Evolution and Historical Biogeography of Pacific Coprosma

EVOLUTION AND HISTORICAL BIOGEOGRAPHY OF PACIFIC COPROSMA

A DISSERTATION SUBMITTED TO THE GRADUATE DIVISION OF THE UNIVERSITY OF HAWAI‘I AT MĀNOA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

BOTANY

MAY 2014

Jason T. Cantley

Dissertation Committee:

Sterling Keeley, Chairperson Vicki Funk Clifford Morden Tom Ranker Lyndon Wester

ACKNOWLEDGMENTS

Without the work of the many funding sources and individuals who helped with

the completion of this dissertation, it simply would not have been possible. First, I would

like to thank the Botany Department at the University of Hawai‘i at Mānoa, the American

Society of Plant Taxonomists, the Lamoureux Fellowship fund for graduate student research, the Graduate Student Organization at the University of Hawai‘i at Mānoa and

Department of Plant Biology at Michigan State University for financial support.

Extensive field work would in New Zealand not have been possible without the help of

Dr. Adrienne Markey and Rewi Elliot at Otari-Wilton’s Bush in Wellington. We are very grateful to the New Zealand Department of Conservation, the Tasmanian Parks and

Wildlife Service, the Dunedin City Council, Otari-Wilton’s Bush, Zealandia, and the

Yellow Eyed Penguin Trust for collection permits and/or access to collection sites, and for the generous assistance of the Allen Herbarium, the Bishop Museum Herbarium,

Tasmanian Herbarium, Auckland War Memorial Museum, the Western Australian

Herbarium, National Tropical Botanical Garden and the Joseph Rock Herbarium for facilitating the exchange of herbarium vouchers. Fieldwork in the Hawaiian Islands would not have been possible without the help of David Lorence and Tim Flynn at

NTBG, Amanda Vernon, Marian Chau, Seana Walsh, Kevin Enriques, Warren Wagner,

Vicki Funk, Butch Hasse with the Moloka’i Land Trust, Bill ‘Wili’ & Audrey Garnett and Russell Kalstrom with The Nature Conservancy. Many need to be thanked for fieldwork which I was not able to fund in locations across the Pacific including Jean-

Yves Meyer at the French Polynesian Delegation for Research (Délégation à la

iii

Recherche) for providing silica dried material of Marquesan species, Art Whistler for

Samaoan species, Gerald McCormack for Cook Islands specimens, Matt Prebble and

Michael Keihn for New Guinean species, Tod Steussy for samples of Juan Fernández

Islands specimens and Adrienne Markey for Tasmanian and Australian Specimens.

A special thanks needs to be extended to my dissertation committee members:

Tom Ranker, Cliff Morden, Vicki Funk, and Lyndon Wester for all of their help with the organization and issues that arose throughout the period of research. I am overly grateful for the guidance of Dr. Sterling Keeley, my dissertation chair person. Without her help, guidance and support, I’m not sure that I would have been able to navigate the time spent as a graduate student. I also am very grateful for many insightful conversations and support from Timothy Gallaher as we were studying very similar evolutionary questions and it enabled us to often work cohesively together to learn phylogenetic software and

DNA protocols. Many thanks to my botanical based friends for moral support and other wise: Marian Chau, Seana Walsh, Joanne Birch, Kim Peyton, Rachael Wade, and Jessica

Neumann. There are many others who were instrumental in helping with my dissertation and I am very grateful for their help.

ABSTRACT

This dissertation investigated the evolutionary history and historical biogeography of Coprosma (Rubiaceae) across the Pacific Ocean. A brief introduction to the status of

Pacific biogeography of angiosperms is discussed in Chapter 1 along with an introduction to the systematic relationships and previous research of the genus Coprosma itself.

Chapter 1 ends with a description of the dissertation aims and hypotheses. The first research based chapter is Chapter 2, which aimed to elucidate the evolutionary history to the Hawaiian Islands using molecular phylogenetic analyses followed by an assessment of phylogeographic patterns within the genus. The analyses inferred two independent colonization events of Coprosma to the Hawaiian Islands: one radiation of 12 orange- fruited taxa from the Marquesas Islands or Austral Islands, and a separate independent colonization for one black-fruited species from an unclear origin. Research of Chapter 3 investigated the historical biogeography of the genus across a wide geographic distribution with many more species represented using molecular clocking techniques.

