.- .s
.l‘

192..
5.1

.5:

 

lllllllllllllllllllllllllllllllllllllllllllllllllllllllll

31293 01019 095

This is to certify that the
thesis entitled

CLADISTIC AND PHENETIC RELATIONSHIPS WITHIN CYPRIPEDIUM
(ORCHIDACEAE) INFERRED FROM FLORAL FRAGRANCE-COMPOUND

DATA
presented by

Todd James Barkman

has been accepted towards fulfillment
of the requirements for

M. S . degree in Botany

/€flfl&w~/

Major professor
Date 117% /7 73
/ I I

0.7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

 

 

LIBRARY
Michigan State
University

 

 

 

PLACE ll RETURN BOX to remove this checkout from your "cord. ‘
TO AVOID FINES mum on or before data duo.

|[—:L__"J
:lr— H

M
fil—jl—j

MSU lsAn Affirmative M0015“ Opportunity lm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

CLADISTIC AND PHENETIC RELATIONSHIPS WITHIN CYPRIPEDIUM
(ORCHIDACEAE) INFERRED FROM FLORAL FRAGRANCE-COMPOUND
DATA

By

Todd james Barkman

A THESIS

Submitted to
Michigan State University" |
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Botany and Plant Pathology

1993

ABSTRACT

CLADISTIC AND PHENETIC RELATIONSHIPS \NITHIN CYPRIPEDIUM
(ORCHIDACEAE) INFERRED FROM FLORAL FRAGRANCE-COMPOUND
DATA

By

Toddjames Barkman

The qualitative fragrance composition of eight species of CWpedium was
compared using phenetic and cladistic techniques to propose relationships within
the genus. The clustering technique used in the phenetic analysis is an effective
way to objectively assess the similarity in floral fragrance composition between
several species, however, phylogenetic inference is limited by problems in
accurately coding the chemical data. In contrast, the cladistic analysis may provide
accurate estimates of phylogeny as the results are corroborated by several recent
independent studies. Characters used in the cladistic analyses were delimited into
states which represented enzymatic steps in proposed biogenetic pathways.
Phylogenetic conclusions include: retention of C. an'etinum within the genus,
monophyly of the C. calceolus complex with a possible inclusion of C. macranthum,
and a lack of resolution for the relative positions of C. acaule, C. guttatum, and C.
reginae. Evolutionary polarity within the genus could not be determined using this

data set.

ACKNOWLEDGMENTS

I am appreciative of financial assistance granted by the Mid-America Orchid
Congress, Education and Research Committee and the William T. Gillis Memorial
Fund.

I thank Dr.John H. Beaman who encouraged me as an undergraduate to
pursue a graduate degree at Michigan State University. His advice and patience
throughout my academic program have helped in my development as a scientist.

I sincerely thank Dr. Douglas Gage, who has guided me in the analytical
chemistry techniques used during my undergraduate and graduate career. The
time and help he has given me is invaluable and this project could not have been
completed without his guidance.

I offer thanks to Dr. Ralph Taggart who served as a member of my
committee and has offered valuable insights to me throughout my career at
Michigan State University. I thank Mr. Fred Case and Mrs. Roberta Case for access
to their cultivated temperate orchids, including species which could not have been
sampled otherwise. Many thanks go to the members of the MSU-NIH Mass
Spectrometry facility for technical and moral support, includingjennifer Johnson,
Kate Noon, Melinda Beming, and Bev Chamberlain. I offer thanks to Mr. Tom
Kalina for encouraging my pursuit of this project.

Finally, I thank Mr. jack Brechtelsbauer for introducing me to the

Orchidaceae and sharing my excitement in this project.

III

TABLE OF CONTENTS

List of tables .

List of figures
Part 1 .
Part 2 .

Part 1. Phenetic relationships within Cypripedium inferred from
floral fragrance-compound data

Introduction . .

Materials 8c Methods

Results

Discussion

Conclusions .

Part 2. Cladistic relationships within Cypn'pedz'um inferred from
floral fragrance-compound data

Introduction . .

Materials 8c Methods

Results

Discussion

Conclusions .

Bibliography .

Page

19
27
35

36
41
55
74
80

81

Table

LIST OF TABLES

Page

Taxa sampled within Cypripedium with compounds listed as

percentages of total composition. . 20

Individual clones sampled from two poulations of CWpedium calceolus

var. pamz'flomm. 22
Jaccard similarity coefficients calculated using coding method 1. 24
Jaccard similarity coefficients calculated using coding method 2. 26

LIST OF FIGURES; PART 1

Page
Character coding for phenetic analyses. . . . . 11
Character matrix with compounds coded according to biogenetic
relatedness. . . . . . . . . 14

Relative abundances of linalool, 1 4-d,imethoxy benzene, and 4-methoxy
benzaldehyde in three clones of Cypripedium calceolus var.

pamiflorum. . . . . . . . 23
UPGMA-derived phenogram based upon Jaccard similarity coefficient
using coding method 1.. . . . . 25

UPGMA-derived phenogram based upon jaccard similarity coefficient
using coding method 2.. . . . . 25

UPGMA-derived phenogram based uponjaccard similarity coefficient
using coding method 2. Analysis includes C. calceolus var. pamiflomm
and C. calceolus var. pubescens sampled by Bergstrom et a1. . 25

UPGMA-derived phenogram including individuals of C. calceolus var.
pamzflomm sampled from two populations using coding method 2. 28

Comparison of UPGMA phenograms based upon floral fragrance and
genetic identities. . . . . . 31

VI

LIST OF FIGURES; PART 2

Figure Page
1 Proposed biogenetic routes for volatile fatty acid derivatives. . 44
2 Proposed biogenetic routes for benzoic acid derivatives. . 46
3 Proposed biogenetic routes for monoterpene hydrocarbons. . 48
4 Proposed biogenetic routes for sesquiterpenes. . . . 51
5 User-defined character-state trees. . . . . . 53
6 Data matrix used in exhaustive search using Wagner parsimony. 54
7 Single most parsimonious tree in which Paphiopedilum is held as an
outgroup search constraint. . . . . . 57
8 One of 119 most parsimonious reconstructions for ijn'pedium and
Paphiopedilum with no outgroup specified. . . . 58
9 Ten equally parsimonious reconstructions for Cypripedium. . 59
10 Consensus tree of ten equally parsimonious reconstructions for
Cypripedium. . . . . . . . . 61
11 Character state changes for the cinnamic acid pathway. . 62

12 Character state changes for biogenesis of 1,4—dimethoxy benzene. 63

13 Character state changes for anthranilic acid pathway. . . 64
14 Character state changes for benzoic acid pathway. . . 65
15 Character state changes for biogenesis of monoterpene

hydrocarbons. . . . . . . 66

VII

16

17

18

19

2O

21

Character state changes for production of monoterpene alcohols. 67
Character state changes for biogenesis of sesquiterpenes. . 68

Character state changes for the biogenetic production of aliphatic
alcohols and acetates. . . . . . . 69

Character state changes for the biogenetic pathway leading to
fatty acid ester production. . . . . . . 70

Tree in which Cfim'pedium candidum and C. macmnthum are included
within the C. calceolus complex. . . . . . 72

Comparison of genetic identities and cladograrn produced from floral
fragrance-compound data for Cypripedium . . . 7 3

VIII

PART 1: PHENETIC RELATIONSHIPS WITHIN CYPRH’EDIW INFERRED
FROM FLORAL FRAGRANCE-COMPOUND DATA

INTRODUCTION

Taxonomic history

Cypripedium is a circumboreal genus of slipper orchids consisting of
approximately 40 taxa. It is one of five genera in the subfamily Cypripedioideae,
all of which are easily recognized by a pouch or slipper-like labellum, which is
unique in the Orchidaceae. Cflm'pedium includes several North American taxa
that are not difficult to distinguish in most cases. These include C. acaule, the
mocassin flower, C. reginae, the showy lady's slipper, C. guttatum, the striped lady's
slipper, C. calceolus, the yellow lady's slipper, C. candidum, the small white lady's
slipper, and C. an’etinum, the rams head lady's slipper. C. kentuckiense is a large
yellow lady's slipper and may be difficult to distinguish from C. calceolus in some
cases. Cypripedium macranthum is native to Asia and is very distinct from the N.
American taxa in terms of floral morphology and color.

Cypripedium has been the subject of numerous studies recently, including those
concerned with pollination biology (Newhouse 1976), population genetic
structure (Case 1993), and species introgression (Klier 1992). In addition, several
species have been the focus of floral fragrance studies (Nilsson 1979; Bergstrom et
a1. 1992; Holman 1984). Interest in the volatiles of Cypripedium is not surprising
due to the unusual pollination system of these species, in which floral fragrance
appears to play an important role. The pouch-like labellum of the flower acts as a
passive trap to insects (primarily bees) which may slip or climb into it (Nilsson

1979). Insects not large enough to escape through the primary orifice of the

1

pouch may pass through the rear of the labellum. The bee's exit is enabled by the
presence of light windows (in some cases), false nectar guides, and trichomes
which are used for traction (Newhouse 1976; Stoutamire 1967). The insect visitors
do not collect nectar or any other known reward from these visits (although this
has been a subject of debate; Newhouse 1976). In such non-rewarding pollination
systems, it is surprising that insects would visit flowers. Indeed, Gill (1988) reports
3% pollination success over ten years in a population of C. acaule. Contrasting this
low figure, Nilsson reports 20% success in C. calceolus subsp. calceolus (1979). In
the same paper, Nilsson (1979) discussed the potential importance of fragrance
compounds as attractants in C. calceolus. Among the components found, many
were similar to those produced by pollinating bees of the genus Andrena as nest or
territory markers. The similarity between the fragrance composition of
Cypripedium to flowers of nearby populations of species with rewarding pollination
systems such as Convallan'a (Nilsson 1979) may also serve to attract pollinators.
The role of volatile compounds to successful pollination in other orchid genera,
particularly tropical euglossine-pollinated species and the importance of these
compounds in maintaining a specific flower/ pollinator interaction has been
reported (Williams 8: Whitten 1992). Although high specificity of pollinators to
flowers due to unique fragrance compositions has not been widely reported in
temperate species, it is common in tropical species (Meeuse 8c Morris 1984). The
putative importance of fragrance compounds for modifying insect behavior in
CWpedium provides a temperate model system for the systematic investigation of
species-specific fragrance composition.

In addition to the widespread interest in the pollination system of Cypfipedium,
the genus has received much taxonomic attention in recent years; however,
circumscription of the genus is presently unclear. Atwood (1984) recommended

removal of Cfim'pedium an'etinum, and placement into a monotypic genus

Cn‘osanthas which had been proposed by Rafinesque in 1818. This suggestion was
based upon the unusual spurred labellum, a staminode which resembles the fertile
stamens, and separate lateral sepals. Recently, however, M. Case (1993) has shown
using allozyme variation data that Cypripedium an'etinum has a relatively high
genetic identity (Nei's genetic identity = .285) with C. calceolus and C. candidum.
Other taxa, including C. reg-mac and C. acaule were shown to have a lower identity
of 0.072 to these taxa. Additionally, Albert (pers. comm.) has suggested that the
taxon remain within Cypripedium on the basis of a cladistic analysis utilizing
molecular, anatomical, and morphological characters. While C. an'etinum appears
to belong within the genus, its relationship among the other taxa is largely
unresolved.

The most complex taxonomic problem within the genus involves the yellow
lady's slipper, C. calceolus. At present, four varieties have been recognized for this
taxon. Cypripedium calceolus subsp. calceolus is restricted in distribution to Europe.
Cflm'pedium calceolus var. planipetalum is a dwarfed plant that has the northern-most
distribution of the three North American taxa (Atwood 1984). These two
infraspecific taxa will not be examined in the present study. The distinctions
between C. calceolus var. paroiflomm and C. calceolus var. pubescem are less obvious,
particularly due to high degrees of morphological variation (F. Case 1987).
Significant overlap in geographic and ecological distribution occurs for these two
varieties (F. Case 1987). In addition, natural hybridization occurs, resulting in
swarms of individuals not easily categorized as either taxon (personal observation).
Recently, on the basis of allozyme variation, M. Case (1993) has recommended the
retention of C. calceolus var. pamiflorum and C. calceolus var. pubescens within C.
calceolus, because genetic divergence was low among the varieties. Few workers
have suggested a specific status for any of these varieties on the basis of

morphological or allozyme allelic differences. Recently, however, Bergstrom et a1.

