<<

Louisiana State University LSU Digital Commons

LSU Historical Dissertations and Theses Graduate School

1995 The olecM ular and Morphological Systematics of (). Malcolm Ray Neyland Louisiana State University and Agricultural & Mechanical College

Follow this and additional works at: https://digitalcommons.lsu.edu/gradschool_disstheses

Recommended Citation Neyland, Malcolm Ray, "The oM lecular and Morphological Systematics of Subfamily Epidendroideae (Orchidaceae)." (1995). LSU Historical Dissertations and Theses. 6040. https://digitalcommons.lsu.edu/gradschool_disstheses/6040

This Dissertation is brought to you for free and open access by the Graduate School at LSU Digital Commons. It has been accepted for inclusion in LSU Historical Dissertations and Theses by an authorized administrator of LSU Digital Commons. For more information, please contact [email protected]. INFORMATION TO USERS

This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely afreet reproduction.

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced at the back of the book.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to .

A Bell & Howell Information Company 300 North Zeeb Road. Ann Arbor. Ml 48106-1346 USA 313/761-4700 800/521-0600

THE MOLECULAR AND MORPHOLOGICAL SYSTEMATICS OF SUBFAMILY EPIDENDROIDEAE (ORCHIDACEAE)

A Dissertation

Submitted to the Graduate Faculty of the Louisiana State University and Agricultural and Mechanical College in partial fulfillment of the requirements for the degree of Doctor of Philosophy

in

The Department of Biology

by Malcolm Ray Neyland B.A., University of W est Florida, 1980 August, 1995 UMI Number: 9609114

UMI Microform 9609114 Copyright 1996, by UMI Company. All rights reserved.

This microform edition is protected against unauthorized copying under Title 17, Code.

UMI 300 North Zeeb Road Ann Arbor, MI 48103 ACKNOWLEDGEMENTS

As an individual with a business background, I am thankful to the

Department of Plant Biology for accepting me as a student and giving me the opportunity to pursue a career in . I am most grateful for the encouragement and guidance offered by my committee members: Meredith Blackwell, Russell

Chapman, Shirley Tucker, Lowell Urbatsch, and Thomas Wendt. Special thanks are extended to my major professor, Lowell E. Urbatsch, who has provided me with unwavering support during my graduate career. I also thank Kevin Jones, William

Rootes, and Cindy Henk for technical assistance. Plant specimens generously supplied by Missouri Botanical Garden, Royal Botanic Gardens, Kew, University of

Connecticut, Selby Gardens, F.L. Stevenson, Sue Grace, and Margaret Oard were vital to this project. The National Science Foundation (LEGSF RD-A-13) and

NSF/LaSER (LASER/EPSCoR 92-96-ADP-02) provided the monetary support necessary to pursue this dissertation project. TABLE OF CONTENTS

ACKNOWLEDGEMENTS...... ii

LIST OF TABLES...... iv

LIST OF FIGURES...... v

ABSTRACT...... ix

CHAPTER

1 INTRODUCTION ...... 1

2 PHYLOGENY OF SUBFAMILY EPIDENDROIDEAE (ORCHIDACEAE) INFERRED FROM CHLOROPLAST ndhF GENE SEQUENCES AND MORPHOLOGICAL CHARACTERS BASED ON PARSIMONY ANALYSIS...... 32

3 PHYLOGENY OF SUBFAMILY EPIDENDROIDEAE (ORCHIDACEAE) INFERRED FROM CHLOROPLAST ndhF GENE SEQUENCES BASED ON A MAXIMUM LIKELIHOOD ANALYSIS...... 112

4 EVOLUTION IN THE NUMBER AND POSITION OF FERTILE ANTHERS IN ORCHIDACEAE INFERRED FROM ndhF CHLOROPLAST GENE SEQUENCES...... 154

5 A TERRESTRIAL ORIGIN FOR THE ORCHIDACEAE INFERRED FROM ndhF CHLOROPLAST GENE SEQUENCES...... 169

6 A PHYLOGENETIC ANALYSIS OF SUBTRIBE (ORCHIDACEAE)...... 180

7 SUMMARY AND CONCLUSIONS ...... 210

APPENDIXES...... 219

VITA 232 LIST OF TABLES

Table 1.1 Recent classification systems of subtribes Epidendroideae and Vandoideae ...... 10

Table 2.1 List of taxa included in this study. Orchidaceae members are arranged according to the of Dressier (1993)...... 37

Table 2.2 Sources of plant material and voucher information for taxa sequenced in this study. Acronyms of herbaria or persons that donated specimens are listed. GenBank accession numbers for ndh? gene sequences are given ...... 40

Table 2.3 Table of informative INDEL characters for the taxa sequenced in this study ...... 51

Table 2.4 Morphological characters and recognized character states for orchid taxa presented in Table 2.2 ...... 53

Table 2.5 Morphological character states for the orchid taxa listed in Table 2 .2 ...... 55

Table 3.1 Log likelihood values and tree topology designations resulting from maximum likelihood searches using the given transversion weighing parameters, with jumble seed (137) parameter in effect versus not in effect ...... 119

Table 3.2 Log likelihood values and designated tree topologies resulting from maximum likelihood searches using the jumble seeds indicated. All tests were performed using the transversion weighing parameter 1.1 ...... 125

Table 6.1 Classification scheme of Pleurothallidinae from Pridgeon (1982b) ...... 184

Table 6.2 Outgroup members examined for characters listed in Table 6.3 ...... 186

Table 6.3 Character states for Pleurothallidinae and outgroup members. Characters and character states are identified in Appendix A. Missing data are indicated by question marks...... 188

iv LIST OF FIGURES

Figure 2.1 Primer sequences used in this study and their approximate positions on the ndhF gene ...... 47

Figure 2.2 Character state changes graph across the entire ndhF gene from single representatives from angiosperm families: Poaceae, Orchidaceae, Caprifoliaceae, and Solanaceae. Each bar represents the number of site changes averaged over 20 base positions ...... 48

Figure 2.3 Rate of change graph by codon position from the taxa used in this study (Table 2.2). Graph restricted to ndhF sequence positions 1072-2304 ...... 57

Figure 2.4 Strict consensus tree of 570 trees discovered from 100 random addition searches using weighted ndhF sequences. Bootstrap values are indicated below each branch ...... 59

Figure 2.5 Strict consensus tree of 25,100 trees discovered from a single parsimony analysis of amino acid data converted from ndhF DNA sequences...... 60

Figure 2.6 Strict consensus tree of 531 trees discovered from 100 random addition parsimony searches using ndhF sequences ...... 61

Figure 2.7 Strict consensus tree of 90 most-parsimonious trees discovered from 100 random-addition searches using ndhF sequences and binary-coded INDELS...... 62

Figure 2.8 Strict consensus tree of 7 most-parsimonious trees discovered from 100 random-addition parsimony searches using a data matrix consisting of 13 morphological characters. Bootstrap values are indicated below each branch ...... 63

Figure 2.9 Single parsimony tree discovered from 100 random addition replications using a data matrix consisting of ndhF sequences and morphological characters ...... 64

Figure 2.10 Summary of deletions and frameshifts in , , Cattteya, , , and sequences...... 72

Figure 2.11 Character-state tree for growth habit for orchid taxa included in this study ...... 82 v Figure 2.12 Character-state tree for type for orchid taxa included in this s tu d y ...... 8 3

Figure 2.13 Character-state tree for infloresence type for orchid taxa included in this study ...... 84

Figure 2.14 Character-state tree for resupination type for orchid taxa included in this study ...... 85

Figure 2.15 Character-state tree for aggregation type for orchid taxa included in this study ...... 86

Figure 2.16 Character-state tree for number of pollinia for orchid taxa included in this study ...... 94

Figure 2.17 Character-state tree for caudicle type for orchid taxa included in this study ...... 95

Figure 2.18 Character-state tree for type for orchid taxa included in this study ...... 96

Figure 2.19 Character-state tree for velamen type for orchid taxa included in this study ...... 103

Figure 2.20 Character-state tree for seed type for orchid taxa included in this study ...... 104

Figure 2.21 Character-state tree for style fusion for orchid taxa included in this study ...... 105

Figure 3.1 Frequency of minimum and maximum base changes between states calculated from a single tree discovered by parsimony analysis using ndhF sequences for all taxa included in this study ...... 116

Figure 3.2 Frequency of base changes between states presented by circular areas calculated from a single tree discovered by parsimony analysis using ndhF sequences for all taxa included in this study ...... 117

Figure 3.3 Maximum likelihood tree (A); log likelihood value = -7,473.95. Branch lengths in terms of expected nucleotide substitutions per site are designated above each branch. Branches also present in the strict consensus parsimony tree (Fig. 2.9) are indicated by an asterisk; those branches that received £ 50% bootstrap support are indicated by a double asterisk. Polymorphic orchid tribes are indicated by shaded regions in the side legend ...... 120 vi Figure 3.4 Maximum likelihood tree (D); log likelihood = -7,730.05 ...... 121

Figure 3.5 Log likelihood values versus transversion weighing parameters graph for maximum likelihood searches. The jumble option is not in effect for these searches ...... 122

Figure 3.6 Maximum likelihood tree (B); log likelihood = -7,479.66...... 126

Figure 3.7 Maximum likelihood tree (C); log likelihood = -7,517.80 ...... 127

Figure 3.8 Log likelihood values versus transversion weighing parameter graph for maximum likelihood searches with the jumble option in effect (seed = 137) and with the jumble option not in effect ...... 128

Figure 3.9 Maximum likelihood tree (E); log likelihood = -7,487.03 ...... 129

Figure 3.10 Character-state tree for growth habit among orchid taxa included in this study ...... 141

Figure 3.11 Character-state tree for leaf type among orchid taxa included in this study ...... 142

Figure 3.12 Character-state tree for infloresence type among orchid taxa included in this study ...... 143

Figure 3.13 Character-state tree for resupination among orchid taxa included in this study ...... 144

Figure 3.14 Character-state tree for pollen aggregation type among orchid taxa included in this study ...... 145

Figure 3.15 Character-state tree for number of pollinia among orchid taxa included in this study ...... 146

Figure 3.16 Character-state tree for caudicle type among orchid taxa included in this study ...... 147

Figure 3.17 Character-state tree for stipe type among orchid taxa included in this study ...... 148

Figure 3.18 Character-state tree for velamen type among orchid taxa included in this study ...... 149

Figure 3.19 Character-state tree for seed type among orchid taxa included in this study ...... 150

vii Figure 3.20 Character-state tree for degree of fusion among orchid taxa included in this study...... 151

Figure 4.1. Designated positions of the liliaceous-type . Diagram is oriented in the resupinate position for comparison to the orchids ...... 156

Figure 4.2. Character-state tree for number of fertile anthers adapted from the strict consensus tree of 531 trees discovered from 100 random addition parsimony searches using unweighted ndhF sequences...... 159

Figure 4.3 Character-state tree for fertile anthers adapted from the maximum likelihood tree with the log likelihood = -7,473.95...... 160

Figure 5.1 Character-state tree for growth substrate adapted from the strict consensus tree discovered from 100 random addition parsimony searchesusing unweighted ndhF sequences ...... 173

Figure 5.2 Character-state tree for growth substrate adapted from the maximum likelihood tree with the log likelihood value = -7,473.95 ...... 174

Figure 6.1 Strict consensus of eight trees discovered from a parsimony search for the terminal taxa included in this study (Table 6.3) ...... 198

Figure 6.2 One of the eight most-parsimonious trees discovered from a cladistic search for the taxa included in this study. The value above each branch represents the number of unambiguous synapomorphies ...... 199

Figure 6.3 Character state tree showing the distribution of states for the number of pollinia (character 44) for the taxa included in this study (Table 6.3) ...... 207

viii ABSTRACT

The current project undertakes the first molecular-based phylogenetic study of subfamily Epidendroideae (Orchidaceae). Approximately 1,200 nucleotide bases (from the

3' half of the ndhF chloroplast gene) for 34 orchid taxa, and a lilioid monocot, Olivia m iniata. (Amaryllidaceae) were subjected to phylogenetic analysis using parsimony methods. Oryza sativa (Poaceae), a nonlilioid monocot was designated as outgroup.

Using unweighted ndhF sequences, the strict consensus cladogram of 531 most- parsimonious trees supports the hypothesis that the large subfamily Epidendroideae is monophyletic, with Listera () as sister. Although subtribal-level relationships in subfamily Epidendroideae are well resolved in this analysis, tribal-level relationships are resolved poorly. A set of 13 morphological characters were combined with unweighted ndhF sequences and used in parsimony analyses. Although the addition of these characters brings an increased level of resolution to the intertribal relationships in

Epidendroideae, branch support for these relationships is weak. Six taxa in this study exhibit deletions that are not evenly divisible by three which results in extensive sequence frameshifts. This suggests that ndhF may be a pseudogene, in these six taxa. A maximum likelihood analysis was also undertaken to infer phylogenetic relationships for the taxa included in this study. To determine the maximum likelihood tree with the greatest log likelihood, an array of transversion weighing parameters and jumble seeds were used in this analysis. A tree with the greatest log likelihood value (-7,473.95) was discovered with a transversion parameter of 1.1. This maximum likelihood tree, suggests that subfamily Epidendroideae is monophyletic with Listera (Neottieae) as sister. Although trees discovered from these two methods of phylogenetic inference are congruent in many respects, topological differences typically occur on branches that define intertribal 116 relationships among the epidendroids. It is hypothesized here that this lack of support is due to a rapid radiation of the epidendroids which coincided with anatomical, morphological, and physiological adaptations that allowed the epidendroids to pioneer xeric epiphytic microhabitats.

x CHAPTER 1

INTRODUCTION

1 The Orchidaceae Juss. is possibly the largest of flowering .

The growing and selling of orchids (primarily as greenhouse ornamentals) is a world

wide business. Vanilla, used to produce flavoring, is grown as a crop especially in

Mexico and Madagascar. According to Atwood (1986), there are over 19,000

known orchid but as many as 23,000 may eventually be discovered.

Orchids are global in distribution (with the exception of Antarctica), but their

greatest diversity occurs in the tropics of southeastern and .

About 73% are epiphytic or lithophytic with the remainder occupying terrestrial

habitats (Atwood, 1986). Although the orchids appear to be an easily recognizable

group, there are few features that consistently distinguish them from other

monocots. Dressier (1981) lists the distinguishing features of the family:

on the abaxial side of the flower; stamen and pistil at least partially united; seeds

tiny and numerous; usually with a lip or and typically resupinate;

pollen usually aggregated into pollinia.

Comprising approximately 80% of all orchid species, the Epidendroideae

(sensu Dressier, 1993) is the largest subfamily. Unlike the remaining

that include mostly terrestrial species, the Epidendroideae are typically epiphytic.

At present, there is no general agreement concerning the higher systematics of

Orchidaceae, especially interfamilial and subtribal relationships. Previous studies

based on anatomical/morphological characters are generally poorly resolved and

considered tentative (Burns-Balogh and Funk, 1986; Dressier, 1993). No molecular study addressing orchid relationships above the tribal level has been published. SCOPE OF STUDY

The present work includes several interrelated phylogenetic studies and addresses the interrelationships of orchid subfamilies (sensu Dressier, 1993); of the tribes and subtribes of subfamily Epidendroideae (sensu Dressier, 1993), and of the genera of subtribe Pleurothallidinae (sensu Pridgeon, 1982). The data consist of ndhF chloroplast DNA sequences, insertion/deletions (INDELS) discovered as a result of the sequence alignment, and anatomical/morphological characters.

Phylogenetic analyses were performed with the data set using parsimony and maximum likelihood methods. Bootstrap confidence values and decay indices

(parsimony methods), and expected nucleotide changes per site (maximum likelihood methods) were used to test for tree robustness.

RESEARCH OBJECTIVES

This dissertation is an exercise in morphological and molecular phylogenetics. The specific objectives of this research were: 1) to test the monophyly and intertribal relationships of the orchid subfamily Epidendroideae with respect to the other orchid subfamilies, 2) to examine the level of congruence between phylogenetic trees derived from parsimony and maximum-likelihood methods, 3) to assess the phylogenetic utility of ndhF sequences, INDELS, and morphological characters, 4) to infer phylogeny in subfamily Epidendroideae, 5) to infer evolution of certain anatomical/morphological characters in Orchidaceae and subfamily Epidendroideae, and 6) to infer a phylogeny of subtribe Pleurothallidinae using anatomical/morphological characters. TAXONOMIC HISTORY

Perhaps the first attempt to subdivide the Orchidaceae was made by Lindley

(1830-1840). For the approximate 2,000 orchid species known at that time, he

recognized seven tribes (, , , ,

Ophrydeae, Neotteae, Cypripedeae) based on anther number and pollinarium

structure. He treated Apostasiaceae as a separate family. Lindley has been named

the father of orchidology and his taxonomic treatment serves as the basis for

modern orchid classification (Burns-Balogh and Funk, 1986). Lindley was the first author to distinguish the vandoid orchids (Vandeae) from the epidendroid orchids

(Epidendreae). The uncertain relationship between the vandoids and the epidendroids has been a taxonomic problem that has been addressed by orchidologists from the time of Lindley to the present.

Bentham's (1881) work is seen to be merely a modification of Lindley’s taxonomic treatment (Dressier, 1981). Like Lindley, Bentham distinguished the

Epidendreae from the Vandeae by: 1) the distinctness of the two anther cells, which are always parallel, or nearly so; 2) margins prominent within the anther case after the discharge of pollinia; 3) pollinia removal without the removal of scale gland or stipes. In contrast, the Vandeae typically have superposed pollinia (if four), weakened anther margins after pollinia removal, and the stipe removed with the pollinia.

Pfitzer (1889) divided the Orchidaceae into monandrous and diandrous groups which he termed the Monandrae and Diandrae respectively. He recognized two diandrous tribes, and divided the monandrous orchids among 29 tribes largely using vegetative and characters. Some orchidologists regard Pfitzer's system as highly artificial (Dressier 1981; Burns-Balogh and Funk, 1986).

Schlechter (1926) relied on anther and pollinarium characters in recognizing four tribes: Cypripedilioideae; Ophrydoideae; Polychrondreae; Kerosphaereae.

Although the names used by Schlechter were Pfitzer's, the names used by Lindley

(1830-1840) and Bentham (1881) had priority under the rules of nomenclature

(Dressier, 1981). Schlechter’s system is still used in some major herbaria (Burns-

Balogh and Funk, 1986).

Mansfeld’s (1937) treatment of the monandrous orchids was viewed by

Dressier (1981) as an improvement of Schlechter's 1926 work. His major groups were basically those of Lindley (1830-1840), but many genera were moved among the different groups (Burns-Balogh and Funk, 1986).

Garay (1960) delimited five subfamilies within Orchidaceae:

Apostasioideae, , Neottioideae, Ophrydoideae, Kerosphaeroideae.

The epidendroids and vandoids were both included in the Kerosphaeroideae. He stated that attempts to assign a definite position and sequence of each of these subfamilies in a linear arrangement was unrealistic since the groups are constantly in "juxtaposition" with each other.

Dressier and Dodson (1960) recognized two subfamilies in Orchidaceae:

Cypripedioideae and Orchidoideae. The monandrous orchids were included in the subfamily Orchidoideae. The Epidendreae was recognized as one of five tribes in the Orchidoideae. Citing previous authors who had distinguished the tribes

Epidendreae and Vandeae based on pollinium structure and presence or absence of a stipe, Dressier and Dodson argued that these characters did not sufficiently

distinguish these groups as separate tribes. Although Dressier and Dodson

segregated the orchids into tribes and subtribes based on their level of

specialization, the classification still contained the principal groups devised by

Lindley (Burns-Balogh and Funk, 1986). According to Dressier (1981), although

Dressier and Dodson's work modified Schlechter's treatment of the orchids

somewhat, it was primarily undertaken to bring Schlechter’s treatment in line with

the rules of nomenclature.

Vermeulen (1966) recognized three families in the Orchidales:

Apostasiaceae, Cypripediaceae, Orchidaceae. The Orchidaceae included the

monandrous orchids and was divided into two subfamilies. The first was the

Orchidoideae and was characterized as having a stamen that is broadly inserted

with two pollinia, each with a caudicle basally connected with the viscidium of the

rostellum. The second was the Epidendroideae and was diagnosed as having a stamen with filament and 2-8 pollinia. The Epidendroideae were not seen by

Vermeulen to form a homogenous group; he described the Neottianthae as mostly

small terrestrial herbs occurring outside the tropics, whereas the Epidendranthae were described as mostly epiphytes from tropical regions.

Dressier (1981) published a comprehensive taxonomic treatment of the

Orchidaceae in which he recognized six subfamilies: Rchb.,

Cypripedioideae Lindl., Orchidoideae, Spiranthoideae Dressier, Epidendroideae Lindl., and Vandoideae Endlicher. He thought the Epidendroideae were less specialized than the Vandoideae and contended that the one feature that united the

Epidendroideae was an "incumbent" anther which is erect in the young flower bud but subsequently bends downward over the apex of the column until it is at a right angle to the column axis at maturity. By swiveling on its point of attachment to the column, this anther type allows for the successful placement of pollen onto a vector. Dressier (1981) believed that the one feature that distinguishes the

Vandoideae from the Epidendroideae is the development of the anther which does not bend down over the apex of the column during development.

Rasmussen (1985) treated the orchids as three distinct families in the order

Orchidales: Apostasiaceae, Cypripediaceae, and Orchidaceae. He did not claim that his classification was phylogenetic in the sense of all taxa being monophyletic, and he considered some groups paraphyletic and defined by shared primitive characters. In that treatment, the monandrous orchids comprised the Orchidaceae and were classified under four subfamilies: Neottioideae, Orchidoideae,

Epidendroideae, and Vandoideae. As in Dressler's 1981 work, Rasmussen (1985) made a distinction between the epidendroid and vandoid orchids. The

Epidendroideae consisted of eight tribes and was distinguished by an "incumbent" anther with a tunica-like connective and operculate anther cap, usually hard pollinia, and possession of at least one vegetative adaptation for epiphytic life. The

Vandoideae was segregated by tegular stipes and even harder pollinia. Rasmussen noted that the inflexion of the anther in Vandoideae taxa began much earlier in the ontogeny of the gynostemium than in the Epidendroideae. Because this inflexion may continue up to 180°, he thought this represented a further specialization.

Burns-Balogh and Funk (1986) performed a cladistic analysis of the

Orchidaceae using 42 floral and vegetative characters. Their classification recognized seven subfamilies: Neuwiedioideae Burns-Balogh and Funk,

Apostasioideae, Cypripedioideae, Spiranthoideae, Neottioideae, Orchidoideae, and

Epidendroideae. Included in the subfamily Epidendroideae were ten tribes and the

Pleurothallis Group. The vandoid orchids were treated as a within the

subfamily Epidendroideae. The Epidendroideae tribes presented some problems to

Burns-Balogh and Funk (1986) because of a lack of information on column

structure. Therefore, they considered their treatment of the Epidendroideae as

tentative.

Dressier (1989) listed the characteristics that had previously been used to

distinguish the vandoid orchids from the epidendroid orchids: 1) early bending of the

anther in floral development, 2) lateral infloresences, 3) reduced number of pollinia

(four or two), 4) superposed pollinia (if four), 5) well developed viscidium, and 6) tegular stipe. Based on KurzweiPs (1987) floral developmental studies that demonstrated anther ontogeny in the vandoid orchids is similar to that in the epidendroid orchids, Dressier (1989) concluded that anther development is not a character that distinguishes the vandoids from the epidendroids. This finding was reflected in Dressler’s (1990) revision in which he eliminated the subfamily

Vandoideae and distributed its tribes within subfamily Epidendroideae. In that revision, Dressier recognized 19 tribes in Epidendroideae that were segregated into three informal categories: primitive tribes, the cymbidioid phylad, and the epidendroid phylad. The vandoid orchids were distributed among both the cymbidioid and epidendroid phylads. Specifically, tribes Malaxideae, Calypsoeae,

Cymbidieae, and Maxillarieae comprised the cymbidioid phylad, whereas Vandeae was placed in the epidendroid phylad. Dressier (1993) contended that the vandoid

orchids as a group are polyphyletic in that the tribe Vandeae is not closely related to

the other vandoid tribes.

Dressier (1993) published his second comprehensive taxonomic treatment

of the orchids and recognized five subfamilies: Apostasioideae, Cypripedioideae,

Spiranthoideae, Orchidoideae, and Epidendroideae. The Epidendroideae consisted of

11 tribes and 38 subtribes. In that revision. Dressier retained the informal

cymbidioid and epidendroid phylads but removed the "primitive" tribes from the

Epidendroideae and referred to them as 'primitive orchids with uncertain affinities in

Orchidaceae'.

As defined by Dressier (1993), the epidendroid phylad is distinguished by the

presence of eight pollinia in its primitive members and includes most taxa with a reed-stem growth habit. Within the epidendroid phylad. Dressier recognized the informal dendrobioid subclade which he characterized as having uniquely derived spherical silica bodies, frequent upper lateral , and chromosome numbers limited to 38, 40, and 42. Dressier characterized most of the cymbidioid phylad members as having vandoid characteristics. Dressler's taxonomic treatment is adopted in this study as it represents the most current revision of the

Orchidaceae and attempts to apply the principles of cladistics to infer relationships in Orchidaceae. Recent classification systems for subfamilies Epidendroideae and

Vandoideae are presented in Table 1.1. 10

Table 1.1 Recent classification systems of subtribes Epidendroideae and Vandoideae.

Taxonomist Epidendroideae Vandoideae

Dressier (1993) Malaxideae Calypsoeae Maxillarieae . Arethuseae Coelogynae Epidendreae (New World) Epidendreae (Old World) Vandeae

Balough & Funk (1986) Malaxideae Maxallarieae Arethuseae Coelogyneae Epidendreae Dendroibieae Vandeae Vanillieae Gastroidieae Triphoreae Pleurothaliis Group

(table cont'd) Rasmussen (1985) Malaxideae Maxillarieae Arethuseae Cymbidieae Coelogyneae Vandeae Epipogieae Polystachyeae Epidendreae Vanilleae

Dressier (1981) Malaxideae Maxillaireae Calypsoeae Cymbidieae Arethuseae Vandeae Coelogyneae Polystachyeae Epidendreae Vanillieae Gastrodieae Epipogieae Cryptarrheneae 12

ORIGINS

There is no general agreement on the time and place of origin for

Orchidaceae, but several authorities have postulated a Cretaceous Southeast Asian origin (Garay, 1972; Benzing, 1987; Arditti, 1992). According to Garay (1960), the association of primitive with advanced characters found in each of the major orchid groups speaks in favor of considering the orchid family as one of those groups established relatively early during the evolution of Angiosperms. The presence of pseudocopulation in some orchid genera was mentioned by Garay as additional support for an early orchid origin. He reasoned that if the orchid family evolved relatively recently, such a precarious adjustment between the flowering time and the emergence of the fertilizing agents could hardly have come about. According to

Raven and Axelrod (1974) the two least specialized subfamilies (Apostasioideae and

Cypripedioideae) are Laurasian and mainly east Asian; which may suggest the area of orchid origin. According to Benzing (1987), tropical forests, with their canopies suitable for epiphytic growth, probably arose in the early Tertiary period. Once epiphytism was possible, orchid diversity increased due to expanding emigration and subsequent radiation. At present, all but one subfamily (Apostasioideae) has a range that extends across several continents. However, tribal and lower taxonomic levels tend to be localized. For example, less than 5% of the approximate 800 orchid genera are found in both the Old and New Worlds (Benzing, 1987).

Although fossils may provide evidence for the early history of a group, the

Orchidaceae has no positive or useful fossil record (Schmid and Schmid,1977; 13

Benzing, 1987; Dressier, 1993). Four macrofossils {Antholithes, Paleorchis,

Protorchis, Orchidacites) have been attributed by various experts to the orchids or

protorchids. According to Benzing (1987), none of the compressed shoots, tubers,

or fruits represented by these fossils can be assigned to any modern orchid form

without serious reservation, nor is their relegation to the Liliopsida beyond

challenge. Schmid and Schmid (1977) stated that these fossils are hardly reliable

records for the Orchidaceae since they were assigned to this family on purely

superficial grounds.

Schmid and Schmid (1977) attributed the lack of fossil evidence to the

following characteristics of most Orchidaceae: 1) predominant occurrence, both in the present and presumably in the distant past, in the wet tropics, which are areas

of rapid decay, 2) herbaceous habit, 3) epiphytic habit, which would generally

preclude orchids from the conditions (usually aquatic) most conductive to fossilization, 4) production of pollinia (usually) rather than individual pollen grains,

and dispersion of these by animal vectors instead of by wind, and 5) minute, easily

degradable seeds. Even if orchid pollen were preserved as fossils, it may not be

recognized as such. Botanists could easily fail to recognize fossilized pollen of

orchidaceous forms if orchids had not yet evolved pollinia. Individual pollen grains

derived from fragmented pollinia might not be identifiable as orchid pollen.

Wolter and Schill (1985) tested the resistance to degradation of several types of orchid pollen packaging (monads, tetrads, massulae, pollinia) to acetolysis.

Tetrads are held together in groups by "elastoviscin," thus forming a pollinium.

Their study showed that elastoviscin is susceptible to acetolysis. Furthermore, 14 their study also suggested that pollen accessory structures (caudicles, viscidia, stipes) are not resistant to acetolysis. They concluded that if microorganisms and time have not destroyed typical features of orchid pollinia, acetolysis, routinely used in processing fossil pollen, would do so.

MYCORRHIZA

Mycorrhiza is a term that refers to the symbiotic relationship between a and the of a . An understanding of orchid reproduction, physiology, anatomy, and ecology requires a basic knowledge of this relationship.

Most orchid seeds cannot germinate without a mycorrhizal infection; however, mature plants seem less dependent on this relationship (Dressier, 1981).

Orchid mycorrhizal fungi have traditionally been classified in the form

Rhizoctonia. Once the sexual stages of these fungi become known, they are typically reassigned to other fungal genera. Orchid seeds cannot use their own lipid reserves, break down starch, nor photosynthesize. Therefore, they are totally reliant on a mycorrhizal infection to foster successful germination. As described by

Arditti (1992), the infection of orchid seeds begins with a fungal penetration of the testa at the suspensor end of the seed and spreads to adjacent cells by means of single hypha. Hyphae subsequently branch and form a network called a "pelton."

These peltons rupture and release their contents into the orchid cells. It is presumed that the pelton contents provide the orchid with nutrients that include sugars, vitamins, and amino acids. After successful infection, the seed develops into a protocorm. As the protocorms grow, rhizoids are produced and hyphae grow 15 into them as well. No specific mechanism is known that suggests the fungus is attracted to the seed. At present, the evidence suggests that the fungus encounters the seed by chance.

Orchid roots are penetrated by fungi through epidermal cells or root hairs.

The roots of terrestrial species are often heavily infected, whereas those of epiphytic species become infected only after coming into contact with a fungus- containing substrate (Arditti, 1992). Digestion of the fungus occurs soon after infection in "digestion" cells. In some orchids, the outer cells of the root act as a host layer where the hyphae are not digested (Dressier, 198; Arditti, 1992).

According to Arditti (1992), phosphatases seem to play a role in the digestion of fungal hyphae, and probably help to localize the fungus and to release its cell contents.

PREVIOUS PHYLOGENETIC ANALYSES

The Orchidaceae has all the characteristics of a family in active evolution; unfortunately, this results in many of its morphological and anatomical characters displaying high levels of homoplasy (Dressier, 1993). To date, few cladistic studies of anatomical-morphological characters have been performed. As previously discussed, Burns-Balogh and Funk (1986) analyzed 42 floral and vegetative characters in a cladistic study of Orchidaceae. A lack of information on the column structure in the largest subfamily, Epidendroideae, renders that study incomplete and tentative. Dressler's 1993 phylogenetic treatment of the Orchidaceae used few characters and was not subjected to parsimony analysis; in that treatment Dressier used only 12 characters to infer a tribal-level phylogeny of the Epidendroideae. No 16 molecular-based study above the tribal level in Orchidaceae has been published.

This present study is the first attempt to resolve the phylogeny of subfamily

Epidendroideae through analysis of molecular characters.

PHYLOGENETIC INFERENCE METHODS

Inferring phylogenetic relationships from molecular data requires the selection of an appropriate method from an array of techniques (Swofford and

Olsen, 1990). Inferring a phylogeny is an estimate of an evolutionary history based on incomplete information. In the context of molecular systematics, direct information about the past is generally unavailable; access is thus restricted to contemporary species and molecules (Swofford and Olsen, 1990).

Two assumptions are necessary in all molecular-based methods of analysis:

1) independence among characters which allows the treatment of each position as a separate entity in computational algorithms (i.e., numbers of substitutions can be minimized separately position by position, and then summed over all positions in a parsimony algorithm, or probabilities can be multiplied over all positions in a maximum likelihood approach), and 2) homology of character states (i.e., a character is defined such that all states observed over all taxa for that particular character have been derived from a corresponding state observed in the common ancestor). If phylogenetic relationships are to be inferred from molecular relationships, the definition of homology must be further restricted to include only orthologous, as opposed to paralogous genes (Swofford and Olsen, 1990).

Of the existing numerical approaches to inferring phylogenies directly from the character data, methods based on the principle of maximum parsimony have been the most widely used by far (Swofford and Olsen, 1990). Inferring phylogeny 17 through parsimony methods operates by selecting trees that minimize tree length

(Felsenstein, 1983; Swofford and Olsen, 1990; Swofford and Begle, 1993). The theory behind parsimony analysis is that a simple hypothesis is preferable to more complicated or ad hoc hypothesis.

As described by Felsenstein (1983), parsimony is a criterion that is ancillary to the Hennigan (1966) system. That is, Hennig gave a clear description for inferring phylogeny if there are no incongruencies among the characters. When faced by a conflict between characters, Hennig did not advocate a method of compromise but recommended that the full set of characters be reexamined until the characters that had been misinterpreted are discovered.

Camin and Sokal (1965) were the first researchers to use parsimony to reconstruct phylogeny (Felsenstein, 1983; Swofford and Olsen, 1990). In that study, Camin and Sokal applied parsimony methods to discretely coded morphological characters to deduce branching sequences. They assumed that the ancestral state was known and that characters could not revert from a derived state back to a plesiomorphic state. It is highly unlikely that the assumption of irreversibility could be justified for most types of molecular data (Swofford and

Olsen, 1990).

Eck and Dayhoff (1966) performed one of the earliest large-scale applications of parsimony methods dealing with protein sequences. Parsimony methods based on their "ancestral sequence method" to infer phylogeny have been widely applied to protein and nucleic acid sequences (Felsenstein, 1983). 18

Formalized by Kluge and Farris (1969), Wagner parsimony assumes that any

transformation from one character state to another follows a defined order. Thus it

is appropriate for binary, ordered, multistate and continuous characters (Swofford

and Olsen, 1990). Fitch (1971) generalized the Wagner method to allow unordered

multistate characters. Thus, this parsimony method is appropriate for nucleotide

and protein sequences. Both Wagner and Fitch parsimony methods allow

reversibility from one character state to another.

According to Felsenstein (1983), unweighted parsimony methods assume that all characters have equal rates of change, a condition that is manifestly not met in much data. However, he advocated that if all the characters change sufficiently slowly, they should be weighted equally, even if they do not change at equal rates. He, therefore, justified the use of equal rates of change employed in most parsimony methods.

Weighted parsimony can be used with molecular data to attach greater importance to transversions than to transitions by assigning costs such that changes between two purines or two pyramidines receive lower weight than transversion changes. Swofford and Olsen (1990), thought that a priori weighting of different kinds of changes does not introduce an unacceptable level of subjectivity into the analysis as an assumption of equal weights is itself a strong assumption. They stated that even when there is no estimate of how much more frequently transitions occur than transversions, a transversion-transition ratio such as 1.1:1 weighting may be desirable. 19

Parsimony methods may yield inconsistent estimates of the evolutionary tree

when amounts of evolutionary change in different lineages are sufficiently unequal

(Felsenstein, 1978; Hasegawa and Fujiwara, 1993). Most data involve moderate to

large amounts of change, and in these cases, parsimony methods can fail

(Felsenstein, 1981). To apply a maximum likelihood approach, a concrete model of

the evolutionary process that converts one sequence to another must be specified.

A maximum likelihood approach to phylogenetic inference evaluates the net

likelihood that the given evolutionary model will yield the observed sequences, and

the inferred phylogenies are those with the highest likelihood (Swofford and Olsen,

1990). The maximum likelihood model accounts for substitutions occurring more frequently between pairs of physicochemically similar sites (Hasegawa and

Fujiwara, 1993), and is suitable in problems where the amount of data is limited

(Goldman, 1990). Assumptions of the maximum likelihood model are: 1) each site evolves independently, 2) different lineages evolve independently, 3) each site undergoes substitution at an expected rate that must be specified, and 4) all sites are included (not just those that have changed) (Felsenstein,1993).

In the present study, both parsimony and maximum likelihood phylogenetic inference methods were used. In an attempt to enhance the accuracy of resulting topologies, ndhF chloroplast gene sequences have been analyzed to determine the respective mutation rates for each codon position to determine whether weighting specific codon positions is appropriate in parsimony analyses. Empirically derived transition-transversion rates were determined for use in parsimony and maximum likelihood analyses. 20

TREE ROBUSTNESS METHODS

As described by Felsenstein (1985), the "bootstrap" can be used to place confidence intervals on phylogenies and involves resampling from a set of data with replacement to create a series of samples of the same size as the original data set.

Felsenstein stated that if a phylogenetic group shows up 95% or more of the time, the evidence for it is considered statistically significant. He contended that the justification for bootstrapping relies on each character evolving independently from the other characters according to a stochastic process that has among its parameters the topology and branch lengths of the underlying phylogeny.

Hillis and Bull (1993) used computer simulations and a laboratory-generated phylogeny to test bootstrapping results of parsimony analyses, both as a measure repeatability (i.e. the probability of repeating a result given a new sample of characters) and accuracy (i.e. the probability that a result represents the true phylogeny). Their simulations indicated that any bootstrap proportion provides an unbiased but highly imprecise measure of repeatability. Under conditions they thought typical of most phylogenetic analyses, however, bootstrap proportions in majority-rule consensus trees provide biased but highly conservative estimates of the probability of correctly inferring the corresponding clades. Specifically, under conditions of equal rates of change and symmetric phylogenies, bootstrap proportions £ 70% usually correspond to a probability of £ 95% that the corresponding clade is real (Hillis and Bull, 1993).