The analyses inferred that Coprosma diverged from Nertera in New Zealand during the

Oligocene. Subsequent diversification of New Zealand taxa was correlated temporally with tectonics of the Miocene. It was also inferred that at least 30 bird mediated long distance dispersal events, primarily within the last 10 Ma, have occurred in order to explain the extant distribution of the genus. Chapter 4 investigated the evolutionary inheritance of morphological traits, which were traced onto phylogenetic reconstructions of the genus. Resulting from this work, a key and new taxonomic description that includes six newly circumscribed subgenera was completed. Chapter 5 synthesizes the

v dissertation research which has elucidated a complex novel evolutionary history of

Coprosma which is one of the most widespread and species rich genera in the Pacific.

TABLE OF CONTENTS

Acknowledgements…………………………………………………………………....…iii Abstract………………………………………………………………………….…..……v List of Tables………………………………………………………………………….….x List of Figures…………………………………………………………………………....xi

CHAPTER 1: INTRODUCTION…………………………………………………….…1

Patterns of Pacific biogeography…………………………………………..….…....1 Geographical distribution of Coprosma…………………………………...... 6 Previous biogeographic hypotheses on Coprosma………………………...... 7 Biogeography of the Hawaiian Islands and Coprosma………………………...... 8 Other remote Pacific localities and Coprosma…………………………………....10 The systematic relationships of Coprosma………………………………………..11 Tribe Anthospermeae……………………………………………..………....11 Subtribe Coprosminae……………………………………………….……....12 Closely related genera to Coprosma……………………………...………....13 The genus Leptostigma……………………………………………………....13 The genus Normandia.……………………………………………………....14 The genus Durringtonia.…………………………………………….……....15 The genus Nertera….………………………………………………..……....15 Morphological circumscription of Coprosma…………………………………...... 17 Floral and fruit morphology………………………………………………....17 Leaves and interpetiolar stipules…………………………….……………....18 Secondary Growth and Habit……………………………………………...... 19 Current systematic treatment of Coprosma…………………………………..…...20 Dissertations aims and hypotheses………………………………………...... …....21 Aim A. To investigate the origins of Hawaiian Coprosma taxa...……….....22 Aim B. To understand the historical biogeography of Pacific Coprosma.....23 Aim C. To understand the morphological relationships among Coprosma taxa and their evolutionary implications……...………………………………….23

CHAPTER 2: BIOGEOGRAPHIC INSIGHTS ON PACIFIC COPROSMA (RUBIACEAE) INDICATE TWO COLONIZATIONS OF THE HAWAIIAN ISLANDS...... 25

Abstract…………….……………………………………………...…………….....25 Introduction…………….……………………………………………………….....26 Materials and Methods………………………………………………………….....30 Sampling…………….……………………………………………...…….....30 DNA extraction, amplification and sequencing…………….……………....31 Phylogenetic and biogeographic analyses…………….…….……………....36 Results…………….…………………………………………………………….....38

vii

Maximum Likelihood and Bayesian Inference Analyses..……….………...38 Biogeographic Relationships………………..……………………………....40 Discussion…………….…………………………………………..…………….....44

CHAPTER 3: BIOGEOGRAPHY AND EVOLUTIONARY DIVERSIFICATION OF INSULAR COPROSMA (RUBIACEAE): ONE OF THE LARGEST AND MOST WIDELY DISTRIBUTED GENERA IN THE PACIFIC OCEAN…………………………………………...... 51

Abstract…..…………………………………………………….……………….....51 Introduction…………………………………………………….……………….....53 Materials and Methods………………………………………………………….....59 Sampling………………………………………………………………….....59 DNA extraction, amplification and sequencing……………………….…....59 Phylogenetic and morphological analyses……………………….……….....60 Ancestral state reconstructions………...………………………………….....62 Molecular dating analyses…………………..…………………………….....63 Results…………………………………………………………………….…….....64 Phylogenetic Analyses…………………………...……………………….....64 Geographic origin(s) of Coprosma s.s. and non-New Zealand Pacific Taxa………………………………………………………………...... 65 Morphological trait evolution in dispersal and establishment………...68 Reproductive Strategy……………………...……………………….....68 Habit………………………………………..……………………….....68 Fruit Color...………………………………..……………………….....68 Discussion……………………………………………………………………….....70 Character evolution aiding dispersal and establishment of Coprosma s.s.....70 Historical biogeography of Coprosma s.s. ……………………………….....73 Time, origin, and extensive radiation in New Zealand……..…………….....73 Dispersals from New Zealand to elsewhere…...…………………………....78 Conclusions……………………………………….…………………………….....81