(1992) have published fragrance chemical data which support the separation of
C. calceolus subsp. calceolus, C. calceolus var. [2011an and C. calceolus var. pubescens
into three species. Atwood (1984) also recognized two species including C.
pubescens (formerly recognized as C. calceolus var. pubescem) and C. pamiflomm
(formerly C. calceolus var. pamzflorum). A third taxon similar to C. pubescms in
morphology, C. kentuckiense, is recognized as a distinct species on the basis of its
geographic distribution and relatively larger size (Atwood 1984). Gradation in
floral morphology and color between these varieties has complicated the
taxonomic distinctions proposed.

Few relationships have been proposed among the morphologically distinct
taxa C. acaule, C. reginae, C. guttatum and C. macranthum. In contrast, the C.
calceolus complex (Atwood 1984) consists of several morphologically similar taxa,
including C. candidum, C. calceolus, and C. kentuckiense. Case (1993) has reported
relatively high genetic identities for these taxa. Although genetic identities were
not calculated between C. kentuckiense and the others within the complex due to a
small sample size, preliminary study indicates a relatively high identity also to C.

calceolus.

Use of fragrance in taxonomy

Floral fragrance compounds have been used a few times previously to help
resolve taxonomic problems in the Orchidaceae. Gregg (1986) and Borg-Karlson
(1990) suggested relationships within complex species groups (Cycnoches and
Ophrys, repectively) based upon evaluation of the patterns of fragrance
composition. Qualitative and quantitative distinctions in fragrance composition
between putative Species corresponded to pollinator differences and floral
morphological distinctions in most cases. The fragrance composition of these

species appears to be very important in modifying insect behavior, as Ophrys

exhibits a pseudocopulatory pollination system in which mimic pheromones are
produced (Borg-Karlson 1990), and Cycnoches is pollinated by Euglossine bees
which collect the fragrance components produced (Gregg 1983). Due to the
influence of quantitative and qualitative variation in fragrance chemistry on
pollinator specificity, the differences were considered useful for taxonomic

distinctions within Cycnochas and Ophrys.

Previous fragrance work in Cypripedium

Cypripedium has been the object of several floral fragrance investigations.
Bergstrom et a1. (1992) sampled floral fragrance compound variation among C.
calceolus subsp. calceolus, C. calceolus var. pannflomm, and C. calceolus var. pubescens.
Cypripdium calceolus subsp. calceolus has a substantially distinct qualitative fragrance
profile from the other two varieties. Cypripedium calceolus var. [Jami/10mm and C.
calceolus var. pubescens differ primarily in relative quantities of many shared
compounds. Holman (1973) identified volatile compounds produced by various
organs in C. reginae, C. acaule, C. calceolus, and C. candidum The results of his study
illustrate that very different volatiles can be produced from various floral and
vegetative organs, with the staminode producing the greatest quantity of volatiles
in the flowers of these species. Significant overlap in fragrance components was
described between these taxa, but the taxonomic importance of the patterns was
not discussed by the author. The previous fragrance work in Cypripedium indicates
that diflerences in profiles exist between species. These compositional differences
may be useful in taxonomic studies, especially due to the importance of the
compounds to the reproductive system of the flowers.

Knowledge of the extent of variation for a potential taxonomic character
within any taxon is necessary for comparative study. Floral fragrance composition

has been shown to vary extensively within a single taxon as Gregg (1986) and

Kaiser (1993) reported qualitative and quantitative fragrance-compound variation
within and between populations of Cycnochas and Gymnadem'a, respectively.
Variation in qualitative fragrance composition between individuals may be
attributable to several factors. Kaiser (1993) demonstrated that floral
development can affect quantitative fragrance composition. In Angmecum
sesquipedale, the degree of maturity of a single flower can affect the fragrance
composition on consecutively sampled nights. Indeed, studies of volatile indole
production in several aroids indicate that the compound is only produced and
volatilized for a single day, after which production is either stopped or it is
converted to non-volatile products (Chen 8: Meeuse 1971). Although it has not
been studied, post-pollination processes may also cause variation in fragrance
production. This would not be surprising as several physiological and chemical
processes such as color changes and senescence of the flower occur after
pollination. I have noted (unpublished results) quantitative differences in the
relative abundances of specific compounds over the life of a flower in
Paphiopedilum delenatii. Finally, the timing of fragrance release can be very specific
for many plant species and appears to be under biorhythmic control (Kaiser
1993). Due to the potential for infraspecific and individual variation in fragrance
composition, particularly quantitative, sampling for taxonomic studies must

attempt to control for this the data are utilized for taxonomic study.

The use of phenetics and fragrance compounds in taxonomy

Explicit phenetic techniques have rarely been utilized in conjunction with
floral fragrance compounds in efforts to determine overall similarity/ dissimilarity
of taxa. This is surprising due to the utility of these methods in taxonomic studies
utilizing morphological as well as chemical information in other groups (Sokal 8c

Sneath 1973; Rodman 1991). Gregg (1986) used ordination methods to

distinguish fragrance chemotypes within a population of Cymoches densz'florum. In
her study principal components analysis (PCA) was used to distinguish between
high and low a—pinene clones of a population. Holman and Hiemermann (1976)
utilized a numerical measure of similarity to compare several species of
Epidendmm.

Despite these early phenetic studies, clustering methods have not been
attempted nor systematically applied in conjunction with fragrance compound
data. The use of fragrance compounds in cluster analyses aimed at proposing
taxonomic relationships may be problematic for two reasons. Many volatile
fragrance compounds are common natural products found in very unrelated
organisms such as gymnosperms, angiosperms and fungi, and may not provide
phylogenetically useful information (Whitten 8: Williams 1992). In addition,
inferring relationships from these analyses may be misleading because
phenograms may reflect pollinator relationships rather than historical
relationships (Whitten 8c Williams 1992). Despite these problems, cluster analysis
may represent an objective method for expressing and evaluating the overall
qualitative chemical similarity between a group of taxa (Sneath 8c Sokal 1973).
The use of these explicit methods is desirable over more subjective methods
involving simple descriptions of compound distributions among taxa.

In the present study nine species of Cypripedium were investigated for floral
fragrance compound composition. The qualitative compound distributions are
analyzed using cluster analysis to examine the taxonomic significance of the
chemical data. Additionally, fifteen individuals from two populations of C.
calceolus var. pann’flomm were sampled for qualitative fragrance-compound
variation. The potential taxonomic importance of relative differences is evaluated.
As fragrance compound data have not been previously utilized in conjunction with

cluster analyses, an examination of the utility and/ or shortcomings of this

technique follows by comparison to results of recent studies utilizing independent

data sets.

Materials and Methods

Plant material

All plants were sampled outdoors. Cypripedium macranthum, C. guttatum, C.
candidum, C. calceolus var. pubescent, and C. kentuckieme were sampled in the garden
of Frederick and Roberta Case (Saginaw Co., MI). The plants were grown under
conditions similar to their natural habitats. Cypripedium calceolus var. pamiflorum, C.
an'etinum, C. acaule, and C. reginae were sampled in natural populations.
Cypripedium. calceolus var. pamiflomm clones were sampled from two populations,
Rose Lake and Williamston, both in Ingham County, MI. Only one individual was
sampled for each taxon except for C. calceolus var. pamflorum, in which 15
individuals were sampled. Voucher specimens are deposited in the Beal-

Darlington Herbarium, Michigan State University (MSC).

Fragrance compound isolation

For field sampling studies, one or two flowers on an intact inflorescence were
enclosed in a Plexiglas® box for approximately one hour prior to sampling. The
fragrance-laden air was drawn through a sampling cartridge using a portable
battery-powered sampling pump. Sorbent tubes containing XAD—2 polymers
(SKC) were used to trap volatile components. External atmospheric samples were
drawn simultaneously during sampling periods as controls. Sampling was initiated
at 9:00 and ended at 18:00 each period with a flow rate of 60 ml/ min. Trapped
fragrance compounds were removed from the sorbent tubes using 1 ml hexane.
The hexane mixture was subsequently concentrated using a slow stream of

compressed N2. The evaporation step was necessary to enable detection of the

small quantities of compounds produced by the taxa studied. Evaporation does
not appear to result in the loss of compounds detectable by our instrument;
sensitivity was only improved by successive concentration steps of dilute samples

(Personal observation).

Fragrance compound identification

Samples were analyzed by gas chromatography-mass spectrometry (GC-MS).
Chromatographic separation of fragrance compounds in the sample eluates were
performed on an HP 5098 using a DB-Wax capillary column (0.32mm id x 30 m,
film thickness 0.25 mm) using a temperature program of 2 minutes at 40° C and a
ramp to 240° C at 7° C/ minute. The final temperature was held for 2 minutes.

The capillary column was directly interfaced into the ion source of an Hewlett-
Packard mass selective detector (MSD). Spectra were acquired from 40 to 300
mass units. The spectra were compared to published spectra available in
computer databases and original literature reports. Compound identification was
based upon the combination of comparisons of relative retention times, Kovats

indices and spectral profiles.

Character coding

Coding of the fragrance compound data is necessary for the calculation of
similarity coefficients to be used in the cluster analysis. Erroneous representation
of the data in character coding could significantly affect numerical analyses. The
use of secondary chemical data in conjunction with clustering techniques is
uncommon in botanical classification, so an explanation follows as to how the data
have been represented. No attempt was made to represent quantitative variation
between taxa in the coding procedure. The quantitative variation which can result
from biological and temporal factors outlined above are difficult to control.

Although quantitative variation may be taxonomically significant in these species,

10

the possibility of variation from sources other than genetic differences between
the individuals sampled justified the exclusion of quantitative data from the
analyses.

Two methods for coding the chemical data were evaluated to examine the
effect this step may have upon the cluster analysis. Figure 1 contrasts the two
coding procedures.

Compounds are coded qualitatively as present or absent using the first coding
method. One shortcoming of this method is that each compound is treated as an
independent character. This may not be a very accurate coding technique
because structurally related compounds such as configurational isomers are not
biogenetically independent. In addition, compounds with the same basic carbon
skeleton but differing only in the nature of the functional group substitution (e.g.
alcohol/ aldehyde of the same skeleton) are scored the same as would be two
compounds from biogenetically distinct classes (e.g. indole/benzaldehyde).
Despite the philosophical inconsistencies of this coding procedure, the data were
scored this way to examine the effects upon the clustering analysis. No data
matrix is presented for this coding procedure because each compound identified
in each species (Table 1) is coded as present or absent.

The second coding method was based on the assumption that the biogenesis
of some volatiles identified are linked. This coding method is similar to that
employed by Seaman and Funk (1983) for sesquiterpene lactone data. In the
second method, isomers, both conformational and optical, were treated as
identical compounds and were coded as a single character. Criteria used to
delimit characters were as follows: fundamental carbon skeletons were coded only
once, present or absent, regardless of the state of associated functional groups.
Thus, the interconversions between aldehydes and associated alcohols of a single

carbon skeleton were not treated as independent characters. In addition, each

11

Method 1: All compounds coded as separate characters.
Benzaldehyde Benzene methanol 4-methoxy benzaldehyde 4-methoxy benzene methanol

‘*’ 3%?

Taxon A 1 l l 1

Taxon B l 1 0 0

Method 2: Only basic skeletal types and substitutions coded.

Benzaldehyde Benzene methanol 4-methoxy benzaldehyde 4-methoxy benzene methanol

. H
‘ > | H
ocrr3 ocn3

Taxon A l 1

Taxon B 1 0

Figure 1. Character coding for phenetic analyses.

12

Four conformational isomers of lilac alcohol

 

a b c d
Taxon A 1 1 l 1 Method 1: All compounds coded as
separate characters.
vs.
a,b,c,d
Taxon A 1 Method 2: Only basic skeleton coded.

Figure 1, cont'd.

13

novel substitution of a fundamental carbon skeleton was coded only once. If a
single taxon produced an array of compounds, all of which were derivatives of a
single carbon skeleton that was not found in any other taxon, then these were all
treated as a single character. This exception is justified because the data are
uninforrnative for comparative purposes. Figure 2 shows the character coding
used in the data matrix based upon the biogenetic assumptions.

The assessment of overall qualitative chemical similarity, which served as the
basis for taxonomic judgement between the taxa in this study, was determined by
cluster analysis. The choice of association measure and clustering algorithm,
including justification for inclusion of each, follows. The use of computer
software facilitated utilization of these techniques. NTSYS version 1.7 (Rohlf
1987) was used for the SAHN (sequential agglomerative hierarchical non-

overlapping) clustering analyses.