According to Trueman (1993), one perspective on the Felsenstein (1985) bootstrap is that it indicates how support for nodes on the most parsimonious tree 21 is distributed across the characters. Nodes that are supported by changes in a larger number of characters are likely to recur frequently in the replicates. Nodes that are supported by only a few characters will not recur often, even when there is no evidence to contradict their existence. Trueman argued that the bootstrap provides a one-sided test of phylogeny: nodes that recur are supported by the data, but nodes that do not recur cannot be taken as rejected. Therefore, as a matter of confidence in the tree as a whole, Trueman concluded that the bootstrap is quite useless.

It is apparent that the use of the bootstrap method for determining how well the particular data set supports the given tree has its detractors. Despite the problems associated with the bootstrap, it is currently widely used as a method for estimating confidence in topologies and is used as an estimator in this study.

The decay index is another method used to evaluate the relative robustness of clades found in parsimony trees. With this method, strict consensus trees are constructed of all trees that are progressively longer (in terms of steps) until each clade in the minimum length tree is no longer equivocally supported (Donoghue et al., 1992). This procedure results in a "decay" index that is equivalent to the number of steps that must be added before each clade collapses into an unresolved bush. Selected analyses in this study have been subjected to decay index evaluation.

Evaluation of branch lengths in maximum likelihood trees is not straight­ forward. Output produced by fastDNAml (Olsen et al., 1994) software includes a table showing the length of each tree segment in units of expected nucleotide 22 substitutions per site. For each branch segment, confidence limits on their length are supplied. According to Felsenstein (1993), these confidence limits are very rough. Because there is a simplification in how the confidence limits are calculated, over-confidence in the existence of the branch is the result (Olsen; pers. comm.,1994). Therefore, supplied confidence limits were not used to evaluate branch lengths in this study. However, the range of maximum likelihood branch length values was compared to branch lengths found in parsimony-based trees in order to estimate the range of maximum likelihood branch length values that provide a level of significant support.

ndhF cpDNA SEQUENCES

The past decade has seen a proliferation of molecular biological approaches to the study of angiosperm phylogeny (Qlmstead and Palmer, 1994), and the primary source for molecular characters has been the chloroplast genome (Clegg and Zurawski, 1992). Two major factors that distinguish chloroplast DNA (cpDNA) evolution from nuclear gene evolution are the lack of clearly documented transposon activity associated with the genome and apparent lack of recombinational potential (Clegg and Zurawski, 1992). Chloroplast gene order is highly conserved; only two differences are known among the 120 genes present in the cpDNAs of tobacco and liverwort which diverged some 400 million years ago

(Palmer et al., 1988). The tobacco gene order is found in most other angiosperms

(e.g. orchids) and in at least one and one gymnosperm (Palmer et al., 1988). A conservative mode of evolution makes chloroplast DNA an extremely valuable molecule for phylogenetic studies; its conservatism can also be a serious drawback 23

as the amount of useful DNA variation may be limited at the intraspecific level

(Palmer et al., 1988). Chloroplast genes can be regarded as uniparentially

transmitted molecules that collect mutations stochastically; average rates of

synonymous substitutions for cpDNA protein coding genes vary from 0.2 to 1.0 X

10'9 substitutions per site per year (Wolfe and Sharp, 1988). Chloroplast protein

coding genes evolve at an average rate that is 5X slower than plant nuclear genes

(Wolfe et al. 1987). On average, the rate of silent substitutions in chloroplast genes is two to three times less than in nuclear genes and 20 times less than in animal mitochondrial genes (Palmer et al., 1988). Transitions outnumber transversions in chloroplast genes by a factor of less than two to one which is a much smaller bias than that found in animal mtDNA (Palmer et al., 1988).

According to Olmstead and Palmer (1994), the chloroplast genome varies little in size, structure, and gene content among angiosperms; it typically ranges from 135 -160 kilobases (kb) and is characterized by a large (25 kb) inverted repeat that divides the remainder of the genome into one large and one small subunit.

Single copy regions have approximately a four-fold synonymous substitution rate compared to the inverted repeat (Wolf and Sharp, 1987). By far the most frequent cpDNA mutations are point mutations and deletions/insertions in noncoding regions; other mutations (inversions, and insertion/deletions of genes and introns ) are rare

(Palmer et al., 1988).

According to Olmstead and Palmer (1994), there are 20 chloroplast genes that are sufficiently large (> 1kb) and widespread to be generally useful in comparing sequence studies. At present, these sequencing studies have centered 24 on the gene for the large subunit of ribulose -1,5 bisphosphate, carboxylase (rbcL).

However, several studies using other cpDNA sequences have recently been used; one such sequence is the ndhF gene. Olmstead et al. (1993) has constructed gene trees for the two chloroplast genes, rbcL and ndh? for the families Solanaceae,

Scrophulariaceae, Bignoniaceae, Acanthaceae, and Lamiaceae. The resulting gene trees were congruent in many respects. According to Olmstead et al., the ndh? gene has the desirable property of identifying more informative characters at lower taxonomic levels because it is longer and has a higher substitution rate than rbcL.

In comparing sequences from the ndhF and rbcL genes in Solanaceae and

Convolvulaceae, Olmstead and Sweere (1994) found that ndh? is a more rapidly evolving gene than rbcL and yields more phylogenetic information than rbcL.

The ndh? gene resides in the small single copy region of the chloroplast genome.

(Hiratsuka et al., 1989). According to Sugiura (1989), six chloroplast DNA sequences resemble those components (ND1, 2, 3, 4, 4L, and 5) of the respiratory chain NADH dehydrogenase from human mitochondria. As these are highly expressed in tobacco chloroplasts, they are likely to be the genes for components of a chloroplast NADH dehydrogenase (ndhA, B, C, D, E, and F). According to

Schluchter et al., (1993), the ndh genes in chloroplasts are transcribed, but their protein products and their putative NADH dehydrogenases which they would form have not been identified. The role of an electron transport chain in chloroplast function is not well understood at present (Schluchter et al., 1993).

The ndh? gene was selected to provide molecular characters in this study because of its purportedly higher mutation rate compared to that of rbcL. Because 25

chloroplast genes are highly conserved, phylogenetic relationships among closely

related taxa may be enhanced by using genes with comparatively higher mutation

rates.

SEQUENCE ALIGNMENT

Sequence alignment is obtained by inserting gaps, which correspond to

insertions or deletions, into the data in order to place positions inferred to be

homologous into the same column of the data matrix. Sequence alignment is

probably the most difficult and least understood component of a phylogenetic

analysis (Swofford and Olsen, 1990). In addition to requiring the use of

homologous molecules, phylogenetic analysis of sequence data requires positional

homology. That is, the nucleotides observed at a given position in the taxa under

study should all trace their ancestry to a single position that occurred in a common

ancestor of those taxa (Swofford and Olsen, 1990). Computerized alignment

algorithms can be used to simplify the task of aligning sequences (Swofford and

Olsen, 1990).

Because ndh? is a conserved gene, sequences typically may be aligned by

sight. Alignment of the ingroup sequences with the previously published Oryza sativa (Hiratsuka et al., 1989) and Nicotiana tabacum sequences allowed for the

identification of codon positions. To help ensure maximum positional homology,

sequences were also aligned by Clustal V software (Higgins, et al. 1991) and

compared to sight-aligned sequences. INDELS

It appears that many insertion-deletion mutations are associated with short

direct repeats and probably arose from slip-strand mispairing during replication

(Zurawski and Clegg, 1987). It is probable that particular noncoding regions

experience higher rates of these mutations; it is also probable that addition-deletion

mutations may recur at specific sites, thus contributing to homoplasy in

evolutionary studies (Clegg and Zurawski, 1992). When sequence alignments are

ambiguous, their use as phylogenetic characters is questionable, but otherwise,

they may be phylogenetically informative (Lloyd and Calder, 1991; Revera and

Lake, 1992; Baldwin, 1993; Baum and Sytsma, 1994). INDELS detected in this

study have been evaluated with respect to whether they are of probable

homologous or homoplastic origin, and whether their use as phylogenetically

informative characters is appropriate.

ANATOMICAL/MORPHOLOGICAL CHARACTERS

The cladistic study of the subfamily Epidendroideae presented here also examines the phylogenetic utility of 13 anatomical/morphological characters. These

particular characters were selected because they are purported to be phylogenetically informative at higher taxonomic levels. That is, they are relatively consistent across subfamilies and tribes of Orchidaceae. Because this study was primarily molecular based, morphological characters used were those that could be readily derived from a literature search or a macroscopic examination of voucher specimens. Therefore, characters requiring microtechniques and scanning electron microscopy (SEM) were not included in the present study of the Epidendroideae. 27

The characters and their recognized character states are discussed in detail in

Chapter 2. However, plant microtechniques and SEM were used to determine character states for the morphological-based systematic study of the subtribe

Pleurothallidinae (Chapter 6). The 45 characters used in the Pleurothallidinae study are those recognized by Pridgeon (1982) and are discussed in Chapter 6.

LITERATURE CITED

Arditti, J. 1992. Fundamentals of Orchid Biology. John Wiley & Sons, New York.

Atwood, J. T. Jr. 1986. The size of the Orchidaceae and the systematic distribution of epiphytic orchids. Selbyana 7: 129-247.

Baldwin, B. G. 1993. Molecular phylogenetics of Ca/ycadenia (Compositae) based on ITS sequences of nuclear ribosomal DNA: chromosomal and morphological evolution reexamined. American Journal of Botany 80: 222-238.

Baum, D. A. and K. J. Sytsma. 1994. A phylogenetic analysis of Epilobium (Onagraceae) based on nuclear ribosomal DNA sequences. Systematic Botany 19: 363-388.

Bentham, G. 1881. Notes on Orchideae. Journal of the Linnean Society, Botany 18: 281-360.

Benzing, D. H. 1987. Major patterns and processes in orchid evolution: a critical synthesis. In: J. Arditti, (ed.). Orchid Biology: Reviews and Perspectives Vol. IV. Comstock Publishing Associates, Ithaca, pp 34-77.

Burns-Balogh, P. and V. A. Funk. 1886. A phylogenetic analysis of the Orchidaceae. Smithsonian Contributions to Botany 61: 1-79.

Camin, J. H. and R. R. Sokal. 1965. A method for deducing branching sequences in phylogeny. Evolution 19: 311-326.

Clegg, M. T. and G. Zurawski. 1992. Chloroplast DNA and the study of plant phylogeny: present status and future prospects. In: P. S. Soltis, D. E. Soltis, and J. J. Doyle (eds.). Molecular Systematics of Plants, Chapman and Hall, New York. pp1-13.

Donoghue, M. J., J. Olmstead, J. F. Smith, and J. D. Palmer. 1992. Phylogenetic relationships of Dipsaca/es based on rbcL sequences. Annals of the Missouri Botanical Garden 79: 333-345. 28

. 1981. The Orchids: Natural History and Classification. Harvard University Press, Cambridge, MA .

. 1990. The Neottieae in orchid classification. Lindleyana 5: 102-109.

. 1993. Phylogeny and Classification of the Orchid Family. Dioscorides Press. Portland, Oregon.

. and C. H. Dodson. 1960. Classification and phylogeny in the Orchidaceae. Annals of the Missouri Botanical Garden 47: 25-68.

Eck, R. V. and M. 0 . Dayhoff. 1966. Atlas of Protein Sequence and Structure. 1966. National Biomedical Research Foundation, Silver Springs, MD.

Farris, J. S. 1977. Phylogenetic analysis under Dollo's Law. Systematic Zoology 26: 77-88.

Felsenstein, J. 1978. Inferring phylogenetic trees from chromosome inversions data. Systematic Zoology 27: 401-410.

. 1981. Evolutionary trees from DNA sequences: a maximum likelihood approach. Journal of Molecular Evolution 17: 369-376.

. 1983. Parsimony in systematics: biological and statistical issues. Annual Review of Ecology and Systematics 14: 313-333.

. 1985. Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783-791.

. 1993. Phylogeny Inference Package (Phylip). Version 3.5c. University of Washington.

Fitch, W. M. 1971. Toward defining the course of evolution: minimum change for a specified tree topology. American Naturalist 20: 406-416.

Garay, L. A. 1960. On the Origin of the Orchidaceae. Botanical Museum Leaflets 19: 57-95.

Garay, L. A. 1972. On the origin of the Orchidaceae, II. J. Arnold Arboretum 53: 202-215.

Goldman, N. 1990. Maximum likelihood inference of phylogenetic trees, with special reference to a poisson process model of DNA substitution and to parsimony analysis. Systematic Zoology 39: 345-361. 29

Hasegawa M. and Fujiwara, M. (1993). Relative efficiencies of the maximum likelihood, maximum parsimony, and neighbor-joining methods for estimating protein phylogeny. Molecular Phylogenetics and Evolution 2: 1-5.

Hennig, W. 1966. Phylogenetic Systematics. University of Illinois Press. Urbana, Illinois.

Higgins, D. G., A. J. Bleasby, and R. Fuchs 1991. CLUSTAL V: improved software for multiple sequence alignment.

Hillis D. M. and J. J. Bull. 1993. An empirical test of bootstrapping as a method for accessing confidence in phylogenetic analysis. Systematic Biology 42: 182-192.

Hiratsuka, J., S. Hiroaki, R. Whittier, T. Ishibashi, M Sakamoto, M. Mori, C. Kondo, Y. Honji, C. Sun, B. Meng, Y. Li, A. Kanno, Y. Nishizawa, A. Hirai, K. Shinozaki, and M.Sugiura. 1989. The complete sequence of the rice (Oryza sativa) chloroplast genome: Intermolecular recombination between distinct tRNA genes accounts for a major DNA inversion during the evolution of cereals. Molecular and General Genetics 217: 185-194.

Kluge, A. G. and J. S. Farris. 1969. Quantative phyletics and the evolution of anurans. Systematic Zoology 18: 1-32.

Kurzweil, H. 1987. Developmental studies in orchid flowers. I: Epidendroid and vandoid species. Nordic Journal of Botany 7: 427-442.

Lindley, J. 1830-1840. The Genera and Species of Orchidaceous Plants. Ridgeways. Picadilly, London.

Loyd, D. G. and V. L. Calder. 1991. Multi-residue gaps, a of molecular characters with exceptional reliability for phylogenetic analyses. Journal of Evolutionary Biology 4: 9-21.

Olmstead, R. G., J. A. Sweere, P. A. Reeves, R. E. Spangler, R. W. Scotland and S. J. Wagstaf. 1993a. Comparing and combining ndh? and rbcL sequences for phylogenetic inference. American Journal of Botany 80: 121.

. J. A. Sweere, and K. H. Wolfe. 1993b. Ninety extra nucleotides in ndhF gene of tobacco chloroplast DNA: A summary of revisions to the 1986 genome sequence. Plant Molecular Biology 22: 1191-1193

. and J. D. Palmer 1994. Chloroplast DNA systematics: a review of methods and data analysis. American Journal of Botany 81: 1205-1224.

. and J.A. Sweere. 1994. Combining data in phylogenetic systematics: a empirical approach using three molecular data sets in the Solanaceae. Systematic Biology 43: 467-481. 30

Olsen, G. J., H. Matsuda, R. Hagstrom, and R. Overbeek. 1994. fastDNAml Version 1.0 University of Illinois, Urbana, Illinois.

Pfitzer, E. 1889. Orchidaceae. In A. Engler and K. Prantel, (eds), Die Naturalichen Pflanzenfamilien, II 4: 52-224.

Pridgeon, A. M.1982. Numerical analyses in the classification of the Pleurothallidinae (Orchidaceae). Botanical Journal of the Linnean Society 85: 103-131.

Raven, P. H. and D. I. Axelrod, 1974. Angiosperm biogeography and past continential movements. Annals of the Missouri Botanical Garden 61: 539-673.

Rasmussen, F. N. 1985. Orchids. In: Dahlgren, R. M. T., H. T. Clifford, and P. F. Yeo.(eds.) The Families of the ; Structure, Evolution, and Taxonomy.Springer-Verlag, Berlin, pp 249-274.

Revera, M. C. and J. A. Lake. 1992. Evidence that eukaryotes and eocyte prokaryotes are immediate relatives. Science 257: 74-76.

Schlechter, R. 1926. Das System der Orchidaceen. Notizblatt des Botanischen Gartens und Museums zu Berlin-Dahlem 9: 563-591.

Schluchter, W. M., J. Zhao and D. A. Bryant. 1993. Isolation and characterization of the ndh? gene of Synechococcus sp. strain PCC 7002 and initial characterization of an interposon mutant. Journal of Bacteriology 175: 3342-3352.

Schmid, R. and M. J. Schmid. 1977. Fossil history of the Orchidaceae. In: J. Arditti (ed.) Orchid Biology: Reviews and Perspectives, vol. 1. In: J. Arditti. Comstock Publishing Associates. Ithaca, pp 23-45.

Sugiura, M. 1989. The chloroplast genome. The Biochemistry of Plants 15: 133-150.

Swofford D. L. and G. J. Olsen. 1990. Phylogeny Reconstruction. In: D. M. Hillis and C. Moritz (eds.) Molecular Systematics. Sinauer Associates, Sunderland, MA. pp 411-501.

Swofford, D. L. and D. P. Begle. 1993. Phylogenetic Analysis Using Parsimony. Illinois Natural History Survey, Champaign, Illinois.

Trueman, J. W. H. 1993. Randomization confounded: a response to Carpenter. Cladistics 9: 101-109.

Vermeulen, P. 1966. The system of the orchidales. Acta Botanica Neerlandica 15: 224 241. 31

Wolfe, K. H., W. H. Li, and P. M. Sharp 1987. Rates of nucleotide substitution vary greatly among plant mitochondrial, chloroplast, and nuclear DNA's. Proceedings of the National Academy of Science USA 84: 9054-9058.

. and P. M. Sharp. 1988. Identification of functional open reading frames in chloroplast genomes. Gene 66: 215-222.

Wolter, M. and R. Schill. 1985. On acetolysis resistant structures in the Orchidaceae - why fossil record of orchid pollen is so rare. Grana 24: 139-143.

Zurawski, G. and M.T. Clegg 1987. Evolution of higher-plant chloroplast DNA-encoded genes: implications for structure-function and phylogenetic studies. Annual Review of Plant Physiology 38: 391-418. CHAPTER 2

PHYLOGENY OF SUBTRIBE EPIDENDROIDEAE (ORCHIDACEAE) INFERRED FROM CHLOROPLAST ndhV GENE SEQUENCES AND MORPHOLOGICAL CHARACTERS BASED ON PARSIMONY ANALYSIS

32 33

The purpose of this study is: 1) to infer phylogenetic relationships of

subfamily Epidendroideae using parsimony methods from ndhF sequences, INDELS,

and morphological characters, 2) to evaluate the relative utilities of these three

types of data to infer phylogeny of these selected taxa, and 3) to discuss the

evolution of certain morphological characters in Epidendroideae.

A list of the orchid taxa included in the phylogenetic study presented here,

arranged in accordance with Dressler’s (1993) taxonomic treatment of the family, is

presented in Table 2.1. In this study, a portion of the chloroplast gene NADH

dehydrogenase (ndh?) was sequenced to provide molecular characters to infer

phylogenetic relationships. According to Olmstead and Palmer (1994), there are 20 chloroplast genes that are sufficiently large (> 1kb) and widespread to be generally useful in comparing sequence studies. At present, these sequencing studies have centered on the gene for the large subunit of ribulose-1, 5 bisphosphate, carboxylase (rbcL). However, several studies using other cpDNA genes have recently been used; one such gene is ndhF. Olmstead et al. (1993a) constructed trees for the two chloroplast genes, rbcL and ndhF for the families Solanaceae,

Scrophulariaceae, Bignoniaceae, Acanthaceae, and Lamiaceae. The resulting gene trees were congruent in many respects. According to Olmstead et al. (1993a), the ndhF gene has the desirable property of identifying more informative characters at lower taxonomic levels because it is longer and has a higher substitution rate than rbcL. In comparing sequences from ndhF and rbcL genes in Solanaceae and

Convolvulaceae, Olmstead and Sweere (1994) found that ndhF is a more rapidly evolving gene than rbcL and yields more phylogenetic information than rbcL. 34

The ndhF gene resides in the small single copy region of the chloroplast

genome. (Hiratsuka et al., 1989). According to Sugiura (1989), six chloroplast

DNA sequences resemble those components (ND1, 2, 3, 4, 4L, and 5) of the

respiratory chain NADH dehydrogenase from human mitochondria. As these are

highly expressed in tobacco chloroplasts, they are likely to be the genes for

components of a chloroplast NADH dehydrogenase ( ndhA, B, C, D, E, and F)

(Sugiura, 1989; Shimada and Sugiura, 1990). According to Schluchter et al.,

(1993), the ndh genes in chloroplasts are transcribed, but their protein products and the putative NADH dehydrogenases which they would form have not been identified. The role of an electron transport chain in chloroplast function is not well understood at present (Schluchter et al., 1993).

In the course of this study, specific regions of ndhF with a relatively high frequency of sequence insertions and deletions (INDELS) have been identified.

When sequence alignments are not certain, their use as phylogenetic characters is questionable, but otherwise, they may be phylogenetically informative (Lloyd and

Calder, 1991; Revera and Lake, 1992; Baldwin, 1993; Baum and Sytsma, 1994).

In this study, the use of INDELS as phylogenetically informative characters is explored.

The phylogenetic study of subfamily Epidendroideae presented here also employs a set of 13 morphological-based characters. These particular characters were selected on the basis that they may be phylogenetically informative at higher taxonomic levels. That is, they are relatively consistent across subfamilies and tribes of Orchidaceae. 35

MATERIALS AND METHODS

Plant Material. To represent the major taxonomic components of the

Orchidaceae (sensu Dressier, 1993), a representative from each of the primarily terrestrial orchid subfamilies is included in this study: Apostasioideae,

Cypripedioideae, Spiranthoideae, and Orchidoideae. Not assigned to a subfamily,

but considered by Dressier (1993) among the 'primitive orchids of uncertain affinity', the terrestrial tribe Neottieae Lindl. is also represented. Twenty-nine taxa representing ten tribes and twenty subtribes of the subfamily Epidendroideae are included in this study. These twenty subtribes encompass >74% of the estimated

14,763 species of the subfamily. Representing a liliaceous monocot, which is regarded by most authorities as the sister taxon to the orchids (Garay, 1960;

Rasmussen, 1985; Benzing; 1987; Dressier, 1993), a member from the wholly terrestrial Amaryllidaceae is included. The list of taxa included in this study arranged according to Dressler's 1993 taxonomy is presented in Table 2.1.

Biochemical Techniques. A segment approximately 1,200 base-pair (bp) in length, from the 3' half of the ndhF gene, was sequenced for the taxa listed in

Table 2.2. The source of DNA sequences for the nonliliaceous monocot, Oryza sativa, is Hiratsuka et al. (1989); GenBank accession number: X15901. Total -

DNAs were extracted from fresh or frozen leaf tissue using the CTAB method of

Doyle and Doyle (1987). DNA was amplified via the polymerase chain reaction

(PCR) (Mullis and Faloona, 1987; Saiki et. al.; 1988). Raw DNA template was amplified with Tfl® enzyme (Epicentre Technologies; Madison), using the following thermocycling protocol: (95 °C, 3 minutes) X 1 cycle; (95 °C 1 minute, 50 °C 1 36 minute, 72 °C 1 minute) X 30 cycles; (72 °C, 7 minutes) X 1 cycle: (4 °C soak).

This typically produces 1-3 ug of double-stranded DNA per 100ui PCR reaction.

Double-stranded PCR product was separated by cutting out the desired band on a horizontal 1% agarose gel. The desired band was identified by using a known

DNA size standard. The excised band was purified using Prep-A-Gene®

DNA purification matrix (Biorad; Melville, New York). In doing so, the manufacturer's suggested protocol was modified in order to increase purified DNA yields to acceptable levels (approximately 50% recovery); the initial melting of the agarose gel segments was hastened by shaking at 320 rpm, adding about twice the recommended amount of matrix, increasing the ethyl alcohol content of the wash buffer to 80% , and substituting distilled water in place of the manufacturer's supplied elution buffer.

Following quantification by visual inspection on an ethidium-stained agarose gel against a known standard, the purified PCR product was double-strand sequenced using Sequenase® (United States Biochemical; Cleveland, Ohio) and 35S used as the labeling agent. Labeled fragments were subjected to electrophoresis on an acrylamide gel and transferred to 3MM Whatman paper, vacuum dried and exposed to autoradiographic film.

By comparing complete ndhF sequences of muscifera

(Orchidaceae), Oryza sativa (Poaceae), Nicotiana tabacum (Solanaceae), and

Viburnum sp. (Caprifoliaceae) it was determined that the 3' half of the gene contains more informative characters than the 5' half as illustrated in the relative sequence changes graph (Fig. 2.1). As indicated by this graph, 67.2% of all base 37 Table 2.1 List of taxa taxa included in this study. Orchidaceae members are arranged according to the taxonomy of Dressier (1993).

A. Family Orchidaceae

1. Subfamily Epidendroideae

a. Epidendroid Phylad Tribe Arethuseae Subtribe: : plicata Bl. vestila Lindl. Subtribe: Chysiinae faevis Lindl. Tribe Coelogyneae Subtribe: Thuniinae alba Rchb. f. Subtribe: cristata L. Tribe Epidendreae I (New World) Subtribe: Sobra/ia powelii Schltr. Subtribe: Arpophylliinae giganteum Hartw. ex Lindl. Subtribe: Meiracyllinae Meiracyllium trinasutum Reichb.f. Subtribe: Coeliinae triptera (Smith) G. Don ex Steudel Subtribe: (L.) Lindl. guttata Lindl. (Lindl.) Small Epidendrum conopseum R. Brown Subtribe: Pleurothallidinae grandiflora Lindl. pidax Luer Restrepia muscifera (Lindl.) Rchb. f. ex Lindl. hirtzii Luer Tribe Epidendreaell (Old World) Subtribe: concreta (Jacq.) Garay & Sweet

(table cont'd) 38

a.1 Dendrobioid Subclade of the Epidendroid Phylad

Tribe Podochileae Subtribe: retisquama Rchb. f. hyacinthoides (Bl.) Lindl.

Tribe Dendrobieae Subtribe: Dendrobiinae scabril/ngue Lindl. Subtribe: Bulbophyllinae biflorum Teijsm & Binn Tribe Vandeae Subtribe: lamellata Lindl. Subtribe: phi/ippinense Ames Subtribe: graminifolia (Kranzl.) Schltr.

b. Cymbidioid Phyiad

Tribe Calypsoeae discolor (Pursh) Nutt. Tribe Cymbideae Subtribe: Cyrtopodiinae atropurpureum (Hook, f.) Rolfe Tribe Maxiliarieae Subtribe: brunnea Linden & Reichb. f. Subtribe: Stanhopeinae Stanhopea frymirei Dodson

2. Subfamily Spiranthoideae Spiranthes odorata (Nutt.) Lindl.

3. Subfamily Orchidoideae Habenaria repens Nutt.

4. Subfamily Cypripedioideae Cypripedium acaule Ait.

5. Subfamily Apostasioideae Neuwiedia veratrifolia Blume

(table cont'd) 39

6. Subfamily undetermined Tribe Neottieae Listera australis Lindl.

B. Family Amaryllidaceae Clivia miniata Regel

C. Family Poaceae Oryza sativa L. 40

Table 2.2 Sources of plant material and voucher information for taxa sequenced in this study. Acronyms of herbaria or persons th at donated specimens are listed. GenBank accession numbers for nd h F gene sequences are given.

Taxon Source and voucher” GenBank

accession

number

ArpophylJum giganteum Cult. CONN; Neyland 90 (LSU) U20533

Hatw.

Pleurothallis pidax Luer Cult. F. L. Stevenson; Neyland 105 U2051 1

(LSU)

Octomeria grandiflora Cult. F. L. Stevenson; Neyland 75 U20538

Lindl. (LSU)

Zootrophion hirtzii Luer Cult. MO; A. Lievens & M. LeDoux U20651

320

Restrepia muscifera Cult. MO; Neyland 88 (LSU) U20549

(Lindl.) Rchb.f. ex Lindl.

Brassavola nodosa Cult. LSU; Neyland 72 (LSU) U20540

(L.) Lindl.

Epidendrum conopseum Wild. Neyland 7 7 (LSU) U20665

R. Brown

Cypripedium acaule Ait. Wild. S. Grace; Neyland 115 (LSU) U20563

(table cont'd) 41

Meiracyl/ium trinasutum Cult. CONN; Neyland 93 (LSU) U20562

Rchb. f.

Dendrobium scabrilingue Cult. LSU; Neyland 63 (LSU) U 20534

Lindl.

Coe/ia triptera (Smith) Cult. LSU; Neyland 122 (LSU) U20632

G. Don ex Steudel

Encyclia tampensis Cult. LSU; Neyland 62 (LSU) U20663

(Lindley) Small

Cattleya guttata Lindl. Cult. LSU; Neyland 94 (LSU) U 20544

Sobralia powelii Schltr. Cult. LSU; Neyland 118 (LSU) U20650

Coelogyne cristata L. Cult. LSU; Neyland 124 (LSU) U20632

Ceratostylis retisquama Cult. LSU; Neyland 123 (LSU) U20631

Rchb. f.

Cymbidium atropurpureum Cult. SEL; Atwood 95-2 (LSU) U20649

(Lindl.) Rolfe

Stanhopea frymirei Cult. SEL; Atwood 95-3 (LSU) U20630

Dodson

Maxillaria brunnea Cult. SEL; Neyland 95 (LSU) U 20597

Linden and Rchb f.

(table cont'd) 42 Thunia alba Rchb. f. Cult. SEL; Atwood 95-4 (LSU) U20598

Spiranthes odorata Wild. Neyland 100 (LSU) U20629

(Nutt.) Lindl.

Tipularia discolor Wild. Neyland 104 (LSU) U20628

(Pursh) Nutt.

Eria hyacinthoides Cult. SEL; Neyland 96 (LSU) U20601

(Bl.) Lindl.

Chysis laevis Lindl. Cult. SEL; Atwood & Neyland U 20602

86-0273 (SEL)

Spathog/ottis plicata Bl. Cult. SEL; Neyland 97 (LSU) U 20607

Bulbophyllum biflorum Cult. SEL; Neyland 103 (LSU) U 20606

Teijsm and Binn

Vanda lamellata Lindl. Cult. SEL; Neyland 101 (LSU) U20600

Polystachya concreta Cult. SEL; Neyland 103 (LSU) U20599

(Jacq.) Garay and Sweet

Angraecum philippinense Cult. LSU; Neyland 106 (LSU) U20662

Ames

Aerangis graminifolia Cult. SEL; Neyland 108 (LSU) U 20664

(Kranzl.) Schltr.

Calanthe vestita Lindl. Cult. SEL; Neyland 98 (LSU) U20605

(table cont'd) 43

Listera australis Lindl. Wild. Neyland 107 (LSU) U20603

Clivia miniata Regel Cult. H. Buras; Neyland 119 (LSU) U20539

Habenaria repens Nutt. Wild. M. Oard; Neyland 125 (LSU) U20604

Neuwiedia veratrifolia DNA ex. Kew# 0-460 U20633

Blume

* Abbreviations used: Cult. = cultivated, Wild = wild collected, DNA ex. = DNA extract, CONN = University of Connecticut, LSU = Louisiana State University, MO

= Missouri Botanucal Garden, SEL = Selby Gardens, Kew = Royal Botanic

Gardens, England. 44

changes occurred in the region sequenced in this study (from base 1,072 in each

taxon until the stop codon was reached) (Fig.2.1). Similar results were reported by

Olmstead and Sweere (1994) for Solanaceae and other groups. Therefore, in order

to obtain sequences with substitution rates useful for phylogenetic reconstruction,

sequencing effort was restricted to an approximate 1,200 bp region from the 3* half

of the ndhF gene. The source of DNA sequences for Nicotiana tabacum is

Olmstead et al. (1993b); GenBank accession number: L14953. The source of DNA

sequences for Viburnum sp. is R. Jansen, University of Texas (unpublished).

Primers for the ndhF gene are those developed by Jansen (1992). Primer numbers 6, 7, 9, 10, and 11 have been modified and primer numbers 7.5, 10.5, and 10.7R have been developed specifically to facilitate the sequencing of orchid taxa. Figure 2.2 shows the relative primer positions on the ndhF gene and the primer sequences.

Data Analysis. Phylogenetic analyses were performed using the heuristic search algorithm in PAUP 3.1.1 (Swofford,1993). Because effectiveness of the heuristic tree-building method is sensitive to the sequence of taxa addition used

(Hibbett and Vilgalys, 1993), these searches employed random stepwise addition replications when appropriate. The tree bisection-reconnection algorithm was used as it is the most effective branch-swapping option available in PAUP (Swofford and

Begle, 1993). To avoid the possibility that the resolution of the polytomies chosen during branch swapping may not lead to all minimal trees in an island, zero-length polytomies were not collapsed. Instead of discarding current trees the instant a shorter one is found during a search, the steepest descent command was employed 45 that directs PAUP to continue to look for even shorter trees and uses the trees that give the most improvement for the next round (Swofford and Beagle, 1993).

As measures of clade stability or robustness, the number of unambiguous character state changes, bootstrap confidence intervals (Felsenstein, 1985;

Sanderson, 1989), and decay indices (Dl) were calculated where appropriate

(Donoghue et al., 1992). One thousand bootstrap replications were used in each analysis.

Because ndhF is a conserved gene, sequences were easily aligned by sight.

Alignment of the ingroup sequences with the previously published Oryza sequence allowed for the identification of codon positions. All sequences begin at position

1,072. When possible, gaps and insertions were positioned so that they began and ended on the first and third codon positions respectively. To help ensure maximum positional homology, these sight-aligned sequences were aligned by Clustal V software (Higgins, et al. 1991) using the software's default parameters. After adjustments were made to compensate for missing data and gap/insertion position errors made by the software, the resulting alignment was exactly the same as that prepared by sight.

As an indicator of the rate of synonymous amino acid substitutions, variation in codon position changes was computed using MacClade software (Maddison and

Maddison, 1992). Codon position rate changes were computed for the taxa included in this phylogenetic study to determine whether weighing of first and second codon positions was warranted.

INDELS were incorporated into the character data base by treating gap positions as missing (GAPMODE = MISSING) and then each INDEL scored and 46

entered as a separate binary character (Swofford and Beagle, 1993). This method

is more appropriate than treating INDELS as a fifth base (GAPMODE = NEWSTATE)

because the latter method weights gaps of more than one base excessively.

Because ndhF sequences are conserved and easily aligned, the use of INDELS as

phylogenetic characters is considered appropriate in this study. Specific INDEL

events were included in this analysis via a separate binary-coded data matrix. The

phylogenetically informative INDELS positions are presented in Table 2.3.

The cladistic study of the subfamily Epidendroideae presented here also

examines the phylogenetic utility of 13 anatomical/morphological characters. These

particular characters were selected on the basis that they may be phylogenetically informative at higher taxonomic levels. That is, they are expected to be relatively consistent across subfamilies and tribes of the Orchidaceae. As this study is primarily molecular based, morphological characters used are those which could be derived from literature searches and macroscopic examinations of voucher specimens. Therefore, characters requiring microtechniques and scanning electron microscopy (SEM) were not included in the study of the Epidendroideae. The list of morphological characters with their recognized character states is presented in

Table 2.4. A data matrix illustrating the character state for each orchid taxon included in this study is presented in Table 2.5.

To determine if a reasonable degree of congruence exists between the organismal history and ndhF gene history of the Epidendroideae, a phylogenetic analysis was performed using morphological characters only. Because only 13 such characters are employed in this analysis, a subset of the orchid taxa sequenced 47

5' ndhF 3' ORF 350

PI > P2* P3> P4> PS> P6» P7» P 7 S > Pft> P1Q> P106> P11> P12> P13>

Primer sequences developed by Janeen (1992) 1. AGGTAAGATCCGGTGAATCGGAAAC &. ATAGATGCGACACATATAAAATGCGGTT 2. CTGTCTATTCAGCAAATAAAT S. TTCTATTCAATATCTCTATGGGGT 3. TACTTCCATGTTGGGATTAGTTACTAG 10. ATCCTTATGAATCGGATAATACTATG 4. TTGGATAACGGGGAGTTTCGAATTT 11. CAGTCAGTATAGCCTCTTTCGGAAT 5. CAATGGTAGCAGCGGGAATTTTTC 12. TATATGATTGGTCATATAATCG 6. GCTTTATTTCATTTGATTACTCATGCT 13. TGGGCGTATTTCTTCTTATCTGTTC 7. AGGTACACTTTCTCTTTGCGGTATTCC 14. ACC AAGTTCAATGTTAGCGA GATT AGTC

Primer sequences as modified (6.7. 9.10.11) or designed (7.5.10.5.10.7) by Neyland 6. TT(T,C) CATTTGAT(T.C)ACTCATGC 10. ATACTATGTTATTTCCTCTC 7. GGTACACTTTCTCTTTG(C,T)GG 105 AAATGGTTAACTCCATCG 7.5 TTTTCGCA(A,C)TAATAGCTTGG 1G7R GAAAAA(A.G)GGTTGTCGATGGAG &. TTCTATTCAATATCTCTATGGG 11. TTTCAGTAAGTATAGCTCTT

Figure 2.1 Primer sequences used in this study and their approximate positions on the ndhF gene. 48

Region Not Sequenced In this Study Region Sequenced In This Study Base Changes * 32.8% Base Changes - 67.2%

1000 1200 1400 1600

Figure 2.2 Character state changes graph across the entire ndhF gene from single representatives from angiosperm families: Poaceae, Orchidaceae, Caprifoliaceae, and Solanaceae. Each bar represents the number of site changes averaged over 20 base positions. A total of 590 base changes occurred in the region sequenced in this project (position 1072 - to stop codon). This represents 67.2% of all changes occurring across the entire gene. A total of 287 (37.8%) base changes occurred in the region not sequenced (positions 1 - 1071). 49

was selected in order to achieve a reasonable level of resolution. Specifically, one

taxon representing each of the ten Epidendroideae tribes sequenced, Listera

(Neottieae), and Cypripedium (Cypripedioideae) were chosen to represent the

ingroup. Neuwiedia (Apostasioideae) was selected as outgroup as it is indicated by

this study to be the most basal orchid taxon.

In this study, the use of INDELS, morphological characters, and nucleic acid

sequences were employed to infer phylogenetic relationships. Where appropriate,

nucleic acid sequences, INDELS, and morphological characters were combined as advocated by the principle of total evidence (Hempel, 1965; Good, 1983; Kluge,

1989; Donoghue and Sanderson, 1992; Bull et al., 1993).

RESULTS

Parsimony Analyses. The percentage of sequence change at codon positions one, two, and three was calculated at 27.9%, 23.4%, and 48.6% respectively (Fig.