CHAPTER 4: MORPHOLOGICAL, ECOLOGICAL AND MOLECULAR SUPPORT FOR NEW SUBGENERIC RELATIONSHIPS OF COPROSMA (RUBIACEAE)...... 83

Abstract…...……………………………………….…………………………….....83 Introduction……………………………………….…………………………….....85 Taxonomic History……………………….…………….…………...…………...... 85 Coprosma and uncertain position of allied genera……….……………….....88 Reproductive and vegetative characters of Coprosma……………………...89 Floral and Fruit morphology………….…………………………….....89 Leaves and Interpetiolar Stipules……………………….……….….....90 Secondary Growth and Habit……………………….…………...….....91 Methods……………………….……………………………………..………….....92 Taxonomic sampling……………………….……………………………...... 92

viii

Molecular data and phylogenetic analyses……………………….………....93 Morphological and taxonomic data………………………………………....95 Evolution of morphological characters……..…………………………….....98 Results…………………………………………….…………………………….....98 Polyphyly of subgeneric and sectional divisions of Heads (1996) ...……....99 Vegetative and chromosome number evolution…………………………..…….....99 Reproductive character evolution………………………………………………...100 Dioecy……………………….……………..………………….…………...100 Inflorescence structure…………………………………...………………...101 Fruit Color…………………….…………….……………………………...101 Male floral characters……………………….……………………………...103 Increase in flower part number……………………….……………...103 Discussion……………………….………….……..……………………………...105 Taxonomic implications………………………………….………………...106 A key for Netera, Leptostigma and subdivisions of Coprosma..………….108 Taxonomic circumscription of subgenus Coprosma Heads…...……….....109 Taxonomic circumscription of subgenus Pseudocuneatae (Allan) J. Cantley stat. nov………………………………………….……………..………….110 Taxonomic circumscription of subgenus Parviflorae (Allan) stat. nov. J. Cantley…………………………………………………………………….111 Taxonomic circumscription of subgenus Pumilae (Allan) J.T. Cantley stat. nov…………………………………………………………………………112 Section Pumilae (Allan) J.T. Cantley comb. nov………………...... 113 Section Putidae Allan………………...……………………………...114 Taxonomic circumscription subgenus Acerosae J.T. Cantley stat. nov…..115 Section Acerosae (Allan) J.T. Cantley comb. Nov…………...…….116 Section Petiolatae (Allan) J. Cantley comb. nov………...……...….117 Section Spathulatae (Allan) J. Cantley stat. cons…...……..………..119 Taxonomic circumscription of subgenus Lucidae (C. J. Webb) J.T. Cantley comb. nov. ……………………………………………………..……….....120 Section Australes J. Cantley nom. nov……………….………...... 122 Section Polynesica J.T. Cantley nom. nov………..……….……...... 122 Section Grandiflorae J.T. Cantley nom nov………………………...124 Section Lucidae J.T. Cantley nom. nov……………...……………...125 Conclusions……………………….……………………………………………...125

CHAPTER 5: CONCLUSION……………………….……………………………...127

Aim A. To investigate the origins of Hawaiian Coprosma taxa………………...127 Aim B. To understand the historical biogeography of Pacific Coprosma……...130 Aim C. To understand the morphological relationships among Coprosma taxa and their evolutionary implications……………………….………………….....133 Synthesis…………………………………………………………………………134

LITERATURE CITED……………………………………………………………….136

LIST OF TABLES

Table. 2.1. Voucher information and GenBank accession numbers……………….…...32

Table 2.2. Number of characters and parsimoniously informative characters (PIC) for each gene region and the combined data………………………………...………………39

LIST OF FIGURES

Figure 2.1. Map of the Pacific Ocean indicating selected island archipelagos and continental landmasses discussed in the text……..……………………...………………28

Figure 2.2. Examples of reproductive and vegetative characters in Hawaiian and closely related Coprosma species. …………………………….………………...………………29

Figure 2.3. Phylogenetic reconstruction of Bayesian Inference showing a 50% consensus topology based on two nuclear (ITS and ETS) and plastid (rps16) DNA region.………41