Association measure
The determination of overall chemical similarity between taxa was calculated
by usingjaccard's coefficient of similarity. This coefficient was chosen over others

because it is appropriate for use in chemical studies (Sneath 8c Sokal 1973).

jaccardSJ= 2a/a+b+c

Where a,b,c, are represented by:

Taxon B
Taxon A 1 0
1 (1,1) =a (1,0) =b
0 (0,1) = c (0,0) = d

1 = presence of compound(s)
0 = absence of compound(s)

14

EoB< u o 3535 u H 635222 0:05on 9 95:88 cocoa 350950 55 x52: 8.83:0 .N 83E

—o—°°°°°°°

:OT
:8
@

—°°O°—O°°

mmoo

 

Om
In

-‘°—~°°°O°
o-n—o-I—OOOa—o
c—ooooooo

mzo
mmoo

O-OOa—OOOO

>0

00000—090
Odo-000000

555.39.-
‘13:.8
~13:
.18-
3313‘
£1.30
5.33.3
3:»..3-m
saga:

Knit-:1-
‘33:;
can!
~13.
85.20.13
8-530
‘32}...
2331‘
KasoSssl

mmoo
Allvi Alli'.
__ o

O

15

E222 u o 3585 u H 585228 unocowoB o. 95:88 wovoo 3:59:00 55 x32: nonoaano .350 .N oSmE

£32.32:-

0 _ o
_ o o $.55:
o c o ‘95-:
o o o .1.qu
o o _ 8539!:
o o o .5353.
o o o 5:33.!»
o o o 3333...;
o O _ ‘lxoga s1
— Sufi-.33.:
o a o o ‘33::
o o a o «3.32
_ o o o :13.
o o o c 25.32.54
0 o c o ‘15:...
o o o o EahQrg
o o o o 3320‘
o m w o Easel
. mo
1

”I 0 m Me he s

\ o

16

“82?. u o 388$ .1. H 68:83»: ouocowofi 2 95:88 88o 35888 6m? 59:: 88220 6.28 .N 8sz

Eu

00000—00—

o
o
o
o
o
o
o
o
a

m

 

I
fl

:6

a;

OOOOO—OOO

loci-3!:
{23:8
.153;
:33.
is}:
£535
‘33?!»
Fuzz-m
‘EoSst—l

835.333..
I 53.:
can:
.13.
85.2035.
8.53:1
8.5.3.3
3333a
{again

39%

M

JCH
:.

17

E88... u 0 ”E83; u H 826332 ococuwoB 2 9:288 838 messed—=8 53, anE 88220 .Ho.E8 .N 95mm

£32.33...-
825.2:
2.5:.
:33-
2.53925.
{ISUZU
Elmira
Poo—coat
Knaogpsx

OOOO—OOOO
fl—OOOOOOO

O:

.3553!
‘52....
55:.
3....»-
8533!:
Slain.
‘a‘itoo
3.3.3.1
ESE-t

OOOOOOO~O
occao—o—o

i

O

o\

\/\/\/\/\\/Vf\

>
\

18

This coefficient is particularly useful for chemical data because the calculation
of similarity is based only upon the shared presence of compounds between two
taxa. It does not include shared absences of compounds in the calculation as in
the simple matching coefficient of Sokal 8c Michener (1958). In the case of
volatile chemical data or other secondary compounds, the absence of a compound
in two taxa should not represent a similarity between them (Stuessey 1990). This
coefficient is highly desirable because of the variable nature of volatile chemical
data. Further, the absence of a compound in a taxon may not represent the lack
of biogenetic machinery to produce it, simply a lack of detectable amounts. The
calculated coefficients between pairs of taxa are measures of the chemical

similarity between them and are utilized by clustering algorithms.

Clustering algorithms

UPGMA (Unweighted pair group method arithmetic average) was chosen as
the clustering algorithm due to its widespread use in phenetic analyses. Admission
of a taxon into a cluster is based upon a comparison of its similarity to the average
similarity measure of the cluster. Single and complete linkage algorithms are
extreme opposites with respect to requirements for admission into clusters and
were investigated and compared with the results obtained by UPGMA.
Cophenetic values were computed to determine the effectiveness of the clustering
as a representation of the original data. This measure allows an objective choice
among the various clustering method results (Sneath 8c Sokal 1973; Duncan 8c
Baum 1981). Taxonomic judgements were based upon the phenogram derived

from the algorithm which displayed the highest cophenetic value.

RESULTS

The list of compounds identified in all nine taxa is provided in Table 1.
Although quantitative information is not utilized in the cluster analysis, relative
ratios of compounds found in taxa are reported. No two taxa are very similar in
overall fragrance composition. Cypripedium reginae exhibits the least diverse
fragrance composition with only two benzoic acid-derived compounds produced,
representing the expression of a single biogenetic pathway. C. an'etinum produced
only four compounds, again from a single pathway. All other taxa produce various
classes of compounds including mixtures of benzenoids, monoterpenoids, and
fatty acid derivatives. All compounds identified are grouped according to
biogenetic relatedness. Proposed biogenetic groupings follow: Benzenoid types
include phenyl propanoids, anthranilic acid derivatives, benzoic acid derivatives,
1,4 -dimethoxy benzene, and phenyl acetaldehyde derivatives. Fatty acid
derivatives include acetates, ethyl and methyl esters, and aliphatic alcohols with
straight chain hydrocarbon skeletons. Terpenoids found include monoterpene
hydrocarbons, monoterpene alcohols, pyranoid monoterpene alcohols, and
sesquiterpene hydrocarbons and alcohols. Cypripedium acaule produced the largest
number of compounds. The fragrance was dominated by fatty acid esters with
decanoic acid methyl ester the most abundant (34.5%). Decanoic acid ethyl ester
was the second most abundant (18%). Cypripedium kentuckieme produced a
fragrance composition in which fatty acid acetates comprise 70% of the mixture.
The fragrance of C. guttatumwas composed of several fatty acid esters, of which
octanoic acid methyl ester was dominant (45.3%). Benzene methanol was the
second most abundant compound (17.2%). Several mono- and sesquiterpenoids
were produced but were minor components. CWpedium macranthum had a
fragrance composition comprised almost entirely of the pyranoid monoterpenes,

lilac alcohols and lilac aldehydes (83%). The two varieties of C. calceolus produced

19

20

Tablet. Tummwmmw
mumbm award-locum

 

21

four compounds in common: benzaldehyde, benzene methanol, 1,4-dimethoxy
benzene, and dodecanol. Cypripedium calceolus var. pamijlorum had a fragrance
composition dominated by linalool (37.7%), alpha-farnesene (15.9%), and 1,4—
dirnethoxy benzene (15.2%). In contrast, C. calceolus var. pubescms produced a
composition in which phenyl acetaldehyde (26.8%) and phenyl ethanol (63.6%)
dominated. The putatively closely related taxon, C. candidum, shared 4
compounds with C. calceolus var. pannflomm, and 4 compounds with C. calceolus var.
pubescens. It is interesting to note that the composition of C. candidum is
qualitatively intermediate between these two taxa with benzaldehyde comprising
28.5% and phenylacetaldehyde comprising 25.4% of the mixture. In total, 66
compounds were isolated and identified in the nine taxa sampled.

Table 2 lists compounds found in 15 individuals of C. calceolus var. pamiflorum
from two populations, Rose Lake and Williamston. Compounds are listed as
present (1) or absent (blank) for each clone sampled. Presence of only trace
amounts were recorded as (1). In 8 of 15 individuals, dodecanol was produced.
Nine of 15 individuals produce cis-linalool oxyd and indole; 11 of 15 individuals
produced 4—methoxy benzene methanol; and 14 of 15 individuals produced 1,4-
dimethoxy benzene, alpha-farnesene, 4-methoxy benzaldehyde. All individuals
sampled produced linalool and benzaldehyde. Figure 3 shows three
chromatograms among the clones R1, R0, and R4 sampled respectively. The
peaks in the chromatogram representing the presences of linalool, 1,4-dimethoxy
benzene, and 4-methoxy benzaldehyde are designated by arrows. Quantitative
variation of these primary compounds is extensive between all individuals and no
attempt was made to represent this information. Qualitative variation between
individuals is evident upon inspection of table however these variant compounds
are only minor components of the fragrance. Individuals in the table are listed in

chronological order with R0 as the first individual sampled on 5/ 20/ 92. Rl-3

Table 2. Individual clones sampled from two populations
(R=Roee Lake W=Wiiliamaton) oi Cypripedium calceolus
var. perviflorum

Compounds listed in order of
relative retention times

 

limonene

1, benzene
linalool oxide cia or

methanol

benzoate

benzene
methanol

23

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

linalool
4-mcthoxy benzaldehyde
; 1,4-dimethoxy benzene
.EiEBj
GEBE
@53{
:d I}, . MJWJ'M
a... v i I I I F I V ' I ' I ' I I T
J
1.51:7:
.aE7§
5.833:
airmen--. --*~r'i
l
1.5E?j
1.eE?{
5.8E6j
MHHHLH H3
ear-

Figure 3. Relative abundances of linalool, 1,4-dimethoxy benzene, and
4-methoxy benzaldehyde in three clones (R1, R0, and R4 respectively) of
C ypripedium calceolus var. panriflorum. '

24

were sampled on 5/ 22/ 92, R4—R6 were sampled on 5/ 23/ 92, and R7-9 were

sampled 5/ 24/ 92. W1-2 were sampled on 5/ 25/ 92, W3—4 were sampled 5/ 26/ 92

and W5 was sampled 5/ 27/ 92.

Table 3. jaccard similarity coefficients calculated using coding method 1.

par
pub

can

ken
aca
reg
gut

mac

par
1.000
0.158
0.250
0.231
0.100
0.051
0.000
0.037
0.038

pub

1.000
0.333
0.091
0.125
0.000
0.000
0.043
0.000

can

1.000
0.429
0.067
0.000
0.000
0.048
0.000

ari

1.000
0.000
0.000
0.000
0.059
0.000

ken

1.000
0.000
0.000
0.000
0.000

aca

1.000
0.000
0.132
0.200

reg gut mac

1 .000

0.000 1 .000

0.000 0.000 1.000

Figure 4 shows the UPGMA phenogram obtained from similarity coefficients

of Table 3. The phenogram is comprised of two primary clusters. The first

includes taxa traditionally considered part of the C. calceolus complex. In this

cluster, C. arietinum and C. candidum are the most similar based upon fragrance

chemistry. Cypfipedium calceolus var. pannflomm is the next most similar to these

taxa. Cypripedium calceolus var. pubescens and C. kentuckiense are contained within

this cluster but are not very similar to each other or to any other taxon. All of

these taxa are similar in basic floral structure, possessing linear-lanceolate lateral

petals and an inflated labellum with an opening at the top of the pouch.

25
laccard similarity
0.0 .12 .24 .36 .48

4 I

parviflorum
r——candidum
L— arietinum
pubescens
kentucldense
acaule
macranthum
guttatum
reginae

 

 

 

 

 

 

 

 

 

 

 

 

r—IH

 

 

 

 

 

Figure 4. UPGMA-derived phenogram based upon Jaccard similarity coeffi-
cient using coding method 1.

Jaccard similarity
0.0 .16 .32 .48 .64

J

parviflorum
i candidum
I arietinum
pubescens

 

 

 

 

 

 

 

 

 

 

 

 

kentuckiense
F acaule

H l guttatum
macran. thum

reginae

 

 

 

 

 

 

 

 

 

Figure 5. UPGMA-derived phenogram based upon laccard similarity coeffi-
cient using coding method 2.

Jaccard similarity
0.0 .2 .4 .6 .8

 

parviflorum

I parviflorum-B
I pubescens-B

J candidum

arietinum

pubescens

_. kentuckiense

A acaule

1 guttatum
macranthum
reginae

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 6. UPGMA-derived phenogram based upon laccard similarity coeffi-
cient using coding method 2. Analysis includes C. calceolus var.
parviflorum and C. calceolus var. pubescens sampled by Bergstrom et a1.
(1992).

26

Cypfipedium arietinum is an exception, however, with a floral morphology distinct

within the genus. The other cluster is comprised of morphologically dissimilar

taxa with C. acaule and C. macranthum most similar in terms of fragrance chemistry.

Cypripedium guttatum is loosely associated with the cluster and C. reginae is separate

from all other taxa. The cophenetic correlation coefficient (r = 0.92311) for the

UPGMA-derived phenogram was the highest obtained for all of the clustering

algorithms used.