2.3). Therefore, mutations occurred about twice as often in the third codon position as in each of the first two positions. By weighing the first and second codon positions in data sets where those positions can be demonstrated to mutate at lower rates relative to the third position, it is contended that "noise" in the form of homoplasy may be reduced, leading to more reliable phylogenies. To this end, a parsimony search using 100 random addition replications with the first and second codon positions receiving twice the weight of the third codon position was undertaken. Of the approximate 1,200 nucleotide characters used in this analysis,

245 were informative. This search discovered 570 trees; each tree is 1,114 steps 50 and each has a Cl = 0.498. A strict consensus tree of the 240 trees is presented in Figure 2.4.

After converting DNA sequence data to amino acid data, a parsimony search resulted in the finding of 25,100 trees, each of 482 steps. The number of informative amino acid characters in this analysis is 136. Because computer storage capacity was exceeded when the 25,100 trees were found, only a single incomplete replication was performed. Each of these trees has a Cl = 0.595. A strict consensus tree of the 25,100 trees is presented in Figure 2.5.

When unweighted sequences were used in parsimony analysis, 531 most parsimonious trees were discovered. Individual trees are 750 steps with a Cl =

0.492. Of the approximate 1,200 nucleotide characters used in this analysis, 245 were informative. A strict consensus tree of the 531 most parsimonious trees is presented in Figure 2.6.

When unweighted ndhF sequence data are combined with binary-coded

INDEL data in parsimony analysis, 90 most parsimonious trees were discovered.

Each tree is 775 steps long with a Cl = 0.492. Of the informative characters used in this analysis, 245 are nucleotides and 12 are INDELS. A strict consensus tree of the 90 most parsimonious trees is presented in Figure 2.7.

In the parsimony analysis of 13 morphological characters for the previously defined subset of orchid taxa, 7 most-parsimonious trees were discovered. Each tree is 31 steps with a Cl = 0.774. A strict consensus of these 7 trees is presented in Figure 2.8.

When ndhF sequences and morphological characters are combined in an unweighted parsimony analysis that included only orchid taxa, a single most 51 Table 2.3 Table of informative INDEL characters for the taxa sequenced in this study.

Taxon Sequence Position

1 1 1 1 1 1 1 1 1 2 2 2 2 3 4 4 4 5 5 7 7 7 2 2 2 2 0 0 8 9 2 6 8 9 9 2 2 6 7 0 5 7 5 8 4 0 5 6 1 4 8 8

Arpophy/lum 0 0 0 0 1 0 0 0 0 0 1 0 1 Pleurothallis 0 0 0 0 1 0 0 0 0 0 1 0 1 Octomeria 0 0 0 0 1 0 0 0 0 0 1 0 1 Zootrophion 0 0 0 0 0 0 0 0 0 1 0 1 Rest re pi a 0 0 0 0 1 0 0 0 0 0 1 0 1 Brassavo/a 1 0 1 0 1 0 1 0 1 0 1 1 0 Epic/endrum 1 0 1 0 1 0 1 1 0 0 1 0 0 Meiracyllium 1 0 1 0 1 0 1 0 1 0 1 1 0 Dendrobium 0 0 0 1 1 0 0 0 0 0 1 0 1 Coelia 0 0 0 0 1 0 0 0 0 0 1 0 1 Encyclia 1 0 1 0 1 0 1 0 0 0 1 0 0 Cattleya 1 0 1 0 1 0 1 1 1 0 1 1 0 Sobraiia 0 0 0 0 1 0 0 0 0 0 0 0 1 Coelogyne 0 0 0 0 1 0 0 0 0 0 0 0 1 Ceratostylis 0 0 0 1 1 0 0 0 0 0 1 0 1 Cymbidium 0 0 0 0 1 0 0 0 0 0 1 0 1 Stanhopea 0 0 1 0 1 0 0 0 0 0 1 0 1 Maxillaria 0 0 0 0 1 0 0 0 0 0 1 0 1 Thunia 0 0 0 0 1 0 0 0 0 0 0 1 Spirant hes 0 0 0 1 1 0 0 0 0 0 1 0 1 Tip uiaria 0 0 0 1 1 0 0 0 0 0 1 0 1 Eria 0 0 0 0 1 0 0 0 0 0 1 0 1 Chysis 0 0 0 0 1 0 0 0 0 0 1 0 1

(table cont'd) 52

Spathoglottis 0 0 0 0 0 0 0 0 0 1 0

Bulbophyllum 0 0 0 0 0 0 0 0 0 0 0

Vanda 0 0 0 0 0 0 0 0 0 1 0

Polystachya 0 0 0 0 0 0 0 0 0 1 0

Angraecum 0 0 0 0 0 0 0 0 0 1 0

Aerangis 0 0 0 0 0 0 0 0 0 1 0

Ca/anthe 0 0 0 0 0 0 0 0 0 0 0

Listera 0 0 0 0 0 0 0 0 0 1 0

Clivia 0 1 0 0 0 0 0 0 1 1 0

Habenaria 0 0 0 0 0 0 0 0 0 0 0

Cypripedium 0 0 0 0 0 0 0 0 1 1 0

Neuwiedia 0 0 0 0 1 0 0 0 0 0 0

Oryza 0 1 0 0 1 0 0 0 0 1 0 53

Table 2.4 Morphological characters and recognized character states for orchid taxa presented in Table 2.2.

Attribute Character State

1. Growth Habit (0) Sympodial (1) Monopodial 2. Leaf Structure (0) Conduplicate (1) Plicate 3. (0) Terminal (1) Lateral 4. Resupination (0) Resupinate (1) Not Resupinate 5. Pollen Aggregation (0) Firm (1) Granulate (2) Sectile (3) Nonaggregated

6. Number of Fertile Anthers (1) = 3 (2) = 2 (3) = 1 7. Number of Pollinia (0) = 8

(1) = 4 (2) = 2 (3) = 0 8. Caudicle Architecture (0) Embedded with pollen (1) Absent (2) Sterile 9. Stipe (0) Absent (1) Tegular (2) Hamular 10. Growth Substrate (0) Epiphytic (1) Terrestrial

(table cont'd) 54

11. Velamen Type (0) Ca/anthe

(1) Cymbidium

(2) Epidendrum

(3) Coelogynae

(4) Dendrobium/Bu/bophyl/um

(5) Vanda

(6) Pleurothallis

(7) Spiranthes (8) Absent (-) Present But An Undefined Type

12. Seed Type (0)

(1) Maxillaria

(2)

(3) Dendrobium

(4) Vanda

(5) Goody era

(6)

(7) Pleurothallis

(8) Epidendrum

(9) Cymbidium

(A) Stanhopea

(B)

(C) Neuwiedia

(D) Orchis 13. Fusion of style 1. Style fused with filaments at bases 2. Style fused with filaments and staminode bases 3. Style fused with fialments, staminodes,and stigma to column 55

Table 2.5 Morphological character states for the orchid taxa listed in Table 2.2.

Taxa Characters

1 2 3 4 5 6 7 8 9 10 11 12 13

Arpophy/lum 0 0 0 1 1 3 0 0 0 0 2 6 3 Pleurothallis 0 0 0 1 0 3 2 0 0 0 6 7 3 Octomeria 0 0 0 0 0 3 0 0 0 0 6 7 3 Zootrophion 0 0 0 0 0 3 2 0 0 0 6 7 3 Restrep ia 0 0 0 0 0 3 1 0 0 0 6 7 3 Brassavo/a 0 0 0 0 0 3 0 0 0 0 2 8 3 Epidendrum 0 0 0 0 0 3 1 0 0 0 2 8 3 Meiracy/lium 0 0 0 0 0 3 0 0 0 0 2 6 3 Dendrobium 0 0 0 0 0 3 1 1 0 0 4 3 3 Co elia 0 0 1 0 1 3 0 0 0 0 ? 8 3 Encyc/ia 0 0 0 0 0 3 1 0 0 0 2 8 3 Cattleya 0 0 0 0 0 3 1 0 0 0 2 8 3 Sobra/ia 1 1 0 0 1 3 0 0 0 1 - 2 3 Coelogyne 0 1 1 0 0 3 1 0 0 0 3 3 3 Ceratostylis 0 0 0 0 0 3 0 0 0 0 0 - 3 Cymbidium 0 0 1 0 0 3 2 0 0 0 1 9 3 Stanhopea 0 1 1 0 0 3 2 2 1 0 1 A 3 Maxiilaria 0 0 1 0 0 3 1 2 1 0 1 1 3 Thunia 0 0 0 0 1 3 0 0 0 1 3 3 3 Spiranthes 0 0 0 0 1 3 2 1 0 1 7 5 3 Tipu/aria 0 1 1 0 0 3 1 2 2 1 0 0 3 Eria 0 0 1 1 0 3 0 0 0 0 0 - 3 Chysis 0 1 1 0 0 3 0 0 0 0 - 6 3 Spathog/ottis 0 1 1 0 0 3 0 0 0 1 0 ? 3 Bu/bophyl/um 0 0 1 0 0 3 2 0 1 0 9 3 3

(table corn'd) 56

Vanda 1 0 1 0 0 3 2 1 2 0 5 4 3

Polystachya 0 0 0 1 0 3 1 1 2 0 - 4 3

Angraecum 1 0 1 0 0 3 2 1 2 0 5 4 3

Aerangis 1 0 1 0 0 3 2 1 2 0 5 4 3

Calanthe 0 1 1 0 0 3 0 0 0 1 0 ? 3

Listera 0 0 0 0 1 3 2 1 0 1 8 B 3

Habenaria 0 0 0 0 2 3 2 2 0 1 8 D 3

Cypripedium 0 1 0 0 3 2 3 1 0 1 8 B 2

Neuwiedia 0 1 0 0 3 1 3 1 0 1 8 C 1 Figure 2.3 Rate of change graph by codon position from the taxa used in this study study this in used taxa the to from position restricted Graph codon by graph 2.2). (Table change of Rate 2.3 Figure step ndhF sqec pstos 1072-2304. positions sequence 57 58

parsimonious tree was found. Neuwiedia (Apostasioideae), indicated to be the most

basal orchid taxon (Fig. 2.6), was designated as outgroup. A total of 256

characters were informative (245 nucleotide, 11 morphological). This tree has 587

steps and a Cl = 0.492 (Fig. 2.9). Decay indices were calculated for trees 10

steps longer than the optimum tree; at the point where trees were retained for 4

extra steps (length = 591), computer storage was exceeded. Therefore, decay

indices calculated from trees four to ten steps longer than the optimal tree were

terminated in each step after 24,500 trees were found.

Systematic Relationships. With Oryza sativa (Poaceae) as outgroup, all trees

support the hypothesis that the Orchidaceae is monophyletic (Figs. 2.4, 2.5, 2.6,

2.7, 2.9). The strict consensus cladogram of trees discovered when ndhF

sequences were unweighted suggests that subfamily Epidendroideae is

monophyletic with Ustera {Neottieae) as sister (Fig. 2.9). Although several clades

are resolved, there is a general lack of resolution among the tribes of subfamily

Epidendroideae (Fig. 2.9). The single parsimony tree found when ndhF sequences

and morphological characters were combined is congruent with the strict consensus

tree (Fig. 2.9) in that subfamily Epidendroideae is monophyletic with Listera

(Neottieae) as sister (Fig. 2.9). These trees also indicate that Sobralia (Sobraliinae,

New World Epidendreae) is the most basally diverged element in Epidendroideae,

and not closely related to the New World tribe Epidendroideae core taxa consisting

of Meiracyliinae, Laeliinae, Arpophyliinae, and Pleurothallidinae subtribes.

Additionally, the monogeneric subtribe Coeliinae (New World Epidendreae) does not

appear closely related to either Sobralia nor the core taxa of the New World tribe

Epidendreae; therefore, the New World tribe Epidendreae appears polyphyletic. 59

Arpophyllum Pleurothallis Zootrophion NW Epidendreae ■G R estrepia 60 89 Octomeris Chysis l| Arethuseae II Brassavola Epidendrum m Encyclla NW Epidendrea Cattleya Meiracyllium Dendrobium Dendrobeae Coelia NW Epidendreae Tipularia Calypsoeae Sobralia NW^Egidendreae Coelogyne 52 Thunia Coelogyneae Ceratostylis Podochileae 97 Eria Cymbidium Cymbideae 73 Stanhopea Maxillarieae 85 Maxillaria ■» Spiranthes Spiranthoideae Spathoglottis 111 Arethuseae] 82 Calenthe Bulbophyllum j/ DendrobieaePv Yanda ■M VandeaeI Polystechya OW Epidendreae Angraecum 82 Vandeae Aerangia Listera Neottieae Habenaria Orchidoideae Cypripedium Cypripedioldeae Neuviedia Apostasioideae Clivia Amaryllidaceae Oryza Poaceae

Figure 2.4 Strict consensus tree of 570 trees discovered from 100 random addition searches using weighted ndhF sequences. Bootstrap values are indicated below each branch. A rpplm nua PlttrttklUi Z ntrtpU M f iu l r t p i O c ta w rii Bm avoU IfUttdrim M slm ^llaa ta y d l* Ulltya Epidendroideae S tu ta p a C|BkUiUB tttid llu li ta ir o tlu m C a llj Sakrslli CttlOjJM 9 TkurU 8- Cm teitgH i S p lru U et Spiranthoideae I b k m rtt Orchidoideae T lpritrit £rta 4 C ip ii Sptltegltttii W u t b Epidendroideae k lb o p h y ta y u m M p lK k g i iuqrttcm A srtoji i U its n Neottieae Cgprtpetiua Cypripedioideae IfciM sJi Apostasioideae a tv u Amarytlidaceae 0Y|» Poaceae

Figure 2.5 Strict consensus tree of 25,100 trees discovered from a single parsimony analysis of amino acid data converted from ndhF DNA sequences. 61

Arpophyllum Pleurothallis Restrepia Zootrophion Octomeria Brassavola SNW Epidendreae Epidendrum Encyclia Cattteya Meiracyllium Dendrobium jrlDendrobieae TJ Coelia NW Epidendreae Coelogyne | Coelogyneae | •gm CL Ceratostylis i Podochileae 10 99 Eria i Cymbidium Cymbideae 1 16 £ Stanhopea 70 Maxillarieae Maxiilaria Thunia | Coelogyneae | Tipularia Calypsoeae Chysis Spathoglottis Arethuseae 27 85 Calanthe Bulbophyilum i^Dendjobieaej/! Vanda isilj Vandeae |M Polys tachya 33 OW Epidendreae Angraecum 79 _6 **j Vandeae s; * 100 Aerangis 37 Sobralia NW Epidendreae 20 Listera Neottieae 64 Cypripedium Cypripedioideae 44 14 Spiranthes Spiranthoideae 87 38 Habenaria Orchidoideae Neuwiedia Apostasioideae Clivia Amaryllidaceae Oryza Poaceae

Figure 2.6 Strict consensus tree of 531 trees discovered from 100 random addition parsimony searches using ndhF sequences. The number of unequivocal synapomorphies are indicated above each branch. Bootstrap values are indicated below each branch. Potentially polyphyletic tribes are shaded in the side legend. 62

Arpophyllum Pleurothallis [531 Restrepia Zootrophion 99 Octomeria NW EDidendrea Brassavola Epidendrum Encyclia 100 Cettleya Meiracyllium Chysis 11 Arethuseae til Dendrobium m T3. T ipularia Calypsoeae Q. 3CD Coelia NWEgidendr^e Q. Pol ystachye 3 OW Epidendreae CL Angraecum 0)CD 80 CD Aerangis Sobralia NW Egldend reae Coelogyne 20 Coelogyneae Thunia Bulbophyllum | | g g | | P 33 Ceratostylis Podochileae 9 8 Eria Cymbidium Cymbldeae Stanhopea Maxillarieae M axillaria Spathoglottis ; I Arethuseae 83 Calanthe Vanda [Vandeae! Listera Neottieae Spiranthes Spiranthoideae 2 5 Habenaria Orchidoideae Neuviedia Apostasioideae Cypripedium Cypripedioideae Clivia Amaryllidaceae Oryza Poaceae

Figure 2.7 Strict consensus tree of 90 most-parsimonious trees discovered from 100 random-addition searches using ndhF sequences and binary-coded INDELS. Bootstrap values are indicated below each branch. 63

Epidendrum

25 Ceratostylis

11 Coelogyne

32 Tipularia

38 Calanthe

Epidendroldeae Cymbidium

52 Maxillaria

Vanda 48

95 Polys tachya

Bulbophyllum

91 Listera Neottieae

Cypripedium Cypripedioideae

Neuwiedia Apostasioideae

Figure 2.8 Strict consensus tree of 7 most-parsimonious trees discovered from 100 random-addition parsimony searches using a data matrix consisting of 13 morphological characters. Bootstrap values are indicated below each branch. 64

g — Arpophyllum — Pteurothallis -S. Restrepia — Zootrophion N.W. Epidandraae — Octomaria — Brassavola — Epidendrum J 56 9 4 r ~ ^ Encydia Cattiaya ^ MeiracyUium -22 Chysis Arethuseael Cymbideae m Maxillarieae T> Maxillaria £ Vandeae( S. 3 Polystachya CL O. W. Epidandraae a Angraecum ®a IQOt—4 Aerangis '19 Bulbophyllum Dendrobieae ~ Tipularia Carypsoeae —10 Coeliao r ■iS. Spathoglottis 14 — Caiantha ( Arethuseae | — Caratostylis Podochileae C- Eria trrrFrtinnrmm Dendrobium 'I DendrobieaeK^ S-12 Coelogyne Coebgyneae ~ Thunia — Sobralia -22 Listera pC Neottieae -=£• Cypripedium Cypripedioideae 25 Spiranthes Spiranlhoideae -22. Habsnaria Orchidoideae Neuwiadia Apostasioideae

Figure 2.9 Single parsimony tree discovered from 100 random addition replications using a data matrix consisting of n d h f sequences and morphological characters. Unambiguous synapomorphies are indicated above each branch; the first value below each branch is the bootstrap confidence interval; the second value below each branch is the decay index (the last replication before each branch collapsed). Resolved branches found in this tree but not found in the strict consensus tree (Fig. 2.6) are bold. Polyphyletic tribes are shaded in the side legend. 65

Both cladograms indicate that the subtribe Laeliinae ( Brassavola, Epidendrum,

Encydia, Cattleya) appears to be a monophyletic group with subtribe Meiracyliinae

(Meiracyllium) as sister, and that subtribe Pleurothallidinae (Pleurothallis, Restrepia,

Octomeria, Zootrophion) is monophyletic with subtribe Arpophylliinae lArpophyllum) as sister (Figs. 2.6, 2.9).

Based on the strict consensus cladogram (Fig. 2.6) the tribe Arethuseae potentially represents a polyphyletic taxon as Spathogiottis and Calanthe

(representing the subtribe Bletiinae) form a clade separate from Chysis (Chysiinae).

The cladogram constructed from combined sequence and morphological characters indicates that the Spathoglottis-Calanthe clade occupies a relatively basal position in the Epidendroideae; whereas, Chysis appears as the sister to the core taxa of the

New World tribe Epidendreae (Fig. 2.9).

Although the cladogram (Fig. 2.9) suggests that Thunia (Thuniinae) and

Coelogyne (Coelogyninae) are sister taxa, Coeiogyne appears more closely related to Dendrobium (Dendrobiinae) than to Thunia. Therefore, the tribe Coelogyneae is indicated to be polyphyletic. The relationships among Coeiogyne, Dendrobium, and

Thunia are unresolved in the strict consensus tree of trees discovered when unweighted sequences were used (Fig. 2.6).

A clade composed of those taxa referred to as vandoids by Dressier (1981)

(see Table 1.1) indicates a close alignment of the tribes Vandeae , Old World

Epidendreae, Cymbideae, Calypsoeae, and Maxillarieae. The alignment of

Bulbophyllum (Bulbophyliinae) with the vandoid tribes suggests that the tribe

Dendrobieae is polyphyletic (Fig. 2.9). The strict consensus tree suggests an alliance among the cymbidioid phylad taxa: Cymbidium (Cymbideae), Stanhopea 66

(Maxillarieae), and Maxillaria (Maxillarieae) (Fig. 2.6). An alliance among

Polystachya (Old World Epidendroideae), Aerangis (Vandeae), and Angraecum

(Vandeae) is supported also by the strict consensus tree (Fig. 2.6). However, this

tree fails to resolve the relationship between these two clades, the remainder of the

vandoid taxa, and the relationship between the vandoid taxa and Bulbophyllum.

DISCUSSION

Phylogenetic Utility of ndhF Sequences. Most-parsimonious trees

discovered from searches where unweighted sequences were used have

consistency indices = 0.492. Giving added weight to first and second codon

positions resulted in only a slightly improved consistency index of 0.498. This may suggest that although the third codon position has mutated at a higher rate than the first and second codon positions, secondary mutations at the third codon position have not occurred at a higher frequency than in the first and second positions for the taxa included in this study. The weighing of the first and second codon position higher than the third codon position in this study results in less resolution than unweighted codon characters and does not appreciably increase phylogenetic signal in these analyses. When amino acid data were used to infer phylogeny for the taxa included in this study, the consistency index for most parsimonious trees discovered improved to 0.595. However, the strict consensus tree of most-parsimonious trees discovered using amino acid data is poorly resolved (Fig. 2.5). Unweighted ndhF sequences resolve phylogenetic relationships among the orchid subfamilies in parsimony analysis (Fig. 2.6). Orchid evolution at the subfamilial level is discussed in more detail in Chapters 4 and 5. 67

The monophyly of subfamily Epidendroideae is supported by this study.

The use of unweighted ndhF sequence data using parsimony methods resolves

subfamilial relationships within Orchidaceae and subtribal-leve! relationships within

Epidendroideae. However, the intermediate tribal-level relationships are poorly

resolved (Fig. 2.6). Using restriction site techniques. Palmer et al. (1988) reported

that they had failed to find cpDNA mutations that could establish relationships

among several well-defined lineages of the orchid subtribe Oncidiinae. They thought it likely that this was due to a large number of lineages diverging in a short

period of evolutionary time from a polymorphic ancestor, and that this process

happened so rapidly that either no mutations exist at this level or the few mutations that do exist convey conflicting ideas of relationships because of homoplasy. It appears that a similar explanation may apply to the evolution of Epidendroideae.

Because subtribal-level relationships are generally well resolved in the strict consensus tree from unweighted ndhF characters (Fig. 2.6), it appears that some factor, other than the conserved nature of the gene, is responsible for the lack of tribal-level resolution. That is, the major tribes of this subfamily diverged in a relatively short period in terms of evolutionary time and experienced a rapid radiation. This hypothesized radiation may have been catalyzed by morphological and anatomical adaptations that allowed the mostly epiphytic epidendroids to pioneer xeric arborescent habitats in tropical regions (Chapter 5).

Phylogenetic Utility of INDELS. In the course of this study, 35 INDELS were identified for the region sequenced. A specific region with a particularly high level of sequence gaps was identified beginning at position 1,495 and ending at approximately position 1,716. Olmstead and Sweere (1994) reported that gaps 68

found in sequences of several Asterideae families were confined primarily to a

region within base pair positions 1,443 - 1,697. The region exhibiting a high

propensity of gaps in Asterideae is thus approximately the same region for which

the same propensity occurs in Orchidaceae.

When sequence alignments are not certain, the use of INDELS as

phylogenetic characters is questionable, but otherwise, they may be

phylogenetically informative (Lloyd and Calder, 1991; Revera and Lake, 1992;

Baldwin, 1993; Baum and Sytsma, 1994). Because ndhF sequences have been

shown to be easily alignable, their inclusion in this study is not prohibited.

However, their value as phylogenetically informative characters in this study is

nevertheless questionable. According to Clegg and Zurawski (1992) it is probable

that addition-deletion mutations may recur at specific sites, thus contributing to

homoplasy in evolutionary studies. Of the 35 INDELS identified in this study, 22 are autapomorphic and, therefore, not informative. Of the remaining 13 informative

INDELS (Table 2.3), four appear to be homoplastic. For example, a six bp insertion

occurring at position 1,528 in Zootrophion (TCAATA) and in Neuwiedia (GATATA) is considered homoplastic because the sequences are dissimilar and the two represent disparate elements as suggested by the cladograms (Figs. 2.6, 2.9, 3.3).

A 39 bp deletion beginning at position 1,495 occurs in the disparate taxa:

Ceratostylis, Spiranthes, Tipularia and Dendrobium (Figs. 2.6, 2.9, 3.3). It is possible that this 39 bp deletion is homologous with respect to Ceratostylis and

Dendrobium which are both members of Dressler's (1993) dendrobioid subclade and suggested to be closely allied by the cladograms (Figs. 2.9, 3.3). However, when all four taxa are considered together, the deletion appears homoplastic. This 39bp deletion is flanked by the short direct repeat sequence:

TTTTTTT C A A A A and appears to belong to a class of deletions as described by

Small (1989) that may account for the formation of most if not all novel direct

repeats and many deletions. More specifically, such deletions may occur when a

homologous recombination between short direct repeats in a progenitor chloroplast genome produces a pair of subgenomic circles. A second independent event resulting in a homologous recombination between a second pair of short direct repeats produces another pair of subgenomic circles. If one from each pair of subgenomic circles recombines, a new chloroplast genome with a duplication or a deletion can result.

A series of INDELS occur in the closely aligned subtribes Laeliinae and

Meiracyliinae (New World Epidendreae) and include a six bp deletion beginning at position 1,300, a one bp deletion at position 1,780, and a 24 bp insertion beginning at position at 2,278. These INDELS are unique to the Laeliinae and Meiracyliinae and are probably homologous. An eight bp deletion beginning at position 1,796, and a 10 bp deletion beginning at position 2,268 in the closely allied Brassavola,

Meiracyl/ium , and Cattleya appear to be homologous. A 9bp deletion beginning at position 1,795 in the closely allied Epidendrum and Encydia is also a likely homologous event.

A 227 bp deletion beginning at position 1,487 is flanked by the short direct repeat sequence: TCAATAGGAATTTCTTTT. and appears to belong to the class of deletions previously described (Small,1989). Although this large deletion is found in all members of the apparently closely related Meiracyliinae-Laeliinae clade (New

World Epidendreae), it also occurs in Stanhopea (Maxillarieae) which is not 70 suggested by the cladograms to be closely related to the Meiracyliinae-Laeliinae clade (Figs. 2.6, 2.9, 3.3). Therefore, it is possible that this deletion has occurred independently at least twice among the taxa included in this study.

The sister relationship between Oryza and CHvia as suggested by the cladograms is supported by a six bp deletion beginning at position 1,405 (Figs. 2.4,

2.5, 2.6, 2.7, 3.3). In Oryza, a 9 bp sequence (AATACAGGA) beginning at position 1,594 appears sufficiently similar to the sequence in Neuwiedia

(AATGTAAGA) for the tw o insertions to be considered homologous.

A six bp insertion beginning at position 2,224 occurs in Sobralia, Thunia,

Calanthe, Coeiogyne, Bulbophyllum, Habenaria ,and Neuwiedia. Trees discovered from parsimony searches suggest that this insertion is homoplastic (Figs. 2.6, 2.9); however, as is shown in Chapter 3, the maximum likelihood tree with the greatest log likelihood suggests this insertion may be homologous (Fig. 3.3).

Although there is no conclusive proof, at least four phylogenetically

"informative" INDELS detected in these sequences appear to be homoplastic.

Because it is probable that addition-deletion mutations may recur at specific sites, and thus contribute to homoplasy in evolutionary studies (Clegg and Zurawski,

1992), the branches resolved as a result of including INDELS with nucleotide characters in parsimony analyses must be viewed with caution (Fig. 2.7).

rtdhF: a pseudogene? Six taxa sequenced in this study (Brassavola,

Meiracyllium, Catt/eya, Epidendrum, Encydia, and Stanhopea) exhibit deletions whose base number is not evenly divisible by three which results in a substantial frame shift of subsequent base positions. In Brassavola, Meiracyllium, and

Cattleya, a 227 bp deletion occurs beginning at position 1,487 is followed by a 70 suggested by the cladograms to be closely related to the Meiracyliinae-Laeliinae clade (Figs. 2.6, 2.9, 3.3). Therefore, it is possible that this deletion has occurred independently at least twice among the taxa included in this study.

The sister relationship between Oryza and Clivia as suggested by the cladograms is supported by a six bp deletion beginning at position 1,405 (Figs. 2.4,

2.5, 2.6, 2.7, 3.3). In Oryza, a 9 bp sequence (AATACAGGA) beginning at position 1,594 appears sufficiently similar to the sequence in Neuwiedia

(AATGTAAGA) for the two insertions to be considered homologous.

A six bp insertion beginning at position 2,224 occurs in Sobralia, Thunia,

Caianthe, Coeiogyne, Bulbophyllum, Habenaria .and Neuwiedia. Trees discovered from parsimony searches suggest that this insertion is homoplastic (Figs. 2.6, 2.9); however, as is shown in Chapter 3, the maximum likelihood tree with the greatest log likelihood suggests this insertion may be homologous (Fig. 3.3).

Although there is no conclusive proof, at least four phylogenetically

"informative" INDELS detected in these sequences appear to be homoplastic.

Because it is probable that addition-deletion mutations may recur at specific sites, and thus contribute to homoplasy in evolutionary studies (Clegg and Zurawski,

1992), the branches resolved as a result of including INDELS with nucleotide characters in parsimony analyses must be viewed with caution (Fig. 2.7).

ndhF: a pseudogene? Six taxa sequenced in this study (Brassavola,

Meiracyllium, Cattleya, Epidendrum, Encydia, and Stanhopea) exhibit deletions whose base number is not evenly divisible by three which results in a substantial frame shift of subsequent base positions. In Brassavola, Meiracyllium, and

Cattleya, a 227 bp deletion occurs beginning at position 1,487 is followed by a 71 sequence frameshift until a 1bp deletion at position 1,780 brings the sequence back into frame (Fig. 2.10). An 8bp deletion beginning at position 1,796 is followed by another frameshift until a 10bp deletion beginning at position 2,268 brings the sequence back into frame (Fig. 2.10). Therefore, 775 bases are either deleted or frameshifted in the region of ndhF sequenced for these taxa.

In Epidendrum and Encydia, a 227 bp deletion beginning at position 1,487

(also found in Cattleya, Brassavola, and Meiracyllium) is followed by a sequence frameshift until a 1 bp deletion at position 1,780 brings the sequence back into frame (Fig. 2.10). Therefore, 294 bases are either deleted or frameshifted in the region sequenced for these two taxa.

In Stanhopea, a 227 base deletion beginning at position 1,487 (also found in the five previously mentioned taxa) is followed by a region of frameshifted sequences, a 7bp deletion beginning at position 1,743, frameshifted sequences, an

8bp deletion beginning at position 1,768 and frameshifted sequences until a 16bp deletion beginning at position 1,899 brings the sequence back into frame.

Therefore, 429 bases are either deleted or frameshifted in Stanhopea.

To ensure the position of the 227 bp deletion was accurately recorded, primer 10.7R (Fig. 2.1) was developed specifically to reverse sequence through this deletion. This sequencing strategy was employed to ensure that no fidelity or compression problems were responsible for either the misrecording or misalignment of sequences that define this deletion. To this end, Meiracyllium, Encydia,

Epidendrum, and Cattleya were reversed sequenced through this deletion with no sequence differences vis-a-vis those from produced from the forward-sequencing primer. Brassavola, Meiracyllium, Cattleya Sequence Position # of bp’s deleted 1487 227 1780 1 1796 8 2268 10 TOTAL 246

1487 1780 1796 2268

Epidendrum, Encydia Sequence Position # of bp's deleted 1487 227 1780 1 TOTAL 228

1487 1780

Stanhopea Sequence Position # of bp's deleted 1487 227 1743 7 1768 8 1899 16 TOTAL 258

1487 1743 1899 — h u t ■ ■ - 1768

Gap 1 1 Frameshift

Figure 2.10 Summary of deletions and frameshifts in Brassavola, Meiracyllium, Cattleya, Epidendrum, Encydia, and Stanhopea sequences. 73

Although ndh? has been sequenced from a wide of angiosperm

families and from a liverwort (Ohyama et al., 1986), it is not present in the

chloroplast of at least one gymnosperm, Pinus thurnbergii (Waksugi et al., 1994).

In that study, it was determined that the chloroplast genome in Pinus thunbergii

(black ) lacks all 11 intact ndh genes found in tobacco, rice and liverwort. Four

g enes (ndhA , ndh?, ndhG, and ndhJ) have been lost completely, and the other

seven remain as pseudogenes (ndhC, ndhE, and ndhK) and as truncated

pseudogenes ( ndhB, ndhD, ndhW, and ndh\). These pseudogenes contain multiple

stop codons and frameshifts; short deletions and short insertions occur throughout

the sequences. According to Waksugi et al., one plausible explanation for the

absence of ndh genes in black pine is that all chloroplast ndh genes have been

transferred to the nuclear genome; however, the possibility that NADH

dehydrogenase is absent at least in black pine chloroplasts was not ruled out. They

suggested that further studies on the function and significance of chloroplast and

nuclear ndh genes are necessary in order to resoive these unanswered questions.

Regardless of whether the ndh? gene in these orchid taxa is a pseudogene or not, this present study indicates that large deletions and frameshifts beginning at approximately position 1,487 have no apparent deleterious effects on these taxa.

Specific base frequencies may also suggest ndh? is a pseudogene.

According to Wen-Hsiung and Graur (1991), pseudogenes are expected to become rich in bases A and T because 64.5% of all mutations result in either an A or T, while random mutation is expected at 50%. Since there is a tendency for G and C to change frequently to A or T, and since A and T are not as mutable as C and G, noncoding regions that are subject to no functional constraint (including 7 4

pseudogenes) are generally found to be AT-rich (Wen-Hsiung and Graur,1991). In

the sequence data for orchid taxa used in this project, the frequency for AT bases is

69% and the frequency for GC bases is 31 % (a greater than two to one ratio

of AT bases over GC bases). This ratio of bases further suggests ndh? m ay be a

pseudogene, even among orchid taxa that have no large gaps or frameshifts.

Combining Morphological and Molecular Data. According to Lutzoni and

Vilgalys (in press), phylogenetic trees derived from different data sets are rarely

identical due to: 1) use of an inappropriate evolution model for a given data set, 2) sampling error, and 3) different evolutionary histories between the organism and the molecule. Phylogenetic history for a given molecule can be different from the organismal phylogeny due to lineage sorting acting on polymorphic characters, lateral gene transfer, and hybridization (Lutzoni and Vilgalys, in press). Bull et al.

(1993) suggested that a combined analysis of diverse data is inappropriate unless it is shown that the different data sets are not significantly heterogeneous with respect to the reconstruction model; that is, when analyzed separately, there should be similarity between the resultant topologies. If data sets are demonstrably heterogeneous, they should not be combined in an analysis that assumes character homogeneity. Comparison of he strict consensus tree of seven most parsimonious trees discovered from an analysis using a subset of orchid taxa and 13 morphological characters (Fig. 2.8), with the strict consensus tree using unweighted ndh? sequences (Fig. 2.6) suggests that the organismal history of these selected taxa is approximately consistent with the gene history. Specifically, Cypripedium and Listera form a grade that is basal to the monophyletic Epidendroideae. 75

Therefore, phylogenetic analyses including both morphological characters and ndhF

sequences is considered appropriate in this case.

Hillis (1987) suggested that some character classes are useful in resolving

certain areas of the tree but are not informative for others. If, for example, one

character set resolves nodes closer to the tips of the tree and another is more

useful for basal branching, combination of the two may substantially improve the

resolution of the full tree (Bull et al., 1993). Their observation is well illustrated in

this study. For example, when unweighted ndhF sequences were used exclusively,

the resulting phylogenetic trees were resolved at the familial, subfamilial, and

subtribal levels, but poorly resolved at the tribal level (Fig. 2.6). However, when the morphological characters are combined in the analysis, a single fully resolved tree is discovered (Fig. 2.9). Similar findings were reported by Hibbett and Vilgalys

(1993) in which molecular and morphological characters were combined in

parsimony analyses of Lentinus (Basidiomycotina). They stated that far from being overwhelmed, morphological characters had a significant impact on the results, despite the fact that they were not given greater weight than the more numerous molecular characters.

Because morphological characters provided the data necessary to bring tribal-level resolution to subfamily Epidendroideae, it is apparent that these characters were not overwhelmed by the more numerous nucleic acid characters.

However, the level of confidence in this particular tree topology (mainly with respect to tribal-level relationships) is not particularly high. That is, these morphological characters do not resolve nodes with a great degree of confidence as 7 6 measured by the number of unambiguous apomorphies, bootstrap values, and decay indices (Fig 2.9).

Phylogenetic relationships in Epidendroideae. Based on the single parsimony tree found when molecular and morphological characters are combined (Fig. 2.9), the cymbidioid phylad (Table 2.1) is polyphyletic. Although the cymbidioid taxa

Cymbidium (Cymbideae), Stanhopea (Maxillarieae), and Maxillaria (Maxillarieae) form a strongly supported clade (16 apomorphies and bootstrap confidence interval of 88, Dl = >10), Tipularia (Calyposeae) appears only distantly related to these cymbidioid core taxa. Although the cymbidioid phylad appears to be a polyphyletic taxon, this cladogram suggests a close alliance among the vandoid taxa (sensu

Dressier, 1981). That is, all vandoid tribes (Cymbideae, Maxillarieae, Vandeae,

Calypsoeae) are part of the same clade, suggesting that they are closely allied.

Because Bulbophyllum (Dendrobieae) (an epidendroid) also occurs in this clade, the

Vandoideae (sensu Dressier, 1981) appears polyphyletic. As Tipularia (Calyposeae) appears more closely aligned with the epidendroids than to the cymbidioids, this suggests that the epidendroid phylad (sensu Dressier, 1993) is paraphyletic (Fig.

2.9). Phylogenies inferred from ndh? unweighted sequences (Fig. 2.6) and from combined data (Fig 2.9) show no particular alliance among the tribes of the dendrobioid subclade Podochileae, Dendrobieae, and Vandeae. Therefore, the dendrobioid subclade (sensu Dressier, 1993) appears polyphyletic.

Morphological Character Evolution in Epidendroideae. Evolution of the morphological characters used in this study is discussed below. Evolution of these characters is inferred from the single tree discovered from the parsimony search in which sequence data and morphological characters were combined (Fig. 2.9). 1. Growth Habit. According to Arditti (1992), two general growth forms can

be distinguished in the orchids. The monopodial form is characterized by a

persistent terminal shoot apex, indeterminate growth, absence of , no new

growth from stem bases, adventitious roots produced from the stem, lateral

inflorescences, and branching between nodes. The sympodial form is characterized

by presence of rhizomes, upright determinate stems, terminal or lateral

inflorescence, and new growth from axillary buds. With respect to growth habit,

present data suggests that the sympodial condition represents the predominant

evolutionary theme within the Epidendroideae. As indicated by the character tree

(Fig. 2.11), the sympodial habit clearly represents the basal character state in the

Orchidaceae. Although the monopodial Sobralia (Sobraliinae) appears as the most

basally diverged taxon among the Epidendroideae, the monopodial state apparently

represents an autapomorphy and not a general evolutionary shift toward

monopodialism. Only among the three Vandeae subtribes (Aeridinae, Angraecinae,

Aerangidinae) does there appear to be a major radiation of taxa exhibiting the

monopodial habit.