Figure 3.1. Map of landmasses of the Pacific. ………………………………….………54

Figure 3.2. Vegetative and reproductive morphology of Coprosma and Nertera………57

Fig. 3.3. Phylogenetic reconstruction of Bayesian Inference showing a 50% consensus topology based on two nuclear (ITS and ETS) and two plastid (rps16 and trnQ) DNA regions……………………………………………….………...…………………………66

Figure 3.4. . Character states of Coprosma and outgroups traced onto the Bayesian 50% phylogenetic reconstructions………………………………...…………………..………69

Figure 3.5. Cladogram from BEAST 1.7.5 analyses……………………………………77

Figure 4.1. Phylogenetic reconstruction of a 50% Bayesian Inference topology indicating taxonomic affiliations of Heads (1996) and the new sectional classifications..…………97

Figure 4.2. Vegetative morphological characters reconstructed on the 50% Bayesian Inference topology………………………………...……………………………………100

Figure 4.3. Reproductive and karyological characters reconstructed on the 50% Bayesian Inference topology. ………………………………...……………………………..……102

Figure 4.4. Reproductive characteristic of the male inflorescence reconstructed on the 50% Bayesian Inference topology……………………………………...………………104

CHAPTER 1

INTRODUCTION

There are over 110 species of Coprosma (Rubiaceae) on islands throughout the

Pacific, but the relationships among species and their biogeographical histories are poorly

understood. The purpose of the research conducted here was to develop a framework in

which the evolution and historical biogeography of Coprosma could be elucidated more

clearly. The introduction is organized into sections, which first introduces biogeographic patterns of Pacific organisms with a special focus on Coprosma. Following this, a morphological circumscription of Coprosma and related genera is discussed highlighting the limitations morphology has on the interpretation of species level relationships within

Coprosma, and the need for a more rigorous investigation in order to understand morphological relationships among species of the genus.

Patterns of Pacific biogeography

Our knowledge of Pacific wide biogeographic patterns of lineages is limited at best, but there are many recent insights aiding our understanding. For instance, the biogeographic origins of the Hawaiian Islands (Wagner & Funk, 1995; Keeley & Funk,

2011) and New Zealand (Winkworth et al., 2002) are well studied, but other insular

localities, such as Samoa or the Marquesas Islands of French Polynesia are much less

well investigated. Pacific wide, a comparison of biogeographic patterns of many lineages

has not yet been produced (Muellner et al. 2008) and this is perhaps because few molecular phylogenies exist for Pacific taxa and even fewer of these lineages are confidently dated.

Despite a limited knowledge, a number of major findings are clear. Previous

research has focused on questions such as the number of colonizations to an insular

locality, mode of dispersal, proximity of source to sink, and biological factors—such as

breeding system type—that aid in establishment of a lineage in a new archipelago. For

many lineages, in order to understand these factors, phylogenetic reconstructions

generated from DNA sequence data are required. For instance, among lineages in which

molecular phylogenies were constructed, much research has focused on the number of

inferred dispersals to a given location. These ‘repeated dispersals’ are well documented

for the Hawaiian Islands where many genera have arrived from two or more independent

lineages. For example, two independent colonization events are inferred for Euphorbia

(Yang & Berry, 2011), and Santalum (Harbaugh & Baldwin 2007; Harbaugh, 2008). In

the case of Hawaiian Rubus, two colonizations were unexpectedly inferred, as the species

are quite similar morphologically. It was not until these two species were placed into a

molecular phylogenetic framework that their different evolutionary history was

understood (Howarth et al., 1997; Alice & Campbell, 1999). More than two colonizations are known to occur, but this is infrequently documented among lineages. Currently, they

are known for only Scaevola (three: Howarth et al., 2003; Howarth & Baum 2005) and

the fern genus Dryopteris (five: Geiger & Ranker, 2005). Although with less

documentation, repeated dispersals are occur elsewhere in the Pacific, such as for both

Coprosma and Metrosideros twice to Lord Howe Island (Papadopulos et al., 2011),

Melicope twice to the Marquesas Islands (Harbaugh et al., 2009), and Santalum twice to

the Hawaiian Islands, twice to the Bonin Islands, and twice to the Austral Islands

(Harbaugh & Baldwin, 2007).

