allocation to attractive structures. Evolution 49:557-564. Scheiner, S. M. and D. Berrigan. 1998. The genetics of phenotypic plasticity. VIII. The cost of plasticity in Daphnia pulex. Evolution 52:368-378. Schoen, D., and A. D. H. Brown 1991. Whole and part flower self-pollination in Glycine clandestine and G. argyrea and the evolution of autogarny. Evolution 45:1651- 1664. Shaw, R. G., and G. A. J. Platenkamp. 1993. Quantitative genetics of response to competitors in Nemophila menziessi. Evolution 47:801-812. Shaw, R. G., G. A. J. Platenkamp, F. H. Shaw, and R. H. Podolsky. 1995. Quantitative 193 genetics of response to competitors in Nemophila menziessi: A field experiment. Genetics 139:397-406. Simons, A. M., and D. A. Roff. 1996. The effect of a variable environment on the genetic correlation structure in a field cricket. Evolution 502267-275. Thiede, D. 1996. The impact of maternal effects on adaptive evolution: combining quantitative genetics and phenotypic selection in a natural plant population. Ph.D. Dissertation. Michigan State University. East Lansing, Michigan, USA. Tucic, B., V. Tomic, S. Avramov, and D. Pemac. 1998. Testing the adaptive plasticity of Iris Pumila leaf traits to natural light conditions using phenotypic selection analysis. Acta Oecologica 192473-481. Uyenoyama, M. K., K. E. Holsinger, and D. M. Waller.1993. Ecological and genetic factors directing the evolution of self-fertilization. Oxford Surveys in Evolutionary Biology 92327-381. van Tienderen, P. H. 1991. Evolution of generalists and specialists in spatially heterogeneous environments. Evolution 45:1317-1331. Via, S., and R. Lande. 1985. Genotype-environment interaction and the evolution of phenotypic plasticity. Evolution 39:505-522. Via, S., and R. Lande. 1987. Evolution of genetic variability in a spatially heterogeneous environment: effects of genotype-environment interaction. Genetical Research 49: 147-1 56. Waitt, D. E., and D. A. Levin. 1998. Genetic and phenotypic correlations in plants: a botanical test of Cheverud's conjecture. Heredity 80:310-319. Willis, J. H., J. A. Coyne, and M. Kirkpatrick. 1991. Can one predict the evolution of quantitative characters without genetics? Evolution 45:441-444. 194 HICH RN S 16 TQTE UNIV. LIBRRRIES 1 |MINIIllIllIllNHIIIUIIINIIIHIIMNIIHIIHIIIIIWI 1293020483990 3