According to Holttum (1955), a sympodial habit in branching is almost

universal in monocotyledons. Variations in detail within the sympodial pattern like those displayed by climbing Araceae, bulbous Amaryllidaceae, and cormous

Iridaceae are all modifications of the same basic pattern of sympodial branching.

Holttum suggested that the lack of cambium in monocots led first to a peculiar type of continuous vegetative growth in taxa native to the moist tropics, namely sympodial growth, and that this type of growth proved itself peculiarly adaptable to the production of resting organs, thus allowing cambium-less plants to spread to 78

seasonal climates. He contended that the sympodial growth habit in orchids was

an adaptation that ensured each new stem had its own set of roots which is clearly

valuable for maintaining a water supply and securing (in epiphytes) attachment to a

supporting tree. The monopodial orchids with stems capable of unlimited growth

overcome this problem by producing roots at all nodes of the stem.

Dressier and Dodson (1960) and Dressier (1993) considered the sympodial

condition to be primitive in the orchids. These authors believed that the monopodial

habit appears to have evolved independently in many groups and that its evolution

has followed different patterns in different cases. For example, in the Vanillinae,

the monopodial habit is achieved simply by the retention of apical growth in the

members of a sympodium with lateral inflorescences. In some Oncidiinae, the

monopodial habit seems to have evolved by the retention of a permanent juvenile

form .

2. Leaf Folding. In Orchidaceae, may be folded in two ways. Leaves

with a single fold at the midvein and V-shaped in transverse section are referred to

as conduplicate. Leaves that are folded or pleated along several prominent parallel

veins are referred to as plicate. According to Arditti (1992), conduplicate leaves are

usually strap-shaped but may be triangular or terrete; they are often thick and

fleshy. Plicate leaves are generally thin. Among the Epidendroideae taxa, leaf type

generally appears correlated to habitat. That is, epiphytic taxa typically exhibit

conduplicate leaves, whereas terrestrial taxa exhibit plicate leaves. However, the

epiphytic taxa Coeiogyne (Coelogyninae), Stanhopea (Stanhopeinae), and Chysis

(Chysiinae) with plicate leaves are the exceptions (Fig. 2.12). With respect to leaf folding, the conduplicate leaf represents the predominant evolutionary theme in the 79

Epidendroideae, but the present data (Fig 2.12) does not indicate which condition is plesiomorphic in the family.

According to Dressier and Dodson (1960), the primitive orchid leaf is probably nonarticulate, wide and plicate; the trend toward an articulate leaf, has probably occurred independently in several phyletic lines as the conduplicate leaf appears to be correlated with the epiphytic habit. Conduplicate leaves were thought to have evolved from plicate leaves (Dressier, 1993).

3. Inflorescences. As defined by Arditti (1992), inflorescences produced at the apex of shoots are called terminal. Those arising from nodes near the base of pseudobulbs, from the sides of stems, or from leaf axils (or opposite them) are lateral. Dressier and Dodson (1960) considered the terminal inflorescence to be ancestral and the lateral derived. Dressier (1993) stated that the terminal infloresence is clearly the ancestral condition; an upper lateral inflorescence occurs occasionally in other groups, but consistently occurs in the dendrobioid subclade.

He considered the basal lateral inflorescence derived but noted that this type of inflorescence also occurs in groups characterized by terminal inflorescences such as the Laeliinae and Pleurothallidinae. Terminal inflorescences occasionally occur in groups characterized by lateral inflorescences, and these may be either ancestral features or reversals (Dressier, 1993).

The terminal inflorescence is clearly indicated as the basal character state in

Orchidaceae in the present study (Fig 2.13). This evolutionary trend continues into the Epidendroideae where the most basal taxa, Sobralia (Sobraliinae) and Thunia

(Thuniinae), exhibit a terminal inflorescence. Although the branch leading to the

Dendrobium (Dendrobiinae)-Coe/o<7y/7e (Coelogyninae) and the Ceratostylis (Eriinae)- 8 0

Eria (Eriinae) clades is equivocal, it is apparent that a major evolutionary shift

toward a lateral inflorescence ensued and became a major evolutionary theme in the

Epidendroideae. Within the grade of taxa exhibiting the lateral inflorescence, only

Pofystachya (Polystachyinae) displays an autapomorphic reversal toward the

terminal inflorescence. Of the 14 taxa in this study exhibiting a lateral

inflorescence, all but four ( Coel/a (Coeliinae), Tipularia (Calypsoeae), Maxitiaria

(Maxillariinae), Stanhopea, Stanhopeinae)) are of Old World distribution. A major

reversal toward the terminal inflorescence among subtribes of the New World

Epidendreae (Meiracyliinae, Pleurothallidinae, Laeliinae, Arpophylliinae) is indicated

by the character state tree (Fig. 2.13). Judging by the large number of species in

Epidendroideae exhibiting either the terminal or lateral inflorescence, both character states may be considered successful adaptations.

4. Resupination. In orchids, the orientation of the lip (labellum) on the lower side of the flower is referred to as resupination, and is achieved by various means. In most cases, the pedicel twists 180° during bud development.

Resupination is also achieved when the pedicel bends down beside the peduncle or bends over the apex of the stem. Resupination is passively achieved by a pendent infloresence. According to Dressier (1993), twisting of the pedicel is mediated by gravity and occurs regardless of the position of the plant or rachis; resupination occurs in all orchid subfamilies and is a basic feature of the family though it may be lost or modified.

As depicted in the character tree (Fig. 2.14), resupination represents a pervasive characteristic in the Orchidaceae. Nonresupination exhibited by the disparate Arpophyllum (Arpophyliinae), Pleurothallis (Pleurothallidinae), Eria 81

(Eriinae), and Polystachya (Polystachyinae) appears as independent parallel specializations. Therefore, resupination does not appear to be a consistent character at the subfamilial, tribal or subtribal level. Its use as a consistent systematic character is probably restricted to lower taxonomic levels.

5. Pollen Aggregation. The predominant theme in orchid evolution may be the union of pollen grains into pollinia (Dressier, 1993). In the Apostasioideae, the pollen grains are free powdery monads not aggregated into pollinia. This type of pollen is the least specialized within Orchidaceae (Chen, 1982; Arditti, 1992), and clearly the primitive condition (Dressier, 1993). Although the pollen grains of most

Cypripedioideae are held in a common viscid mass, they typically are not aggregated into pollinia. In most monandrous orchids, the pollen grains cohere into definite masses. Burns-Balogh and Funk (1986) recognized, three types of pollen aggregates occurring in the monandrous orchids: 1) granulate or soft pollinia that may be composed of tetrads or rarely monads, 2) sectile pollinia composed of individual pollen packets called massulae, and 3) hard compact pollinia. Soft pollinia may be granulate or sectile, with sectile representing the derived condition

(Dressier, 1993). Dressier concluded that hard pollinia exhibited by most

Epidendroideae represents a more derived state. As indicated by the character tree

(Fig. 2.15), granulate pollen, represents the plesiomorphic condition in the

Epidendroideae as exhibited by Sobralia (Sobraliinae) and Thunia (Thuniinae). This tree suggests that the condition of firm pollinia represents the derived condition and is exhibited by the majority of taxa representing the subfamily. Reversals occur in the Arpophylliinae and Coeliinae. Arpophyllum Pleurothallis Restrepia Zootrophion Octomeria Brassavola Epidendrum Encydia Cattleya Meiracyllium Chysis Cymbidium Stanhopea M axillaria Vanda Polystachya Angreecum Aerangis Bui bophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Eria Dendrobium Coeiogyne Habit Thunia Sobralia Sympodial H H Listera Monopodial 1 I Cypripedium Spiranthes Habenaria Neuwiedia

Figure 2.11 Character-state tree for growth habit for orchid taxa Included In this study. Arpophyllum Fleurothallis Restrepia Zootrophion Octomerla Brassavola Epidendrum Encydia Cattleya Meiracyllium Chysis Cymbidium Stanhopea Maxillaria Vanda Polystachya Angraecum Aerangi3 Bulbophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Eria Dendrobium Coeiogyne Thunia Conduplicate B H Sobralia Plicate m Listers Equivocal Cypripedium Spiranthes Habenaria Neuwiedia

Figure 2.12 Character-state tree for leaf type for orchid taxa included in this study. Arpophyllum Pleurothallis Restrepia Zootrophion Octomeria Brassavola Epidendrum Encydia Cattleya Meiracyllium Chysis Cymbidium Stanhopea M axillana Vanda Polystachya Angraecum Aerangis Bulbophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Eria Dendrobium Coeiogyne Infloresence Thunia Terminal MW Sobralia Lateral [~ ~ l Equivocal E = i Lister a Cypripedium Spiranthes Habenaria Neuwiedia

Figure 2.13 Character-state tree for infloresence type for orchid taxa included in this study. Arpophyllum Pleurothallis Restrepia Zootrophion Octomeria Brassavola Epidendrum Encyclia Cattleya Meiracyllium Chysis Cymbidium Stanhopea Maxillaria Yanda Polystachya Angraecum Aerangis Bulbophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Eria Dendrobium Coeiogyne Thunia Resuplntlon Sobralia Resupinate B H Listera Nonresupinate I 1 Cypripedium Spiranthes Habenaria Neuwiedia

Figure 2.14 Character-state tree for resupination type for orchid taxa included in this study. Arpophyllum Pleurothallis Restrepia Zootrophion Octomeria Brassavola Epidendrum Encyclia Cattleya Meiracyllium Chysis Cymbidium Stanhopea Maxlllaria Vanda Polystachya Angraecum Aerangls Bulbophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Eria Dendrobium Coeiogyne

Pollen Aggregation Thunia Sobralia Firm Granulate Li s ter a Sectile Cypripedium Nonaggregated Spiranthes Equivocal Habenaria Neuwiedia

Figure 2.15 Character-state tree for pollen aggregation type for orchid taxa included in this study. 87

Dressier (1993) believed that the Epidendroideae should not be limited to

those orchids with hard pollinia as there is a continuous gradation from the loose

sticky pollen grains exemplified by Vanilla to the very hard pollinia exemplified by

Vanda. He stated that if an arbitrary line were drawn between taxa with hard and

soft pollinia in an attempt to separate the Epidendroideae from the Neottioideae, the

Epidendroideae so delimited would probably be polyphyletic. The results of this

study support Dressler's contention that the Epidendroideae should not be delimited

by the hardness of pollinia as those taxa with soft pollinia (Sobralia, Thunia,

Arpophyllum, and Coefia) are clearly indicated to be epidendroids (Fig. 2.9).

6. Number of Fertile Anthers. Traditionally, the orchids have been separated into two major groups based on the number and position of fertile anthers (Burns-

Balogh and Funk, 1986). In the first group are the monandrous orchids which produce a single fertile anther and encompass approximately 20,000 species. The second group is composed of those taxa with two or three fertile anthers, and encompasses about 200 species in the Cypripedioideae and Apostasioideae.

Historically, the diandrous and triandrous orchids have been considered more primitive than the monandrous orchids largely based on this character (Raven and

Axelrod, 1974; Burns-Balogh and Funk; 1986). The evolution of this character is discussed in Chapter 4.

7. Number of Pollinia. The form and number of partitions of the anther of monandrous orchids determines the shape and number of the pollinia. In most orchids there are four pollinia which represent the four anther cells (Dressier, 1993).

According to Rasmussen (1985), the basic number of pollinia in Orchidaceae is four, representing each pollen sac (microsporangium). The interior strands of the 88

anther tissue wall remain as septa between the pollinia and form the anther

partitions. Through reduction of these partitions, the pollinia may be only two and

by the formation of secondary partitions, the number of pollinia may increase up to

twelve. Dressier (1993) stated that eight pollinia probably represents the primitive

condition in the epidendroid phylad, and within the tribe Epidendreae, reduction to

six, four, and two pollinia occurred independently in the Laeliinae and

Pleurothallidinae subtribes. Pridgeon (1982) supported the hypothesis that the

Pleurothallidinae shows a reduction series in the number of pollinia from eight to six

to four to two with eight representing the primitive condition. However, Luer

(1986) apparently believed that the direction of pollinia evolution in Pleurothallidinae

may have proceeded in the opposite direction from two to eight.

The most parsimonious explanation for the evolution of pollinium number in

Epidendroideae is depicted in the character state tree (Fig. 2.16). This tree suggests that the eight pollinium condition represents the basal condition in the subfamily. The four pollinium state exhibited by Dendrobium (Dendrobiinae) and

Coelogyne (Coelogyninae) may be considered autapomorphic. Because more than

one character state provides an equally parsimonious explanation, the branch leading to the vandoid clade is equivocal. However, it is apparent from the character state tree (Fig. 2.16) that the ancestral precursors to the vandoids exhibited eight pollinia.

Along the axis of evolutionary descent toward the New World tribe

Epidendreae, the eight pollinium character state remained unchanged along one branch to the Meiracyliinae-Laeliinae clade and along the other branch to the

Arpophyliinae-Pleurothallidinae clade. Octomeria with eight pollinia occupies the 89 basal position within subtribe Pleurothallidinae. The reduction in number of pollinia in Restrepia (four) and in Pleurothallis (two) and Zootrophion (two) is consistent with the hypothesis of Pridgeon (1982) and Dressier (1993) that the

Pleurothallidinae has undergone an evolutionary reduction in the number of pollinia.

A similar trend is also noted in the Laeliinae as Brassavola (eight pollinia) occupies the basal position with respect to Epidendrum, Encydia, and Cattleya (each exhibiting four pollinia).

8. Caudicles. The majority of orchids have some type of accessory structure accompanying the pollinia and viscidium. These structures may take the form of pollen-embedded caudicles or sterile caudicles with a stipe. A caudicle may be defined as a slender, mealy, or elastic extension of the pollinium produced within the anther. The caudicle functions as a weak point that permits the pollinia to break away from a and attach onto the stigma of a conspecific flower

(Yeung, 1987).

As described by Dressier (1993), Apostasioideae and the Cypripedioideae generally produce neither aggregated pollen nor caudicles. According to Rasmussen

(1986), the most primitive caudicle is merely an apical extension of the pollinia as exhibited in Ludisia discolor (Spiranthoideae).

The majority of Epidendroideae have some form of caudicle. The vandoids usually exhibit a stipe that is accompanied by a sterile caudicle; the epidendroids typically produce a pollen-embedded caudicle without a stipe. Caudicles are lacking in the Malaxideae and Dendrobieae.

Blackman and Yeung (1983) performed developmental studies on

Epidendrum ibaguense and described the caudicle as beginning as a mass of 9 0

meristematic cells in the microsporangium. The central cells in the mass enter a

division cycle which produces linear tetrads. These tetrads subsequently form thick

secondary walls that form the main structural element of the caudicle at maturity.

The cells on the periphery of the mass are largely thin-walled and produce large

quantities of a lipid polymer. At maturity, these thin-walled cells undergo autolysis

and release their lipid masses that form the elastic joining compound between

adjacent caudicles and between caudicles and pollinia.

Dressier (1993) recognized three types of caudicles: 1) hard or bony, 2)

granular, and 3) hyaline (elastoviscin). Burns-Balogh and Funk (1986) recognized

two broad types of caudicles: 1) embedded with pollen; 2) sterile (elastoviscin).

Elastoviscin is defined as a translucent elastic material lacking cellular composition

and is produced by the tapetum. According to Dressier (1993), mealy pollinia with distinct caudicles is the primitive condition. However, he considered the lack of

caudicles in the Malaxideae and the Dendrobieae to be derived; he also considered the massive caudicles of some Coelogyneae to be derived.

As illustrated by the character state tree (Fig. 2.17), the fertile caudicle

appears as the basal and most pervasive type of pollinium accessory structure in

Epidendroideae. The fertile caudicle character state remains unchanged along the axis of evolutionary descent from the most basal taxon, Sohra/ia (Sobraliinae) to the derived Arpophyliinae-Pleurothallidinae clade. As contended by Dressier (1993) and suggested by the character state tree (Fig. 2.17) the absence of caudicles in

Dendrobium (Dendrobiinae) probably represents an autapomorphy. That is, lack of caudicles in the Dendrobiinae with hard pollinia is probably not homologous with 91 those taxa with granulate pollinia that also lack fertile caudicles (e.g. Listera

(Listeriinae) and Spiranthes (Spiranthoideae).

Sterile caudicles associated with a stipe appear in the character state tree

(Fig. 2.17) as a major evolutionary trend in the vandoid clade. Based on this tree,

Bulbophyllum (Bulbophyliinae) descended from ancestors exhibiting a sterile caudicle

(and an implied stipe). Therefore, the absence of caudicles in Bulbophyliinae may be considered a reversal from a sterile caudicle condition and not a reversal from the fertile caudicle condition (as is the case in Dendrobium). The contention that the Bulbophyliinae evolved from taxa exhibiting sterile caudicles and stipes is supported by the fact that a few species of Bulbophyllum exhibit hamular stipes

(Rasmussen, 1985). However, a view contrary to this evolutionary scenario is discussed in the following section on stipe evolution. The fertile caudicles exhibited in Cymbidium (Cyrtopodiinae) are considered a reversal form the sterile caudicle- stipe condition. It is cautioned that this reversal is not necessarily representative of the Cyrtopodiinae as other genera in this subtribe exhibit a sterile caudicle with stipe.

9. Stipes. As noted above, the majority of orchids have some type of accessory structure accompanying the pollinia and viscidium. For a minority of

Epidendroideae, these accessory structures take the form of sterile caudicles with a strap-shaped stipe. Most orchids with stipes are vandoid-like taxa (sensu Dressier,

1993).

Unlike the caudicle that is formed within the anther, stipes are stigmatic in origin, developing from the rostellum (Rasmussen, 1986). Depending on their ontogenetic development, two types of stipes are recognized. The tegular stipe is 92

the more common of the two and is described by Rasmussen (1985) as a ridge or

strap-like plate of thick-walled cells that becomes freed from the abaxial side of the

rostellum by disintegration of the anticlinal cell walls. After anthesis, the plate of

cells is glued to the pollinia by viscid sterile material. Dressier (1993) described the

tegular stipe as formed by the epidermis of the rostellum and clinandrium (anther

bed). He stated that the tegular stipe is said to be rostellar in origin but may be

stylar as often as stigmatic, so columnar may be a better term. The tegular stipe is

usually a single cell thick.

The hamular stipe develops differently from the tegular stipe. As described

by Dressier (1993), the hamular stipe develops as an extension of the rostellum and

forms a connection between the pollinia. Rasmussen (1985) described the hamular stipe as a hook-like structure formed by the apical growth of the rostellum, and stated that the pollinarium stalks provide diagnostic characters in orchid classification. Dressier (1993) believed that each type of stipe evolved independently in several groups and that caudicles are usually a prerequisite for stipes.

As is evident in the character-state tree (Fig. 2.18), stipes represent a major morphological component in the vandoid taxa. The tegular stipe is consistent with the allied taxa: Stanhopea (Stanhopeinae); MaxiHaria (Maxillariinae); Vanda

(Aeridinae); Po/ystachya (Polystachyinae); Angraecum (Angraecinae); Aerangis

(Aerangidinae). The lack of a stipe in Cymbidium (Cyrtopodiinae) may be viewed as a reversal in the genus but not the subfamily as most Cyrtopodiinae exhibit stipes.

The hamular stipe exhibited in Tipularia (Calypsoeae) represents an autapomorphy. 9 3

As illustrated by the character state tree (Fig. 2.18), the lack of a stipe in the

nonvandoid Bulbophyllum (Bulbophyliinae) did not occur by a reversal from the stipe

character state. That is, the Bulbophyliinae evolved from a nonstipe precursor.

This evolutionary scenario is incompatible with that suggested by the caudicle character-state tree (Fig. 2.17) where the absence of a caudicle exhibited in

Bulbophyllum appears as a reversal from the sterile caudicle state. As an alternative explanation for caudicle evolution that would be compatible with stipe evolution in the Epidendroideae, the caudicle character-state tree should be amended (Fig.

2.18). Specifically, the branch leading to the vandoid -Bulbophyllum clade would be changed to the "fertile caudicles" state which would imply that the sterile caudicle associated with the hamular stipe of Tipularia would represent an autapomorphy.

10. Substrate. Most authorities have held that the terrestrial habit represents the basal condition in Orchidaceae and the epiphytic condition is derived

(Schimper, 1888; Goebel, 1922; Garay, 1960; Mulay and Deshpande, 1961;

Garay, 1972; Ackerman, 1983; Benzing and Atwood; 1984; Benzing, 1987).

However, the contention that the Orchidaceae had a terrestrial ancestry was challenged by Robinson and Burns-Balogh (1982) on the basis of microspermy, floral modifications leading to specialized mechanisms, and root velamen.

The evolution of epiphytism is discussed in Chapter 5.

11. Velamen. Pridgeon (1987) attributed much of the success of the

Orchidaceae in colonizing most of the world’s land masses to root anatomy. He defined velamen as that tissue which arises from root dermatogen; it consists of dead cells at maturity and is bordered internally by an exodermis provided with passage cells capable of absorbing water and dissolved substances. Noel (1974) Arpophyllum Pleurothallis Restrepia Zootrophion Octomeria Brassavola Epidendrum Encyclia Cattleya Meirecyllium Chysis Cymbidium Stanhopea Maxillaria Vanda Polystachya Angraecum Aerengis Bulbophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Eria Dendrobium Coelogyne Thunia Number of Pollinia Sobralia Listera Cypripedium Spiranthea Habenaria Equivocal Neuwiedia

Figure 2.16 Characater-state tree for number of pollinia for orchid taxa Included in this study. Arpophyllum Pleurothallis Restrepia Zootrophion Octomeria Brassavola Epidendrum Encyclie Cattleya Mei racyllium Chysis Cymbidium Stanhopea Maxillaria Vanda Pol ystachya Angraecum Aerangis Bui bophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Eria Dendrobium Coelogyne Thunia Caudicles Sobralia With Pollen Listera None Cypri pedium Sterile Spiranthes Equivocal Habenaria Neuwiedia

Figure 2.17 Character-state tree for caudicle type for orchid taxa included in this study. Arpophyllum Pleurothallis Restrepia Zootrophion Octomerla Brassavola Epidendrum Encyclia Cattleya Meiracyllium Chysis Cymbidium Stanhopea M axillaria Vanda Polystachya Angraecum Aerangis Bulbophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Eria Dendrobium Coelogyne Thunia Sobralia Absent Listers Tegular Cypripedium Hamular F!?r1 Spiranthes Habenaria Neuwtedia

Figure 2.18 Character-state tree for stipe type for orchid taxa Included In this study. 97 described cell wall structure and development of velamen in gigantea. That study demonstrated that velamen originates by periclinal divisions from the root dermatogen. Immediately after divisions leading to the formation of the velamen have ceased, there follows a phase of cell vacuolation and enlargement. At this time, the cell walls become impregnated with lignin as helical wall thickenings. At a level approximately 12 mm proximal to the apical meristem, velamen cells lose their cytoplasm and die. Although this result in a loss of turgor pressure, collapse of the velamen is prevented by the helical thickenings in the cell wall. Perforations appear in the radial and tangential walls soon after the death of the cells. These perforations represent the final stages in the breakdown of the primary pit fields.

Experiments using 32P have demonstrated that velamen acts to absorb water (Sheehan et al., 1967; Barthlott and Capesius, 1975). These experiments demonstrated that the highest absorption capacity occurs in younger roots but that older living roots were still capable of absorption. According to Benzing et al.

(1982), velamen seems designed to mediate moisture exchange and mineral absorption. They believed that the velamen acts as a sponge allowing the root to immobilize a reservoir of moisture and minerals from precipitation or canopy leachates. Once the velamen is fully imbibed, the aqueous solution lies close to the absorptive exodermal passage cells; at this point, water and salt enter the cortex.

In Sobralia and other orchids that possess fibrous bodies, moisture and solutes must traverse interstitial spaces before becoming accessible to passage cells. Benzing et al. (1982) believed that the fibrous body provides a barrier to transpiration by lengthening the path water must traverse to breach the exodermis velamen barrier. They speculated that if the fibrillar components of the fibrous body alternately compact and swell upon desiccation and hydration, its mass would function like a one-way valve. That is, access to underlying passage cells would be maximal when the velamen is engorged and transpiration impeded when the velamen is dry. Pridgeon et al. (1983) termed these fibrous bodies "tilosomes" and defined them as lignified excrescences from the walls of cells of the innermost velamen cell layer adjacent to thin-walled passage cells of the exodermis. In a survey of 350 orchid species in 175 genera conducted by Pridgeon (1987), tilosomes were found to occur in 95 species of 39 genera; the presence of tilosomes is almost exclusively a Neotropical phenomenon. Pridgeon (1987) found that the presence and type of tilosomes could be useful for taxonomic purposes at the generic and subtribal levels. He viewed multiseriate velamen as an evolutionary specialization designed to diminish transpiration rates and to allow for longer access to moisture and dissolved minerals. Pridgeon (1987) considered the number of cell layers as having little systematic utility above the species level as it is more a measure of adaptation to different habitats and substrates.

As the result of an examination of 344 species of 262 genera, Porembski and Barthlott (1988) developed a classification scheme based on velamen type for

Orchidaceae, naming the different types after genera exhibiting the syndrome. A brief description of the velamen types relevant to this study is listed in Appendix A.

These velamen types were based on the following six characters: 1) number of cell layers, 2) stratification of layers, shape, and size of single cells, 3) helical wall thickenings (which often allow the delimitation of tribes) and the number and size of pores perforating the dead cell walls (which may be specific for particular taxa). 99

4) Stabkorper (tilosomes) which exhibit a variety of structures, 5) size and

thickness of exodermal cells, and 6) types of idioblasts in the cortex.

As indicated by the character-state tree (Fig.2.19), velamen type has

undergone several major evolutionary lines of descent in the Epidendroideae. The

velamen type in the most basal epidendroid taxon Sobralia (Sobraliinae) is

undefined, therefore, the plesiomorphic velamen type is unresolved. Of the recognized types, the Coelogyne velamen type as exhibited in Thunia and

Coelogyne appears as the plesiomorphic condition.

The Calanthe velamen type next appears along the axis of evolutionary descent as exhibited in Eria and Ceratostylis (Eriinae); in Spathoglottis and Calanthe

(Bletiinae), and Tipularia (Calyposeae). As indicated by the character-state tree (Fig.

2.19), the Epidendrum velamen type appears along the axis of evolutionary descent as exhibited in the New World Epidendreae subtribes (Meiracyliinae, Laeliinae, and

Arpophyliinae). Within the New World tribe Epidendreae, the Pleurothallis velam en type (exhibited by all members of subtribe Pleurothallidinae) represents the derived sta te .

The Vanda velamen type is consistently exhibited by the three Vandeae subtribes Aeridinae, Aerangidinae, and Angraecinae. With the exception of

Tipularia (Calypsoeae), the three representatives of the cymbidioid phylad:

Cymbidium (Cymbidieae), Stanhopea (Maxillarieae), Maxillaria (Maxillarieae) consistently exhibit the Cymbidium velamen type. As suggested by the character state tree (Fig. 2.19), velamen type appears as a consistent phylogenetically informative character since no reversals nor parallelisms appear to have occurred in the Epidendroideae. 1 00

12. Seeds. Orchid seeds are very small, dust like and consist of a tiny rudimentary embryo suspended in a reticulate or net-like testa and surrounded by a large volume of space (Arditti, 1992). They range from 3 to 14ug in weight and from 0.4 to 1.25mm long. Some species may produce as few as 1,300 seeds per , whereas others may produce as many as 4 million. Most species lack an endosperm and cotyledon.

In the Apostasioideae and Cypripedioideae, the testa is opaque, sculptured, and derived from the inner and outer integuments of the ovule. All other orchid seeds have transparent testae derived from the outermost ovule integument. The testae become transparent as the outermost cell layer of the integument loses its protoplasts (Garay, 1960). Because only the outer layer of the outer integument persists in the mature orchid seed, Barthlott (1976) believed that this morphology may have adaptive significance with regard to aerodynamic and wettability properties. Seeds of the Apostasioideae and the Cypripedioideae are generally considered to represent the primitive state in Orchidaceae (Garay, 1960; Chen,

1982; Arditti, 1992).

Until recently, the use of seeds in orchid classification has been neglected.

Scanning electron microscopy (SEM) has allowed botanists to survey a larger selection of orchid seeds (Dressier, 1993). For example, Barthlott (1976) used SEM to survey the seeds of 150 species of 58 orchid genera. He concluded that seed morphology can be used to separate genera is most important at the subtribal and tribal levels. Dressier (1993) has summarized and modified seed types as defined by Ziegler (unpublished). These seed types are defined in Appendix B. 101

Due to undefined seed types in Ceratostylis (Eriinae) Eria (Eriinae), and undetermined seed types of Spathoglottis (Bletiinae) and Calanthe (Bletiinae), an overall evolutionary trend in the Epidendroideae cannot be determined by the character state tree (Fig. 2.20). However, several major trends in seed evolution are evident. For example, the closely aligned Dendrobium (Den6rob\\nae)-Coelogyne

(Coelogyninae) clade and Thunia (Thuiinae) share the Dendrobium seed type. Based on the character tree (Fig. 2.20), the Dendrobium seed type exhibited in

Bulbophyllum (Bulbophylliinae) appears to represent parallel or convergent evolution in this seed type. The Po/ystachya (Polystachyinae) -Angraecum (Angraecinae)-

Aerangis (Aerangidina e)-Vanda (Aeridinae) association exhibits the Vanda seed type.

As this seed type displays no reversals or parallelisms in the character state tree

(Fig. 2.20), the Vanda seed type appears as a consistent homologous character.

Although the Pleurothallis seed type exhibited by over 3,000 species in the

Pleurothallidinae and the Epidendrum seed type exhibited by over 1,400 species in the Laeliinae are the predominant seed types found in the New World tribe

Epidendreae, the character state tree (Fig. 2.20) suggests that the EHeanthus seed type represents the plesiomorphic seed type in this largest of all orchid tribes. From

Chysis (Chysiinae) along one branch toward the Arpophyliinae and along another branch toward Meiracyliinae the EHeanthus seed type appears to have been the precursor that led to both the Pleurothallis and Epidendrum seed types.

13. Fusion of Anthers to Style. The gynostemium (column) is the central structure of an orchid flower and is variously formed by the fusion of style, stigma, filaments, and staminodia. The basic condition occurs in Neuwiedia with only basal fusion of style and filaments. The length of the style varies according to species; 102

however, the three anthers are always below the stigma base at anthesis. The

filaments of the two fertile anthers in Cypripedioideae are appressed to the style.

The anther lobes are large and rounded; the staminode is prominently expanded. In

the monandrous orchids, the style is fused with the filaments, staminodia, and

stigma; they exhibit the most extreme level of fusion between the gynostemium

and androecium.

Burns-Balogh and Funk (1986) considered the progression in fusion of filaments

to be one of the major trends in the evolution of Orchidaceae. They suggested that

the fusion of filaments resulted in the evolution of three pollination systems. One

described system occurs in the Apostasioideae where filament fusion and anther

expansion have led to the formation of a typical "buzz-pollination" flower. The

second occurs in the Cypripedioideae where the fusion of filaments in the

androecium along with two fertile anthers on either side of a wide staminode has

produced a trap blossom in conjunction with an inflated labellum. The third system

occurs in the monandrous orchids where the fusion of filaments has resulted in the

reduction of spatial isolation between stigma and anther. According to Burns-

Balogh and Funk (1986), this reduction in distance has permitted the evolution and

diversity of the pollinarium; severe spatial isolation between anther and rostellum

would be disadvantageous as it would preclude the connection between dry pollinia

and the viscidium.

The root for the character-state tree (Fig. 2.21) may be considered to represent the evolutionary line of descent of the ancestral orchids. Because the ancestral state for the degree of fusion of anthers and style in ancestral orchids is unknown, it is unresolved whether basal fusion of style and filaments as exhibited 103 Arpophyllum Pleurothallis Restrepia Zootrophion Octomeria Brassavola Epidendrum Encyclia Cattleya Melracyllium Chysis Cymbidium Stanhopea Maxillaria Vanda Pol ystachya Angraecum Aerangis Bui bophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Erie Dendrobium Coelogyne Thunia S eed s Sobralia Neuwiedia liMa Eulophia Listers Orchis im Vanda Cypripedium Goodyera PTX1 Maxillaria Limodorum EBB Cymbidium Splranthes Bletia Stanhopea Hebenaria Dendrobium EHeanthus Neuviedia Undefined EasaS Pleurothallis KTiiTJ Unknown P73 Equivocal fZZJ Epidendrum i I

Figure 2.20 Character-state tree for seed type for orchid taxa included in this study. 1 0 4

Arpophyllum Pleurothallis Restrepia Zootrophion Qdomeria Brassavola Epidendrum Encydia Cattleya Meiracyllium Chysis Cymbidium Stanhopea Maxillaria Vanda Pol ystachya Angraecum Aerangis Bulbophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Eria Dendrobium Coelogyne Thunia Velamen Sobralia Absent Vanda E23____ Listers Spiranthes Cymbidium Bffla Undefined l l Epidendrum Cypripedium Coelogyne Dendrobium E-E3 Spi ranthes Pleurothallis htrii Calanthe Habenaria BulbophyllumESS Equivocal EZ23 Neuviedia

Figure 2.19 Character-state tree for velamen type for orchid taxa included in this study. Arpophyllum 105 Pleurothallis Restrepia Zootrophion Oetomeria Brassavola Epidendrum Encyclia Cattleya Meiracyllium Chysis Cymbidium Stanhopea Maxillaria Vanda Polystachya Angraecum Aerangis Bulbophyllum Tipularia Coelia Spathoglottis Calanthe Ceratostylis Erie Dendrobium Coelogyne Thunia Sobralia Listera Cypripedium Spiranthes Style fused with filaments Habenaria Neuviedia at b ases and staminode bases F^l staminodes, and stigma to coiumnHH Equivocal

Figure 2 .21 Character-state tree for style fusion for orchid taxa included in this study. 106 in Neuwiedia represents the plesiomorphic state in the extant orchids or merely an autapomorphy. The character-state tree (Fig. 2.21) illustrates that the condition where filaments of the two fertile anthers are appressed to the style (as exhibited in

Cypripedium) arose from the monandrous state that is characterized by a style that is fused with filaments, staminodia, and stigma. The fusion of anthers and style as exhibited in Cypripedium may be explained most parsimoniously as an autapomorphy and not a reversal along the predominant line of orchid evolution that is characterized by the fusion of style, filaments, staminodia, and stigma exhibited in the monandrous orchids.

As previously stated, the 13 morphological characters used in this study bring a substantial degree of resolution to the phylogenetic relationships of the

Epidendroideae, albeit branch support for these relationships is typically weak in terms of bootstrap values and decay indices. In the aggregate, these characters show little homoplasy and offer new insights into the morphological evolution of

Epidendroideae. With the exception of the character for resupination, these characters appear to be useful in higher level systematic studies of the Orchidaceae.

LITERATURE CITED

Ackerman, J. D. (1983). On the evidence for a primitively epiphytic habit in orchids. Systematic Botany 8: 474-476.

Arditti, J. 1992. Fundamentals of Orchid Biology. John Wiley & Sons, New York.

Atwood, J. T. Jr. 1984. The relationships of the slipper orchids (Subfamily Cypripedioideae). Selbyana 7: 129-247.

Baldwin, B. G. 1993. Molecular phylogenetics of Calycadenia (Compositae) based on ITS sequences of nuclear ribosomal DNA: chromosomal and morphological evolution reexamined. American Journal of Botany 80: 222-238. 107

Barthlott, W. 1976. Morphologie der Samen von Orchideen im Hinblick auf taxonomische und funktionelle Aspekte. Proceedings of the 8th World Orchid Conference, Frankfurt, 444-455.

. and I. Capesius. 1975. Mikromorphologische und funktionelle Untersuchungen am Velamen Radicum der Orchideen. Berichte der Deutschen Botanischen Gesellschaft 88: 379-390.

Baum, D. A. and K. J. Sytsrna. 1994. A phylogenetic analysis of Epilobium (Onagraceae) based on nuclear ribosomal DNA sequences. Systematic Botany 19: 363-388.

Benzing, D. H. 1987. Major patterns and processes in orchid evolution: a critical synthesis. In: J. Arditti, (ed.). Orchid Biology: Reviews and Perspectives Vol. IV. Comstock Publishing Associates, Ithaca, pp 34-77.

Benzing, D. H., D. W. Ott, and W. E. Friedman. 1982. Roots of (Orchidaceae): structure and function of the velrnen-exodermis complex. American Journal of Botany 69: 608-614.

Benzing, D. H. and J.T. Atwood Jr. (1984) Orchidaceae: Ancestral habitats and current status in forest canopies. Systematic Botany 9: 155-165.

Blackman, S. J. and E. C. Yeung. 1983. Structural development of the caudicle of an orchid ( Epidendrum ). American Journal of Botany 70: 97-105.

Bull, J. J., J. P. Huelsenbeck, C. W. Cunningham, D. L. Swofford, and P. J. Waddell. 1993. Partitioning and combining data in phylogenetic analysis. Systematic Biology 42: 384-397.

Burns-Balogh, P. and V. A. Funk. 1886. A phylogenetic analysis of the Orchidaceae. Smithsonian Contributions to Botany 61: 1-79.

Chen, S. C. 1982. The origin and early differentiation of the Orchidaceae. Acta Phytotaxonomica Sinica 20: 1-22.

Clegg, M. T. and G. Zurawski. 1992. Chloroplast DNA and the study of plant phylogeny: present status and future prospects. In: P. S. Soltis, D. E. Soltis, and J. J. Doyle (eds.), Molecular Systematics of Plants. Chapman and Hall, New York, pp 1-13.

Donoghue, M.J. and M.J. Sanderson 1992. The suitability of molecular and morphological evidence in reconstructing plant phylogeny. In: P.S. Soltis, D.E. Soltis, and J.J. Doyle (eds.). Molecular Systematics of Plants. Chapman and Hall, New York, pp 340-368. 108

Donoghue, M. J., J. Olmstead, J. F. Smith, and J. D. Palmer. 1992. Phylogenetic relationships of Dipsacales based on rbcL sequences. Annals of the Missouri Botanical Garden 79: 333-345.