Repeated dispersals to the same locality can be referred to as a ‘biogeographic

sink’ but some areas also serve commonly as a ‘biogeographic source.’ For instance,

three independent colonizations of Bidens from North America are inferred, one each to the Hawaiian Islands plus French Polynesia, to Starbuck Island of the Line Islands, and to

Socorro Island (V. Funk, pers. comm.). A similar situation occurs for Astelia, where New

Zealand was a biogeographic source area for two Pacific lineages of Astelioid taxa: an eastern Pacific and a separate western Pacific lineage (Birch et al., 2013). Extrapolating this pattern on a much larger scale, continents often act as a biogeographic source. Such is known for the Hawaiian Islands where the origin of the flora is an assemblage of many lineages arriving from the Americas, Asia, Australia and from other oceanic localities of the Pacific (Funk & Wagner, 1995; Keeley & Funk, 2011).

In order for organisms to arrive from a biogeographic source area to a sink, long distance dispersal must occur. Most Pacific islands, which are the product of de novo volcanic activity, are separated from other localities by thousands of kilometers and were never connected to each other (Clague et al., 2010). Therefore, some mechanism of long distance dispersal must explain their distributions across the Pacific. Vicariance is largely irrelevant to the biogeographic histories of Pacific wide lineages. One commonly inferred pattern of long distance dispersal is the distribution of organisms from source to sink area, utilizing a number of intermediate areas as stepping-stones. Stepping stone dispersal receives much attention as a common explanation for the distribution of Pacific organisms, but direct dispersals without intervening archipelagos are also indicated.

Stepping-stone dispersals are known for many genera, such as in Melicope (Harbaugh et al. 2008), Metrosideros (Wright et al., 2001; Percy et al., 2008), Tetramolopium (Lowry,

1995) and Santalum (Harbaugh & Baldwin, 2007). Wright et al. (2001) suggest that the

Hawaiian lineages of Metrosideros originated in New Zealand, but utilized the

Marquesas Islands as a stepping-stone prior to reaching the Hawaiian Islands via a second long distance dispersal event. Conversely, direct dispersals from distant locations without intervening stepping-stone archipelagos are documented in Astelia (New Zealand to the Hawaiian Islands; Birch et al., 2013) and Hesperomannia (Africa to the Hawaiian

Islands; Kim et al., 2007). However, even as direct dispersals are certainly possible, one needs to remain cautious in their inference as extinct taxa might represent an intermediate locality of historical distribution.

The distribution of organisms across the Pacific is directly influenced by biological factors concerning dispersal and establishment limitations. The three major mechanisms aiding dispersal of angiosperms among archipelagos are wind, water or internally or externally by birds (Carlquist, 1967). Molecular phylogenetic reconstructions indicate wind dispersal was important in the establishment of

Tetramolopium (Lowry, 1995) and Metrosideros (Wright et al., 2001; Percy et al., 2008), on archipelagos across the Pacific, which correlates with their small dispersal units that are easily carried in the air. Wright et al. (2001) suggest wind patterns of the Pacific are a possible explanation for the dispersal event that establish Metrosideros in the Hawaiian

Islands from the Marquesas Islands. A similar pattern is found among organisms in which long distance dispersal is mediated by water currents. For example, the Pacific distribution of Pandanus (Tim Gallaher, pers. comm.) and Hibiscus tiliaceus (Takayama et al., 2006) are at least partly the result of oceanic currents, which have distributed floating diaspores to predictable locations. The most common dispersal mechanism is

thought to be that of bird-mediated dispersal (Carlquist, 1967), and such is the hypothesized mechanism for the widespread species of Pacific Coprosma.

The biogeographic patterns of bird dispersed lineages, have a less predictable pattern than that of wind or water dispersed lineages. For instance, few migratory birds exist between remote insular localities and many of these are sea birds with limited contact to lineages not adapted to coastal/maritime environments. More often, it is thought that lineages that are bird dispersed arrive to a novel location via unusual weather conditions, which perhaps force birds to new locations. In theory, this makes it harder to predict a bird dispersed lineage’s biogeographic pattern than a wind or water dispersed lineage which largely rely on predicable air and water currents such as typhoons or large storms.

In the case of Coprosma, all dispersals of the Pacific are thought to be mediated by internal ingestion by birds to new insular localities because of their fleshy fruit, but few biogeographical relationships among these species were ever postulated. Despite the fact that Coprosma is one of the most widespread and species rich genera in the Pacific, the pathways of radiation and the factors favoring its movement across the Pacific are unknown. To date, phylogenetic hypotheses for Coprosma are based only on morphology characters, but the morphology of Coprosma is complex and, as Gardner (2002) argues, incompletely assessed. The generation of a dated molecular phylogeny from this dissertation will aid in understanding the relationships of Coprosma taxa across the

Pacific. An implication of understanding Coprosma relationships influences our knowledge of the wider scale patterns of Pacific angiosperm lineages.