Table 4. jaccard similarity coefficients calculated using coding method 2.

par
pub

can
ken
aca

reg

mac

par
1.000
0.273
0.375
0.286
0.200
0.077
0.000
0.083
0.000

pub

1.000
0.222
0.125
0.200
0.000
0.000
0.083
0.091

can

1.000
0.500
0.125
0.000
0.000
0.111
0.000

ari

1.000
0.000
0.000
0.000
0.143
0.000

ken

1.000
0.000
0.000
0.000
0.111

aca

1.000
0.000
0.182
0.091

reg gut mac
1 .000

0.000 1 .000

0.000 0.000 1.000

Figure 5 represents the results of a cluster analysis using the UPGMA

algorithm for the same group of taxa as in Figure 4 using similarity coefficients

based upon the second coding method. The tree topology is identical with the

exception of C. acaule now represented as more similar to C. guttatum than to C.

macranthum The similarity coefficients are higher between other taxa due to the

27

differences in coding of the data matrix. The cophenetic correlation coefficient is
r = 0.91372 and was higher than that obtained by either single or complete
linkage.

Figure 6 shows the results of an analysis which included the taxa sampled by
Bergstrom et a1. (1992). Cypripedium calceolus var. pann'flomm is represented by
parviflorum-B and C. calceolus var. pubescens is represented by pubescens-B. The
most interesting result obtained in this cluster analysis is the separation of the two
individuals of C. calceolus var. pubescens in the phenogram. The individuals
sampled by Bergstrom et al. (1992) cluster with C. calceolus var. pann'flomm sampled
here. The similarity coefficients were calculated from compounds coded
according to biogenetic relatedness. The cophenetic correlation coefficient is r =
0.94024

Figure 7 shows the results of the cluster analysis including all previously
analyzed taxa with the addition of all individuals of C. calceolus var. pamiflomm
sampled from the Rose Lake (R) and Williamston (W) populations. There does
not appear to be a separation of the two populations from each other. It is clear
that none of the individual clones of C. calceolus var. parviflorum sampled are more
distinct from each other so as to cluster with any other taxon in the study. While
the variation between individual clones appears great from inspection of Table 2,
when compared to the variation among other taxa within the genus, the
differences are not phenetically significant. The cophenetic correlation
coefficient is r = 0.97246.

DISCUSSION

Comparison of coding methods
A comparison of the phenograms based upon the two coding methods

indicates the effect that the coding procedure can have upon the clustering results

28

Jaccard similarity

0.0 0.25 0.5 0.75 1.00

 

 

 

 

 

 

 

 

 

 

 

, parviflorum-B
pubescens-B
R7

R8
ir—: wz
q W1
W3
W4
1 candidum
arietinum
pubescens
kentuckiense
acaule
guttatum

macranthum
re ginae

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

r—II—q

 

 

Figure 7. UPGMA-derived phenogram including individuals of C. calceolus var. parviflorum
sampled from two populations (R=Rose Lake, and W=Williamston) using coding method 2.

29

of an analysis. The two phenograms (Figures 4 and 5) exhibit similar topologies
with a single difference being the relative placement of C. guttatum and C.
macranthum The first coding method, in which isomers were treated as
independent characters, resulted in an artificially high measure of similarity for C.
acaule and C. macranthum These taxa share only a single pyranoid monoterpene
pathway, which produces lilac alcohol and lilac aldehyde. However, the
expression of this pathway results in the production of four isomers for the alcohol
and four isomers for the aldehyde. Coding these eight compounds as
independent characters between C. acaule and C. macranthum resulted in
erroneously high similarity values for these taxa and obscured the similarity
between C. acaule and C. guttatum. The results of the second coding method, in
which isomers were treated as single dependent characters, indicate the similarity
that is present between C. acaule and C. guttatum. These two taxa share several
independent pathways including the fatty acid-ester pathway, and an unidentified
oxygenated monoterpene pathway. The co-occurrence of two independent
pathways between C. guttatum and C. acaule is interpreted as a better indicator of
relationship rather than the co-occurrence of eight isomers, produced by the same

biogenetic pathway.

Taxonomic Implications for Cypripedium arietirmm

Taxonomic relationships, as indicated by the clustering analysis utilizing the
coding method 2, will be discussed and compared with other recent evidence from
independent studies. The segregation of C. arietinum into Criosanthas (Atwood
1984) is not supported by the cluster analysis performed here. Cypripedium
arietinum clusters closely to members of the C. calceolus complex. The clustering of
this taxon with members of the complex is congruent with results obtained by

Case (1993). Figure 8 shows a comparison of the phenograms of Case (1993) and

30

of that obtained in this study. Although the relative position of C. arietinum differs
between the studies, and although the taxon is distinct with respect to its floral
morphological features, many independent data sets including those constructed
from plastid DNA, allozyme variation, and secondary metabolites, support its

inclusion in Cypripedium.

Erroneous similarity between Cypripedium candidum and Cypripedium arietinum
The relatively high similarity displayed between C. arietinum and C. candidum,
shown in the Figure 3 phenogram, requires an explanation, because, on the basis
of floral morphology, these taxa would not be predicted to be more similar than C.
candidum is to C. calceolus var. pamzflorum or C. calceolus var pubescens. The affinity
between C. arietinum and C. candidum is due to the co-occurrence of benzaldehyde,
benzene methanol, 4-methoxy benzaldehyde, and 4-methoxy benzene methanol.
These four compounds are likely produced via the same single biogenetic
pathway. This is compared to the co—occurrence of these same four compounds
plus linalool, representing two independent biogenetic pathways, between C.
calceolus var. pamiflomm and C. candidum Calculation of the jaccard similarity
coefiicient includes mismatches (0,1) and (1,0) which lower the similarity between
any two taxa producing different compounds. Although C. arietinum and C.
candidum only share four biogenetically related compounds representing the
expression of a single pathway, they do not possess many differences in overall
composition. In contrast, C. calceolus var. pannflomm and C. candidum share two
biogenetic pathways yet they also express several unshared pathways. For this
reason, two taxa with very different gross morphologies are represented as
chemically similar although they share only one major biogenetic pathway. In
spite of the different pathways not shared between C. candidum and C. calceolus var.

pann'florum, the two taxa are biogenetically more related to each other than C.

31

.8833:
82:83 238m new 0638 £53 32286 8:88:

:o .8me 2:98:23 <28: U6 cemtmanu .w 23m:

 

858.. 82.63

 

$252282.

8832.5 -
638.. £88 _

 

 

 

 

 

8826829.

 

 

2.8893 8., .28

2.52.: min 8:58.81]—

Eabeemo i
_ E3388 L
Flfl 2.52....qu ems .28

accommnaq ...w> .uHmo EEoSEmQ ems .onu

 

 

 

 

 

 

 

 

 

_ H H H H H H
NB. tum. 4mm. mvm. m. m. 0.0
8:582 288m 962 382.5 288..

32

an’etinum is to either of them, yet in terms ofjaccard similarity, C. arietinum and C.

candidum are the most alike.

Taxonomic implications for C. calceolus complex

The taxa composing the cluster of C. candidum, C. calceolus var. pann‘florum, C.
calceolus var. pubescens, and C. kentuckiense (with the exception of C. arietinum) are
recognized as the C. calceolus complex (Atwood 1984). This grouping of taxa
within the complex is generally congruent with relationships presented by Case
(1993). While the delimitation of the entire complex is not in question, the
specific relationship between C. calceolus var. pannflomm and C. calceolus var.
pubescens remains ambiguous. The phenogram in Figure 5 supports Atwood
(1984) in recognizing C. calceolus var. pubescens and C. calceolus var. pamiflomm as
distinct species, but the phenogram of Figure 6 does not. The separation of C.
calceolus var. pubescens from C. calceolus var. pannflorum is ambiguous because the
clone sampled by Bergstrom et al. clusters with C. calceolus var. pamzflomm, while
the clone sampled in this study was distinct from it. The fact that the C. calceolus
var. pubescens individual sampled here is very different in composition from the
individual sampled by Bergstrom et a1. indicates that extensive qualitative
fragrance variation exists within this taxon. This pattern should be examined
further before taxonomic judgment is reached concerning the separation of C.

calceolus var. pamiflomm and C. calceolus var. pubescens into distinct species.

Taxonomic implications for other taxa

The species in the genus not included in the C. calceolus complex form a loose
cluster distinct from this complex and are not very similar to each other with
respect to fragrance composition. This is concordant with results obtained by
Case (1992). Although she did not study C. macranthum and C. guttatum, C. reginae

and C. acaule only possess a genetic identity of 0.247 with each other. The

33

relationship displayed between C. acaule and C. guttatum found in this study is
surprising as the taxa are very dissimilar in floral morphology. The placement of
C. macranthum is distant from all other taxa investigated. Stoutamire (1967) noted
a fragrance for this taxon which is similar to that found in C. acaule and C.
an'etinum, yet it has a shared similarity coefficient less than 0.1 with C. acaule and C.
arietinum. Little can be proposed for the relationships among these taxa on the
basis of floral fragrance alone, yet it appears that none of them are very closely

related to each other.

Population Variation

Fragrance profiles for individuals within the same population differ
qualitatively and quantitatively. Entire classes of compounds were found to vary
between individuals both between and within populations. Monoterpene
hydrocarbons, cinnamic acid derivatives, straight chain alcohols, sesquiterpene
alcohols, and tri-methoxy benzene are only expressed in a few individuals.
Although these variant compounds were often minor components of the
fragrance profiles, they do represent variant expressed pathways and hence
potentially different genotypes.

The results of the cluster analysis in which all individuals of C. calceolus var.
pamzflomm are included indicate that the qualitative variation observed is not
taxonomically important, as all individuals cluster more closely with each other
than with other taxa in the analysis. This is an important result which
demonstrates that although fragrance variation may appear high between
individual clones of a taxon, these differences may not have taxonomic
importance when compared against variation displayed among closely related
taxa. Other taxonomic studies utilizing fragrance data (Gregg 1984; Bergstrom et

a1. 1992) have not compared fragrance composition across several related species

34

within the genus sampled. Knowledge of the extent of variation across
congenerics allows a more effective evaluation of "significant" fragrance

compound variation.

Lack of concordance

The individual of C. calceolus var. pubascem sampled here differ greatly in
qualitative composition in comparison to that sampled by Bergstrom et al. (1992).
The diflerences in trapping techniques prohibit strong inferences from the
comparisons. It should be noted, however, that most of the compounds isolated
and identified by Bergstrom et al. were also isolated and identified by the
technique employed here in different taxa. Thus, differences in affinity of sorbent
tubes are assumed to be minimal. The differences in fragrance composition
between the individuals of C. calceolus in the two studies indicate that extensive
genetic variation may exist for the production of fragrance compounds, especially
when quantitative differences are considered. In addition, comparison of results
obtained by Holman (1983) indicate substantial variation in fragrance
composition between the same taxa sampled here. The wide variation in reports
of fragrance composition for this taxon emphasizes the need for larger sample
sizes if the data are to be used in taxonomic studies.

Figure 8 shows a comparison of the UPGMA-derived phenograms of Figure 5
and Case (1993). While the distinction of two primary groups within the genus is
concordant between the studies, specific relationships are not. The relationships
proposed by Case (1993) between C. calceolus, C. candidum, and C. arietinum are
very different from ours. Due to limitations in the coding procedure and in the
calculation of similarity values for floral fragrance data, as discussed above, the
specific relationships indicated in the Figure 5 phenogram are not strongly

suggestive of evolutionary relationships. The clustering of taxa may more

35

accurately represent pollinator relationships as suggested by Whitten 8c Williams

(1992). This hypothesis remains to be tested.

Conclusion

The use of floral fragrance-compound data in combination with objective
measures of comparison result in clusters of taxa which reflect general taxonomic
relationships recently proposed for Cypfipedium The "broad" resolution of this
technique in the identification of general evolutionary relationships is supported
by other data sets. However the "fine" resolution produced by this technique is
not. The cluster analysis has been useful in delimiting the C. calceolus complex as
separate from all other taxa. Relationships among other taxa within Cypfipedz’um
are diflicult to infer given the data accumulated thus far, although the retention of
C. arietinum within Cypn'pedium appears well supported. The phenetic methods
used have several shortcomings when used in conjunction with volatile compound
data for the purpose of estimating phylogenetic relationships. The treatment of
isomers as independent compounds for use in calculation of similarity coefficients
will result in erroneously high values. In addition, the similarity values utilized do
not accurately reflect the importance of shared biogenetic pathways if multiple
unshared pathways are present between two taxa. The systematic investigation of
several putatively related taxa is necessary before taxonomic judgments can be
effectively made because the variation between taxa can only be evaluated once

variation across other related taxa is considered as well.