Doyle, J. J. and J. L. Doyle. 1987. A rapid DNA isolation procedure for small quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11-15.

Dressier, R. L., and C. H. Dodson. 1960. Classification and phylogeny in the Orchidaceae. Annals of the Missouri Botanical Garden 47: 25-68.

. 1981. The Orchids: Natural History and Classification. Harvard University Press, Cambridge, MA .

. 1993. Phylogeny and Classification of the Orchid Family. Dioscorides Press. Portland, Oregon.

Felsenstein, J. 1985. Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39:783-791.

Garay, L. A. (1960). On the Origin of the Orchidaceae. Botanical Museum Leafllets 19: 5 7-95.

Garay, L. A. 1972. On the origin of the Orchidaceae, II. Journal of the Arnold Arboretum 53: 202-215.

Goebel, K. 1922. Erdwurzeln mit Velamen. Flora 115: 1-26.

Good, I. J. 1983. Good Thinking: The Foundations of Probability and Its Applications. Univ. Minnesota Press, Minneapolis, MN.

Hempel, C. G. 1965. Aspects of Scientific Explanation. Free Press, N.Y.

Hibbett, D. S. and R. Vilgalys (1993). Phylogenetic relationships of Lentinus (Basidiomycotina) inferred from molecular and morphological characters. Systematic Botany 18: 409-433.

Higgins, D. G., A. J. Bleasby, and R. Fuchs 1991. CLUSTAL V: improved software for multiple sequence alignment.

Hillis, D. 1987. Molecular versus morphological approaches to systematics. Annual Review of Ecology and Systematics 18: 23 -42.

Hiratsuka, J., S. Hiroaki, R. Whittier, T. Ishibashi, M Sakamoto, M. Mori, C. Kondo, Y. Honji, C. Sun, B. Meng, Y. Li, A. Kanno, Y. Nishizawa, A. Hirai, K. Shinozaki, and M.Sugiura. 1989. The complete sequence of the rice (Oryza sativa) chloroplast genome: Intermolecular recombination between distinct tRNA genes accounts for a major plastid DNA inversion during the evolution of cereals. Molecular and General Genetics 217: 185-194. 109

Holttum, R. E. 1955. Growth-habits of monocotyledons-variations on a theme. Phytomorphology 5: 399-423.

Jansen, R. K. 1992. Current research. Plant Molecular Evolution Newsletter 2: 13-14

Kluge, A. G. 1989. A concern for evidence and a phylogenetic hypothesis of relationships among Epicrates (Boidae, Serpentes). Systematic Zoology 38: 7-25.

Loyd, D. G. and V. L. Calder. 1991. Multi-residue gaps, a class of molecular characters with exceptional reliability for phylogenetic analyses. Journal of Evolutionary Biology 4: 9-21.

Luer, C.A. 1986 leones Pleurothallidinarum I. Systematics of the Pleurothallidinae. Annals of the Missouri Botanical Garden, St. Louis, Missouri.

Lutzoni, F. and R. Vilgalys. In press. Integration of morphological and molecular data sets in estimating fungal phylogenies. Canadian Journal of Botany.

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

Mulay, B. N. and B. D. Deshpande. 1961. Velamen in terrestrial monocots - role of velamen tissue in taxonomy and phylogeny of monocotyledons. Proceedings of the Rajasthan Academy of Science 8:115-120.

Mullis, K. B. and F. A. Fallona. 1987. Specific synthesis of DNA in vitro via a polymerase chain reaction. Methods in Enzymology 155: 335-350.

Noel, A. R. A. 1974. Aspects of cell wall structure and the development of velamen in Ansellia gigantea Reichb. f. Annals of Botany (London) 38: 495-504.

Olmstead, R. G., J. A. Sweere, P. A. Reeves, R. E. Spangler, R. W. Scotland and S. J. Wagstaff. 1993b. Comparing and combining ndh? and rbcL sequences for phylogenetic inference. (Abstract) American Journal of Botany 80: 121.

. J. A. Sweere, and K. H. Wolfe. 1993b. Ninety extra nucleotides in ndhF gene of tobacco chloroplast DNA: A summary of revisions to the 1986 genome sequence. Plant Molecular Biology 22: 1191-1193

. and J. D. Palmer 1994. Chloroplast DNA systematics: a review of methods and data analysis. American Journal of Botany 81: 1205-1224.

. and J.A. Sweere. 1994. Combining data in phylogenetic systematics: a empirical approach using three molecular data sets in the Solanaceae. Systematic Biology 43: 467-481. 110

Palmer, J. D., R. K. Jansen, H. J. Michaels, M. W. Chase, and J. R. Manhart. 1988. Chloroplast DNA variation and plant phylogeny. Annals of the Missouri Botanical Garden 75: 1180-1206.

Porembski, S., and W. Barthlott. 1988. Velamen radicum micromorphology and classification of Orchidaceae. Nordic Journal of Botany 8: 117-137.

Pridgeon, A. M.1982. Numerical analyses in the classification of the Pleurothallidinae (Orchidaceae). Botanical Journal of the Linnean Society 85: 103-131.

Pridgeon, A. M., W. L. Stern, and D. H. Benzing. 1983. Tilosomes in roots of Orchidaceae: morphology and systematic occurrence. American Journal of Botany 70: 1365-1377.

Pridgeon, A. M. 1987. The velamen and exodermis of orchid roots. In: J.Arditti, (ed.). Orchid Biology: Reviews and Perspectives Vol. IV. Comstock Publishing Associates, Ithaca, pp 140-192.

Rasmussen, F. N. (1985). Orchids. In: Dahlgren, R. M. T., H. T. Clifford, and P. F. Yeo (eds.) The Families of the Monocotyledons; Structure, Evolution, and Taxonomy.Springer-Verlag, Berlin, pp 249-274.

Rasmussen, F. N. 1986. On the various contrivances by which pollinia are attached to viscidia. Lindleyana 1(1): 21-32.

Raven, P. H., and D. I. Axelrod. 1974. Angiosperm biogeography and past continental movements. Annals of the Missouri Botanical Garden 61: 539-673.

Revera, M. C. and J. A. Lake. 1992. Evidence that eukaryotes and eocyte prokaryotes are immediate relatives. Science 257: 74-76.

Robinson, H. and P. Burns-Balogh. 1982. Evidence for a primitively epiphytic habit in Orchidaceae. Systematic Botany 7: 353-358.

Saiki, R. K., D. H. Gelfand, S. Stoffel, S. J. Scharf, R. Huiguchi, G. T. Horn, K. B. Mullis, and H. A. Erlich. 1988. Primer-directed enzymatic amplification of DNA with thermostable DNA polymerase. Science 239: 487-491.

Sanderson, M. J. 1989. Confidence limits on phylogenies. Cladistics. 5: 113-129.

Schluchter, W. M., J. Zhao and D. A. Bryant. 1993. Isolation and characterization of th e rtdhF gene of Synechococcus sp. strain PCC 7002 and initial characterization of an interposon mutant. Journal of Bacteriology, 175: 3 3 4 2 -3 3 5 2 . 111

Sheehan, T.J., N. Joiner, and J. K. Cowart. 1967. Absorption of 32P by Cattleya Trimos' from foliar and root applications. Proceedings of the Florida State Horticultural Society 80: 400-404.

Shimada, H. and M. Sugiura. 1991. Fine structural features of the chloroplast genome: comparison of the sequenced chloroplast genomes. Nucleic Acids Research 19: 983-994.

Schimper, A. F. W. 1888. Die epiphytische Vegetation Amerikas. Botanische Mitteilungen aus den Tropen. II, 162 pp. Jena (Gustav Fisher).

Small. J. 1989. Evolution of plant mitochondrial genomes via substoichiometric intermediates. Cell 58: 69-76.

Sugiura, M. 1989. The chloroplast genome. The Biochemistry of Plants 15: 133-150.

Swofford, D.L. 1993. Phylogenetic Analysis Using Parsimony. Version 3.1.1. Champaign Illinois: Illinois Natural History Survey.

Swofford D. L. and Begle, D. P. 1993. Phylogenetic Analysis Using Parsimony. Illinois Natural History Survey, Champaign, Illinois. de Vogel, E. F. 1969. Monograph of the tribe Apostasieae (Orchidaceae). Blumea 17: 313-350.

Wakasugi, T, J. Tsudzuki, S. Ito, K. Nakashima, T. Tsudzuki, and M. Sugiura. Loss of all ndh genes as determined by sequencing the entire chloroplast genome of the black pine, Pinus thunbergii. Proceedings of the National Academy of Science 91: 9794- 9798.

Wen-Hsuing, L. and D. Graur. 1991. Fundamentals of Molecular Evolution. Sinauer Associates, Inc., Sunderland, MA.

Yeung, E. C. 1987. Development of pollen and accessory structures in orchids. Pp 195-226 in J .Arditti, (ed.). Orchid Biology: Reviews and Perspectives Vol. IV. Comstock Publishing Associates, Ithaca. CHAPTER 3

PHYLOGENY OF SUBFAMILY EPIDENDROIDEAE (ORCHIDACEAE) INFERRED FROM CHLOROPLAST ndliF GENE SEQUENCES BASED ON A MAXIMUM LIKELIHOOD ANALYSIS

112 113

The purpose of this study was to employ a maximum likelihood method using ndhF sequences: 1) to infer a phylogeny of the taxa listed in Table 2.2,

2) to evaluate the effects of differing transversion weighting parameters and jumble seeds used in the fastDNAml algorithm on resulting tree topologies and log likelihood values, and 3) to compare the phylogeny inferred from the maximum likelihood tree with the highest log likelihood value to phylogenies inferred from parsimony analyses.

Inferring phylogenetic relationships from molecular data requires the selection of an appropriate method from an array of techniques (Swofford and

Olsen, 1990). Inferring a phylogeny is an estimation procedure producing a best estimate of an evolutionary history based on incomplete information. In the context of molecular systematics, direct information about the past is generally unavailable; access is thus restricted to contemporary species and molecules (Swofford and

Olsen, 1990).

Two assumptions are necessary in all character-based methods of analysis:

1) independence among characters which allows the treatment of each position as a separate entity in time-consuming computational algorithms (e.g. numbers of substitutions can be minimized separately position by position and then summed over all positions in a parsimony algorithm or probabilities can be multiplied over positions in a maximum likelihood approach), and 2) characters are homologous

(e.g. a character must be defined such that all states observed over all taxa for that particular character must have been derived from a corresponding state observed in the common ancestor). If phylogenetic relationships are to be inferred from molecular relationships, the definition of homology must be further restricted to 114 include only orthologous, as opposed to paralogous genes (Swofford and Olsen,

1990).

Of the existing numerical approaches to inferring phylogenies directly from the character data, methods based on the principle of maximum parsimony have been the most widely used by far (Swofford and Olsen, 1990). Inferring phylogeny through parsimony methods operate by selecting trees that minimize tree length (ie. the total number of steps) (Felsenstein, 1983; Swofford and Olsen, 1990; Swofford and Begle, 1993). The theory behind parsimony analysis is that a simple hypothesis is preferable to more complicated or ad hoc hypothesis.

Parsimony methods may yield inconsistent estimates of the evolutionary tree when amounts of evolutionary change in different lineages are sufficiently unequal

(Felsenstein, 1978; Hasegawa and Fujiwara, 1993). Most data involve moderate to large amounts of change, and in these cases, parsimony methods can fail

(Felsenstein, 1981). To apply a maximum likelihood approach, a concrete model of the evolutionary process that converts one sequence to another must be specified.

A maximum likelihood approach to phylogenetic inference based on molecular data evaluates the net likelihood that the given evolutionary model will yield the observed sequences, and the inferred phylogenies are those with the highest likelihood

(Swofford and Olsen, 1990). The maximum likelihood model accounts for substitutions occurring more frequently between pairs of physicochemically similar ones (Hasegawa and Fujiwara, 1993), and is suitable in problems where the amount of data is limited (Goldman, 1990). Assumptions of the maximum likelihood model are: 1) each site evolves independently, 2) different lineages evolve independently, 115

3) each site undergoes substitution at an expected rate that must be specified, and

4) all sites are included (not just those that have changed) (Felsenstein, 1993).

MATERIALS AND METHODS

Sequences from taxa used in parsimony analyses (Table 2.2) were used in

this maximum likelihood analysis. DNA extraction, purification, and sequencing

techniques are discussed in Chapter 2.

Phylogenetic analyses were performed using the maximum likelihood search

algorithm fastDNAml (version 1.0, Olsen, 1994). This program is a

computationally faster version of Felsenstein's DNAML, and is coded in the

language "C" (Felsenstein, 1993). Trees were sought using the following options:

1) F: empirical base frequencies of input sequences were used, 2) T: various

transversion-weighing parameters were tested to determine their effect on tree topology and log likelihood values, 3) 0: outgroup was specified (Oryza sativa), and

4) J: input order of sequence was randomized (jumble seeds) in some cases and not randomized in other cases to determine the effect this option has on tree topology and log likelihood values.

RESULTS

Sequences for the 36 taxa included in this study (Table 2.2) were tested using MacClade software (Maddison and Maddison, 1992) to determine the estimated ratios of transitions to transversions. Using one of the 531 trees discovered using parsimony methods with unweighted ndhF sequences (Chapter 2), the minimum and maximum frequency of base changes between states was calculated, and are represented numerically and pictorially (Figs. 3.1 and 3.2).

These figures suggest a general level of symmetry between complementary base 116

A C G T

A 75 69 32

C 57 15 93

From:

G 123 24 73

T 39 92 46

Figure 3.1 Frequency of minimum and maximum base changes between states calculated from a single tree discovered by parsimony analysis using ndh? seq u en ces for all taxa included in this study. 117

To:

c

From:

G

(||) Transversions

^ Transitions

Figure 3.2 Frequency of base changes between states presented by circular areas calculated from a single tree discovered by parsimony analysis using n d h f seq u en ces for all taxa included in this study. 118 changes. For example, the frequency of changes from base G to base C (a transversion) is nearly equal to the frequency of change from base C to base G (Figs 3.1; 3.2). From these analyses, the total number of unambiguous transversions were calculated at 361, and the total number of unambiguous transitions were 377. Using these values, the ratio of transitions to transversions is 1.04 : 1.

To find trees with the highest log likelihoods, an array of transversion weighing parameters were designated in separate maximum likelihood analyses. For this series of analyses, the J option (randomization of input data) was not in effect.

The parameters used in these analyses were: 1.0 (no weight given to transversions); 1.04 (the empirically derived weight from parsimony analysis); 1.1

(minimal weight given to transversions); 2.0 (default weighing in fastDNAml); 3.0,

4.0, and 5.0 (progressively heavier weighing of transversions). The log likelihood values for trees produced using these parameters are listed in Table 3.1. All weighing parameters (with the exception of 5.0) resulted in the same topology (here designated as topology "A"); (Fig 3.3). The parameter of 5.0 resulted in topology

"D" (Fig 3.4). A graph of the effect of transversion weighing parameters on the log likelihood values for trees produced is presented in Figure 3.5.

The results of these analyses suggests that topology A (Fig. 3.3) is stable through a broad range of transversion weighing parameters. Only at the extreme weighting of 5.0 does the topology experience change. Although the results suggest that the transversion weighing of 1.1 produces the tree with the greatest log likelihood (-7,473.95), the empirically derived parameter (1.04) produces only a 119

Table 3.1 Log likelihood values and tree topology designations resulting from maximum likelihood searches using the given transversion weighting parameters, with jumble seed (137) parameter in effect versus not in effect.

Transversion Jumble Seed (137) Log Likelihood Topology weighting Used? (Yes / No) Designation param eter

1.0 No -7 ,4 7 6 .1 5 A 1.0 Yes -7,482.57 B 1 .0 4 No -7,474.90 A 1.04 Yes -7 ,4 8 1 .0 9 B 1.1 No -7 ,4 7 3 .9 5 A 1.1 Yes -7 ,4 7 9 .6 6 B 2.0 No -7 ,5 1 3 .0 0 A 2.0 Yes -7,517.80 C 3.0 No -7,588.40 A 4 .0 No -7,662.86 A 5.0 No -7 ,7 3 0 .0 5 D 120

i2 i£ Spathoglottls

: ] A reth u seae | , C alanthe

-000 Vanda - V andeael .032 . U1J. rnmnmn Cymbidium C ym bidoae

Stan ho pea

Maxlllarieae Matdllaha .002 _ . . u Polyslachya OW Epktendreae

V an d eae Angraecum

afiiSi Ootomeria

Zootrophton

| Restrepla

Pleurothallis

sS oti Arpophyllum

* ™ * C h y sis A reth u seae

B m ssavola

Epidendrum Encydia NW EpkJendreae iC attleya

------1 Meiracyllium

• f l & O M l i .

° l aSA&i Tipularia C alyp aoaae

•SSw Dendrobium

Cer&tostylis Podochlleae Ena

' Bulbophyllum f c T e ^ d o * o t p | .015 - , C oelogyne Coelogyneae >013 - . ■ ■ " T hunia

®£*i SobraJia NW Epidendreae 1 .039 , t Listera N eottteae .091 .. . '1 1 Neuwiedia Apostasioideae

H abenaria Orchidoideae

Spiranthes Spiranthoideae

Cypripedium Cyprfpedloideae

Cirvia Amaryllidaceae

Oryza P o a c e a e

Figure 3.3 Maximum likelihood tree (A); log likelihood value = -7,473.95. Branch lengths in terms of expected nucleotide substitutions per site are designated above each branch. Branches also present in the strict consensus parsimony tree (Fig. 2.9) are indicated by an asterisk; those branches that received > 50% bootstrap support are indicated by a double asterisk. Polyphyletic orchid tribes are indicated by shaded regions in the side legend. 121 Clivio

------Nauwiadia +—33 -31 +------Habenaria I +------Spiranthes

+— Zootrophio ♦— 2 ! ! + Restrepia -1 +—3 I + Pleurattal + - I 1 ♦— Octane ria -21 I t ♦— Arpophyllu I Chysis +— Tipularia I I ♦ ------Stanhopea 1 I ♦-19 +—15 Meirccylli —34 I ! I I I I ♦-—6 + Epidendrua I ! I ! ♦-25 I I +—5 + Cattleya ♦—I I +-10 ! +—9 + Encyclia ! 1 1 + Brassavola + -3 0 I +— Coalia + - U + Aerangis ♦ —27 +-26 + Angraacua

+-13 Polystochy

♦— Dendrobiua

+ Eria +—20 +-22 ♦ Ceratostyl

+ Vanda ♦-24 -32 ! ! +— M axillaria ! +—16 +-12 +-14 ♦------Cynbidiun ! I + Calanthe ♦—28 ♦— Spathoglot

+ -1 S +— Bulbophyll I +-17 ! I +— Coelogyne I +-23 +-29 +— Thunia I I +-— Sobralia ! +• L istera Cypripediu

Oryza

Figure 3.4 Maximum likelihood tree (D); log likelihood = -7,730.05. xmu ieiod erhs Tejml oto i nt n fet o hee searches. ese th for effect in not is option jumble The searches. likelihood um axim m Figure 3 .5 Log likelihood values versus transversion weighting param eters graph for for graph eters param weighting transversion versus values likelihood Log .5 3 Figure

Log Likelihood -7700 -7600 -7500 -7400 : 21 : 41 5:1 4:1 3:1 2:1 1:1 Transition:Transversion Ratio Transition:Transversion 122 123 slightly less likely result (-7,474.90). The 1.04 : 1 ratio of transitions to transversions is consistent with the contention that transitions outnumber transversions in chioroplast genes by a factor of less than two to one

(Palmer et al., 1988).

To examine the effect of the jumble option on topology and log likelihood values, maximum likelihood searches were made using transversion weighing parameters: 1.0, 1.04, 1.1 and 2.0. The jumble seed value of 137 was used in all analyses. It is noted that there is nothing special about the jumble seed value of

137 as any value would have served the same purpose in this test. The log likelihood values for trees discovered using these parameters are listed in Table 3.1.

All transversion weighing parameters (with the exception of parameter 2.0) resulted in the same topology (here designated as topology "B") (Fig 3.6). The transversion weighing parameter of 2.0 resulted in the topology designated here as "C" (Fig.

3.7). As in the previous test with the jumble option not in effect, the tree resulting from this test (jumble option set at 137) indicates that the transversion parameter of 1.1 produced the tree with the highest log likelihood (-7,479.67) (Table 3.1).

The log likelihood values for trees produced from weighing parameters 1.04 and 1.0 are only slightly less likely than that for the 1.1 parameter at -7,481.09 and -

7,482.57 respectively. Only at the weighing of 2.0 does the topology change with a log likelihood value of -7,517.80 (Fig. 3.6).

As illustrated in Figure 3.8, the log likelihood values for trees produced with or without the jumble option in effect are similar with respect to the transversion weighing parameter specified. However, in all cases, the log likelihood values are 1 24 higher with respect to transversion weighing parameters when the jumble option is not in effect (Fig. 3.8).

According to Olsen et al. (1994), when sequence addition is replicated with different jumble random seeds, they have about the same probability of finding the best tree, but any given seed might give a different tree. To test the effect of the jumble option seed number on phylogeny and log likelihood values, an additional test was undertaken. Specifically, 15 additional maximum likelihood searches were employed with the transversion weighing parameter set at 1.1. Five searches were made with seed number static at 137, and single searches made with the seeds:

17, 37, 47, 57, 67, 87, 97, 101, 117,127. The results of these searches suggest that the jumble seed number has an effect on topologies discovered (Table 3.2).

For example, these tests suggest that using the same jumble seed (137) in five separate searches yielded the same tree (topology "B") with the same log likelihood value (-7,479.66) in each case. Therefore, when the jumble seed option and transversion weighting parameter are used in repeated searches, the same topology with the same log likelihood value are produced from each search.

Jumble seed numbers: 57, 101, 117, and 137 produced trees of topology

B; all with a log likelihood value of -7,479.66 (Table 3.2). Jumble seed numbers:

17, 37, 67, 97, and 127 produced trees of topology "A," all with a log likelihood value of -7,473.95. Jumble seed number 47 produced a different tree (topology

"E") with a log likelihood value of -7,487.03 (Fig. 3.9).

Of the 26 maximum likelihood searches made in this study, topology "A"

(Fig. 3.3) has the highest log likelihood value and is, therefore, selected as the tree most likely to represent the true phylogeny for the taxa included in this study. 125

Table 3.2 Log likelihood values and designated tree topologies resulting from maximum likelihood searches using the jumble seeds indicated. All tests were performed using the transversion weighting parameter 1.1.

Jumble Seed Log Likelihood Topology Designation

17 -7 ,4 7 3 .9 5 A 37 -7 ,4 7 3 .9 5 A 47 -7,487.03 E 57 -7 ,4 7 9 .6 6 B 67 -7,473.95 A 87 -7,473.95 A 101 -7,479.66 B 117 -7 ,4 7 9 .6 6 B 127 -7,479.66 A 137 -7 ,4 7 9 .6 6 B 137 -7 ,4 7 9 .6 6 B 137 -7 ,4 7 9 .6 6 B 137 -7,479.66 B 137 -7,479.66 B 137 -7,479.66 B ♦------Cypriptdiua 126

Listera

+— Coalogyna +—9 ♦ ~ 2 ♦— Thunia I ♦— Bulbophyllua Stonhopea +—31 + Cottleya I I +—3 + Brassavola I +-17 I I I ♦— Meiracylliua +-21 +— 4 +-13 I + Epidendrua +-12 ! ♦-33 + Encyclia + Vanda ♦—8 I | +— Maxilloria I +—32 +-18 +------Cyabidiua I I +— Spathoglottis +— 1 + Calanthe

+ Eria +—23 + Ceratostylis

+— Dendrobiua ♦—6 +-16 +— Coelia +-19 +— Tipulario

+-28 + Plaurothallis +--11 —30 +-25 +-24 ♦ Restrepia ! I ♦-14 ♦------Zootrophion I I +■ 34 Octoaeria ♦-22 ! I +-10 +— Arpophyllua I ♦-— Chysis + Polystachy +-26 I + Aerangis ♦—20 + Angroecxa

♦— Sobralia Neisriedia ♦—27 +-29 +- Habenaria I + - — - - Spirant has G iv ia

Oryza

Figure 3.6 Maximum likelihood tree (B); log likelihood = -7,479.66. ♦--— — Listera 127 ♦ -— Coalogyn* ♦ —9 +—2 ♦— Thunie I +— Bulbophyll Stanhopes 1 ♦—31 ♦— Cattleya I I +—3 + Brassavola I +-17 +-21 1 ! I +— Meiracylliua +-—4 +-13 +—6 ! + Epidendrua I ♦-33 + Encyclia + Colonthe +—1* 1 ♦— Spathoglottis + — 1 I + Vanda + -8 ! +—- Maxillaria ♦ -3 2 +------CymbidiiB + Eria +—23 + Ceratostylis +— Dendrobiia +-12 +-16 +— Coelia +-19 +— Tipulario

+-28 + Pleurothallis +-11 ♦-25 24 + Restrepia I ! ♦-14 +— Zootrophion ! I <—34 +— Octoaeria +-22 I ! I ♦-10 ♦ Arpophyllui I +— Chysis ♦-as + Polystachya +-26 ! + Aerangis + -2 0 + Angraecun

-27 +—- Sobralia

— Cypripediua

+------Habenaria —30 ♦—29 +------Spiranthes Neimedia

+------Clivia ! ♦ ------Oryza

Figure 3.7 Maximum likelihood tree (C); log likelihood = -7,517.80. maximum likelihood searches w ith the jum ble option in effect (seed = 137) and and 137) = (seed effect in option effect. in ble jum not the option ith w jumble the searches ith w likelihood maximum Figure 3 .8 Log likelihood values versus transversion weighting param eter graph for for graph eter param weighting transversion versus values likelihood Log .8 3 Figure

Log Likelihood -7520 -7510 -7480 -7500 -7490 -7470 treesdiscovered withthe Jumbleoption In effect D ( treesdiscovered with the optionjumble not ineffect 9 A,B,C = designated = A,B,C topologiestree 1:1 Transltion:TransversionRatio 1.04:1 1 . 1:1 2:1 128 Clivia 129 +— Bulbophyll +-18 ! ! + Polystochy + - -B ! + Angraecun +—24 +--2 + Aerangis +— Coelio +-29 ! +— Tipulario +-16 ! +— Chysis +-33 ! +----- Arpophyllu + -2 8 ! +— Octaneria +-10 +-19 ! +------Zootrophio +-22 ! + Restrepia +—30 + Pleurothal +— M axillaria +— 8 +-31 +------Cymbidium ; ; +—5 +--1 + Vanda —12 ! +— Spathoglot +—34 + Calanthe

+— Thunia +-26 +-13 +— Coelogyne ; | ! +— Dendrobiun +-11 +------Stanhopea ! ! + Encyclia +— 6 ! +—25 +— Meiracylli ! ! +-27 ! +-20 + Epidendrum +--4 ! +—9 ! + Brassavola i + Cattleya

+ Eria +—14 +--15 + Ceratostyl -- Sobralia +-21 Listera

------Spiranthes -17 +—32 ! ! + — ■ Neuwiedia ! +—23 ! +------Habenaria

Cypripediu

Oryzo

Figure 3.9 Maximum likelihood tree (E); log likelihood = -7,487.03. 1 3 0

Evaluation of branch lengths in maximum likelihood trees is not straight

forward. Output produced by fastDNAml (Olsen et al., 1994) software includes a table showing the length of each tree segment in units of expected nucleotide substitutions per site. For each branch segment, confidence limits on their length are supplied. According to Felsenstein (1993), these confidence limits are very rough. Because there is a simplification in how the confidence limits are calculated,

overconfidence in the existence of the branch is the result (Olsen; pers. comm.,

1994). Therefore, supplied confidence limits were not used to evaluate branch lengths in this study.

In parsimony and maximum analyses, branch length is considered an estimated measure of clade stability. Branch length in parsimony analyses represents the number of synapomorphies, whereas branch length in maximum likelihood analyses represents expected nucleotide substitutions per site; thus, the two measurements are not synonymous. Therefore, using branch lengths to evaluate clade stability in maximum likelihood analyses is problematic. Specifically, in any given maximum likelihood topology, what criterion should be used to evaluate resulting branch lengths with respect to clade stability?

Branch lengths in the maximum likelihood tree ranged from .050 to .000

(Fig. 3.3). To provide an estimate of the branch length value that may be considered to represent a "significant" level of clade support, all branches found in the strict consensus parsimony tree (Fig. 2.6) that also occurred in the maximum likelihood tree were identified. Those branches with bootstrap values < 50% are designated by an asterisk, and those branches with bootstrap values >50% are designated by a double asterisk on the maximum likelihood tree (Fig. 3.3). Because 131

these branches are present in both trees, their respective branch lengths (in terms

of expected nucleotide substitutions per site) may be considered to represent an

estimate of clade support that is more "significant" than those branches not

present in both trees. It is acknowledged that the term "significant" in this sense

has no statistical basis; however, it does provide a crude estimate of clade support

in the maximum likelihood analysis. Using this criterion, maximum likelihood branch

length values that range as low as .001 are considered to provide this defined level

of "significant" clade support.

The maximum likelihood tree suggests that Sobralia (Sobraliinae, New World

Epidendreae) is the most basally diverged element in subfamily Epidendroideae and not closely allied with the New World tribe Epidendreae core taxa (Meiracyliinae,

Laeliinae, Arpophyliinae and Pleurothallidinae) (Fig. 3.3). Therefore, the New World tribe Epidendreae appears polyphyletic. This tree also suggests that subtribe

Laeliinae, represented by Brassavola, Encyclia, Cattleya, Epidendrum (New World tribe Epidendreae), is monophyletic and forms a clade with Meiracyliinae (New

World tribe Epidendreae), and Chysis (Chysiinae, Arethuseae); therefore, the New

World tribe Epidendreae is suggested to be paraphyletic. This tree also suggests that subtribe Pleurothallidinae, represented by Pleurothallis, Octomeria, Restrepia,

Zootrophion (New World tribe Epidendreae), is monophyletic with Arpophylliinae

(New World tribe Epidendreae) as sister (Fig. 3.3).

Based on the maximum likelihood tree, the tribe Arethuseae is polyphyletic a s Spathoglottis and Cafanthe (representing subtribe Bletiinae) form a clade distinct from Chysis (Chysiinae). This tree suggests the Spathoglottis-Calanthe clade 132 occupies a relatively basal position in Epidendroideae, whereas Chysis appears as sister to the derived Arpophyliinae-PIeurothallidinae clade.

Aerangis and Angraecum (Vandeae), and Po/ystachya (Old World tribe

Epidendreae) form a clade distinct from the clade composed of Stanhopea and

Maxillaria (Maxillarieae), Cymbidium (Cymbideae), and Vanda (Vandeae). Therefore, the Vandeae tribe appears polyphyletic and, the vandoid characteristics (Chapter 1) exhibited by the taxa in these two separate clades may represent parallel evolution.

As Bulbophyllum and Dendrobium (both members of Dendrobieae) appear in the maximum likelihood tree as not closely allied, tribe Dendrobieae is also suggested to be polyphyletic.

DISCUSSION

Maximum Likelihood Parameters. For the data set used in this study, a range of maximum likelihood transversion weighting parameters tested had minimal effect on topologies produced (Fig. 3.8). Determining the rate of transitions to transversions from parsimony trees is shown by this study to provide a close estimate of the maximum likelihood transversion weighing parameter that yields the tree with the highest log likelihood. The maximum likelihood jumble seed value is shown by this study to have a significant impact on resulting tree topologies and their respective log likelihoods (Table 3.2). This study suggests that several tests using different jumble seeds are necessary when performing maximum likelihood analyses before the topology with the highest log likelihood value is discovered.

Systematic Relationships. Phylogeny inferred from parsimony methods and maximum likelihood methods is compared by examining the congruence among trees produced from each. In this discussion, the maximum likelihood tree with the 133

highest log likelihood value (Fig. 3.3) is compared to the strict consensus parsimony

tree where unweighted ndhF sequences were used in the analysis (Fig. 2.6) and to

the single parsimony tree discovered when DNA sequences and morphological

characters were combined (Fig. 2.9). Both the strict consensus parsimony tree (Fig.

2.6) and the maximum likelihood tree (Fig. 3.3) suggest that the subfamily

Epidendroideae is monophyletic with Listera (Neottieae) as sister.

Within the subfamily Epidendroideae, neither the cymbidioid nor epidendroid

phylads (sensu Dressier, 1993) is suggested to be monophyletic by either the maximum likelihood tree (Fig. 3.3) nor the parsimony tree in which DNA sequences and morphological characters were combined (Fig. 2.9). Although the cymbidioid tax a Stanhopea (Maxillarieae), Maxillaria (Maxillarieae), and Cymbidium (Cymbideae) form a clade in both parsimony trees (Figs. 2.6, 2.9) and the maximum likelihood tree (Fig. 3.3), in none of these trees does the other cymbidioid taxon, Tiputaria

(Calypsoeae) appear closely allied with this clade. Therefore, the cymbidioid phylad appears to be polyphyletic.

Vanda (Vandeae) appears as sister to the core taxa of the cymbidioid phylad in both the maximum likelihood tree (Fig. 3.3) and the parsimony tree (Fig. 2.9).

Also in both trees, Aerangis (Vandeae), Angraecum (Vandeae) and Polystachya (Old

World tribe Epidendreae) form a clade distinct from the cymbidioid core taxa -Vanda clade which suggests the tribe Vandeae is polyphyletic.

Both the maximum likelihood tree (Fig. 3.3) and the parsimony trees (Figs.

2.6, 2.9) suggest that the New World tribe Epidendreae is polyphyletic. In all trees

Sobratia (Sobraliinae) represents the most basal element in Epidendroideae and, therefore, not closely allied with the core taxa of the New World tribe Epidendreae 134

(Arpophyliinae, Laeliinae, Pleurothallidinae, and Meiracyliinae). The phylogenetic

relationship of Coelia (Coeliinae, New World tribe Epidendreae) is ambiguous. In the

parsimony tree (Fig. 2.9), Coelia does not appear closely related to the core taxa of

the New World tribe Epidendreae; however, the maximum likelihood tree (Fig. 3.3)

su g g e sts Coelia forms a clade with Tipularia (Calypsoeae) that is sister to the New

World tribe Epidendreae core taxa. In the parsimony tree (Fig. 2.9), Chysis

(Chysiinae, Arethuseae) appears as sister to the core taxa of the New World tribe

Epidendreae. In the maximum likelihood tree (Fig. 3.3) Chysis appears as sister to the Pleurothallidinae-Arpophyliinae clade. Whether Chysis represents a basal taxon

in the New World tribe Epidendreae or sister to the tribe, both trees suggest this taxon is closely allied with the New World tribe Epidendreae and not with the

Spathoglottis-Ca/anthe (Arethuseae) clade.

In the parsimony tree (Fig. 2.9), Coelogyne (Coelogyneae) forms a clade with

Thunia (Coelogyneae). Therefore, the parsimony tree suggests that although Thunia and Coelogyne are closely allied, the tribe Coelogynae is polyphyletic. However, the maximum likelihood tree indicates that Coelogyne and Thunia form a clade suggesting the monophyly of the tribe Coelogyneae (Fig. 3.3). In neither the parsimony tree (Fig. 2.9) nor the maximum likelihood tree (Fig. 3.3) are Dendrobium

(Dendrobieae) and Bulbophyllum (Dendrobieae) suggested to be closely allied which suggests that the tribe Dendrobieae is polyphyletic.

All Epidendroideae clades found in the strict consensus parsimony tree

(Fig. 2.6) are also found in the parsimony tree where DNA sequences and morphological characters were combined (Fig. 2.9) and the maximum likelihood tree

(Fig. 3.3). Branches leading to these clades are generally well supported in terms of 135

unambiguous synapomorphies and bootstrap values (parsimony trees), and expected

nucleotide substitutions per site (maximum likelihood tree). Therefore,

Epidendroideae clades with strong branch support recur in all analyses. The

variability between the parsimony tree (Fig. 2.9) and the maximum likelihood tree

(Fig. 3.3) is found where branch support is weak.

Morphological Character Evolution In Epidendroideae. In this section,

anatomical/morphological character states defined in Chapter 2 are mapped onto

the orchid taxa from the maximum likelihood tree with the greatest log likelihood

(Fig. 3.3). Evolution of these characters in terms of this maximum likelihood tree

are discussed below.

1. Growth Habit. As suggested by the character tree (Fig. 3.10), the

sympodial habit represents the basal character state in Orchidaceae. Although the

monopodial Sobralia (Sobraliinae) appears as the most basal taxon among the

Epidendroideae, this character state apparently represents an autapomorphy and not

a general evolutionary shift toward monopodialism. Only among the Vandeae

subtribes (Aeridinae, Angraecinae, Aerangidinae) does there appear to be a major radiation of taxa exhibiting the monopodial habit. With respect to growth habit, character evolution as suggested by the maximum likelihood tree (Fig. 3.10) is not substantially different than that depicted in the parsimony tree (Fig. 2.11).

2. Leaf Folding. Character evolution with respect of leaf folding in subfamily

Epidendroideae, as depicted in the maximum likelihood tree (Fig. 3.11), is similar to that suggested by the parsimony tree (Fig. 2.12). That is, leaf type generally appears correlated to microhabitat with epiphytic taxa typically exhibiting conduplicate leaves, and terrestrial taxa exhibiting plicate leaves. However, the 136 epiphytic taxa Coelogyne (Coelogyneae), Stanhopea (Maxillarieae), and Chysis

(Arethuseae) with plicate leaves are exceptions (Fig. 2.12). With respect to leaf folding, the conduplicate leaf represents the predominant evolutionary theme in the

Epidendroideae, but the present data (Fig 2.12) does not indicate which condition is plesiomorphic in the family.

3. Inflorescences. As in the parsimony tree (Fig. 2.13), the terminal inflorescence is suggested by the maximum likelihood tree (Fig. 3.12) as the basal character state in Epidendroideae. The maximum likelihood tree (Fig. 3.12) suggests that a major evolutionary shift toward a lateral infloresence ensued early in the history of Epidendroideae which is also suggested in the parsimony tree (Fig.

2.13). Both trees suggest that a major reversal toward the terminal infloresence among subtribes of the New World tribe Epidendreae (Meiracyliinae,

Pleurothallidinae, Laeliinae, Arpophylliinae) subsequently occurred in the subfamily.

4. Resupination. As depicted in the character state tree (Fig. 3.13), resupination represents a pervasive character in subfamily Epidendroideae.

Nonresupination exhibited by the disparate Arpophy/lum (Arpophylliinae),

Pleurothallis (Pleurothallidinae), Eria (Eriinae), and Polystachya (Polystachyinae) appears as independent parallel specializations. The evolution of resupination as depicted in the maximum likelihood tree (Fig. 3.13) is essentially the same as in the parsimony tree (Fig. 2.14).

5. Pollen Aggregation. As indicated by the parsimony tree (Fig. 2.15) and the maximum likelihood tree (Fig. 3.14), granulate pollen, represents the plesiomorphic condition in subfamily Epidendroideae as exhibited by Sobralia

(Sobraliinae). These trees suggests that the condition of firm pollinia represents the 137 derived condition and is exhibited by the majority of taxa representing the subfamily. Reversals occur in the Arpophylliinae and Coeliinae and Thunia

(Thuniinae). The results of this study support Dressler's contention that the

Epidendroideae should not be delimited by the hardness of pollinia as those taxa with soft pollinia (Sobralia, Thunia, Arpophyiium, and Coelia) are clearly indicated to be epidendroids (Fig. 2.9).

6. Number of Pollinia. The maximum likelihood tree (Fig. 3.15), suggests that the eight pollinia condition represents the basal state in subfamily

Epidendroideae. This agrees with the parsimony tree that also suggests the eight pollinia condition is basal (Fig. 2.16). The eight pollinia character state undergoes no reversals in the subfamily; four and two pollinia states appear as either shared derived characters or autapomorphies in the subfamily (Fig. 3.15).

7. Caudicles. As illustrated by the maximum likelihood character-state tree

(Fig. 3.16), the fertile caudicle appears as the basal and most pervasive type of pollinium accessory structure in Epidendroideae. The fertile caudicle character state remains unchanged along the axis of evolutionary descent from the most basal taxon, Sobralia (Sobraliinae) to the derived Arpophyliinae-Pleurothallidinae clade.

The absence of caudicles in Dendrobium (Dendrobiinae) probably represents an autapomorphy. That is, lack of caudicles in the Dendrobiinae with hard pollinia is probably not homologous with those taxa with granulate pollinia that also lack fertile caudicles (e.g. Listera (Listeriinae) and Spiranthes (Spiranthoideae).

Unlike the parsimony tree (Fig. 2.17) that suggests the sterile caudicle associated with a stipe appears in a single clade of vandoid taxa, the maximum likelihood tree (Fig. 3.16) suggests the sterile caudicle has evolved independently at 138 least twice in Epidendroideae. Specifically, one clade composed of Vanda

(Vandeae), and Stanhopea and Maxillaria (Maxillarieae) all exhibit sterile caudicles.

Cymbidium (Cyrtopodiinae) without a sterile caudicle also appears in this clade.

Lack of a sterile caudicle in Cymbidium may represent an autapomorphic reversal as most members of the Cyrtopodiinae produce a sterile caudicle. All members of a second clade composed of Polystachya (Old World tribe Epidendreae), and Aerangis and Angraecum (Vandeae) exhibit tegular stipes.

8. Stipes. Unlike the parsimony tree (Fig. 2.18) that suggests that the tegular stipe has arisen only once among the vandoid taxa, the maximum likelihood tree suggests that the tegular stipe has arisen independently at least twice in subfamily Epidendroideae (Fig. 3.17). Specifically, one clade composed of Vanda

(Vandeae), and Stanhopea and Maxillaria (Maxillarieae) all exhibit the tegular stipe.

Cymbidium (Cyrtopodiinae) without a tegular stipe also appears in this clade. Lack of a stipe in Cymbidium may represent an autapomorphic reversal as most members of the Cyrtopodiinae produce a stipe. All members of a second clade composed of

Polystachya (Old World tribe Epidendreae), and Aerangis and Angraecum (Vandeae) exhibit tegular stipes. Therefore, this tree suggests that the presence of a tegular stipe does not consistently delimit vandoid or cymbidioid taxa (Fig. 3.17). The hamular stipe in Tipuiaria (Calypsoeae) appears as an autapomorphy.

9. Velamen. As indicated by the character-state tree (Fig.3.18), velamen type has undergone several major evolutionary lines of descent in the

Epidendroideae. The velamen type in the most basal epidendroid taxon Sobralia

(Sobraliinae) is undefined, therefore, the plesiomorphic velamen type is unresolved. 139

Of the recognized types, the Coelogyne velamen type as exhibited in Thunia and

Coelogyne appears as the plesiomorphic state.

The Calanthe velamen type appears as a consistent character among Eria

and Ceratostylis (Eriinae), and Spathoglottis and Calanthe (Bletiinae). The Calanthe

velamen type exhibited by Tipuiaria (Calyposeae) is suggested by this tree to be a

parallelism. As indicated by the character-state tree (Fig. 3.18), the Epidendrum

velamen type appears as a consistent character in the New World Epidendreae

subtribes Meiracyliinae, Laeliinae, and Arpophyliinae. Within the New World tribe

Epidendreae, the Pleurothallis velamen type (exhibited by all members of subtribe

Pleurothallidinae) represents the derived state within the tribe.

The Vanda velamen type is consistently exhibited by the three Vandeae

subtribes Aeridinae, Aerangidinae, and Angraecinae. Because the maximum

likelihood tree suggests the polyphyly of the Vandeae, this velamen type appears to

have arisen independently at least twice in the Epidendroideae. With the exception

of Tipuiaria (Calypsoeae) having a Calanthe velamen type, the remaining

representatives of the cymbidioid phylad: Cymbidium (Cymbidieae), Stanhopea and

Maxillaria (Maxillarieae) consistently exhibit the Cymbidium velamen type.

Therefore, velamen type provides additional support that the Cymbidioid Phylad

represents a polyphyletic taxon based on the maximum likelihood method.

10. Seeds. Due to the numerous equivocal branches displayed in the maximum likelihood tree (Fig. 3.19), an overall evolutionary trend in the

Epidendroideae cannot be determined for seed type. However, several major trends in seed evolution are evident. For example, the Dendrobium seed type is consistently exhibited by the clade composed of Buibophyiium (Dendrobieae), and 140

Coelogyne and Thunia (Coelogyneae). Based on the character tree (Fig. 3.19), the

Dendrobium seed type exhibited in Buibophyiium (Dendrobieae) appears to represent

parallel evolution of this seed type. Members from the Polystachya (Old World tribe

Epidendroideae), and Angraecum and Aerangis (Vandeae) clade exhibit the Vanda seed type. Because Vanda (Vandeae) does not appear in this clade, the character tree (Fig. 3.19) suggests that the Vanda seed type has arisen at least twice in the

Epidendroideae.

Although the Pleurothallis seed type exhibited by the Pleurothallidinae and th e Epidendrum seed type exhibited by the Laeliinae are the predominant seed types found in the New World tribe Epidendreae, the character state tree (Fig. 3.19) suggests that the E/leanthus seed type represents the plesiomorphic seed type in this largest of all orchid tribes. Therefore, the E/leanthus seed type is suggested by the parsimony tree (Fig 2.20) and the maximum likelihood tree (Fig. 3.19) to have been the precursor that led to both the Pleurothallis and Epidendrum seed types.

11. Fusion of Anthers to Style. Although the maximum likelihood tree with the highest log likelihood suggests that Cypripedium (Cypripedioideae), characterized as having styles fused with filaments and staminode bases, represents the most basal taxon in this study, (Fig. 3.20), an alternative maximum likelihood tree (Fig. 3.7) suggests Neuwiedia (Apostasioideae), characterized as having style with filaments at bases only, is basal. The latter tree is congruent with the strict consensus parsimony tree discovered using unweighted ndh? sequences

(Fig. 2.6) in that Neuwiedia is basal and Cypripedium is derived with respect to the nonepidendroid taxa included in this study. Although this conflict exists between these two maximum likelihood trees, neither topology suggests that fusion of style Spathoglottis

Calanthe

Vanda Cymbidium

S tan h op ea

Maxillaria

Polystachya A erangis

Zootrophion R estrepia

Pleurothallis Arpophyllum

C hysis

B rassavola

Epidendrum Encyclia

Cattleya Meiracyllium

C oelia

Tipuiaria

Dendrobium

Ceratostylis

Eria

Buibophyiium

C oelogyne

Thunia

Sobralia

Listera

H abenaha

Spiranthes

Cypripedium

______t M i l______Sympodial ■111 ...... Monopodial r r * *

Figure 3.10 Character-state tree for growth habit among orchid taxa included in this study. 142

Spathoglottis

CyroUdum + * * + Stanhopea

Polystachya Aerangis

Angraecum

aOctomarta Zootrophion Ha strop la

PidurothalUs Arpophyiium

Chyslo Braaaavola r f l - * — Encyctla

Bimuhw Cattloya Melracy ilium Cool la

Tipuiaria Dendrobium

CeratostyU

Buibophyiium

Cosiogyna m Thunia Sobralia

Habanana

Splmnthes

'. Cypdpedum

■laayaa CondupUcata PUcate

Figure 3.11 Character-state tree for leaf type for orchid taxa included in this study 143

* Spathoglottia

r 4 r .Calonlhe '4 ^ ^ Vanda '4r ^CymUdUim ^ ^ r 4> stanhopea * *4 *** <* + htexUl&rtA

■ Polyatachya A e ra n g it i -»J 4 * * Angraecum Octomerla

Zootrophion Rastrepia

PteurothalUs Arpophylluni

******* Chysla f asamt Braesavola

li Epidendrum Sa«™» Encyclla

L m C ath ey a inramtmiwiniiim MelracyUlum f * -CoaUa

* u Tipuiaria — Dendrobium

' Ceratostylla ,T **************** 1 & * * Erie r * * * •• Buibophyiium ■ 4 4 K *• ‘ Coelogyne ~ L _ Thunia Sobralia Ustera Neuwleda Habenarta

HZ ■ Splranthea

i Cyprlpedum

Inflorescences Terminal Lateral

Figure 3.12 Character-state tree for inflorescence type among orchid taxa included in this study 144

m SpathogloW*

Cwantns Vanda Cymbidum

Maxillaria

r. * w * » Polystachya

k rAngraecum * Octomeria r—i t—■— "Zootrophkm Rastrepla

i h : PlaurothalUi Arpophyllum

Chysle

yu m Bressavola

*“ | winiim Epidandrum ^ JP™— Encvdla SnsraCattloya MelracyUlum • Ooelia

•Tipuiaria • Dendrobium

>|“ Ceralostylls

0 «Eiia Buibophyiium

Coelogyne

Thunia Sobralia lists ra

Neuwleda Habenarla k f Splran tries

i Cyprlpedum

Resuulnatlon Ftesuplnate ■» Not Ftesuplnate

Figure 3.13 Character-state tree for resupination for orchid taxa included in this study. 145

SpathoglottU

Calanthe Vanda Cymbkflum

Polystachya Asrangls

Angraecum Octomeria

Zootrophion Rastrspla

PtaurothalUs Arpophyllum

Bmssavola

Epidendrum Encydla

Catseya MelracyUlum

Tipuiaria Dendrobium

Ceratostytis

Buibophyiium

Coo logy no

Thunia SobraJla Lists ra

Neuwiedia Habonaria Splmnthes

Cypripodum

Pollen Aggregation Firm Soctlie Granulate Not Aggregated Equivocal

Figure 3.14 Character-state tree for pollen aggregation type among orchid taxa included in this study. 146

SpadlOQlOtUt & 4 Calanthe Vanda CymUdum Stanhopes

Polystachya Aerangis

H i Angraecum y * * * * * * Octomeria

f * r^ f L j" " " " " * ™ Zoo,roPhlon t I I*™* Re strep la

t Pleurothallis Arpophyllum

( '' BI Epidendrum ' „ J §*»«-.Encydla * ; IsnCattleya ^ & *r ^ * jr & MolracylUuin J

r

Coelogyne

Thunia • Sobralia Ustera Neuwiedia Habenaria

Spiranthes

i Cypripedium

Number ol Pollinia, Eight r * +

Figure 3.15 Character-state tree for number of pollinia among orchid taxa included in this study. 147

Spathoglottla

Calanthe Vanda Cymbidium Stanhopea

Polystachya Aatangl*

Angraecum Oclomeria

Zootrophion Rastrepla

PtourothalUa Arpophyllum

Chysla Braasavoia

Epidendrum Encyclia

Cattlsya MelracyUlum

Tipuiaria o' < Dendrobium

Ceralostyus

■ / / / « Buibophyiium

1 1— Coelogyne

Kmum Thunia ■ Sobralia

' ■*> 4 Listers r » » -Neuwleda S'*...... Habenaria * * J Sp Iran the*

Cypripedium

C au dld aa With caudlclea i 11 ■" Sterile Absent * * * *

Figure 3.16 Character-state tree for caudicle type among orchid taxa included in this study. 148

i Spalhoglottl*

n Calanthe P- L f * s. • Vanda K# 4T^ ■ Cymbldum r Stanhopea

1# & A b

r a + a Polystachya ’t f a AerangJs t* a * • a * Angraecum Odomeria r L j — Zootrophion Hestrapia

Pleurothallla Arpophyllum

Chyala Braasavola

Epldandrum Encyclia n£Biiinnw Catllaya MelracyUlum Coaila ■c Tipuiaria Oondroblum

MHB Ceratostylia iZ Eria Buibophyiium

Coelogyne

Thunia Sobralia Uatara ■Nauwladia , Habenaria

• Splranthes

Cypripedium

Sllpa Absent Hamular Tegular * ■ » * . Equivocal

Figure 3.17 Character-state tree for stipe type among orchid taxa included in this study. 149 SpathogtotUa

f w v r v r v u v r u Vanda Cymbidium « Stanhopea

Podyatachya Aaranpto

**»•-»• Angraecum yrnw m im m ii QotoflWla

j f m tm m a a Zootrophion I £■«■* Raatrepta

l»»Pburo(hftUa Arpophyllum

Epidendrum S bbw Encycfla

L m C ettleya UcJracylfium

Tlpuia/ia

Caraloatytla Die ** Buibophyiium

»«*»» Coelogyne

Item* Thunia

Sobrafla Uatara Neuwiedia « Habenaria

Spnnthea

+ Cypripedium

Velamen ______

Pieurothaifia **■***■*«*” Equivocal m J L m m UndeflnedUnitnown m m m

Bufeophytlum Splrarthaa ■■■■■■ Cymbidium * • •

Epidarxbum CoUnlhc Vanda > v.

Aim** r ~ * Dendrobium ■— Coriogyna

Figure 3.18 Character-state tree for velamen type among orchid taxa included in this study. 150

•»»!>•»•*•%»%•»• Vanda mmmaammamm, Cymbidium

a te ii Maxllaria

PoJyaiachya *•»« Aarangia

Epidandrum EneycSa

CotUeya V w w w w w MalracyUium Kp» i Coa8a

Thunia

i

..J

• Cypripadium

Soada

PtourathaUa Equivocal Undafhad/Unknown MaxJUarie

Elaanthua Goodyara Eulophla Vanda

Epidandrum Orchla Nauwiadla Cy rubidium Umodorum Blotla ftanhopaa Dandroblum

Figure 3.19 Character-state tree for seed type among orchid taxa included in this study. 151

'Calanthe Vanda CymbkJum Stanhopea

Angraecum

Tipuiaria Dendrobium

Erta

Thunia Sobralia Ustera

Splranthea

Cypripedium

Style (used with filaments... at bases ■ * - and atamtnode bases stamlnodia, and stigma to column iw m Equivocal

Figure 3.20 Character-state tree for degree of fusion between style and filaments among orchid taxa included in this study. 152 and filaments has followed a unilinear path towards increasing adnation in

Orchidaceae.

Within the subfamily Epidendroideae, evolution of these selected anatomical/morphological characters as mapped on the maximum likelihood tree

(Fig. 3.3) appears generally consistent with that depicted on the parsimony trees

(Chapter 2). That is, evolutionary trends with respect to habit, leaf folding, inflorescences, pollen aggregation, number of pollinia, velamen, and seeds suggested by either method of phylogenetic inference are not significantly different.

However, the maximum likelihood tree suggests that the hamular stipe and sterile caudicle arose independently at least twice in the subfamily, whereas the parsimony trees suggest a single origin for these two morphological structures.

LITERATURE CITED

Dressier, R. L. 1993. Phylogeny and Classification of the Orchid Family. Dioscorides Press. Portland, Oregon.

Felsenstein, J. 1978. Inferring phylogenetic trees from chromosome inversion data. Systematic Zoology 27: 401-410.

. 1981. Evolutionary trees from DNA sequences: a maximum likelihood approach. Journal of Molecular Evolution 17: 369-376.

. 1983. Parsimony in systematics: biological and statistical issues. Annual Review of Ecology and Systematics 14: 313-333.

. 1993. Some available phylogeny programs (Long). [email protected]

Goldman, N. 1990. Maximum likelihood inference of phylogenetic trees, with special reference to a poisson process model of DNA substitution and to parsimony analysis. Systematic Zoology 39: 345-361.

Hasegawa M. and Fujiwara, M. (1993). Relative efficiencies of the maximum likelihood, maximum parsimony, and neighbor-joining methods for estimating protein phylogeny. Molecular Phylogenetics and Evolution 2: 1-5. 153

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

Olsen, G. J., H. Matsuda, R. Hagstrom, and R. Overbeek. 1994. fastDNAml Version 1.0 University of Illinois, Urbana, Illinois.

Palmer, J. D., R. K. Jansen, H. J. Michaels, M. W. Chase, and J. R. Manhart. 1988. Chloroplast DNA variation and plant phylogeny. Annals of the Missouri Botanical Garden 75: 1180-1206.

Swofford D. L. and G. J. Olsen. 1990. Phylogeny Reconstruction, p. 411-501 In Molecular Systematics. Hillis, D. M. and C. Moritz (eds.) Sinauer Associates, Sunderland, MA.

Swofford D. L. and Begle, D. P. 1993. Phylogenetic Analysis Using Parsimony. Illinois Natural History Survey, Champaign, Illinois. CHAPTER 4

EVOLUTION IN THE NUMBER AND POSITION OF FERTILE ANTHERS IN ORCHIDACEAE INFERRED FROM ndhF CHLOROPLAST GENE SEQUENCES

1 5 4 155

Traditionally, the orchids have been separated into two major groups based

on the number and position of fertile anthers (Burns-Balogh and Funk, 1986). In the

first group are the monandrous orchids which produce a single fertile anther per

flower and encompass over 19,000 known species. The second group is composed

of those taxa with two or three fertile anthers per flower, and includes about 200

species in the Cypripedioideae and Apostasioideae. Historically, the diandrous and triandrous orchids have been considered basal to the monandrous orchids largely

based on this character (Raven and Axelrod, 1974; Burns-Balogh and Funk, 1986).

However, preliminary evidence from rbcL sequences using parsimony techniques indicates the cypripedioids are embedded between two single-anther clades suggesting parallel evolution of that character or a reversal to two anthers (Cameron et al. 1994) .

Each flower in the monocotyledons has six stamen positions, only three of which bear structures in orchids (two known exceptions exist and are discussed subsequently). As described by Garay (1960), orchid flowers are organized on a trimerous pattern and are mere modifications of the liliaceous type. The essential deviation from the liliaceous pattern is illustrated in the staminal whorls (Fig. 4.1).

In the liliaceous pattern, two staminal whorls are fertile and fully developed. In the orchids, only the three abaxial anthers (parts of both whorls) generally can be fertile. The anther position designated A, occupies the outer staminal whorl adjacent to the medial ; the anther positions a, and a2 occupy the epipetalous or inner staminal whorl (Vermeuien, 1966). The adaxial anthers are generally lacking but their vestiges have been reported from floral developmental studies

(Kurzweil 1987a, 1987b, 1988, 1992). 156

Abaxial Adaxial

Figure 4.1. Designated stamen positions of the liliaceous-type flower. Diagram is oriented in the resupinate position for comparison to the orchids. 157

In Neuwiedia (Apostasioideae), all three abaxial anthers are fertile. In the

Cypripediaceae anther positions a, and a2 are fertile, whereas the A, position is reduced to a staminode (Garay, 1960; Vermeulen, 1966; de Vogel, 1969; Atwood,

1984). In the apostasioids, the anther filaments are present, whereas in the cypripedioids they are fused with the style. These differing floral morphological features have resulted in distinct pollination system s (Burns-Balogh and Bernhardt,

1985). In the apostasioids, lily-like anthers and powdery pollen result in a typical

"buzz-pollination" flower which has evolved independently in many monocot and dicot families. The fusion of filaments in the androecium of the cypripedioids, with two fertile anthers on either side of a wide staminode, in conjunction with an inflated pouch-like labellum, forms a sophisticated trap blossom (Burns-Balogh and

Bernhardt, 1985).

Burns-Balogh and Bernhardt (1985) stated that the extreme reduction in the number of fertile anthers in monandrous orchids has permitted the evolution of the modern pollinarium. The wing-shaped staminodes have evolved to control the movement of the pollinator's body, proboscis, or bill so that the pollinarium is properly positioned as it is removed from the flower and that it is subsequently deposited onto the stigma of a conspecific.

In this chapter, an hypothesis is proposed concerning the evolution of stamen number in the Orchidaceae. This hypothesis is inferred from a gene tree derived from molecular sequence characters of the chloroplast gene ndhF.

MATERIALS AND METHODS

In this chapter, the strict consensus parsimony tree of most-parsimonious trees discovered in the analysis using unweighted ndhF sequences (Fig. 2.6) and 158 the maximum likelihood tree with the highest log likelihood value (Fig. 3.3) are used to discuss character evolution of fertile anthers in Orchidaceae. Materials and methods used to derive the parsimony and maximum likelihood trees are discussed in Chapters 2 and 3 respectively.

RESULTS

The number of fertile anthers (one, two, three, six) is mapped onto the strict consensus cladogram of trees discovered from a parsimony search using unweighted ndh? sequences (Fig 4.2). These character states are also mapped onto the maximum likelihood tree with the highest log likelihood tree (Fig 4.3).

Both trees are congruent with respect to the basal position of Clivia

(Amaryllidaceae) as the most basal taxon among the ingroup members. However, there is conflict between the two trees with respect to the most basal orchid taxon.

That is, the parsimony tree (Fig. 4.2) suggests that Neuwiedia (with three fertile anthers) is basal and the maximum likelihood tree (Fig 4.3) suggests that

Cypripedium (with two fertile anthers) is basal. Neither topology suggests that

Neuwiedia and Cypripedium are sisters.

DISCUSSION

Through floral developmental studies of the diandrous Cypripediaceae,

Kurzweil (1992) found the staminal organs in the adaxial half of the flower are presumed lost or strongly reduced. Based on its position and early initiation, a small bulge on the ventral base of the gynostemium is interpreted as the a3 stamen.

Stamens A2 and A3 are presumed lost.

In floral studies of the monandrous Epidendroideae and the Vandoideae

(sensu Rasm ussen,1985), Kurzweil (1987a) paid special attention to the early 159

Arpophyllum Pleurothallis Restrepia Zootrophion Octomeria Brassavola Epidendrum Encyclia CatUeya Meiracy Ilium Dendrobium c tKEtfa Coelogyne Ceratostyiis Eria Cymbidium Stanhopea Maxillaria Thunia Tipuiaria Chysis Spathoglottis Calanthe Buibophyiium Vanda Polystachya Angraecum Aerangis Sobralia Ustera Cypripedium Spiranthes Habenaria Neuwiedia Clivia Oryza

-Fertile Antfiers- Monandrous Hexandrous Diandrous Equivocal Triandrous 111 • • • • a i

Figure 4.2 Character-state tree for number of fertile anthers adapted from the strict consensus tree of 531 trees discovered from 100 random addition parsimony searches using unweighted ndh? sequences. Spalhogtolto 160

Calanthe

Cymbidium

Rolyataehya Aarangis

Angraecum Oclomarta

Zootrophion Ha Drop la

Pteurothalll* Arpophyllum

Chysia Brassavola

Epidendrum Encyclia

CatUeya MelracyUlum Cod la

Tipuiaria Dendrobium

Ceratostytl*

Buibophyiium

Coelogyne

Ustera * •Nouwleda Habenaria

Splranthes

a Cypripedum

anCUvla

««Oryia

Monandrous mmmmmm Haxandroua Dlandroua ■•••••■ Equivocal Trtandrous v m m *

Figure 4.3 Character-state tree for fertile anthers adapted from the maximum likelihood tree with the log likelihood value = -7,473.95. 161

development of the gynostemiun and its appendages. He concluded that the lateral

appendages of the mature gynostemium are homologous with the two lateral

stamens (a, and a2). In fact, he observed staminal primordia in species that lack

prominent appendages in the mature gynostemium. Kurzweil reported that a small

number of taxa exhibited ventral (adaxial) processes of different sizes on the

gynostemium. These processes initiated during early development, became

indistinct and were incorporated into the gynostemium during later development.

Kurzweil suggested that these appendages are likely to represent vestiges of the

adaxial stamens. He considered the Epidendroideae and Vandoideae (sensu

Rasmussen, 1985) advanced orchid groups that exhibit plesiomorphic floral

development features, in particular the occurrence of prominent staminodes in early

development.

In the monandrous subfamily Orchidoideae (sensu Rasmussen,1985),

Kurzweil (1987b), found that development in the Orchis and Habenaria groups were

remarkably uniform with respect to the form and position of the early anther and to

the presence of auricles and basal bulges. These studies suggested that auricles

(unique to the monandrous orchids) originate from the dorsal side of the young

fertile anther and are merely appendages of it. Based on the early initiation of these

auricles and their position superposed to the , Kurzweil suggested that they

may represent vestiges of the stamens a, and a2. In late development, the basal

bulges fuse with the auricles and with the lateral margins of the median stigma

lobe. Therefore, they can hardly be distinguished in the mature flower.

In subfamily Neottioideae (sensu Rasmussen, 1985), Kurzweil (1988), examined early morphogenesis of the gynostemium in nine species of the tribe 162

Epipactieae and ten species from tribe Neottieae. Kurzweil found the gynostemium

development in tribe Epipactieae similar to that of Epidendroideae and Vandoideae.

In the tribe Neottieae, he observed a progressive reduction and delayed initiation of the staminodes forming part of the gynostemium. In most Neottieae examined, staminode primordia are obtuse vestiges in early development.

Kurzweil's floral developmental studies suggest that the adaxial stamens in the Orchidaceae have either been lost or greatly reduced. However, the potential to

produce staminodia and fertile anthers at these sites has apparently not been lost completely. As reported by Chen (1982), the genus Tangtsinia produces a single fertile median abaxial anther (A,) and staminodia at all of the five remaining positions (A2, A3, a,, a2, a3). Chen also reported that in the genus Diplandorchis, both the median abaxial anther (A,) and the median adaxial anther (a3) are fertile.

Chen considered these taxa not derivative but surviving ancient types. He stated that the flowers in these genera have several quite primitive states: the mostly regular, stigma terminal, pollinia naked, and rostellum usually absent.

However, Robert Dressier (pers. comm. 1994) believes that these two taxa may be derived because they are saprophytes. lindenii has an extra stamen in the inner whorl (a3) with an elongated filament that causes contact of the anther with the stigma with consequent autogamy (Atwood 1984).

In the case of Diplandorchis, the fertility of the a3 anther may represent a primitive state as argued by Chen (1982) or may represent a desuppression of the a3 anther if this taxon (and this character) are derived as suggested by Dressier. In the case of Phragmapedium lindenii the fertility of the a3 anther (which promotes autogamy) most likely represents a derived state. According to Stebbins (1974), 1 63

species that move along an evolutionary path from obligate outcrossing to

predominantly self fertilization usually represent the ends of evolutionary lines.

Whether these examples of a3 fertility represent plesiomorphic or derived

characters, it is clear that the ability of the orchids to produce fertile adaxial

anthers has not been lost totally.

With the basal position of Clivia (Amaryllidaceae) (Fig. 4.2), the topology of

the parsimony tree is consistent with the hypothesis that the orchids evolved from

lily-like ancestors that probably exhibited six anthers in two whorls.

Among the orchid taxa selected for inclusion in this study, it is not

surprising that Neuwiedia (Apostasioideae) occupies the basal position on the

parsimony tree (Fig. 4.2). It has often been hypothesized to represent the most

"primitive" of the approximately 800 orchid genera (Garay, 1960; de Vogel, 1969;

Rasmussen, 1985). Dressier (1993) stated that the apostasioids are of special

interest as a link between the orchids and more ordinary lily-like plants, and that

they appear to be a sister to the remaining orchids either as a subfamily or a distinct

family. Other characters besides the presence of three abaxial anthers have been

used to justify Neuwiedia as the most basal taxon in the orchids: 1) anthers joined

to the gynostemium by visible filaments, 2) pollen in monads, 3) distinct style with

an apical stigma, 4) three-chambered with axile placentation, and 5)

unspecialized "buzz-pollination" flower type. In addition, unpublished results using rbcL chloroplast gene sequences and parsimony methods suggest that the subfamily

Apostasioideae is basal in Orchidaceae (Mark Chase, pers. comm., 1994).

The character state for the branch leading from Clivia to Neuwiedia is equivocal, as one, three, and six anthers all provide an equally most parsimonious 164

presentation for the character state along this branch (Fig 4.2). Although this

branch may be considered to represent the evolutionary line of descent to the

ancestral orchids or " protoorchids," the character state or states along this branch

will probably never be known. Distinguished from their lilioid precursors, the

ancestral orchids probably exhibited some suppression of adaxial anthers. It is

possible that these ancestral taxa assumed the condition as in Neuwiedia with the

three adaxial anthers absent or possibly with four fertile anthers (the A2 and the A3

anthers absent) as suggested by Chen (1982). Because the ancestral states for

anther number and position in orchids are unknown, it remains unresolved whether

the presence of three abaxial anthers in Neuwiedia represents the plesiomorphic

condition in extant orchids or an autapomorphy.

An interesting result of this study is the position on the cladogram (Fig. 4.2)

of Cypripedium (Cypripediaceae) between the two monandrous taxa: the Habenaria

(Orchidoideae) - Spiranthes (Spiranthoideae) clade and Listera (Neottieae). As has

been previously discussed, the cypripedioids were often thought to represent one of the tw o most basal orchid groups largely due to the number of fertile anthers.

Aside from their nonmonandrous condition, they share a purportedly plesiomorphic

character with the Apostasioideae: pollen in monads. Although exhibiting these two characters, the sophisticated zygomorphic trap flower can hardly be considered

plesiomorphic. In his discussion of whether the cypripedioids should be considered

primitive and relictual, Atwood (1984) suggested that the presence of tw o fertile stamens merely indicates that the cypripedioids diverged very early during the evolution of orchids. Although the cypripedioids are often considered primitive.

Dressier noted that they are specialized in a way different from the other orchids. 165

Furthermore, he suggested that the cypripedioids are clearly more closely allied to

the monandrous orchids than are the apostasioids. This parsimony tree (Fig. 4.2) is

consistent with the preliminary rbcL sequencing results of Cameron, et al. (1993) that suggest the cypripedioids are embedded between two monandrous clades in their cladograms.

The most parsimonious presentation for the evolution of diandry in the cypripedioids is depicted in the character tree (Fig. 4.2). This tree suggests that diandry arose from the monandrous state and implies a monandrous ancestor for the cypripedioids. As suggested by Kurzweil’s floral development studies, the two median abaxial anther positions in a broad spectrum of monandrous orchids are represented by staminodia. Because these positions have not been lost through the course of evolution, it is plausible that a desuppression from staminode to fertile anther could have occurred in the cypripedioids. The purported reversal of the a3 anther in Phragmipedium lindenii suggests that fertile anthers may appear from previously sterile positions. Judging by the great success of the monandrous orchids in terms of species number and cosmopolitan range, it follows that selection pressure is strongly skewed toward monandry at the A, position. With this apparent advantage, it appears unlikely that one monandrous phyletic line would suppress this character state and revert to diandry at the a, and a2 positions.

Therefore, diandry in the cypripedioids is considered an autapomorphy (possibly occurring under relaxed selection pressure) and not reflective of a wholesale reversal along the monandrous line of descent. This study suggests that the monandrous orchids comprise a polyphyletic group, an interpretation consistent 166

with the findings of Cameron, et al. (1993) that the Cypripedioideae should not be

distinguished from the monandrous orchids as a separate family.

The maximum likelihood tree (Fig. 4.3) is consistent with the parsimony tree

(Fig. 4.2) with respect to the basal position of Clivia (Amaryllidaceae) among the

ingroup members. Unlike the parsimony tree, Cypripedium (Cypripedioideae)

represents the most basal orchid taxon and Neuwiedia (Apostasioideae) is derived

(Fig. 4.3). The position of Neuwiedia as sister to Habenaria (Orchidoideae), in the maximum likelihood tree is unexpected. Such a relationship has not been suggested by any previous molecular- or anatomical/morphological-based studies. As suggested by the maximum likelihood tree topology, the triandrous character state in Neuwiedia represents a reversal from the monandrous state. Based on the previous discussion, it is understandable how such a reversal could have occurred.

That is, anther positions a, and a2 would have undergone desupression. However, when the host of other reversals that would have had to occur are considered, the derived relationship of Neuwiedia in this maximum likelihood tree is suspect. That is, specialized pollination associated with monandry became unspecialized in the form of a "buzz-pollination" flower, specialized anthers became unspecialized, aggregated pollen became free, pollen in tetrads became monads, parietal placentation became axile, and almost complete adnation of style with a single filament in the form of a column reverted to an almost completely free style with three filaments.

The basal position of Cypripedium and the derived position of Neuwiedia as suggested by the maximum likelihood tree (Fig. 4.3) is further weakened by an alternative maximum likelihood tree that suggests Neuwiedia is basal and 167

Cypripedium is derived. That is, the maximum likelihood tree discovered when the

jumble seed of 137 is used with the default transversion weighting parameter of

2.0 (Fig. 3.7) suggests a topology very similar to that of the parsimony tree (Fig.

4.2). Therefore, the relationships of Cypripedium and Neuwiedia are shown to

conflict by the maximum likelihood method of phylogenetic inference when different

transversion weighting parameters are used.

Notwithstanding the basal position of Cypripedium and the derived position

of Neuwiedia in the maximum likelihood tree with the highest log likelihood value

(Fig. 4.3), the preponderance of evidence appears to suggest that this relationship

is unlikely. That is, when rbcL and ndhF chloroplast gene sequences are separately

subjected to parsimony methods of inference, Neuwiedia (Apostasioideae) is

suggested to represent the most basal taxon in Orchidaceae and that Cypripedium

(Cypripedoideae) is derived. Additionally, the number of anatomical/morphological

character reversals required to position Neuwiedia as derived with respect to a

monandrous taxon far exceeds that of Cypripedium. Therefore, the topologies that

suggest Neuwiedia (Apostasioideae) represents the most basal taxon in Orchidaceae

appears to be more credible than that which suggests Neuwiedia is derived.

LITERATURE CITED

Atwood, J. T. Jr. 1984. The relationships of the slipper orchids (Subfamily Cypripedioideae). Selbyana 7: 129-247.

Burns-Balogh, P. and P. Bernhardt. 1985. Evolutionary trends in the androecium of the Orchidaceae. Plant Systematics and Evolution 149: 119-134.

Burns-Balogh, P. and V. A. Funk. 1886. A phylogenetic analysis of the Orchidaceae. Smithsonian Contributions to Botany 61: 1-79.

Cameron, K. M., D. Jarrell, and M. W. Chase. 1994. Evidence from rbcL sequences and phylogenetic relationships of major lineages within Orchidaceae (Abstract). American Journal of Botany 81: 145. 168

Chen, S. C. 1982. The origin and early differentiation of the Orchidaceae. Acta Phytotaxonomica Sinica 20: 1-22.

Dressier, R. A. 1981. The Orchids. Harvard University Press, Cambridge, MA and London, England

. 1990. The Neottieae in orchid classification. Lindleyana 5: 102-109.

. 1993. Phylogeny and Classification of the Orchid Family. Dioscorides Press, Portland, Oregon.

Garay, L. A. (1960). On the Origin of the Orchidaceae. Botanical Museum Leaflets 19: 57-95.

Kurzweil, H. 1987a. Developmental studies in orchid flowers. I: Epidendroid and vandoid species. Nordic Journal of Botany 7: 427-442.

. 1987b. Developmental studies in orchid flowers II: Orchidioid species. Nordic Journal of Botany 7: 443-451.

. 1988. Developmental studies in orchid flowers III: Neottioid species. Nordic Journal of Botany 8: 271-282.

. 1992. Developmental studies in orchid flowers IV: Cypripedioid species. Nordic Journal of Botany 13: 423-430.

Rasmussen, F. N. (1985). Orchids. In: Dahlgren, R. M. T., H. T. Clifford, and P. F. Yeo. (eds.) The Families of the Monocotyledons; Structure, Evolution, and Taxonomy. Springer-Verlag, Berlin, Heidelberg, New York, Tokyo, pp 249-274

Raven, P. H., and D. I. Axelrod. 1974. Angiosperm biogeography and past continental movements. Annals of the Missouri Botanical Garden 61: 539-673.

Stebbins, G. L. 1974. Flowering plants, evolution above the species level. The Belknap Press of Harvard University Press. Cambridge, MA.

Vermeulen, P. 1966. The system of the Orchidales. Acta Botanica Neerlandica 15: 224-241. de Vogel, E. F. 1969. Monograph of the tribe Apostasieae (Orchidaceae). Blumea 17: 313-350. CHAPTER 5

A TERRESTRIAL ORIGIN FOR THE ORCHIDACEAE INFERRED FROM ndhF CHLOROPLAST GENE SEQUENCES

169 Most authorities have held that the terrestrial habit represents the basal

condition in Orchidaceae and that the epiphytic condition is derived (Schimper,

1888; Goebel, 1922; Mulay and Deshpande, 1960; Garay, 1960; Garay, 1972;

Ackerman, 1983; Benzing and Atwood, 1984; Benzing, 1987). Upon finding

velamen in the roots of the terrestrial Liliaceae and Amaryllidaceae, Goebel (1922)

concluded that velamen was not acquired with the epiphytic habit but was already

present. According to Mulay and Deshpande (1961), the occurrence of a simple

velamen type suggests that the Liliaceae forms a basic stock for the evolution of

monocots. Structure of the velamen becomes more modified and complex along

one line towards Amaryllidaceae (a wholly terrestrial family) and along another line towards the Araceae and Orchidaceae where velamen in epiphytic taxa may be highly specialized. They believed that terrestrial forms of Orchidaceae and Araceae with velamen were preadapted to an epiphytic habit.

Garay (1960) proposed that the orchid prototype may be visualized as a terrestrial with an inferior ovary, numerous minute seeds with a rudimentary embryo; stamens and pistils not yet aggregated into a column, and flowers fertilized by . Garay (1972) stated that the family Orchidaceae originated in the area during the Cretaceous period when most angiospermous families became differentiated. He hypothesized that all species were geophytes and that the epiphytic mode of life was a more recent development, dating from the Pleistocene. The epiphytic mode of life in orchids evolved, in his opinion, through a secondary differentiation which began in the subtribe Neottioideae and subsequently gave rise to the Epidendroideae. Benzing 171

(1987) suggested that the earliest orchids were probably autotrophic perennials

native to mesic acidic substrata.