PART 2: CLADISTIC RELATIONSHIPS WITHIN CYPRH’EDIUM INFERRED
FROM FLORAL FRAGRANCE-COMPOUND DATA

INTRODUCTION

The genus ijm'pedium is composed of approximately 40 species having diverse
floral morphologies. The species are fairly similar vegetatively, so little has been
proposed about phylogenetic relationships due to a lack of polarizable
reproductive character trends. Several species are native to North America and
have received considerable taxonomic attention in the last ten years. One group
of taxa that is morphologically coherent is the C. calceolus complex (Atwood 1984),
which includes C. calceolus var. pannflomm, C. calceolus var. pubescens, C. calceolus var.
planipetalum, C. kaztuckiense, C. montanum, and C. candidum This complex shares
presence of linear-lanceolate lateral petals, an inflated pouch with an orifice on
the top, and a white or yellow labellum (Atwood 1983). Cypripedium calceolus is
noted for extreme morphological variation, and the designated varieties are
difiicult to distinguish in many cases. In addition, hybridization between the
named varieties of C. calceolus as well as C. candidum results in populations which
are taxonomically ambiguous. Introgression confounds taxonomic inference and
has been demonstrated between C. calcealus var. pubascens and C. candidum using
isozyme data (Klier 1992). While membership of these taxa within the complex
seems reasonably certain, it is not known which ones are more evolutionarily
derived than others.

Other taxa within Cypripedium are a diverse assemblage with much variation in
labellum morphology and color. ijm'pedium reginae has an inflated labellum

similar to that found in the C. calceolus complex yet differs in that the lateral petals
36

37

are oblong-lanceolate and the pouch is pink. Cypripedium acaule is distinct in that
it has an anterior suture in the pink labellum and has only two leaves. This
labellum architecture is very different from any other taxon in the genus.
CWpedium arietinum has a diminuitive flower characterized by a spurred pouch
and free lateral sepals. The lateral sepals in all other taxa are fused into a
synsepal, although various degrees of splitting of the synsepal are found in C.
calceolus (personal observation). Cypfipedium gutattum, native to Alaska and japan,
has an unusual striped flower with a tubular labellum similar to that found in the
tropical genus Paphiopedz’lum. Cypripedium macranthum is an Asian species
characterized by an inflated pouch which has a raised ridge around the orifice.
The flower color is burgundy and the lateral petals are oblong-lanceolate. Its
relationships to other taxa within the genus are unknown, although artificial
hybrids have been reported between it and C. cakeolus (Stoutamire 1967).
Phylogenetic relationships remain to be suggested for the morphologically
divergent taxa included here.

The first phylogenetic treatment for several North American species of
Cypfipediumwas that by Atwood (1984) , which also included the tropical slipper
orchids, Paphiopedilum, Phragmipedium, and Selenipedium. Atwood proposed
phylogenetic relationships within the subfamily Cypripedioideae based upon
morphological, anatomical, and karyological data. In his treatment of Cypripedium,
C. arietinum, the rams head lady slipper, is placed in a segregate genus, Cn'osanthes.
This decision was based upon unique floral morphological features including a
spurred pouch, a staminode which nearly resembles a fertile anther, and free
ventral sepals. These characters were interpreted by Atwood as primitive and
divergent for the genus. Relationships of the remainder of the genus include a
group treated as "most" Cypripedium (C. calceolus complex, C. macranthum, and C.

reginae) which were interpreted as intermediately advanced and somewhat

38

separate from the primary evolutionary line that culminates in the most highly
derived taxa, C. acaule and C. guttatum. Atwood (1984) urges: "If this work
stimulates botanists into seeking ways for presenting falsifiable hypotheses
concerning their [the lady slippers] origin and evolution it shall have served its
purpose." This pioneering study of the phylogenetics of the slipper orchids has
indeed spurred several other studies within the subfamily and especially of
Cypripedium.

The most recent work addressing evolutionary relationships of several North
American taxa was that by Case (1993) . She utilized allozyme-frequency
information to show that C. calceolus and C. candidum are closely related, sharing a
genetic identity of 0.794. An interesting result is the relatively high genetic
identity of C. arietinum shared with C. calceolus and C. candidum (Nei's genetic
identity = 0.247). This value is higher than that obtained for C. acaule and C.
reginae with the other three taxa (0.072). While her study proposes potentially
evolutionarily related taxa, it does not distinguish phylogenetically advanced taxa
from primitive taxa. Case also investigated evolutionary relationships for
infi‘aspecific taxa of C. calceolus. A genetic identity of 0.945 was reported between
C. calceolus var. pannfiorum and C. calceolus var. pubescens. This is comparable to
identities reported for infraspecific taxa in other genera. Although other workers
have proposed recognition of two distinct species, Case recommends their
inclusion in a single species.

Albert (1993) has investigated systematic relationships for Cypripedium based
upon rbcL plastid variation, floral morphology, and vegetative morphology and
anatomy. Within the genus, C. plectrochilon (putatively related to C. arietinum) is
most primitive and C. reginae is relatively advanced, although the data sets used do
not provide good resolution for these relationships within the genus (Albert pers.

comm.) .

39

The use of chemical data in conjunction with cladistic methods to infer
phylogeny at lower taxonomic levels has proven successful in Tetragvnotheca and
Iva (Asteraceae) (Seaman 8c Funk 1983). Biogenetic transformation series for
sesquiterpene lactones were constructed based upon skeletal and substitutional
features. Outgroup analysis was used to determine the plesiomorphic and
apomorphic states. Synapomorphies were reconstructed using parsimony and
resulting cladograrns were compared with results from analyses derived from
independent data sets such as those based upon morphology.

Although floral fragrance compounds have not been utilized previously for
reconstructing phylogeny, there are several advantages to using this type of data
over others. Volatile compounds can be objectively collected and their structure
unambiguously identified using gas chromatography-mass spectrometry. This
analytical technique is relatively easy to use with the only shortcoming being the
ambiguity in positively identifying isomers. However, isomers of a particular
compound are often chromatographically distinct, making them useful for
comparative studies. Compounds can be described, when pathway information is
sufficient, as biogenetically simple or complex with ordered enzymatic steps
elucidated in many cases. Floral fiagrance compounds are often structurally
diverse, produced via several independent primary biogenetic pathways, offering
many characters for systematic evaluation. Although use of floral fragrance data
has many advantages, a major disadvantage is the widespread distribution of
identical secondary metabolites in phylogenetically unrelated taxa. Whitten and
Williams (1992) report a limited use of fragrance-compound data in
reconstructing phylogenies for euglossine-pollinated orchids, due to the
widespread distribution of terpenes, which are found to be produced in many

plants and fungi.

40

Volatiles in orchids are ultimately derived via three pathways, mevalonic acid,
shikimic acid, and acetic acid (Bergstrom 1991). The mevalonic acid pathway can
yield one of two volatile classes of compounds, mono- and sesquiterpenoids. The
shikimic acid pathway, via phenyl alanine or not, can yield diverse groups of
compounds including benzaldehydes, benzoic acids, phenyl acetic acids,
anthranilic acids and cinnamic acids. The acetate pathway yields derivatives such
as straight-chain alcohols, acetates, or esters. Although the mevalonic, shikimic,
and acetic acid pathways are ubiquitous in distribution, the distribution of their
derivative pathways are not and may provide useful phylogenetic information.

Fragrance chemistry within Cypripedium has been reported for several North
American taxa. In no cases, however, have the data been utilized in a cladistic
analysis. This may be due to the complexity in coding fragrance data. The
previous studies have compared quantitative and qualitative variation for
taxonomic purposes (Bergstrom et al. 1992; Gregg 1984). Taxonomic judgement
has been based upon perceived importance of this variation to the pollination
system of these taxa. Bergstrom et al. (1992) have discussed the variation in
fragrance chemistry in three varieties of C. calceolus. They recommended the
recognition of three distinct species on the basis of extensive qualitative and
quantitative variation. The results of cluster analyses provided in Chapter 1 show
that qualitative distributions of biogenetically related floral fragrance compounds
are useful in delimiting broadly related groups of taxa only. A different approach
is taken here to analyze the distribution of biogenetic pathways and enzymatic
steps as potentially informative synapomophies for phylogenetic reconstruction.

The present study is an attempt to suggest relationships within the genus
Cypripedium, including C. arietinum, and to suggest phylogenetic placement for
these taxa using floral fragrance data alone. Floral fragrance compounds have

been isolated and identified and are used as an independent data set. The coding

41

step is the most important part of the process in any cladistic analysis. The
eflective coding of secondary chemical-compound data is dependent upon a
knowledge of biogenetic interrelatedness of the compounds to ensure coding of
independent characters. Fragrance compounds should be arranged according to
biogenetic relatedness as is stressed by Salatino and Gottlieb (1981 ). Enzymatic
steps required for interconversions between compounds are coded as present or
absent and were utilized as the data matrix for the analysis. In this coding
method, basic biogenetic pathways are treated as characters and specific
transformations between compounds are character states. This coding method
represents the dependent nature of the data. Parsimony analyses are used to
discern cladistic relationships among the taxa. Relationships among the
morphologically divergent taxa C. acaule, C. reginae, C. guttatum, and C. macranthum
will be explored and potential associations proposed based upon considerations of
other recent findings. Other proposed relationships to be investigated include the
controversial position of C. arietinum and the questionable hierarchical placement

of designated varieties of C. calceolus.

MATERIALS AND METHODS

Lists of taxa sampled and compounds identified for each are provided in

Chapter 1.

Biogenetic schemes

Schrier (1981) and Croteau (1991) summarize the results of various
biosynthetic studies to propose general pathways for the production of most
volatile compounds. Most of the pathways have been found to occur in higher
plants, although occasionally the only direct evidence derives from experiments

with fungi. Despite this work, which addresses the biogenesis of major skeletal

42

types, little is known about pathways leading to substitutions and skeletal
modifications such as reduction or acetylation of functional groups. Although
pathways have not been elucidated for these modifications of basic skeletons,
pathways have been assumed in most instances. Support for the assumed pathways
is based upon the presence of all related compounds in one species. For
example, in C. arietinum, benzaldehyde, benzene methanol (the reduced
aldehyde), 4-methoxy benzaldehyde, and 4-methoxy benzene methanol (methoxy
substituted alcohol) are all produced (See figure 2 for assumed biogenetic

pathway) .

Fatty acids

The biogenesis of fatty acid derivatives is widely reported (Croteau, 1991;
Shreier, 1983). Catabolism of fatty acids such as linoleic acid by beta-oxidation
can lead to methyl and ethyl esters of mono and dienoic acids (including 010 and
C-12 chains) in ripening Bartlett pear tissue. This is a likely pathway responsible
for the derivatives found in C. acaule and C. guttatum. Many reaction sequences
could result in the formation of dodecanol and various aliphatic acetates found in
C. calceolus and C. kentuckiense respectively. Labeling experiments in banana slices
have shown fatty acids, such as hexanoic acid, to be reduced to 1-hexanol (Tressl
8c Drawert, 1973 in Schreier 1984). Analogous experiments have shown the same
reaction to occur in labeled decanoic acid. Aliphatic alcohols may also be formed
during hydroperoxidation reactions in which fatty acids, liberated from
membrane glycerolipids, are cleaved to yield aldehydes of various lengths which
may undergo further reduction. Acetylation of the alcohol products could explain
the occurrence of aliphatic acetates found in C. kentuckiense. Beta-oxidation may
produce the aliphatic alcohols and acetates, found in these species of Cypripedium

Evidence for this exists as the distribution of volatiles included chain lengths of 8,

43

10, and 12 which may represent products formed at various stages of catabolism of
a single long-chain precursor. Due to the ambiguity in the origin of the fatty acid
compounds isolated in several species, an independence of biogenetic pathways
was assumed between the two primary types, methyl and ethyl esters, and alcohols
and acetates. Figure 1 shows a proposed biogenetic scheme which served as the

basis for our coding procedure.