The contention that the Orchidaceae had a terrestrial ancestry was

challenged by Robinson and Burns-Balogh (1982) on the basis of small seed size

(microspermy), floral modifications leading to specialized pollination mechanisms,

and root velamen. Their hypothesis was based on work they performed in the

primarily terrestrial subtribes Spiranthoideae, Neottioideae, and Orchidioideae. They

noted that certain specializations found in the Spiranthoideae and shared with the

predominantly epiphytic Epidendroideae are likely in plants having an epiphytic

ancestry. To help resolve the controversy as to whether the Orchidaceae had an

epiphytic or terrestrial origin, a phylogeny of the Orchidaceae is here inferred from

gene trees discovered using molecular sequence characters of the chloroplast gene

ndhF.

MATERIALS AND METHODS

In this chapter, the strict consensus parsimony tree of most-parsimonious

trees discovered in the analysis using unweighted ndhF sequences (Fig. 2.6) and the maximum likelihood tree with the highest log likelihood value (Fig. 3.3) are used

to discuss character evolution of epiphytism and terrestrialism in Orchidaceae.

Materials and methods used to derive the parsimony and maximum likelihood trees are discussed in Chapters 2 and 3 respectively.

RESULTS

Predominant growth substrate characters (epiphytic and terrestrial) for each genus are mapped onto the strict consensus cladogram of trees discovered from a parsimony search using unweighted ndhF sequences (Fig 5.1). These character 172 states are also mapped onto the maximum likelihood tree with the highest log likelihood tree (Fig 5.2). Both trees suggest that epiphytism is derived with respect to terrestrialism among the taxa selected for inclusion in this study.

DISCUSSION

The contention that the Orchidaceae had a terrestrial ancestry was challenged by Robinson and Burns-Balogh (1982) on the basis of microspermy, floral modifications leading to specialized pollination mechanisms, and root velamen.

They noted that orchids produce small nonendospermous seeds in quantities from

1,300 to 4,000,000 per capsule, and concluded that the orchids sacrifice the advantages of supplying maternal nutrients in their propagules for the benefit of producing a large quantity of seeds. Robinson and Burns-Balogh (1982) believed that because the terrestrial habitat is widely available, there is little pressure to produce numerous small seeds. By contrast, this specialization is advantageous in an epiphytic habitat where seeds are wind dispersed and unlikely to land in a place suitable for germination.

Countering the hypothesis of Robinson and Burns-Balogh (1982), Benzing and Atwood (1984) contended that small seed size in orchids was attributable to specialized nutritional strategies involving a mycorrhizal relationship with fungi. As carbon mycotrophs, orchids exhibit "mycotrophic microspermy" that is advantageous in dark environments (e.g. forest floors) where it probably evolved.

Additionally, fungal symbiosis appears diminished among epiphytes and probably has never been as pronounced there as in the terrestrials. Benzing and Atwood also challenged the contention that a terrestrial habitat is widely available by stating that such habitats are often characterized by patchiness and disturbance. In these ArpophyUum Plaurothaliis Restrepia Zootrophion Octomeria Brassavola Epidendrum Encyclia Cattieya Meiracyllium Dendrobiiun Coalia Coalogyne Caratostyiis Eria Cymbidium Stanhopea Maxillaria Thunia Tipularia Chysis Spathoglottis Calantha Bulbophyllum Vanda p Potystachya Angraecum Aerangis Sobralia Listara Cypripedium Spiranthes t Habenaria - . J • i Neuwiedia J t. Clivia Oryza

Growth Substrate Epiphytic Equivocal Terrestrial

Figure 5.1 Character-state tree for growth substrate adapted from the strict consensus tree of 531 trees discovered from 100 random addition parsimony searches using unweighted ndhF sequences. Spathoglottis

• Calanthe Vanda Cymbldlunt Stanhopea

Maxlllarta

Potystachya Asrangls

Angraecum Octomeria

Zootrophion Restrspla

Pleurothallii Arpophytlum

Chysis Brassavola

Epidendmm Encyclia Cattieya Meiracylllum CoeUa

Tipularia Dendroblum

Ceratostylls

Bulbophyllum

Coetogyno

r Thunia 4 4 . Sobralia r J4 Lisle ra 4 4 Neuwfedta 4 s Habenaria 4_ Spiranthes

a 4' * * * * ^ 4 4 > Cypripedium r 4 * Clivia * i m r . ' Oryza

Substrata Epiphytic "■■"i*”""’ Terrestrial r r * *

Figure 5.2 Character-state tree for growth substrate adapted from the maximum likelihood tree with the log likelihood value = -7,473.95. 175 habitats, selection will favor high fecundity and dispersability because populations must constantly recruit scattered safe sites.

The second point in the Robinson and Burns-Balogh (1982) hypothesis is that the high degree of floral specialization found in orchids is designed to achieve pollination between isolated plants or small populations in an epiphytic environment.

The development of pollinia seems a byproduct of the many-seeded fruit that compensates for rare acts of successful pollination (Robinson and Burns-Balogh,

1982). They contended that such floral specializations would be prohibitively unlikely in the terrestrial environments of most Spiranthoideae where they form comparatively dense populations.

In response, Benzing and Atwood (1984) contended that some terrestrial genera such as Ophrys, Calochilus, and Cryptostylis rival or exceed most epiphytic relatives in their floral specializations and reliance of wide-ranging pollination vectors. They noted that low fruit production is common in both terrestrial and epiphytic orchids and thus pollination mechanisms do not argue for either a terrestrial nor an epiphytic origin.

Robinson and Burns-Balogh (1982) suggested that early orchids developed resupinate flowers to facilitate pollination by having the labellum act as a landing platform. In the early epiphytic orchids, this was more readily achieved by producing pendent inflorescences rather than by twisting the pedicel.

Ackerman (1983) argued that a nodding inflorescence is compatible with either an epiphytic or terrestrial habit and noted that some primitive terrestrial orchids have nodding inflorescences. Ackerman contended that it was not 176

necessarily true that the pollination of ancestral orchids required resupinate flowers;

for example, flowers are nonresupinate in the apparently primitive genus .

Robinson and Burns-Balogh (1982) noted that some Spiranthoideae species

are facultatively epiphytic and many others in the subtribe are capable of being

grown on substrates that closely approach the conditions of epiphytism. They

stated that a major reason for this is the presence of roots approaching the type of

specialization exhibited in the epiphytic Epidendroideae. Robinson and Burns-Balogh

(1982) speculated that such terrestrial orchids may be the only nonepiphytic plants

among the monocotyledons having a two-layered velamen.

Benzing and Atwood (1984) agreed that several monocotyledonous groups

other than orchids produced velamentous terrestrial species; however, they

questioned Robinson and Burns-Balogh's speculation that some Spiranthoideae may

be the only nonepiphytic monocotyledons without a two-layered velamen by noting that Goebel (1922) had found multiserate velamen in some species of Liliaceae and

Amaryllidaceae. They stated that until there is evidence to the contrary, it must be

assumed that the velamen of orchids and nonorchids differ in no way that would

differentiate the two groups by function. Benzing and Atwood (1984) noted that

many terrestrial orchids have no velamen. They speculated that these orchids may

never have had such root specializations and their purported root degeneracy is the

result of mycorrhizal association and not a shift in substratum. According to Arditti

(1992), morphological modifications leading to the formation of velamen made

possible the evolution of epiphytes. Because a simple velamen has also been observed in the Liliaceae, he found that Robinson and Burns-Balogh's (1982) line of reasoning could lead to the suggestion that lilies also had an epiphytic origin. In order to establish a link between the terrestrial orchids and the nonorchid monocots, Benzing and Atwood (1984) generalized that the four predominantly terrestrial subfamilies Apostasioideae, Cypripedioideae, Spiranthoideae, and

Orchidoideae were similar to the nonorchid monocots by the following: 1) sympodial shoots equipped with mesomorphic leaves and C3-type , 2) anomocytic stomata, 3) flowers with a more generalized liliaceous morphology

(especially in the Apostasioideae), 4) less fusion of sexual appendages, and 5) larger and presumably less specialized seeds. The authors also noted that the few epiphytic members of these typically terrestrial subfamilies barely qualify as epiphytes as most are restricted to humus-filled crevices, bark fissures and lichen- covered limbs in wet forests. They contended that this lack of specialization for epiphytic life relative to the more advanced stress-tolerant epidendroids is consistent with Schimper’s (1888) hypothesis that most epiphytes arose from mesic terrestrial ancestors native to moist shady forest understories and eventually moved into the forest canopy. Among the orchid subfamilies, the Apostasioideae has no epiphyte species, the Cypripedioideae has about 29% epiphytic species, the

Orchidoideae and Spiranthoideae each have < 1 % epiphytic species, whereas approximately 90% of the Epidendroideae species are epiphytic (Atwood, 1986).

Based on the taxa selected in this study, the most parsimonious and maximum likelihood explanations for the evolution of epiphytism in the Orchidaceae are depicted in the character trees (Figs. 5.1, 5.2). Both trees suggest that the orchids may have arisen from terrestrial lily-like ancestors and maintained terrestrialism as a dominant evolutionary theme until sufficient specializations permitted large scale radiations into arborescent habitats. Although a few taxa in the Cypripedioideae, Spiranthoideae, and Orchidoideae are weakly epiphytic

(Benzing and Atwood, 1984) most are poorly adapted to the stressful xeric sites of tropical forest canopies. By contrast, many Epidendroideae are adapted to these xeric conditions through the following specializations: thickened conduplicate leaves that are heavily coated with wax and cutin (Dressier and Dodson, 1960);

Crassulacean acid metabolism (CAM)-type photosynthesis that allows for efficient recycling of C 02 in hot dry microclimates (Taiz, 1991); multilayered velamen designed to diminish transpiration rates and allow for longer access to moisture and dissolved minerals (Pridgeon, 1987). The specializations that have allowed for large-scale radiations of epiphytic orchid taxa have been shown to be especially successful in the primarily epiphytic Epidendroideae which contains more species than all other orchid subfamilies combined. The results of this study suggest that terrestrialism represents the plesiomorphic condition and epiphytism (especially the type found in the xeric-tolerant Epidendroideae) represents the derived condition in

Orchidaceae. Epiphytism in the predominantly terrestrial subfamilies may be viewed as having evolved through several independent parallel evolutionary events.

Therefore, it is unlikely that the anatomical/morphological characters necessary for epiphytism in the family followed an unbroken line of increasing specialization that fostered the large-scale radiation of epiphytic taxa into xeric arborescent microhabitats.

LITERATURE CITED

Ackerman, J. D. (1983). On the evidence for a primitively epiphytic habit in orchids. Systematic Botany 8: 474-476.

Arditti, J. 1992. Fundamentals of Orchid Biology. John Wiley & Sons, New York, Chichester, Brisbane, Toronto, Singapore. 179

Benzing, D. H. 1987. Major patterns and processes in orchid evolution: a critica synthesis. In: J. Arditti, (ed.). Orchid Biology: Reviews and Perspectives Vol. IV. Comstock Publishing Associates, Ithaca, pp 34-77.

Benzing, D. H. and J.T. Atwood Jr. (1984) Orchidaceae: Ancestral habitats and current status in forest canopies. Systematic Botany 9: 155-165.

Cameron, K. M., D. Jarrell, and M. W. Chase. 1994. Evidence from rbcL sequences and phylogenetic relationships of major lineages within Orchidaceae (Abstract). American Journal of Botany 81: 145.

Dressier, R. A. 1990. The Neottieae in orchid classification. Lindleyana 5(2): 102-109.

Dressier R. A. 1993. Phylogeny and Classification of the Orchid Family. Dioscorides Press Portland, Oregon.

Garay, L. A. 1960. On the Origin of the Orchidaceae. Botanical Museum Leaflets 19: 57-95.

. 1972. On the origin of the Orchidaceae, II. J. Arnold Arboretum 53: 202-215.

Goebel, K. 1922. Erdwurzeln mit Velamen. Flora 115: 1-26.

Mulay, B. N. and B. D. Deshpande. 1961. Velamen in terrestrial monocots - role of velamen tissue in taxonomy and phylogeny of monocotyledons. Proceedings of the Rajasthan Academy of Science 8: 115-120.

Pridgeon, A. M. 1987. The velamen and exodermis of orchid roots. In: J.Arditti, (ed.). Orchid Biology: Reviews and Perspectives Vol. IV. Comstock Publishing Associates, Ithaca, pp 140-192.

Robinson, H. and P. Burns-Balogh. 1982. Evidence for a primitively epiphytic habit in Orchidaceae. Systematic Botany 7: 353-358.

Schimper, A. F. W. 1888. Die epiphytische Vegetation Amerikas. Bot. Mitt. Tropen. II, 162 pp. Jena (Gustav Fisher).

Taiz, L. and E. Zeiger 1991. Plant Physiology. The Benjamin/Cummings Publishing Company, Red Wood City, CA. de Vogel, E. F. 1969. Monograph of the tribe Apostasieae (Orchidaceae). Blumea 17: 313-350. CHAPTER 6

A PHYLOGENETIC ANALYSIS OF SUBTRIBE PLEUROTHALLIDINAE (ORCHIDACEAE)*

•Reprinted by permission from the Botanical Journal of the Linnean Society. 181

Pleurothallidinae Lindl. (Orchidaceae) is a Neotropical subtribe included in

tribe Epidendreae (Dressier 1993). Its members range from Brazil and

northward through into and southern Florida; the greatest

diversity of species occur in cloud forest habitats (Pridgeon 1982a,b). An

articulation associated with an abscission zone between ovary and pedicel

distinguishes the subtribe (Garay,1956; Luer, 1986a; Dressier, 1993; Neyland &

Urbatsch, 1993). In all other orchids the articulation occurs at the base of the

pedicel (Dressier 1981).

Recent revisions by Luer (1986a, 1991) attributed 29 genera and as many as

4,000 species to the subtribe. Luer acknowledged that his treatment w as artificial and that he was unable to bring phylogenetic order to the group. An earlier study by Pridgeon (1982a) used 45 anatomical/morphological characters to undertake a numerical analysis of the subtribe and suggested that the phenetic relationships identified in his study represented seven major phylogenetic lines.

An important character in Pridgeon's analysis is number of pollinia. He supported the hypothesis by Dressier (1981, 1993) that the Pleurothallidinae shows a reduction series in the number of pollinia from eight to six to four to two with eight representing the plesiomorphic condition. However, Luer (1986a) apparently believed that the direction may have proceeded in the opposite direction from two to eight. Most genera and an overwhelming number of species bear two pollinia; therefore, the relatively few members with more than two pollinia are considered pivotal in this study. For example, Octomeria bears eight pollinia, whereas

Restrepia, , Restrepiopsis, , and Dresslere/la each bear four. 182

According to Garay (1956), has six; however, Luer (1986a) included

the eight-pollinium species of the invalid genus, Yolanda, in Brachionidium.

According to Luer (1986c), 29 subgenera comprise the very large genus

Pleurothallis to which over 2,000 epithets have been attributed. In Pridgeon's

(1982a) numerical analysis, two subgenera and 10 species complexes were

recognized. Dressier predicted (pers. comm. 1992), that Pleurothallis eventually

may be divided into a several distinct genera corresponding with their three basic

habits. Most species have the habit with short stems and oblanceolate

leaves that taper basally. Some have the habit with long stem s and

relatively wide leaf bases. The Pleurothallis cardiothallis group has long stem s,

cordate leaves, and fascicled flowers.

The primary objective of this study is to infer a hypothesis of phylogenetic

relationship within Pleurothallidinae by combining the character-state data matrix

constructed for taxa studied in Pridgeon's (1982a) phenetic analysis (here

designated as the ingroup) with a data matrix constructed for an appropriate

outgroup and subjecting the combined data sets to cladistic analysis. The resulting cladograms are used to: 1) discuss the cladistic relationships among the genera and

Pleurothallis species complexes of Pleurothallidinae; 2) evaluate whether the genus

Pleurothallis represents a monophyletic group or is polyphyletic and thus can be divided into several genera; 3) discuss the evolutionary trends in pollinium number.

MATERIALS AND METHODS

The ingroup members in this study are those used as terminal taxa in

Pridgeon's (1982a) phenetic analysis. These taxa consist of 24 genera;

Pleurothallis consists of two subgenera further divided into 10 species complexes. 183

For a detailed list of ingroup material examined see Pridgeon (1982a). Pridgeon's

classification scheme is presented in Table 6.1.

Three taxa were selected as outgroup members based on their status as

representing possible sisters to subtribe Pleurothallidinae. These include

Arpophyl/um giganteum (Arpophyllinae); Brassavola nodosa and Epidendrum ciliaris,

(Laeliinae). Voucher specimens are housed at Louisiana State University (LSU)

(Table 6.2). According to Dressier (1960), Arpophyllum appears to represent a

survivor of the ancestral stock from which the Pleurothallidinae were derived.

Pridgeon (1982a) stated that it is plausible that Octomeria (an ingroup member) and

Arpophyllum have a common ancestor and represent divergent evolutionary lines from that progenitor. According to Dressier (pers. comm. 1992) and Pridgeon

(pers. comm. 1992), subtribe Laeliinae may also be considered a likely outgroup to the Pleurothallidinae. Furthermore, a close relationship with Laeliinae is also suggested by the cladistic analysis of Burns-Balogh and Funk (1986), that suggested the tribe Epidendreae is sister to the "Pleurothallis Group." It is noted that the selection of the outgroup members was made prior to the DNA sequencing studies discussed in the previous chapters. However, the parsimony tree (Fig. 2.9) and the maximum likelihood tree (Fig. 3.3) both indicate that the selection of members from the Arpophyliinae and Laeliinae were appropriate outgroup members in this study.

Forty-five multi-state and binary characters were used in this analysis

(Appendix C). Characters and character states assigned to the ingroup members are those used in Pridgeon's (1982a) numerical analysis. These characters were expected to yield phylogenetically significant information for the reasons detailed in 184 Table 6.1. Classification scheme of Pleurothallidinae from Pridgeon (1982b)

Number of Pollinia Taxon

Octomeria

Pleurothallopsis

Brachionidium

Chamelophyton

Barbose/ia

Dresslerella

Restrep ia

Restrepiella

Restrepiopsis

2 Acostaea

Andreettaea

Condylago

Crocodeiianthe

Cryptophoranthus

Dracula

Dryadella

Lepanthes

Lepanthopsis

Masdevaiiia

Ph/oeophila

Physothallis

Platysteie

(table cont'd) Pleurothallis subg. Pleurothallis

P. alexandrae complex

P. blaisdellii complex

P. dura complex

P. loranthophylla complex

P. peduncularis complex

P. ruscifolia complex

P. scoparum complex

P. tuerckheimii complex

P. subg. Specklinia

P. grobyi complex

P. sclerophylla complex

Porrog/ossum

Sal pi stele

Scaphosepa/um

Ste/is

Trisetella 186

Table 6.2. Outgroup members examined for characters listed in Table 6.3.

Taxon Source Accession Number

Arpophyllum giganteum Hartw. University of Connecticut Neyland 90

Brassavola nodosa (L.) Lindl. Louisiana State University Neyland 72

Epidendrum ciliare L. Louisiana State University Neyland 81 187

Pridgeon's (1982b) diagnosis. Character states were determined for the outgroup

members using light microscopy, scanning electron microscopy (SEM), direct

observation, and literature sources. Character states for all are summarized in Table

6.3. Cladistic analyses were performed using PAUP (version 3.0s, Swofford 1991).

Multi-state characters were unordered using Fitch criterion (Fitch 1971). Most

parsimonious trees were sought from a heuristic stepwise addition search

employing 100 replications. MacCiade (version 3.0, Maddison and Maddison 1992)

was used to explore hypothesized phylogenetic trees further and to visualize

character evolution.

RESULTS

Using the three designated taxa as outgroup, eight most parsimonious

cladograms were discovered. The data matrix included 36 phylogenetically

informative characters. All cladograms have 230 steps and a consistency index (Cl)

of 0.27. A strict consensus tree is presented in Figure 6.1. In six of the parsimony trees, the Pleurothallis blaisdellii, P. sclerophylla, P. tuerckheimii, and P. loranthophylla complexes form a clade that is distinct from the clade composed of

Cryptophoranthus, the Pleurothallis dura and P. ruscifolia complexes, , and Restrepia. In the remaining two trees, the two clades are combined into a single clade. Figure 6.2 shows one most-parsimonious tree representing the topology where the two clades are distinct (reflecting six of the eight topologies discovered).

DISCUSSION

The consistency index of 0.27 represents a high level of homoplasy and may be a function of the rapid rates of evolution believed to have occurred in 188

Table 6.3. Character states for Pleurothallidinae and outgroup members. Characters and character states are identified in the Appendix C . Missing data are indicated by question marks.

Taxa 1 2 3 4 5 6 7 8 9

Ingroup

Acostaea 2 0 0 0 0 2 2 0 1

Barbose/la 2 0 0 0 0 0 1 0 0

Brachionidium 1 0 0 0 0 2 2 1 0

Condy/ago 2 0 0 0 0 2 0 1 1

Cryp toph or an th us 2 0 0 0 1 1 2 1 1

Dracu/a 2 0 0 0 0 2 2 1 1

Dresslerella 2 1 0 0 0 2 2 0 0

Dryadella 2 0 0 0 0 1 2 1 1

Lepanthes 2 0 0 0 0 2 2 0 1

Lepanthopsis 2 0 0 0 1 2 0 1 1

Masdeval/ia 2 0 0 0 1 2 2 0 1

Physosiphon 2 0 0 0 1 2 2 0 1

Physothallis 2 0 0 0 0 1 1 0 1

P/atyste/e 2 0 0 0 0 2 2 1 1

Pleurothallis ruscifolia complex 2 0 0 0 0 1 1 1 1

Pleurothallis alexandrae complex 0 0 1 0 0 0 1 0

Pleurothallis blaisdel/ii complex 2 0 0 0 1 1 0 0 1

Pleurothallis dura complex 2 0 0 0 1 1 2 1 1

Pleurothallis grobyi complex 2 0 0 0 0 1 0 1 1

Pleurothallis loranthophy/la complex 2 0 0 0 1 1 1 1 1

Pleurothallis peduncularis complex 0 0 0 1 1 1 1 1

Pleurothallis sclerophylla complex 2 0 0 0 0 2 0 1 1

Pleurothallis scoparium complex 2 0 0 0 0 1 2 1 1

(table cont'd) 189

Pleurothallis subgen. Speck/inia 2 0 0 0 0 2 1 1 1 Pleurothallis tuerckheimii complex 2 0 0 0 0 2 1 1 1

Porroglossum 2 0 0 0 0 1 2 0 1

Restrepia 2 0 0 0 0 1 1 0 1

Restrepiella 0 0 0 0 1 0 1 1 0

Restrepiopsis 1 0 0 0 0 1 2 1 0

Salpiste/e 2 0 0 0 1 2 2 1 1

Scaphosepalum 2 0 0 0 0 2 2 1 1 Ste/is 2 0 0 0 1 2 2 1 1

Trisetella 2 0 0 0 1 2 2 1 1 Octomeria 0 0 0 0 0 0 0 0 0 Outgroup

Arpophyllum giganteum 0 0 0 0 0 0 1 0 0

Brassavola nodosa 0 0 0 0 1 1 2 0 1

Epidendrum ciliaris 0 0 0 0 1 1 0 1 1

(table cont'd) 190

Taxa 1 1 1 1 1 1 1 1 1 0 1 2 3 4 5 6 7 8

Ingroup

Acostaea 1 0 1 0 0 1 1 1 3

Barbosella 1 0 1 0 0 2 1 0 1

Brachionidium 1 1 3 0 0 0 0 0 3

Condylago 1 0 1 0 0 1 1 1 3

Cryp toph or an thus 1 0 0 0 0 3 1 1 0

Dracula 1 0 1 0 0 0 0 3

Dresslere/Ja 0 2 3 0 0 3 1 0 0

Dryade/la 1 0 3 0 0 2 1 2 1

Lepanthes 1 0 0 0 0 2 1 0 0

Lepanthopsis 1 0 1 0 0 2 1 1 0

Masdevallia 1 1 2 0 0 3 1 2 1

Physosiphon 1 1 2 0 0 2 1 1 2

Physothal/is 1 2 0 0 0 2 1 1 3

Platystele 1 0 1 0 1 2 1 1 3

Pleurothallis ruscifolia complex 1 0 1 0 0 2 1 1 0

Pleurothallis aiexandrae complex 0 2 3 0 0 2 1 0 0

Pleurothallis blaisdellii complex 1 1 1 0 0 1 1 1 0

Pleurothallis dura complex 1 1 0 0 0 2 1 1 0

Pleurothallis grobyi complex 1 0 0 0 0 2 1 2 2

Pleurothallis loranthophylla complex 1 1 0 0 0 3 1 1 1

Pleurothallis peduncularis complex 1 1 2 1 0 2 1 0 0

Pleurothallis sclerophylla complex 1 1 1 0 0 2 1 1 1

Pleurothallis scoparium complex 1 1 1 0 0 3 1 2 2

Pleurothallis subgen. Specklinia 1 0 0 0 0 2 1 0 1

Pleurothallis tuerckheimii complex 1 1 2 0 0 2 1 1 0

(table cont' 191

Porrog/ossum 1 1 2 0 0 3 1 2 0

Restrepia 1 0 1 0 0 1 1 1 0

Restrepiella 1 0 0 0 0 2 1 0 0

Restrepiopsis 1 0 0 0 0 2 1 1 0

Sal pi stele 1 0 1 0 0 1 1 1 3

Scaphosepa/um 1 0 1 0 0 3 1 2 1

Ste/is 1 0 2 0 0 2 1 1 0

Trisetella 1 1 2 0 0 1 1 2 2

Octomeria 1 1 2 0 0 2 1 0 2 Outgroup

Arpophyllum giganteum 0 1 3 0 0 0 0 0 0

Brassavola nodosa 0 2 3 0 0 1 1 0 0

Epidendrum ciliaris 0 1 3 0 0 0 0 0 0

(table cont'd) Taxa 1 2 2 2 2 2 2 2 2 9 0 1 2 3 4 5 6 7

Ingroup

Acostaea 0 1 0 1 0 0 1 1

Barbosella 0 1 1 0 1 0 1 0

Brachionidium 0 1 0 2 1 0 ? ?

Condylago 0 1 0 2 1 0 1 1

Cryp tophoran th us 2 1 0 2 0 0 1 0

Dracula 0 1 1 2 1 0 1 1

Dresslere/la 0 1 0 2 1 0 1 0

Dryadella 1 1 1 2 0 0 1 1

Lepanthes 0 1 0 2 1 0 1 1

Lepanthopsis 2 1 0 2 1 0 ? ?

Masdevallia 2 1 1 2 1 0 1 1

Physosiphon 1 1 0 2 1 0 1 1

Physotha/lis 0 1 0 2 0 0 0 1

P/atyste/e 1 1 0 0 0 0 1 1

Pleurothallis ruscifolia complex 2 1 0 2 1 0 1 0

Pleurothallis alexandrae complex 0 0 0 2 1 0 0 0

Pleurothallis blaisdellii complex 1 1 0 2 1 0 0 0

Pleurothallis dura complex 2 1 0 2 0 0 ? ?

Pleurothallis grobyi complex 0 1 0 2 0 0 1 1

Pleurothallis loranthophyl/a complex 2 1 0 2 1 0 0 1

Pleurothallis peduncularis complex 0 0 0 2 1 0 1 0 0

Pleurothallis sclerophylla complex 1 1 0 2 1 0 0 1 0

Pleurothallis scoparium complex 1 1 0 2 1 1 1 1 1

Pleurothallis subgen. Speck/inia 0 1 0 2 0 0 1 1 1

(table cont'd) 1 9 3

Pleurothallis tuerckheimii complex 1 1 0 1 1 0 1 1 0

Porrog/ossum 1 1 1 2 0 0 1 1 1

Restrepia 2 1 0 2 1 0 1 0 1

Restrepiel/a 2 1 1 2 1 0 0 1

Restrepiopsis 2 1 0 1 1 0 ? ? ?

Sa/piste/e 0 1 0 1 0 0 ? ? ?

Scaphosepa/um 1 1 0 2 1 0 1 1 1

Stelis 2 1 1 2 0 0 1 1 1

Trisetella 1 1 1 2 1 0 1 1 1

Octomeria 0 0 0 2 0 0 0 0 0 Outgroup

Arpophy/lum giganteum 0 0 0 2 1 0 0 0 0

Brassavo/a nodosa 0 0 0 2 0 0 0 1 0

Epidendrum ciliaris 2 0 1 2 1 0 0 0 0

(table cont'd) 194

Taxa 2 2 3 3 3 3 3 3 3 8 9 0 1 2 3 4 5 6

Ingroup Acostaea 1 1 1 0 1 0 2 1 0 Barbosel/a 1 1 0 0 1 0 2 1 2

Brachionidium ? 7 7 7 7 7 7 ? 1

Condy/ago 0 1 ? 7 7 7 7 ? 0

Cryp tophoran th us 0 1 2 0 1 0 1 1 0

Dracula 1 1 0 0 1 0 0 1 0

Dresslerella 0 0 1 3 1 1 2 1 0

Dryade/la 1 1 2 2 1 0 2 1 0

Lepanthes 1 1 ? 7 ? ? ? ? 1

Lepanthopsis 7 7 7 7 ? ? 7 ? 0

Masdevallia 1 1 3 0 1 0 1 1 0

Physosiphon 0 1 2 0 1 0 2 1 0

Physothallis 0 1 ? 7 7 7 7 ? 0

PIaty stele 1 1 2 0 1 0 2 1 0

Pleurothallis ruscifolia com plex . 0 1 0 3 0 1 1 1 1

Pleurothallis a/exandrae com plex 0 1 0 1 1 0 0 0 0

PI euro thallis blaisde/lii com plex 0 1 ? 7 7 7 ? ? 0

Pleurothallis dura complex 7 7 7 7 7 7 ? ? 0

Pleurothallis grobyi complex 1 1 ? 7 7 7 ? ? 0

Pleurothallis loranthophylla com plex 0 1 ? 7 7 7 7 ? 0

Pleurothallis peduncularis com plex 0 1 3 0 1 0 0 0 0

Pleurothallis sclerophylla com plex 0 1 ? 7 ? ? 7 ? 0

Pleurothallis scoparium com plex 0 1 ? 7 ? ? 7 ? 0

(table cont'd) 195

Pleurothallis subgen. Specklinia 1 1 0 0 0 2 0 1

Pleurothallis tuerckheimii complex 0 1 0 1 0 2 1 0

Porrog/ossum 1 1 3 3 0 1 1 1

Restrepia 0 1 3 0 1 1 1 0

Restrepiella 0 1 0 2 0 0 1 0

Restrepiopsis ??? ? ? ?? 0

Sa/piste/e ? ??? ? ? ? 0

Scaphosepa/um 1 1 3 3 1 2 1 2

Ste/is 0 1 0 2 0 2 1 1

Trisetella 1 1 2 0 0 2 1 0

Octomeria 0 1 1 0 0 0 0 0 Outgroup

Arpophyl/um giganteum 0 0 0 0 0 0 0 2

Brassavola nodosa 0 1 0 0 0 0 0 0

Epidendrum cifiaris 1 1 0 0 0 2 0 0

(table cont’d) 196

Taxa 3 3 3 4 4 4 4 4 4 7 8 9 0 1 2 3 4 5

Ingroup Acostaea 2 0 2 1 0 2 0 3 0 Barbosel/a 2 0 2 0 0 2 0 2 0 Brachionidium 0 0 2 0 0 0 0 1 1 Condy/ago 2 0 2 1 0 2 0 3 0 Cryp toph or an th us 2 1 2 0 0 2 0 3 0 Dracula 2 0 1 0 1 2 0 3 0 Dresslerella 2 0 2 0 0 2 0 2 0 Dryadel/a 2 0 2 0 0 2 0 3 0 Lepanthes 2 0 0 0 0 0 2 3 0 Lepanthopsis 2 0 1 0 0 0 0 3 1 Masdevallia 2 0 2 0 0 2 0 3 0 Physosiphon 2 0 2 0 0 2 0 3 0 Physotha/lis 2 0 2 0 0 2 0 3 0 2 0 1 0 0 1 0 3 1 Pleurothallis ruscifo/ia complex 1 0 1 0 0 1 0 3 0 Pleurothallis a/exandrae complex 2 0 2 0 0 2 0 3 0 Pleurothallis blaisde/lii complex 2 0 2 0 0 2 0 3 0 Pleurothallis dura complex 2 0 2 0 0 2 0 3 0 Pleurothallis grobyi complex 2 0 2 0 0 2 0 3 0 Pleurothallis loranthophylla complex 0 0 2 0 0 2 0 3 0 Pleurothallis peduncularis complex 2 0 2 0 0 2 0 3 0 Pleurothallis sclerophylla complex 2 0 2 0 0 2 0 3 0 Pleurothallis scoparium complex 2 0 2 0 0 2 0 3 0 Pleurothallis subgen. Specklinia 2 0 2 0 0 2 1 3 0

(table cont'd) 197

Pleurothallis tuerckheimii complex 2 0 2 0 0 2 0 3 0

Porroglossum 2 0 2 1 0 2 1 3 0

Restrepia 2 0 0 0 0 0 0 2 0

Restrepiella 2 0 2 0 0 2 0 2 0

Restrepiopsis 2 0 2 0 0 2 0 2 0

Sa/pistele 2 0 0 0 0 0 3 3 0 2 0 2 0 0 2 0 3 0

Ste/is 2 0 2 0 0 1 1 3 1

Trisetella 2 0 2 0 0 2 0 3 0

Octomeria 0 0 1 0 0 1 0 0 0

Outgroup

Arpophyllum giganteum 0 0 0 0 0 2 0 0 0

Brassavola nodosa 2 0 0 0 0 0 0 0 0

Epidendrum ciliaris 2 0 0 0 0 0 0 2 0 198

Acostaea Sal pi stele Brachionidium Dracula Condglago Dryadella Masdevallia Porroglossum Physosiphon Trisetella P. scoparium Scaphosepalum Platg3tele Lepanthe3 P.subgen.Specklinia Steli3 P. grobgi Phgsothalli3 Cryptophoranthus P. dura Lepanthopsis P. ruscifolia

Figure 6.1 Strict consensus tree of eight trees from a parsimony search for the taxa included in this study. 199

Aco3taea Sal pi 3tele Brachionidium Dracula Condylago Dryadella Masdevallia Porroglossum Trisetella Physosiphon P. scoparium Scapho3epalum Platystele Lepanthes P.subgen.Specklinia P. grobyi Physothallis P. blaisdelii P. sclerophylla P. tuerckheimii P. loranthophylla Cryptophoranthus P.dura Lepanthopsis P. ruscifolia Restrepia Re3trepiop3is Restrepiella Barboaella Dresslerella P. alexandrae P. peduncularis Octomeria Arpophyllum Bras3avola Epidendrum

Figure 6.2 One of the eight most parsimoniuos trees discovered from a parsimony search for the taxa included in this study. The value above each branch represents the number of unambiguous synapomorphies. 2 0 0

Orchidaceae (Dressier, 1993). However, the use of characters with continuous

states (characters: 6,7,11,12,15) and polymorphic states (characters:

15,17,18,19,30,36,39) used in this study may also have contributed to this level of homoplasy (Appendix C). Branch lengths representing unambiguous synapomorphies are generally considered a measure of clade stability. As indicated

by one of the most parsimonioius cladograms, nodal support as defined by the number of synapomorphies is often weak (Fig. 6.2). This level of support is due in part to the low ratio of terminal taxa (34) to characters (45). Despite the high level of homoplasy and generally weak branch support, the data set used in this study brings a substantial level of resolution to the ingroup and provides a cladistic framework for a phylogenetic discussion of Pleurothallidinae.

Octomeria and the Pleurothallis peduncularis complex form the most basal clade in this study (Fig. 6.2). The position of Octomeria is consistent with the hypothesis that this taxon represents the most basal element in Pleurothallidinae.

The apparently close relationship between Octomeria and the P. peduncularis complex was suggested by Pridgeon (1982a). The P. alexandrae complex appears as sister to the Octomeria-P. peduncularis clade. Pridgeon (1982a) suggested a close relationship among these taxa and distinguished the P. alexandrea complex from the P. peduncularis complex by floral specializations that led to fly pollination.

Ascending the cladogram, a grade of genera exhibiting four pollinia

(Dresslerella, Barbosella, Restrepiella , Restrepiopsis) are next encountered (Fig. 6.2).

A close association among all Pleurothallidinae exhibiting four pollinia was suggested by Pridgeon (1982a). However, Restrepia (also exhibiting four pollinia) 201

appears in a separate clade and thus appears not closely related to taxa in this

grade (Fig. 6.2).

The Cryptophoranthus, Lepanthopsis, Pleurothallis dura complex,

Pleurothallis ruscifolia complex, and Restrepia clade is distinguished in this study by

stem cuticles <5/vm (character 25), and a variable number of root protoxylem poles

(character 34). The position of Lepanthopsis in this clade suggests that this taxon

is not closely allied with Lepanthes as was suggested by Pridgeon (1982a), who

distinguished Lepanthes from Lepanthopsis primarily by the loss of spirally thickened idioblasts. The two genera are vegetatively similar with respect to

lepanthiform (tubular) sheaths of the ramicaul. The Pleurothallis blaisde/ii, P. sclerophylla, P. tuerckheimii, and P. loranthophylla species complexes form a clade distinguished in this study by the absence of a column foot (character 42) and by a variable degree of fusion between the labellum and column (character 39). A single synapomorphy in six of the eight most parsimonious trees discovered separates these two clades. Therefore, this distinction between the two clades is considered weak.

Consisting of two terrestrial species Physothallis appears as sister to the

Pleurothallis grobyi complex. Claiming that the deeply connate dorsal sepal is insufficient for defining this taxon at the generic level, Luer (1986a) relegated

Physothallis to subgeneric status in Pleurothallis. The clade composed of the

Pleurothallis subgen. Specklinia and Stelis is distinguished in this study by differentiated leaf mesophyll (character 18), incumbent anther (character 43), and variability in resupination (character 36). Although the large genus Stelis 0 500 species) appears to be a natural genus, the great similarity in floral morphology has 2 0 2

made taxon distinction at the species level difficult. Pridgeon (1982a) suggested a

close relationship existed between Stelis and Brachionidium on the basis that both taxa share a footless and bibrachiate column. However, this cladistic analysis suggests that the two taxa represent disperate elements in Pleurothallidinae

(Fig. 6.2).