Benzenoids

The biogenesis of benzenoids is well elucidated for higher plants. The
anthranilic acid pathway results in the formation amino benzaldehyde and of
indole ultimately, both of which are present in C. calceolus and C. acaule. The
dehydroshikimic acid pathway results in the formation of benzoic acid. From
benzoic acid, the benzaldehyde and benzene methanol derivatives are easily
formed by reduction. Many of the species produce derivatives from benzoic acid,
a non-volatile precursor. The cinnamic acid pathway from phenylalanine results
in the formation of cinnamaldehyde derivatives and eugenol. Cypripedium
macranthum and C. guttatum express this pathway. It is not clear, however, if the
phenyl ethyl derivatives found in C. candidum, C. calceolus, and C. kentuckiense are
formed by decarboxylation and reduction of cinnamic acid or if they can derive
directly from phenylalanine transformation. For this analysis these are assumed to
be derived from cinnamic acid as suggested by Croteau (pers. comm.) . Benzyl
benzoate biogenesis is uncertain because it may derive from a benzaldehyde
precursor or directly from benzoic acid. Due to this ambiguity, this compound has
been coded as directly arising from benzoic acid. Likewise, 1,4-dimethoxy
benzene biogenesis is unknown and has necessitated a coding procedure which
treats it as independently derived from any other compound. Four multistate

characters were constructed for the benzenoid compounds; benzoic acid,

44

.8362: Eon E 5.6% 93 8280250 ouaEmuco maucomoae 883 58820

88958 28 b5 oHHHmHo> SH 889 ouocowoB 888$ .H 8:me

IIIY 28 220::
8

8528.83 H /

/\/\/\/\/\/\
\/\/\/\/\ o\

coHHmeuom

/\/\/\/\/\/xVT EQK/ \/\//\<fi\

All
88288 +

8828-58 H N

o
o It 2... as :K/

45

cinnamic acid, anthranilic acid, and 1,4-dimethoxy benzene. Figure 2 shows the

biogenetic scheme which served as the basis for this coding.

Monoterpenoids

Monoterpenoid biogenesis has received much attention, and many of the
reactions involved in the formation of cyclohexanoid monoterpenes have been
elucidated (Croteau 8c Karp 1991). However, exclusive pathways for particular
terpenoids do not appear to exist. Geranyl pyrophosphate (GPP), formed by the
condensation of dimethyl allyl pyrophosphate (DMAPP) and isopentyl
pyrophosphate (IPP), is the ultimate precursor of monoterpene hydrocarbons and
alcohols, including acyclic, monocyclic or bicyclic skeletons of each. Linalool is
reportedly formed from GPP although it can arise from its isomer linaloyl
pyrophosphate (LPP) as well. LPP can produce other mono- and bicyclic terpenes
also. Due to the ambiguity of a common precursor which can produce this wide
diversity of derivatives, all monoterpenes identified in this study were artificially
divided into monterpene hydrocarbons and monoterpene alcohols. A common
precursor of GPP is assumed for all derivatives in each of these two classes. Figure

3 shows the proposed biogenetic scheme for both hydrocarbons and alcohols.

Sesquiterpenoids

Sesquiterpene biogenesis is poorly known. Famesyl pyrophosphate is reported
to give rise to all sesquiterpenes, although any of four configurational isomers can
give rise to derivative sesquiterpenes. Farnesene results from the reduction of
famesol, a primary product of FPP. Nerolidol results from the dephosphorylation
of neryl perphosphate, which is an isomer of FPP. Caryophyllene arises by
transformation from the original precursor FPP without isomerization to NPP

(Croteau 8c Karp 1991). Figure 4 shows this basic biogenetic scheme.

$522 Eon E 550% 2a 2065250 23835 watcomofia 885 88820

.8383me Box 28:8 H8 858 unocomofi 3835 .N Emmi

oo

\ h

of

All: Eon eggs". All
o
\ <
a
Ally .\ .m
m u 7
m
mmoo Mao
3/ .\m /
mu.
.m
I

:0 I o\ . .\mm0.

47

.
W
W

.mEocoNcon 8m £9553 ozocowoB @83on .350 .N 8:me

ocEmgxcoza
4

9% 28586

mmmo

 

 

U56 UZMSEECG
_ ll .

«I ~ o

\ .

£0
05209 xxofioemvéq A: .. u u

.5

£022 Eon 5 E505 8a 206528 238.35 managed»: 8.9.. 68820

.mconaoofizn 0592808 W8 858 unocomoE @38on .m Emmi

0:258 8.3 as m6

9
l/ c
08.383058 153% Alll
\ < an: + AESZQ

ococoE: m

802,953

49

.2050on 0588958 H8 388 0:08on @83on 6.88 .m 083m

an: + nE<EQ All

c
323:: 5:: .550
0862525 «3 “ED 80h <
568250 88% 2838
m encamofiea 358%
/ 822620
was u p a 6:82 2:: 60:2: \\m G
0.. O
\AIV _ _ __ \{Iom _ mo
2
0 288“ 322556
was 3.2.: Since? 2:: Awash do .28

   
   

\\

x O “.6me 60:3: v.98 Boas: b.8152:
lb _ _ _ _ A E E m 28 2:285
l 030 $00

mmnVfi
o

50

Unknown biogenesis

In some cases compounds could be derived via alternate pathways. One class
of compounds, the lilac alcohols and aldehydes, are of unknown biogenetic origin.
The position of the epoxide bridge, between 2-C and 7-C, however allows an
inference into its biogenetic origin. This compound may be derived from
monoterpene polyols. The position of alcohol substitutions determine where the
epoxide bridge may form. Linalool has an alcohol substitution at the 2-C position.
No other monoterpene alcohol has alcohol substitution at this position. Despite a
potential derivatization from linalool, the pyranoid monoterpenes had to be
coded as biogenetically independent from it. The taxa which produce the lilac
alcohols do not produce linalool so no biogenetic association was assumed. Figure

3 shows the placement of the enigmatic alcohols within the proposed pathways.

User-defined characters

Figure 5 shows the user-defined character types used in the cladistic analysis.
PAUP and MacClade allow these character types, which could not be effectively
coded otherwise. These allow the representation of proposed biogenetic pathways
as character-state transformations. The transformations between states are
ordered, in that the formation of a particular compound is dependent upon the
formation and modification of its precursor first. Each character-state represents
an enzymatic step in a pathway. The user-defined character states are then
reconstructed for the taxa within CWpedium in the most parsimonious manner.
The most parsimonious topology represents the distribution of enzymatic steps for
the taxa such that independent origins of biogenetic pathways are minimized.
Figure 6 shows the data matrix used in the analysis in which the character types

are represented by the user-defined characters of Figure 5.

51

.mcoznx Eon E 5505. 8: 32395: 05635 waccomoao: 833 88820

$582698 :8 858 unocowoB 883$ .v 2:me

ocomofifiénfla an: + AEO
<

22:8:quan 185$

PEWmo NNV U

$co§fi0§9afi3 m

use .A 8.98
0538:; = a 6208:
m
/ AllY \

52

Cladistic Analysis
PAUP (Swotford 1991) was used to generate trees representing cladistic
relationships of the taxa. Several assumptions were used to generate the most

parsimonious trees.

Character types

Unordered character states were assumed for all binary coded characters.
Multistate characters with states that could be organized into linear
transformations were treated as ordered characters. The ordered characters were
coded according to the biogenetic pathways by which they are formed. Hence, in
a multistate character, with states 0-1-2, the formation of 2 is dependent upon the
initial formation of 1. Figure 1, representing the proposed biogenesis of fatty acid
derivatives, illustrates the coding of linearly ordered character states. The
characters representing biogenetically related compounds which could only be
represented by branching transformation series were treated as user-defined
character types. The character states are essentially ordered in the sense that the

specified path must be followed for transitions between various states.

Rooting

An hypothetical ancestor served as the root of the analysis. All states were
assigned as unknown, which is the default setting of PAUP. The assignment of all
unknown states allows the most parsimonious assignment of character states to the
ingroup under all circumstances. This analysis was undertaken to achieve locally
most parsimonious resolutions for the ingroup. The unrooted trees obtained can
be rooted anywhere without a change in tree length.

Outgroup analysis was carried out using members of Paphiopedilum sampled by

Barkman (unpublished). The rooted trees obtained allow hypotheses about

53

benzaldehyde

 

 

Figure 5. User-defined character-state trees.

54

 

 

 

 

 

 

 

 

 

 

bark.an copy 2 1 1 2 3 4 5 6 7 8 9
Cinn dime anth benz mon mon sesq fatt) fattL
1 parviflorum O 1 1 E B C B 1 O
2 pubescens 2 1 O D B O O 1 O
3 candidum 1 O O E 0 8 O O 0
4 arietinum 0 O O E O O O O O
5 kentuckiense 1 O O O 8 B B 2 O
6 acaule O 0 1 C D E O O 2
7 reginae 0 O O B O O 0 0 O
8 guttatum B 0 O D O D D O 1
9 macranthum C O O F B E C O O

 

 

 

 

 

 

 

 

 

 

Figure 6. Data matrix used in exhaustive search using Wagner parsimony.

55

ancestral and advanced taxa within the genus. Our outgroup includes
representatives from several different sections of Paphiopedilum (sensu Cribb
1986). Sections of Paphiopedilum included Brachypetalum, Pamisepalum,
Cochlopetalum, Cmyopetalum, and meata. Species chosen include P. emmonii, P.
niveum, P. pfimulinum, P. kolopakingii, P. argus and P. ameniacum. The choice of
several outgroup taxa may allow a better resolution of plesiomorphic states to be
used in the analysis and avoid the bias of choosing a single taxon. The ordering of
these taxa within the outgroup was based upon the results of a phylogenetic
analysis by Albert (1992). The ordering of outgroup taxa is an important
consideration in a cladistic analysis because of the effect it can have on character
states of the outgroup node.

The initial analyses in which character states were reconstructed with no
specification of outgroup taxa resulted in topologies which placed members of
Paphiopedilum within Cypripedium. This analysis was undertaken to see if members
of Paphiopedilum would be resolved separately from Cfim’pedium. Due to these
preliminary findings, Paphiopedilum could not be specified as an outgroup to
CWpedium with the monophyly of Cypripedium remaining intact. For this reason,
topological constraints were specified under the search options in PAUP. The
topology chosen for the outgroup was used as a constraint for the exhaustive
searches of the ingroup. No topological constraints were placed on the taxa
within CWpedium, however.

RESULTS
Outgroup analysis
A single tree resulted from the outgroup analysis in which the topology of

Paphiopedilumwas used as a search constraint. Figure 7 shows the results of the

analysis. A clade composed of C. arietinum and C. reginae as sister taxa was basal to

56

the entire genus. Cypripedium acaule and C. macranthum were sister taxa in a clade
which was in a derived position relative to the C. an’etinum/ C. reginae clade. The
topology supported the monophyly of the C. calceolus complex, which was placed
in a derived position in the genus. Basal to the complex and monophyletic with it
was C. guttatum. Although the outgroup analysis resulted in a single most
parsimonious topology, the results were abandoned due to the nature of the
fragrance compound distributions, which resulted in no distinction between the
out- and ingroup. Figure 8 shows the results of a single tree of 119 most

parsimonious trees obtained when no outgroup taxa were specified.

Ingroup analysis

Analyses of the ingroup alone were performed to achieve local parsimony. No
hypotheses of plesiomorphic and apomorphic charater states could be assumed,
however. An exhaustive search was carried out in which all possible topologies
were evaluated for most parsimonious reconstructions of the character states. Ten
equally parsimonious trees were obtained, all of 45-step length. Figure 9 shows all
10 unrooted trees obtained. Several trends are clear from comparisons of all trees.
CWpedium calceolus var. pann'flomm, C. calceolus var. pubescens, and C. kentuckiense
are consistently associated with each other and are supported by seven
synapomorphic gains. Cypripedium reginae and C. arietinum are found together in
seven trees although only one synapomorphic gain supports the clade. However,
the clade is supported by eight synapomorphic losses. Cypn'pedium acaule and C.
macranthum are found as sister taxa in a clade in only three of the trees. The
position of C. guttatum is unstable, because it is occupies a different position in six
of the trees. All other trees differ in the relative positions of C. candidum and C.
macranthum. Four of the topologies support the resolution of C. candidum with the

rest of the members of the C. calceolus complex. Alternatively, three of the

parviflorum

kentuckiense

S7

 

pubescens
candidum
guttatum
macranthum
arietinum
emersonii
armeniacum
niveum
kolopakingii

reginae

acaule

 

argus

primulinum

 

Figure 7. Single most parsimonious tree in which
Paphiopedilum is held as an outgroup search
constraint. (Paphiopedilum enclosed within box)

 

pawifiorum

58

primulinum ‘——
armeniacum ‘—

kentuckiense
kolopakingii
macranthum
guttatum
pubescens
niveum
candidum
arietinum
acaule
reginae

Figure 8. One of 1 19 most parsimonious reconstructions for
Cypripedium and Paphiopedilum with no outgroup specified.
(Paphiopedilum is recognized by arrows)

emersonii

l ..

 

parvilbmm
candidum
80'9an
acaule

 

I
933735”.

 

Wham
kentuckiense

 

L.)

 

parvmorun

 

guttalun

ll

 

 

kerfludu’eme

01

 

 

 

 

 

V

 

 

 

 

 

‘ '0

 

 

 

 

 

 

parvfllorum

 

 

pwescens

candidum
alielmwn
reginae

galalum
;—-— acaule

macranthum
kentuckiense

 

 

pan/1110mm

 

 

 

pwesoens
candidum
aneMum

r mae
99.3mm
acaule
naaanmum
kentuckiense

 

 

 

 

anentm

acaule
mm

 

 

WWW

 

 

 

 

 

 

 

pam? loam
pwesoens

 

arietinun
acaule
maaarzmun

 

r .
gutialum

 

kernuckiense

Figure 9. Ten equally parsimonious reconstructions for Cypripedium.

6O

topologies support C. macranthum with the C. calceolus complex. The other three

trees do not support any single taxon as sister to the C. calceolus complex.

Orientation of unrooted tree

Orientation of the unrooted consensus tree was done with the C. calceolus
complex kept separate from other taxa. Figure 10 shows the majority rule
consensus tree. This orientation follows the results presented in Chapter 1 for a
clear separation of these two groups of taxa. Trees are oriented such that the C.
calceolus complex is monophyletic. MacClade (Maddison 8c Maddison 1992) was
used to trace character evolution on the consensus tree. The nature of the
compound distributions is such that few synapomorphic gains have defined any
specific lineages. In addition, the entire genus is not supported by any common
synapomorphy.
Character state reconstructions and distribution of synapomorphies

Figures 11-19 show character state reconstructions for each character on the
consensus topology which includes C. candidum within the C. calceolus complex.
The legend accompanying each reconstruction lists states defined for each
character. The enzymatic conversions to which each state corresponds can be
found in figures 1—4, representing the proposed biogenetic pathways. The clade
composed of C. reginae and C. arietinum are united by the lack of expression of all
biogenetic pathways found in the genus excepting the benzoic acid pathway. The
synapomorphic biogenetic pathway shared between them is not very strong as it is
only the formation of a common precursor, benzoic acid.

The specific topology of the C. calceolus complex is clear in that C. kentuckiense
and C. pamijlomm are sister taxa, supported by a synapomorphy of alpha-famesene
and linalool. Cypripedium calceolus var. pubescens is sister to these two taxa.

Together, the three taxa share synapomorphic pathways leading to the production

parviflorum

 

 

100
100

61

candidum
arietinum
reginae

7O

guttatum

 

 

20

 

 

macranthum

acaule

7O

 

 

40

 

45 steps

Figure 10. Consensus tree of ten equally parsimonious reconstructions for

Cypripedium.

62

Eaficfione

038m

Eamfiam

8:68

6:53:

6:923

«583::

85503:}

520528

I ‘ may...
.3.
Jun.
f”?
#9:.
. _ .
. «A .
v . )5.
I. ../.
.. u.
v 2...
A. . 3......
w a...

 

 

Figure 11. Character state changes for the cinnamic acid pathway.

parviflorum

63

0

m E
“C, m 3
~- 8 E E E 5
in) U 3 3 q, 3 C
u 0 '0 .4 _§ .2 3 U
C D c Q 0'! u m m
0 3 (g C Q 3 U

x O. U N L. 0’ N E
I I

 

dimethoxy benzene
unordered

1:1 0
- l
Eequivoal

Figure 12. Character state changes for biogenesis of
1,4-dimethoxy benzene.

64

E:£:m._omE

2:8:

8335:

8:62

83:50::

636:8

«5802::

85:63:»...

820228

 

anthranilic acid

unordered

[:0

-1

Figure 13. Character state changes for anthranilic acid pathway.

pawiflorum

kentuckiense

Figure 14.

pubescens
candidum

arietinum

65

reginae

E
3
‘J
to
H
‘J
3
U)

v‘

A I
,4!)
.-V
N
'V

acaule

macranthum

‘-l!.
IL

benzaldehyde
benzaldehyde

[:1 0
::-?.::;:.-;: 1
A
s
m C
D
- E
- F

Character state changes for benzoic acid pathway.

66

8:55:88

0.38:

8333a

8:62

83:32.5

8.66:3

«:33: 3:

3:35:83

83.6523

 

Figure 15. Character state changes for biogenesis of

monterpene hydrocarbons.

67

83:88:18

238:

83225

$89:

8386::

885:8

mcoomonsa

3:210:83

8233.3

.3

- E
=eQLIvocal

D a .w. .H.
Y9 \

‘04 5..

0‘ .NM
‘ p . .
. .
‘ ‘
‘
‘ ‘
‘
m ......... \‘
...... .v.. .‘
....- ‘v

u u. ‘

lg .IHI. I “ ‘ ‘

.u.- ‘
‘
H and. III
. 3...: Ma. .1. .

a . w

I w

Figure 16. Character state changes for production of monoterpene

alcohols.

parviflorum

68

8 8
S ‘8 8 E g
5 8 3 g Q) 3 E
g u: 9 .E “5 ‘5 2 S
C .0 C 0 u (3
co 3 a "C 0') 3 U m
.2 o. o m 3 a: co E
:I l I [j I Q ~_'-
4.; J I / ‘9 g 3] V5!
I. u? 9 .5]
‘.| :
4| J

\g u

 

.‘J sestherpene
sestherpene

 

Figure 17. Character state changes for biogenesis of
sesquiterpenes.

69

parviflorum
macranthum

0

U)

5 '3 8 E E

32 ‘” 3 3 q, 3

g 8 .9 s m a; 2

u .8 “O H ,E u 3

c: c 2 o; H N

0 3 N u 0 3 U

.x O. U N L. O) N
I g I

BI

BN3

2%.

   

fatty acid ester
ordered

[:10

:1 as,

 

Figure 18. Character state changes for the biogenetic production
of aliphatic alcohols and acetates.

7O

3 E
E S z-z E E 2
3 '— E u
.5 "5 8 3 3 iv 3 c
._ ._ _ L
M 0 'O H .E H 3 U
E c 8 s. 2 m s 3 m
o. 3 o. u «'3 3 o: a: E
u c: D a -_>< I
‘ / ‘9 .‘J “.4? 3 :1
o .595
J V” 3“"
94"

 

fatty acid
31 ordered

 

Figure 19. Character state changes for the biogenetic pathway
leading to fatty acid ester production.

71

of cis-ocimene and the fatty acid alcohols and acetates. Strengthening the cladistic
association of these taxa is the shared presence of the cinnamic acid pathway,
leading to phenyl ethyl derivatives, between C. calceolus var. pubescens and C.
kentuckiense. In addition, the shared presences of the biogenetic capability to form
1,4-dimethoxy benzene and benzaldehyde derivatives between C. calceolus varieties
adds support to the lineage. The basal-most member of the complex, C. candidum,
derives support for inclusion in the lineage by the shared ability to produce phenyl
ethyl derivatives, benzaldehyde derivatives, and linalool.

Synapomorphic gains of biogenetic capability supporting the clade composed
of C. acaule and C. macranthum include the formation geranyl pyrophosphate
resulting in lilac alcohol and monoterpene hydrocarbons. In addition, the
capability to produce the general precursor, benzoic acid, is shared as well. Only
the single synapomorphic gain of the biogenetic capability to produce benzoic

acid supports the clade in which C. guttatum is basal to C. arietinum and C. reginae.

Alternative topologies

The primary topological difference between the three general types of trees is
that C. candidum is excluded from the C. calceolm clade in six of the ten trees. If C.
candidum is included in the C. calceolus clade, rather than C. macranthum, more
synapomorphies support the clade. The relatively higher number of
synapomorphic biogenetic steps and the similarity of these taxa with respect to
floral and vegetative morphology are the bases for choosing the topologies which
include C. candidum within the C. calceolus clade. Three trees supported the
inclusion of C. macranthum within the C. calceolus complex. Trees were imported
into MacClade to evaluate the inclusion of both C. macranthum and C. candidum
within the complex. Tree length was increased by a single step with this new

topology (see Figure 20).

pawiflorum

72

kentuckie nse
pubescens
candidum
macranthum
arietinum
guttatum
acaule

reg inae

 

 

 

 

 

 

 

 

 

Figure 20. Tree in which Cypripedium candidum and
C. macranthum are included within the C. calceolus complex
(47 steps).

73

 

 

 

 

 

 

 

 

 

 

.072
Nei's genetic identities I
.247 .247
.794
I
.945 7
I kentuckiense I candidum macranthum
parviflorum pubescens arie tinum acaule reginae guttatum

One of ten most
parsimonious
reconstructions
(45 steps)

Figure 21. Comparison of genetic identities (Case 1993) and cladogram
produced from floral fragrance data for Cypripedium.

74

Phylogenetic interpretations are derived from tree 10 due to a high degree of
concordance with results obtained by UPGMA of Case (1993). Figure 21 shows a

comparison of the two topologies.

DISCUSSION

Outgroup analysis

The results of the outgroup analysis must be carefully interpreted. Although a
single most parsimonious topology was obtained, it may not accurately reflect
phylogenetic relationships among taxa in Cypripedium The topology of the
outgroup had to be held as a search constraint due to the similarity of biogenetic
pathways for many fragrance compounds between Paphiopedilum and Cypripedium
The initial search, in which no constraining topology was imparted on
Paphiopedilum resulted in species of the outgroup occupying sister taxon positions
to five different species in the ingroup. The implications of this result are such
that if each taxon in the outgroup were used independently as an outgroup, a
difierent topology would result for the ingroup in each analysis. Due to the lack of
consistency among the outgroup taxa for character states with respect to the
ingroup, the implications of the analysis, a basal most position of C. arietinum and

C. reginae, are not very strongly interpreted.

Phylogenetic Implications
CWpedium calceolus/ C. kentuckieme

The results of the parsimony analysis indicate several trends for evolutionary
relationships within Cypripedium. First, the C. calceolus complex is monophyletic,
including C. kentuckiense, C. calceolus, C. candidum, and possibly C. macranthum

Cypripedium candidum is basal within the lineage. This early evolutionary

75

divergence from the rest of the C. calceolus lineage is supported by the genetic
identities obtained by Case (1993). Within C. calceolus, C. calceolus var. pann'florum
and C. calceolus var. pubescm appear to be separate entities. In all trees C. calceolus
var. pannflomm is sister to C. kentuckiense. Case (1993) proposed a potential
progenitor-derived relationship for C. kentuckiense from C. calceolus var. pubescems on
the basis of preliminary allozyme data. However, she does not recognize C.
calceolus var. pubescens and C. calceolus var. pannflorum as distinct species. The
derived sister positions for C. calceolus var. pamiflomm and C. kentuckiense as
compared to C. calceolus var. pubescens in all topologies obtained in this study
necessitates the same treatment for both taxa. Either both are varieties of C.
calceolus or both are distinct species. Although taxonomic judgement is reserved
in this study, due to the use of a single data set, the topology for these two taxa is
such that the same status needs to be applied to both.

Atwood (1984) recommended the recognition of three distinct species on the
basis of relative size of the plants and floral features, although much overlap in
quantitative characters occurs. Likewise, Bergstrom et al. recommended the
recognition of C. calceolus var. pame and C. calceolus var. pubescens as distinct
species on the basis of differences in floral fragrance composition. One problem
with the conclusion reached by Bergstrom et al. (1992) is that only one individual
of each taxon was sampled (see Chapter 1). A consideration of the apparently
variable nature of floral fragrance-compound data in Cypripedium (consider
differences between Bergstrom et al. (1992), Holman (1983) and results reported
in Chapter 1) and the high genetic identity shared by C. calceolus var. pubescens and
C. calceolus var. pannflomm obtained by Case (1993) , does not support the
recognition of separate species. It follows, on the basis of this study, that C.

kentuckiense should be considered for inclusion within C. calceolus.

76

Cfln'ipedium candidum

The monophyly of the C. calceolus complex including C. candidum is not clearly
resolved. The synapomorphies which link it to the complex or to certain taxa
within it are the benzaldehyde pathway, the linalool pathway, and the expression
of phenyl ethanol. Additionally, in considering the pathways which separate it
from the complex, one is due to the lack of monoterpene hydrocarbon biogenesis
in C. candidum The expression of these compounds is reportedly variable within
populations (Chapter 1), and Holman (1983) has indeed reported the presence of
a monoterpene (probably terpinene) in C. candidum Finally, the overwhelming
morphological similarity between C. candidum and C. calceolus justifies its inclusion

within the monophyletic lineage.