The large clade composed of Dryadella, Masdeva/lia, Porroglossum,

Trisetella, Physosiphon, Scaphosepalum, Platystele, and the Pleurothallis scoparium complex is distinguished in this study by uniseriate velamen (character 30) and variabile presence of spirally thickened idioblasts (character 19). Platystele appears as the most basal taxon in this clade. Consisting of approximately 73 species, this genus is distinguished by Luer (1986a) as having a lateral racemose inflorescence and ramicauls shorter than the leaves. Luer (1986a) considered Platystele closely related to Pleurothallis. Dryadella consists of about 40 species that formally resided in Masdevallia based on superficial similarities (Luer, 1986a). As suggested by the cladogram, Dryadella does not appear closely allied with Masdevallia.

Scaphosepalum (approximately 30 species), distinguished by Luer (1988) as having a pair of cushion-like osmophores positioned on each of the lateral , appears as sister to the Pleurothallis scoparium complex (Fig. 6.2). According to Luer

(1986a), Trisetella (approximately 15 species), distinguished by a column with a hooded ventral anther, was previously treated as a section of Masdevallia. Results from this study indicate that Trisetella and Masdevallia are closely allied, and the treatment of Trisetella as a genus is justified. Masdevallia (approximately 350 species) is distinguished by Luer (1986b) as having calliferous petals and a lip hinged to a free incurved extension from the apex of the column foot. Luer (1987) 2 0 3

stated that Porroglossum resembles Masdevallia by having short ramicauls and

sepals with long tails; however, its sensitive labellum sufficiently distinguishes it

from Masdevallia. The results of this study indicate that Masdevallia and

Porroglossum are sister taxa.

The clade composed of Acostaea, Salpistele, Brachionidium, Dracu/a, and

Condylago is distinguished in this study by having 1-3 adaxial hypodermal layers in

the leaf (character 1 5). Pridgeon (1982a) suggested that Brachionidium was closely

allied with Octomeria and is distinguished from Octomeria by the loss of two

pollinia. However, the relatively derived position of Brachionidium suggests the two

taxa are disparate elements in Pleurothallidinae (Fig. 6.2). Pridgeon (1982a)

considered Dracula to be closely allied with Brachionidium and distinguished Dracula

from Brachionidium by the loss of four pollinia, perianth specialization, and presence

of glandular trichomes on leaf surfaces. This cladistic analysis indicates that

Dracula and Brachionidium are sister taxa (Fig. 6.2). Acostaea, Condylago, and

Porroglossum each exhibit sensitive labella. According to Luer (1987), because the

labellum in each of these genera is structurally and functionally distinct, these taxa

are not closely related. As suggested by the results of this study, the sensitive

labellum arose independently within each of these three taxa (Fig. 6.2). Luer

(1991), stated that Salpistele may be closely allied with Lepanthes ; however, he considered the trumpet-shaped column and lack of lepanthiform sheaths sufficient to distinguish Salpistele as a genus. This cladistic analysis suggests that Salpistele is sister to Acostaea and that it is not closely related to Lepanthes (Fig. 6.2).

The results of this cladistic study suggests that the large genus Pleurothallis is polyphyletic and, therefore, may be divided into several genera. One genus 2 0 4

would be represented by the P. peduncularis complex that was segregated as

Myoxanthus by Luer (1992). The members of this genus are characterized by well-

developed ramicauls with scrufy sheaths and short rhizomes. A second genus may

be represented by the P. alexandrae complex which was elevated to

Si/enia by Luer (1992). Both complexes occupy a relatively basal position

within the subtribe and are far removed from the more derived Pleurothallis

complexes. The basis for separating these two closely related complexes is that

this analysis indicates that the P. peduncularis complex appears more closely related

to Octomeria than to the Pleurothallis alexandrae complex (Fig. 6.2).

Among the more derived Pleurothallis complexes, P. dura and P. ruscifolia

share a close affinity and occupy the same clade. Luer (1986a) elevated the P.

dura complex to subg. Tubella. Members of this subgenus produce

short ramicauls, connate lateral sepals, and ciliate petals. Luer (1986c) placed the

P. ruscifolia complex in Pleurothallis subg. Pleurothallis sect. Pleurothallis subsect.

Pleurothallis series Pleurothallis. Members of this series produce flowers borne in

short-fascicled racemes. Because this complex includes the type species, P.

ruscifolia (Jacq.) R. Br., the name Pleurothallis must be retained. This cladistic

analysis indicates that both complexes are more closely related to other genera than

to each other. Therefore, Luer's placement of the P. dura complex within

Trichosalpinx seem s justified (Fig. 6.2).

The Pleurothallis loranthophylla, P. tuerckheimii, P. sclerophylla, and P. blaisdellii complexes constitute a single clade. Luer (1986c) placed the P. loranthophylla complex in Pleurothallis subg. Rhynchopera. Luer characterized this subgenus as an easily recognizable homogenous group with well-developed 2 0 5

ramicauls that bear a petiolate leaf; the simple lip is firmly united to the short,

pedestal-like foot of the column. The P. tuerckheimii complex is included in

Pleurothallis subg. Dracontia in Luer's treatment. Its well-developed ramicauls bear

a tubular sheath near the middle and are subtended by two or three other sheaths

at the base; the lip is thick with erect, thin, wing-like basal lobes. Luer placed the

P. sclerophylla complex in Pleurothallis subg. Specklinia sect. Acuminatae. In that section, as in Pleurothallis subg. Dracontia, tubular sheaths also are inserted on the ramicaul. However, the lip is oblong-ligulate often with a pair of calli; the base is simply hinged to the stout column foot. The P. blaisdellii complex is placed in

Trichosalpinx subg. Trichosalpinx by Luer (1986a). That subgenus is characterized by stout ramicauls, free lateral sepals and petals with entire margins. Because these four complexes are grouped in the same clade, justification for their taxonomic separation into discrete genera is uncertain. If considered a single taxonomic unit, the clade appears to warrant elevation to generic status (Fig. 6.1).

Moving toward the most derived taxa, the Pleurothallis grobyi complex,

P.subg. Specklinia, and the P. scoparium complex show little affinity and are thus logical candidates for generic status (Fig. 6.2). The P. grobyi complex was placed within Pleurothallis subg. Specklinia sect. Hymenodanthae by Luer (1986c). Luer described that section as having numerous, closely allied caespitose species with an abbreviated ramicaul and a racemose inflorescence that is usually longer than the leaf. Pleurothallis subg. Specklinia includes numerous species characterized by the lip hinged to the column foot. The column is well developed; its apex covers the anther, rostellum, and stigma. Luer placed the P. scoparium complex within

Pleurothallis subg. Scopu/a. He attributed four species to the subgenus; they are 2 0 6 characterized by elliptical leaves with a tuft of single-flowered peduncles emerging from what appears to be the midrib near the leaf apex.

The number of pollinia (character 44) is plotted onto one of the most parsimonious trees to illustrate character state changes (Fig. 6.3). Equivocal branches were resolved by the MacClade (Maddison & Maddison, 1992) equivocal cycling feature. This analysis suggests that although the eight-pollinium state may represent the primitive condition as espoused by Dressier and Pridgeon, reduction in the number of pollinia did not follow a unilinear series. Instead it appears that the eight-pollinium state represented by Octomeria, the two-pollinium state represented by the P. peduncularis and P. alexandrae complexes, and the four-pollinium state represented by the grade comprising Restrepiopsis, Restrepiella, Barbose/la, and

Dresslerel/a all arose early in the phylogeny of the subtribe. The two-pollinium state reappears as a parallelism and is represented by most of the genera and a vast majority of species within the subtribe. Because Restrepia appears in a clade removed from the other four-pollinia taxa, its character state also is considered a parallelism. The six-pollinium condition of Brachionidium represents an autapomorphy.

This study represents the first attempt to bring cladistic resolution to the subtribe Pleurothallidinae. Because the characters used in this study exhibit a high level of homoplasy and branch nodes on cladograms are often weakly supported, the results of this analysis are considered tentative. Other data to include molecular, cytological, and palynological characters should be evaluated in order to further clarify the phylogeny of this vast subfamily. 207 iQ Brassavola iQ Epldendrum IH

Pollinium Number

Figure 6.3 Character-state tree showing the distribution of states for the number of pollinia (character 44) for the taxa included in this study. 2 0 8

LITERATURE CITED

Burns-Balogh, P. and V. A. Funk. 1886. A phylogenetic analysis of the Orchidaceae. Smithsonian Contributions to Botany 61: 1-79.

Dressier R. L. 1960. The relationships of Meiracyllium (Orchidaceae). Brittonia 12: 222-225.

Dressier R. L. 1981. The Orchids: Natural History and Classification. Cambridge, Massachusetts: Harvard University Press.

Dressier R. L. 1993. Phylogeny and Classification of the Orchid Family. Portland, Oregon: Dioscorides Press.

Fitch W. M. 1971. Toward defining the course of evolution: minimal change for a specific tree topology. Systematic Zoology 20: 406-416.

Garay L. A. 1956. Studies in American orchids. II. The genus Brachionidium Lindl. Canadian Journal of Botany 34: 721-743.

Luer C. A. 1986a. leones Pleurothallidinarum I. Systematics of the Pleurothallidinae. St. Louis: Missouri Botanical Garden.

. 1986b. leones Pleurothallidinarum II. Systematics of Masdevallia St. Louis: Missouri Botanical Garden.

. 1986c. leones Pleurothallidinarum III. Systematics of Pleurothallis. St. Louis: Missouri Botanical Garden.

. 1987. leones Pleurothallidinarum IV. Systematics of Acostaea Condylago and Porroglossum. St. Louis: Missouri Botanical Garden.

.1988. leones Pleurothallidinarum V. Systematics of Dresslerel/a and Scaphosepalum. St. Louis: Missouri Botanical Garden.

. 1991. leones Pleurothallidinarum VIII. Systematics of Lepanthopsis, Octomeria subgenus , Restrepiella, Restrepiopsis, Salpistele, and . St. Louis: Missouri Botanical Garden.

. 1992. leones Pleurothallidinarum IX. Systematics of Myoxanthus. St. Louis: Missouri Botanical Garden.

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

Neyland R., Urbatsch L. E. 1993. Anatomy and morphology of the articulation between ovary and pedicel in Pleurothallidinae. Lindleyana. 8(4): 189-192. Pridgeon A. M. 1982a. Numerical analyses in the classification of the Pleurothallidinae (Orchidaceae). Botanical Journal of the Linnean Society 85:103-131.

Pridgeon A. M. 1982b. Diagnostic anatomical characters in the Pleurothallidinae (Orchidaceae). American Journal of Botany 69: 921-938.

Swofford D. L. 1991. Phylogenetic Analysis Using Parsimony. Version 3.0s. Champaign Illinois: Illinois Natural History Survey. CHAPTER 7

SUMMARY AND CONCLUSIONS

2 1 0 211

The dissertation presented here offers new insights into the systematics of

Orchidaceae and particularly the subfamily Epidendroideae. Cladograms discovered from both parsimony and maximum likelihood searches are congruent in that subfamily Epidendroideae is monophyletic with the tribe Neottieae as sister.

When unweighted ndhF sequences are used in parsimony analyses, subfamilies of the Orchidaceae and subtribes of the Epidendroideae are well resolved; however, tribal-level relationships within Epidendroideae are weakly resolved (Fig. 2.6). This phenomenon may be due to a large number of lineages diverging in a short period of evolutionary time. This process may have happened so rapidly that either no mutations exist at this level or the few mutations that do exist convey conflicting ideas of relationships due to homoplasy. This hypothesized radiation may have been catalyzed by morphological, physiological, and anatomical adaptations that allowed the epidendroid orchids to pioneer xeric arborescent habitats in tropical regions.

The strict consensus tree using just morphological characters suggests that the organismal history is not significantly different from the gene history for the

Orchidaceae (Fig. 2.8). Therefore, following the criterion of Bull et al. (1993), parsimony analyses using combined morphological characters and ndh? sequences is considered appropriate for this assemblage of taxa. When these two data types are combined and subjected to parsimony analysis, a single tree was discovered that resolved intertribal relationships. As morphological characters provided the data necessary to bring tribal-level resolution to the subfamily Epidendroideae, it is apparent that these characters are not overwhelmed or "swamped" by the more numerous DNA sequences characters. However, the level of confidence in clades 2 1 2

resolved by these data is not high. That is, tribal-level nodes are not resolved with

a substantial degree of confidence as measured by the number of unambiguous

synapomorphies, bootstrap values, or decay indices. (Fig. 2.9).

Of the 35 INDELS identified in this study, 22 are autapomorphic and, therefore, not informative. Of the remaining 13 informative INDELS (Table 2.3), at least four appear to be homoplastic. Because it is probable that addition-deletion mutations may recur at specific sites, and thus contribute to homoplasy in evolutionary studies (Clegg and Zurawski, 1992), the branches resolved as a result of adding INDELS to nucleotide characters must be viewed with caution (Fig. 2.7).

Sequence data for the taxa included in this study were also used to infer phylogeny by the maximum likelihood method. In order to find trees with the highest log likelihoods, it was necessary to test an array of different transversion weighing parameters in separate analyses. For this series of analyses, the jumble option (randomization of input data) was not used. For the data set used in this study, a range of maximum likelihood transversion weighting parameters tested had minimal effect on topologies produced (Fig. 3.8). To test the effect of the jumble option on topologies discovered and their respective log likelihood values calculated by the algorithm, maximum likelihood searches were made using a series of jumble seeds. The maximum likelihood jumble seed value is shown by this study to have a significant impact on resulting tree topologies and their respective log likelihoods.

The results of this study also suggest that several tests using different jumble seeds are necessary when performing maximum likelihood analyses before the topology with the highest log likelihood value is discovered. 2 1 3

Output produced by fastDNAml (Olsen, 1994) software includes a table

showing the length of each tree branch in units of expected nucleotide substitutions

per site. For each branch segment, confidence limits on their length are supplied.

According to Felsenstein (1993), these confidence limits are very rough. Because

there is a simplification in how the confidence limits are calculated, over confidence

in the existence of the branch is the result (Olsen; pers. comm., 1994). Therefore,

supplied confidence limits were not used to evaluate branch lengths in this study.

In parsimony and maximum likelihood analyses, branch length is considered

an estimated measure of clade stability. Because branch length in parsimony

analyses represents the number of synapomorphies, whereas branch length in

maximum likelihood analyses represents expected nucleotide substitutions per site, the two measurements are not synonymous. Therefore, using branch lengths to evaluate clade stability in maximum likelihood analyses is problematic.

In order to provide an estimate of the branch length value that may be considered to represent a "significant" level of clade support, all branches found in the strict consensus parsimony tree (Fig. 2.6) that also occurred in the maximum likelihood tree were identified. Because these branches are present in both trees, their respective branch lengths (in terms of expected nucleotide substitutions per site) may be considered to represent an estimate of clade support that is more

"significant" than those branches not present in both trees. It is acknowledged that the term "significant" in this sense has no statistical basis; however, it does provide a crude estimate of clade support in the maximum likelihood analysis. Using this criterion, maximum likelihood branch-length values that range as low as .001 are considered to provide this defined level of "significant" clade support. 2 1 4

Phylogeny inferred from parsimony methods and maximum likelihood

methods is compared by examining the congruence among trees produced with

each method. In this discussion, the maximum likelihood tree with the highest log

likelihood value (Fig. 3.3) is compared to the strict consensus parsimony tree where

only unweighted ndh? sequences were used (Fig. 2.6), and to the single parsimony tree discovered when DNA sequences and morphological characters were combined

(Fig. 2.9). Using Oryza (Poaceae) as outgroup, both the strict consensus parsimony tree (Fig. 2.6) and the maximum likelihood tree (Fig. 3.3) suggest that the subfamily

Epidendroideae is monophyletic with Listera (Neottieae) as sister.

Within the subfamily Epidendroideae, neither the cymbidioid nor epidendroid phylads (sensu Dressier, 1993) is suggested to be monophyletic by either the maximum likelihood tree (Fig. 3.3) nor the parsimony tree in which DNA sequences and morphological characters were combined (Fig. 2.9). Although the cymbidioid taxa: Stanhopea and Maxillaria (Maxillarieae), and Cymhidium (Cymbideae) form a clade in both parsimony trees (Figs. 2.6, 2.9) and in the maximum likelihood tree

(Fig. 3.3), in none of these trees does the other cymbidioid taxon, Tipularia

(Calypsoeae) appear closely allied with this clade. Therefore, the cymbidioid phylad appears polyphyletic.

Vanda (Vandeae) appears as sister to the core taxa of the cymbidioid phylad in both the maximum likelihood tree (Fig. 3.3) and the parsimony tree (Fig. 2.9).

Also in both trees, Aerangis (Vandeae), Angraecum (Vandeae) and Polystachya (Old

World tribe Epidendreae) form a clade distinct from the cymbidioid core taxa-Vanda clade which suggests the tribe Vandeae is polyphyletic. 2 1 5

Both the maximum likelihood tree (Fig. 3.3) and the parsimony trees (Figs.

2.6, 2.9) suggest that the New World tribe Epidendreae is polyphyletic. In all trees

Sobralia (Sobraliinae) represents the most basal element in Epidendroideae and, therefore, not closely allied with the core taxa of the New World tribe Epidendreae

(Arpophyliinae, Laeliinae, Pleurothallidinae, and Meiracyliinae subtribes). The phylogenetic relationship of Coelia (Coeliinae, New World tribe Epidendreae) is ambiguous. In the parsimony tree (Fig. 2.9), Coelia does not appear closely related to the core taxa of the New World tribe Epidendreae; however, the maximum likelihood tree (Fig. 3.3) suggests Coelia forms a clade with Tipularia (Calypsoeae) that is sister to the New World tribe Epidendreae core taxa. In the parsimony tree

(Fig. 2.9), Chysis (Chysiinae, Arethuseae) appears as sister to the core taxa of the

New World tribe Epidendreae. In the maximum likelihood tree (Fig. 3.3) Chysis appears as sister to the Pleurothallidinae-Arpophyliinae clade. Whether Chysis represents a basal taxon in the New World tribe Epidendreae or sister to the tribe, both trees suggest this taxon is closely allied with the New World tribe Epidendreae and not with the Spathog/ottis-Calanthe (Arethuseae) clade.

In the parsimony tree (Fig. 2.9), Coelogyne (Coelogyneae) forms a clade with

Dendrobium (Dendrobieae). Therefore, the parsimony tree suggests that although

Thunia (Coelogyneae) and Coelogyne are closely allied, the tribe Coelogyneae is polyphyletic. However, the maximum likelihood tree indicates that Coelogyne and

Thunia form a clade suggesting the monophyly of the tribe Coelogyneae (Fig. 3.3).

In neither the parsimony tree (Fig. 2.9) nor the maximum likelihood tree (Fig. 3.3) are Dendrobium (Dendrobieae) and Bulbophyllum (Dendrobieae) suggested to be closely allied which suggests that the tribe Dendrobieae is polyphyletic. 2 1 6

Although congruent in many respects, topological conflicts between the completely resolved parsimony tree (Fig. 2.9) and the maximum likelihood tree (Fig.

3.3) are found where branch support is weak.

The approximate 1,200 bases sequenced from the 3' half of the ndhF gene bring a substantial degree of resolution at the subfamilial level within Orchidaceae and at the subtribal level within subfamily Epidendroideae using parsimony and maximum likelihood methods of phylogenetic inference. However, most tribal-level relationships within Epidendroideae are weakly resolved when measured by the number of unambiguous synapomorphies, bootstrap values, decay indices

(parsimony methods) or by the number of expected nucleotide changes per site

(maximum likelihood methods). To bring a more robust resolution to the tribal relationships in subfamily Epidendroideae, future studies may rely on analyses where complete ndhF sequences are subjected to phylogenetic inference methods.

However, this approach may prove problematic as the 5’ half of the gene is highly conserved and may provide few informative characters (Fig. 2.2). The sequencing of different genes may provide the best option for resolving these tribal relationships. According to Olmstead and Palmer (1994) there are 20 chloroplast genes that are sufficiently large (> 1 kb) and widespread to be generally useful in this type of study.

An unpublished study using the rbcL gene for about 70 orchid taxa has been undertaken by Mark Chase at the Royal Botanic Gardens, Kew (pers. comm.,

1993). His preliminary results indicate that branches are so short for the taxa in his study that internal support is poor in many cases; therefore, rbcL appears not to bring much resolution to phylogenetic relationships within Orchidaceae . At 2 1 7

present, another molecular study has been initiated for Orchidaceae at Kew in

which (maturase K) matK sequences are used for phylogenetic inference purposes

(Mark Chase pers. comm., 1994). Although no results from this project have been

published, the m atK gene may provide a better degree of internal support in

Orchidaceae than either the rbcL or ndhF genes. According to Johnson and Soltis

(1994), m atK evolves faster than rbcL and appears to have more phylogenetic

potential for such a study.

Not withstanding the generally weak resolution ndhF sequences bring to

tribal relationships within subtribe Epidendroideae, this gene is shown by this study

to bring new insights into the phylogenetic relationships in Orchidaceae and

subfamily Epidendroideae. Due to the conserved nature of ndhF sequences, it

appears that this gene may be valuable in resolving angiosperm phylogenies at the

interfamilial level. Although the ndhF gene has been discovered in Marchantia po/ymorpha, a liverwort, (Ohyama et al., 1986), preliminary evidence suggests that this gene is absent in at least one gymnosperm, Pinus thunbergii, (Wakasugi,

1994). Therefore, use of ndhF to help resolve the relationships between the angiosperms and gymnosperms is problematic. More experiments designed to detect whether the ndhF gene is present in other gymnosperms are required. If ndhF proves to be consistently absent in the gymnosperms, an interesting question arises. What is the evolutionary significance of ndhF occurring in liverworts and angiosperms but not in gymnosperms? Although this is one of the first studies to use ndhF sequences to infer plant phylogeny, it is apparent that studies using this gene have potential to yield more answers to questions concerning the evolution of plants. 2 1 8

LITERATURE CITED

Bull, J. J., J. P. Huelsenbeck, C. W. Cunningham, D. L. Swofford, and P. J. Waddell. Partitioning and combining data in phylogenetic analysis. Systematic Biology 42: 384-397.

Clegg, M. T. and G. Zurawski. 1992. Chloroplast DNA and the study of plant phylogeny: present status and future prospects. In: P. S. Soltis, D. E. Soltis, and J. J. Doyle (eds.). Molecular Systematics of Plants. Chapman and Hall, New York, pp 1-13.

Dressier, R. L. 1993. Phylogeny and Classification of the Orchid Family. Dioscorides Press. Portland, Oregon.

Felsenstein, J. 1993. Phylogeny Inference Package (Phylip). Version 3.5c. University of Washington.

Ohyama, K., H. Fukuzawa, T. Kohchi, H. Sanyo, T. Sano, S. Sano, K. Umesono, Y. Shiki, M. Takeuchi, Z. Chang, S. Aota, H. Inokuchi, and H. Ozeki. 1986. Chloroplast gene organization deduced from complete sequence of liverwort Marchantia po/ymorpha chloroplast DNA. Nature 322: 572-574.

Olsen, G. J., H. Matsuda, R. Hagstrom, and R. Overbeek. 1994. fastDNAml Version 1.0 University of Illinois, Urbana, Illinois.

Olmstead R. G. and J. D. Palmer 1994. Chloroplast DNA systematics: a review of methods and data analysis. American Journal of Botany 81:1205-1 224.

Wakasugi, T, J. Tsudzuki, S. Ito, K. Nakashima, T. Tsudzuki, and M. Sugiura. 1994. Loss of all ndh genes as determined by sequencing the entire chloroplast genome of the black pine, Pinus thunbergii. Proceedings of the National Academy of Science 91: 9794 - 9798. APPENDIXES 2 2 0

APPENDIX A

Velamen types relevant to this study as defined by Porembski and Barthlott

(1988):

1) Calanthe type. Velamen usually 3-4 layered without helical thickenings. Cell walls often show relatively small pores. Stabkorper lacking. Exodermis cells frequently large and only slightly thickened. Root cortex always parenchymatous.

2) Bulbophyllum type. Velamen one-layered consisting of rectangular radially elongated cells. Helical thickenings absent. Cell edges partially thickened.

Stabkorper sometimes present. Walls of the exodermis usually strongly thickened.

Tracheoidal idioblasts are rare.

3) Pleurothallis type. Velamen one to three layered. In case of multi-layered velamen, a small-celled epivelamen is present. In transverse section, cells somewhat extended in radial direction. Helical thickenings absent. Exodermis slightly thickened; usually the outer walls, rarely their whole cell wall. Tracheoidal idioblasts sometimes present in the cortex.

4) type. Velamen one or two layers without helical thickenings. Pores of different sizes occur. Exodermis cells often large. Cortex transformed into a pseudovelamen with cell walls reduced to ledge-like remains. In some cases, a normal parenchymatous cortex is present, but it contains a high number of tracheoidal idioblasts.

5) Spiranthes type. Usually one or two-layered velamen, not differentiated in epi- and endovelamen. Helical thickenings rather fine; small pores in unthickened wall portions. Cells in transverse section not conspicuously elongated. Exodermis with slightly thickened walls. 6) Coelogyne type. Usually four to six layered velamen, no clear differentiation in

epi- and endovelamen. Walls with relatively massive thickenings arranged as helical

bands. Nonthickened portions show relatively small pores. Granular Stabkorper

often present. Walls of the exodermis either partially or totally thickened.

Tracheoidal idioblasts frequently found in the cortex.

7) Dendrobium type. Distinction between epi-and endovelamen often not clear.

Cells in transverse section radially elongated; stabilized by helical thickenings of

different diameter. Relatively small pores in unthickened portions. Stabkorper and

trachoidal idioblasts absent.

8) Epidendrum type. Epivelamen cells are smaller than those of the endovelamen;

their latter extended in radial direction, but not as strongly elongated as the cells of the Cymbidium type velamen. They possess thickenings which are frequently fused

into composed ledges. These ledges may form a sinuate pattern over the entire cell

wall, but often they are characteristically restricted to the cell edges. Large pores

occur and may occupy large portions of the wall. Stabkorper forming ribbed half­ spheres are occasionally present. Exodermis cells are commonly slightly thickened.

The cortex sometimes contains tracheoidal idioblasts.

9) Cymbidium type. Distinction between epi- and endovelamen not as clear as in the Vanda type. Cells of the epivelamen usually smaller than the endovelamen cells; the latter extended in radial direction. They show a characteristic shape in transverse section. The velamen cell walls show thin long-sinuate wall thickenings; their regular arrangement is always disturbed by large pores occurring in the nonthickened wall portions. These pores may occupy > 50% of the cell wall.

Stabkorper are rare. Exodermis cells are frequently smaller than velamen cells and possess slightly thickened outer tangential walls. The cortex frequently contains

tracheodial idioblasts or is occasionally totally transformed into a pseudovelamen.

10) Vanda type. Velamen usually three- to five-layered; differentiated in large-celled epivelamen and an endovelamen with cuneate cells in transverse section. Massive anastomosing helical thickenings dominate in the epivelamen, in particular the radial walls. Unthickened wall portions show relatively small pores. The walls of the endovelamen cells are mostly thicker than those of the epivelamen cells. Cell size decreases toward the exodermis. Stabkorper always absent. Exodermis cells are often much larger than the velamen cells. Their outer tangential walls and parts of the radial walls are slightly thickened. Tracheoidal idioblasts in the cortex are occasionally present.

11) Unspecified velamen. In a few genera, velamen characters are unspecific; a coordination with any of the previously described types is not feasible.

1 2) Velamen absent. Some taxa do not develop any velamen. They are characterized by a one-layered living rhizodermis and the absence of an exodermis 2 2 3

APPENDIX B

Seed type descriptions relevant to this study as developed by Ziegler

(unpublished) and modified by Dressier (1993):

1. Eulophia type. Club-shaped balloon seeds to thread seeds; whitish to light

brown; the testa cells are distinctly elongate; intercellular spaces are never present;

the cell border is never covered by a folded "cuticle," and may be both raised and

sunken in the same seed; it may be present only as a small three-layered star at the

cell corners; definite thickenings are present on both anticlinal and periclinal walls;

the marginal ridge may be warty.

2. Maxillaria type. Dust seeds or oblong dust seeds, 250-500 pm long; yellowish to

brownish; the terminal testa cells are isodiametric or nearly so; although the cells of

the median sector are strongly elongate, this sector is usually not more than two

cells long; a marginal ridge is present and is round in section; the cell border ridge is

fine and concealed by the marginal ridges; the periclinal walls have reticulate or

longitudinal thickenings; fine warts or micropapillae may be present.

3. Bletia type. Dust seeds to long balloon seeds, sometimes almost thread-like; the testa cells are flat trough-shaped in cross section, show no thickenings, and are

always elongate and compressed laterally so that the transverse anticlinals are

bowed and project outward; resembles the Limodorum type (#12).

4. Dendrobium type. Short or oblong dust seeds, with a total length of 300-500 p m ; yellow to orange in color, sometimes yellowish green or red-orange, rarely brownish; the testa cells all of the same size and elongate; the transverse anticlinal walls are strongly bowed so that the ends of the cells are broadly rounded; the surface of the seed is always dull and velvety, that is, covered by very fine warts;

larger warts may also be present.

5. Vanda type. Dust seeds or oblong dust seeds; may be yellowish, but usually

brown or blackish brown; length from 300-500 /urn; the testa cells are always so

strongly elongate that the longitudinal anticlinal walls are in contact; the surface of

the seed is made up of marginal ridges; a cell border ridge is present but may be

covered by the marginal ridges; periclinal thickenings are not seen; the surface may

bear either micropapillae or warts.

6. Goodyera type. Generally dust seeds about five cells wide; cells about the same

size throughout; isodiametric or slightly elongate; intercellular spaces are prominent,

especially the cell corners which may be rounded.

7. Elleanthus type. Dust-like kernel seeds about 200 pm long. Basal and apical

testa cells slightly elongate; medial cells strongly elongate and twisted, deeply

trough-like in section with clear cell-border ridges; periclinal walls with strong,

longitudinally reticulate thickenings.

8. Pleurothallis type. Dust seeds from 150-300 pm long; yellowish or brownish; the

testa ceils are all the same size and distinctly elongate; seeds almost always 2-3

cells long; flat marginal ridges are present and always overtopped by a distinct cell

border ridge; the anticlinal walls have prominent thickenings; the seeds may be

covered by warts or scales that are easily soluble.

9. Epidendrum type. Seeds oblong to elongate dust-like 500-1000 pm long; usually

pale yellow; all cells similar and distinctly elongate; cell corners are not rounded, but

prominently acute-angled; anticlinals high, narrow, and sharp-angled; cell border not visible; with or without anticlinal thickenings; may be verrucose. 2 2 5

10. Cymbidium type. A very distinct seed type; always small balloon seeds from

500 to 1000 pm long; white, with the yellow embryo visible through the testa; the

testa cells are polygonal or slightly elongate; the cell border is not visible; the cell

corners are strongly raised, and each is covered with a small or large wax hood; the

anticlinal and periclinal walls are densely covered by parallel, longitudinal

thickenings.

11. Stanhopea type. White or brownish balloon seeds more than 500//m in

diameter; both the basal and apical sectors may project as tiny stalklets; the testa

cells in the balloon-like medial sector are all of the same size and isodiametric-

polygonal, but the terminal cells are more or less strongly elongate; the testa cells

are not or are incompletely collapsed so that the outer periclinal wall stretches over

the cell lumen; the anticlinal walls are reticulate; a smooth marginal ridge is present,

as are cell borders.

1 2. Limodorum type. Colorless or brownish balloon seeds about 1.0 mm long and

0.1 mm wide; the cells are of the same size throughout, isodiametric or slightly

elongate; the seeds are about 1 5-30 cells long and 10 cells wide, thus relatively

many celled.

13. Neuwiedia type. Media! portion of seed coat pitted, base and apex of seed coat inflated, balloonlike, shiny, and light brown; seed about 600/ym long; cell walls of

balloon sectors thin.

14. Orchis type. Dust seeds, light to dark brown; the medial cells distinctly elongate, the medial sector 1-2 cells long; the terminal cells more or less isodiametric; distinct parallel or reticulate thickenings are present on the anticlinal walls. 2 2 6

APPENDIX C

Attribute Character State

1. Glandular trichomes (leaf) (0) Absent (1) Limited to one surface (2) Present on both surfaces 2. Epidermal papillae (leaf) (0) Absent (1) Present 3. Multicellular mucilage trichomes (leaf) (0) Absent (1) Present 4. Hispid leaf sheaths (leaf) (0) Absent (1) Present 5. Cuticle surface (leaf) (0) Smooth (1) Papillose 6. Cuticle thickness (leaf) (0) 15-22/vm (1) 6-14//m (2) <3//m 7. Epidermal radial wall thickness (leaf) (0) > 5-7/ym (1) 3-5/ym (2) <3//m 8. Epidermal cell wall thickening (0) Differentially thickened (leaf in transection) (1) Uniformly thickened 9. Epidermal cell shape (0) Dome-shaped (1) Not dome-shaped 10. Level of stomatal apparatus (leaf) (0) Raised above the epidermis (1) Flush with the epidermis 2 2 7

11. Mean guard cell length (leaf) 0) 24.92 - 3 1 .22jt/m 1) 32.39 - 40.46//m 2) > 40.46/um 12. Mean stomatal length/width ratio 0) 0.96:1.0 - 1.04:1.0 (leaf) 1) 1.06:1.0 - 1.15:1.0 2) 1.17:1.0 - 1.25:1.0 3) 1.27:1.0 - 1.36:1.0 13. Epidermal raphide clusters (leaf) 0) Absent 1) Present 14. Epidermal oil droplets (leaf) 0) Absent 1) Present 15. Number of adaxial hypodermal layers 0) 0 (leaf) 1)1-3 2) >3 3) Variable number of layers 16. Abaxial hypodermis (leaf) 0) Absent 1) Present

17. Hypodermal thickenings (leaf) 0) Absent 1) Present 2) Variable presence of thickenings 18. Mesophyll cell differentiation into 0) Differentiated palisade and mesophyll cells (leaf) 1) Scarcely differentiated 2) Variably differentiation 3) Not differentiated 2 2 8

19. Spirally thickened idioblasts (leaf) (0) Absent (1) Presence of thickenings variable (2) Present

20. Number of vein series (leaf) (0 ) 2

( 1 ) 1 21. Parenchymatous bundle sheath (0) Absent (1) Present 22. Sclerotic bundle sheath (leaf) (0) Absent (1) Presence of bundle sheath variable (2) Present 23. Marginal bundle rotation (leaf) (0) Rotated (1) Not rotated 24. Embedded peduncle (0) Absent (1) Present 25. Cuticle thickness (stem) (0) 5 - 15//m (1) <5/vm 26. Epidermal cell shape (stem in (0) Dome-shaped transection) (1) Not dome-shaped 27. Sclerotic epidermis (secondary stem) (0) Present (1) Absent 28. Sclerotic subepidermal layers (0) Present (secondary stem) (1) Absent 29. Vascular cylinder (stem) (0) Eccentric (1) Not eccentric

30. Number of velamen layers (root) (0 ) > 2

( 1 ) 2

( 2 ) 1 (3) Variable number of layers 2 2 9

31. Exodermal thickenings (root) 0) U-shaped 1) O-shaped 2) U- and O-shaped 3) Variably shaped 32. Cortex (root) 0) Central large layer 1) Uniform 33. Endodermal thickenings (root) 0) O-shaped 1) Variably shaped 34. Number of protoxylem strands (root) 0) >9 1) Variable in number

2 ) <8 35. Sclerotic pith 0) Absent 1) Present 36. Resupination (flower) 0) Resupinate 1) Presence of resupination variable 2) Nonresupinate 3 7 .Relative shape of perianth (flower) 0) Parts similar 1) Presence variable 2) Parts dissimilar 38. Sepal fusion (flower) 0) Apices fused 1) Apices not fused 39. Fusion of labellum and column (flower) 0) Firmly fused 1) Presence of fused labellum variable 2) Articulate 40. Labellum sensitivity (flower) 0) Not sensitive 1) Sensitive 41. Labellum divided into epichile and 0) No hypochile (flower)

(1) Yes 2 3 0

42. Column foot (flower) (0) Absent (1) Presence of column foor variable (2) Present 43. Anther position (flower) (0) Incumbent (1) Dorsal (2) Apical (3) Trumpet-shaped

44. Number of pollinia (flower) (0 ) 8

( 1) 6 (2) 4 (3) 2

45. Number of stigma lobes (flower) (0 ) 1

( 1) 2 231

APPENDIX D

HARCOURT Harcourt Brace &; Company Limited 24-28 Oval Road BRACE London NWl ? n \ 'I'd 017!-267 4n66 Lux O P 1-482 22')i & 0171-485 4“o;

16 June 1995

Ray Neyland Louisiana State University 502 Life Sciences Building Baton Rouge Louisiana 70803-105 USA

Dear Ray Neyland

Re. Botanical Journal Society of the Linnaean society, "A Phylogenetic analysis of subtribe Pleurothallidinae (Orchidaceae) 117:13-28

Thank you for your letter of 7 April requesting permission to use the above material from the Journal of Molecular Biology.

We are happy to grant permission for this use of your material provided that (1) complete credit is given to the source, including the Academic Press copyright notice (2) the material to be used has appeared in our publication without credit or acknowledgment to another source and (3) if commercial publication should result, you must contact Academic Press again.

We realize that University Microfilms must have permission to sell copies of your thesis, and we agree to this. We would point out, however, that this does not apply to separate sale of your article.

Thank you for approaching us in this matter.

Yours sincerely

CATHERINE JOHN (Ms) Permissions Manager AP London and Bailliere Tindall VITA

Ray Neyland was born in Harris County, Texas on 12 December 1963, the child of Malcolm and Lillian Neyland. He received his high school diploma from

Bolton High School in Alexandria, Louisiana and graduated from the University of

West Florida with a B. A. in Accounting. Working as a computer auditor in

Jacksonville, he decided that a career in botany would be more satisfying than remaining in a business-related field. Therefore, in 1990 he returned to Louisiana to begin undergraduate science course work at Louisiana State University at

Alexandria. In 1991, he was accepted into the graduate program in botany at

Louisiana State University in Baton Rouge.

232 DOCTORAL EXAMINATION AND DISSERTATION REPORT

Candidate: Malcolm Ray Neyland

Major Field: Botany

Title of Dissertation: The Molecular and Morphological Systematics of Subfamily Epidendroideae (Orchidaceae)

Approved: iJUP tliUrc.L Major Professor and Chairman

Graduate School

EXAMINING COMMITTEE:

C 1

/J, y 7

Date of Examination:

May 1, 1995