CWpedium macranthum

There is some evidence to suggest the inclusion of C. macranthum within the
monophyletic C. calceolus complex, excluding C. candidum This taxon is not
native to North America although it is cultivated by orchid growers. It does not
share any striking morphological features with the members of the complex,
although the labellum is of the same general construction in that the orifice is
dorsal on the inflated pouch. Stoutamire (1967) reported that C. calceolus and C.
macranthum may artificially hybridize. Case (pers. comm.) found C. macranthum to
share several alleles in common with C. calceolus, although this evidence is based
upon investigation of a single individual. In spite of these similarities, the shape of
the lateral petals and the flower color is very different from any taxon within the
complex. Although the inclusion of C. macranthum within the clade is suggested
by three of the topologies, the primary synapomorphy supporting its inclusion is
the monoterpene hydrocarbon ocimene. Other synapomorphies include
biogenesis of general precursors shared with certain taxa within the complex, but

no identical compounds. This is in contrast to C. candidum which shares as many

77

synapomorphic pathways, and in addition the fragrance compounds produced are
identical to those produced by other taxa within the complex. The presence of
lilac alcohol in C. macranthum may represent shared biogenetic expression with
the complex, but the origin of the epoxide alcohol is unknown. The character
states between these two taxa appear to be incompatible because the two are never
found together with the C. calceolus complex. In fact, a topology which includes
both of these taxa requires at least two extra steps to accomodate the character-
state changes. While support is available for the monophyly of C. macranthum with
the C. calceolus complex, it probably is not more closely related to C. calceolus than
is C. candidum

The alternative position of C. macranthum as sister to C. acaule is unlikely. Only
one synapomorphy supports the clade, presence of lilac alcohol biogenesis. The
two taxa are very different in floral morphology and the presumed relationship
based upon fragrance biogenesis may represent a parallelism.
Cfim’pedium arietinum

The placement of C. arietinum with C. reginae is supported in seven of the trees.
The segregation of C. arietinum into Criosanthes (Atwood 1984) is not supported by
this analysis. The derived sister-taxon placement of C. arietinum and C. reginaewas
not expected, however. Little confidence is placed in this sister-taxon relationship
due to the support for it deriving primarily from the synapomorphic absence of
nearly all fragrance biogenetic expression. Although this topology was supported
by seven of the trees, it is not supported by tree 10 which is most congruent with
the results of Case (1993). The possibility that the most common reconstructed
placement of these taxa is not correct, illustrates the importance of discussing
alternative most-parsimonious topologies, not simply relying upon the consensus

tree.

78

Other species

Phylogenetic interpretations for the rest of the morphologically diverse taxa in
the genus, based upon the topology most congruent with the results of Case
(1993), include a basal position for C. guttatum to the rest of the species.
Cypripedium arietinum is the next taxon to diverge from these other species. The
three most derived species within the genus are C. reginae, C. acaule and C.
macranthum The advanced position of C. macranthum among these taxa may be
artificial due to the possibility of its inclusion within the C. calceolus complex.
Albert (1993) reports phylogenetic relationships within the genus based upon
rbcL variation. He did not sample the C. calceolus complex, C. acaule or C.
guttatum. Of the taxa sampled, C. plectrochilon (putatively closely related to C.
arietinum) was basal in the genus, with C. reginae more derived (Albert pers.

comm.). The topology of tree 10 supports this proposed evolutionary history.

Orienting the tree without an outgroup

The lack of resolution for the ingroup using the specified outgroup leaves the
rooting of the tree, based solely on the data at hand, a diflicult task. The ad hoc
assumption that substituted skeletons represent enzymatic complexity and perhaps
an evolutionarily derived state could justify the rooting of the trees with taxa which
do not contain substituted compounds. No rooting could be performed in this
manner because all taxa appear to possess fragrance compounds which are
enzymatically advanced as well as some which are enzymatically simple. Another
ad hoc assumption might be that loss of biogenetic complexity is an advanced state
(Seaman 8: Funk 1983). The results of this analysis would indicate that C.
arietinum and C. reginae are the most advanced within the genus using this
criterion. This alternative is unlikely given the results of Albert (1993) in which C.

arietinum and C. reginae occupy basal and derived positions respectively.

79

The distribution of character states upon the branches of the trees of Figures
11—19 illustrates two points. First, the most parsimonious reconstruction of the
character states represents the simplest distribution of enzymatic pathways
exhibited by these taxa for total fragrance biogenesis. Steps on the cladogram
represent the expression of an entire, in the case of binary coded characters, or
part of a biogenetic pathway, in the case of multistate characters. Second, the
topology chosen to represent the distribution of biogenetic pathways within
Cypripedium suggests that parts of, or entire pathways can be gained or lost fairly
easily within and between lineages.

The initial outgroup analyses using Paphiopedilum with CWpedium yielded
important information for the use of fragrance data in reconstructing evolutionary
history. A limitation to the use of floral fragrance-compound data for evolutionary
reconstruction is in the widespread nature of the compounds. The biogenetic
pathways for monoterpene hydrocarbon and alcohols are found in many groups of
unrelated plants. Within the Orchidaceae only four of 154 taxa sampled in 60
genera by Kaiser (1993) did not produce either of these two classes of volatiles.
Additionally, approximately 30 out of 150 taxa did not produce the cinnamic acid
derivatives and 15 out of 150 did not produce benzoic acid derivatives. Because of
the ubiquitous distribution of these compounds between genera, it is not
surprising that the data are of little importance as phylogenetic markers above the
level of species. In contrast, the distribution of fatty acid derivatives such as
straight chain alcohols and acetates, greater than 10 carbons in length, are very
restricted. Only 35 of 150 species sampled produced compounds of these types.
The relative distributions of these compounds between taxa are predicted to be

more reliable indicators of phylogenetic relationships.

CONCLUSIONS

The use of floral fragrance-compound data appears to be usefiil in
phylogenetic reconstruction. Effective representation of the biogenetic pathways
which result in the diversity of fragrance compounds found in Cypripedium can be
achieved using the user-defined characters of PAUP. This is a better method for
representing the data than was the coding procedure utilized for the phenetic
analysis of Chapter 1. The relative advantage lies in the ability to express
biogenetic dependency between character states in the cladistic analysis.
Additionally, the distribution of Character states on a most parsimonious
reconstruction are easily interpreted as parts of biogenetic pathways in which each
step represents a biochemical conversion. The phylogenetic interpretations of the
cladistic analysis are highly congruent with those obtained by other recent
workers. Results include the monophyly of the C. calceolus complex, with the
possible inclusion of C. macranthum, a relatively basal position for C. arietinum and
its inclusion within the genus, and derived positions for C. acaule and C. reginae.
The position of C. guttatum is unresolved using this data set. The primary
shortcoming of the cladistic analysis using floral fragrance data is that outgroup
analysis is not possible because of the widespread nature of the biogenetic
pathways to produce these compounds in many flowering plants. While
conservative phylogenetic interpretations have been made using this single data
set, broader conclusions may be possible from an analysis involving many
independently derived data sets, including information from morphology,

anatomy, cytology, and DNA sequences.

80

BIBLIOGRAPHY

BIBLIOGRAPHY

Albert, V. A. 8: Chase, M. W. 1992. Mexipedium: A new genus of slipper orchid.
Lindleyana 7(3): 172-176.

Atwood,J. T. 1984. The relationships of the slipper orchids (subfamily
Cypripedioideae, Orchidaceae). Selbyana 7: 129—247.

Atwood,]. T. 1984. The Cypripedium calceolus L. complex in North America. In
ee in : Thel orl r i confer nce. Ed. Kiat Tan, International
Press Co. Singapore.

Bergstrom, G., Birgersson, G., Groth, 1. 8c Nilsson, LA. 1992. Floral fragrance
disparity between three taxa of lady's slipper Cypripedium caleeolus.
Phytochemistry 31 (7): 2315-2319.

Borg-Karlson, A.-K. 1990. Chemical and ethological studies of pollination in the
genus Ophrys (Orchidaceae). Phytochemistry 29(5): 1359-1387.

Case, F.W. 1987. Orchids of the Western Great La_k_es Region. Cranbrook

Institute of Science, MI.

Case, M. A. 1993. Genetic diversity and evolutionary relationships within and
among five Cypripedium species: Inferences from allozyme electrophoresis.
Ph.D. thesis, Michigan State University, East Lansing, MI.

Chem]. 8c Meeuse,J. D. 1971. Production of free indole by some aroids. Acta
Bot. Neerl. 20(6): 627-635.

Croteau R. 8c Karp, F. 1991. Origin of natural odorants. In Perfumes; Art, Science,
and Technology. Ed. P. M. Muller and D. Lamparsky. Elsevier Applied Science.
NY.

Duncan, T. 8c Baum, B. R. 1981. Numerical phenetics: Its uses in botanical
systematics. Ann. Rev. Ecol. Syst. 12: 387-404.

Erman, W. F. 1985. Chemistry of the monoterpenes; an encyclopedic handbook.

Part A. In Studies in Qrggic Chemistry Ed. Paul G. Gassman. Marcel
Dekker, Inc. NY, NY.

81

82

Gregg, K. B. 1983. Variation in floral fragrances and morphology: Incipient
speciation in Cycnochas? Bot. Gaz. 144(4): 566-576.

Holman, R. T. 1983. Studies on the fragrances of North American terrestrial
orchids. In E. H. Plaxton [ed.], Ncgh Amcri ncg lferrestriag Orchids;
WW; Michigan Orchid Society, Livonia.

Holman, R T. 8c Heimermann, W. H. 1976. First symposium on the scientific
aspects of orchids. eds. H. H. Szmant and]. Wemple, University of Detroit
Publ. pp. 75-89.

Kaiser, R. 1993. The Scent 9f Orchids. Amsterdam, Elsevier.

Klier, K., Leoschke, M.J., 8c Wendel,J. F. 1991. Hybridization and introgression in
white and yellow ladyslipper orchids (Cypripedium candidum and C. pubescens).
]. Heredity 82: 305-318.

Maddison, W. P. and Maddison, D. R. 1992. MacClade: Analysis of phylogeny
and character evolution. Version 3.0. Sinauer Associates, Sunderland,
Massachusetts.

Meeuse, B. 8c Morris, S. 1984. Il_1_c Sex Lifc cf Elma, Oxford Scientific Films
Ltd. New York, NY.

Muller, P. M. Sc Lamparsky, D. 1991. Pcrfumcs; A13 Scicnce and Icchnology.
Elsevier, New York, NY.

Newhouse, C._]. 1976. Pollination biology in seven taxa of Michigan Orchidaceae
and a study of Cypn'pedium calceolus in Michigan based on living plants and
herbarium specimens. M. S. Thesis, Michigan State University.

Nilsson, LA. 1979. Anthecological studies on the Lady's slipper, Cypripedium
calceolus. Bot. Not. 132: 329-347.

Rodman,J. E. 1991. A taxonomic analysis of glucosinolate-producing plants, part
1: Phenetics. Syst. Bot. 16(4): 598-619.

Rohlf, F.]. 1992. Numerical Taxonomy and Multivariate Analysis System. Version
1.70. Exeter Software. New York.

Salatino, A. 8c Gottlieb, OR. 1981. Quinolizidine alkaloids as systematic markers
of the Genisteae. Biochem. System. Ecol. 9: 267-273.

Schreier, P. 1984. Chrgmatogcaphic Studics of Biogcncsis cf Elan; yclag'lcs. Dr.
Alfred Huthig Verlag, GmbH Heidelberg.

83

Schreier P. 1986. Biogeneration of plant aromas. in Developments in Food
Flavours Ed. Birch G. G. and Lindley M.G. Elsevier Applied Science. NY.

Seaman, F. C. 8c Funk, V. A. 1983. Cladistic analysis of complex natural products:

Developing transformation series from sesquiterpene lactone data. Taxon
32 ( l ) : 1-27.

Sneath, P. H. 8c Sokal, R. R. 1973. Principlcs of Numerical Taxonomy. San
Francisco, W. H. Freeman and Co.

Stoutamire, W. P. 1967. Flower biology of the lady's-slippers. Michigan Bot. 6(1):
159—175.

Stuessy, T. F. 1990. Plant Igonomy; Thc Systcmatic Muaticg of Comparagy' e
Dag. Columbia University Press. New York.

Swofford, D. L. 1991. PAUP: Phylogenetic analysis using parsimony, Version 3.05.
Computer program distributed by the Illinois Natural History Survey,
Champaign, Illinois.

Whitten, W. M. Sc Williams, N. H. 1992. Floral fragrances of Stanhopea. Lindleyana
7(3): 130-153.

MICHIGAN STnTE UNIV. LIBRARIES
IIHIIHNIIHMIMImmIllWIII!IIHNHIIHIIIIJHWII
31293010190951