Geographical distribution of Coprosma

The genus Coprosma is the largest genus of the Tribe Anthospermeae:

(Rubiaceae) containing over 110 species with a geographic center of diversity in New

Zealand (Oliver, 1935; Heads, 1996). In Oceania, Coprosma is one of the most species rich genera and species are located on nearly every major island archipelago of the

Pacific, east to west from the Juan Fernández Islands near Chile to Borneo, and south to north from the subantarctic islands of New Zealand to the Hawaiian Islands. Its centre of diversity and hypothesized origin is New Zealand where there are at least 55 species

(Oliver, 1935; Gardner 2002b) and three taxonomic entities, which are not formally described (Eagle 1982; Glenny et al., 2010). New Zealand represents the primary center

of diversity and the taxa are distributed from coastal to alpine. The highest amount of this

taxon diversity is on the South Island where many species are restricted to montane and subalpine habitats (Wardle, 1991). A secondary centre of diversity is recognized in the

Hawaiian Islands where there are 13 species (Wagner et al., 1999). New Guinea may also

represent a secondary diversity center, with species forming a large part of the subalpine montane flora, but the number of taxa is disputed (5-15 species; Gardner 2002a;

Utteridge, 2002). Smaller concentrations of species occur in the Marquesas Islands (6 species; Wagner & Lorence, 2011), Lord Howe Island (5 species; Papadopoulos et al.,

2011) and numerous other locations across the Pacific. Australia has 8 species

(Thompson, 2010), but this area is not considered to be an important area of species diversity as taxa are closely related to those in New Zealand and are thought to be unrelated to one another. Few species of Coprosma are widespread, andmany are single island endemics. Coprosma forms a significant part of many plant communities in New

Zealand, New Guinea, Lord Howe Island and in the Hawaiian Islands, but it is of only local importance on other islands. Notably, Coprosma is absent from New Caledonia, but the outgroup taxon Normandia neocaledonica occurs here (van Balgooy, 1966), perhaps representing a case of niche competition, which has prevented the establishment of

Coprosma.

Previous biogeographic hypotheses on Coprosma

Heads (1996) provided a biogeographic hypothesis that Coprosma consists of two subgenera divided along an east-west boundary from an ancient vicariant split of a previously widespread Pacific centered taxon. The delineation of his two subgenera was based on morphology where species with large leaves and complex inflorescences were of subgenus Lucidae and small leaved species with solitary female flowers were of subgenus Coprosma. Species of subgenus Coprosma are found in New Zealand, New

Guinea, Australia and high islands of the Indo-Pacific region. Subgenus Lucidae overlaps with species in New Zealand, but is otherwise found throughout the remote Pacific where subgenus Coprosma is not found. Alternatively, Gardner (2002a) argued that the observed east-west pattern of Coprosma might reflect a biological reality rather than an ancient vicariant event as proposed by Heads (1996). Gardner (2002a) noted that most species of subgenus Coprosma are adapted to montane temperate environments and these species are found in all areas with these habitats. As it stands, there is no available montane habitat in the western distribution of the genus (where subgenus Lucidae is primarily found) as many of the islands such as Samoa, Fiji and the Marquesan Islands are not high enough currently to support a montane subalpine flora. However, the

Hawaiian Islands do reach high elevatons and the species C. ernodeoides of subgenus

Coprosma is found in the respective montane habitats found these islands. Gardner

(2002a) therefore refuted the subgeneric classification of Coprosma by rejecting the rationale of Heads (1996). However, even though Gardner (2002a) believed the conclusions of Heads (1996) were based on an incomplete analysis of characters, he did recognize that an east-west division could in fact be a reality at a lower taxonomic level for some sections of the genus.

Apart from the biogeographic discussion of Coprosma by Heads (1996) and

Gardner (2002a), few hypotheses were formulated for the specific relationships of

Coprosma taxa among islands of the Pacific. Morphological characters that are particularly useful for the delimitation of New Zealand species maintain a high level of overlap among taxa occurring elsewhere and confounds a clear interpretation of the biogeographic history. Because of this, a molecular phylogeny is needed in order to test these hypotheses regarding the relationships within Coprosma taxa.

Biogeography of the Hawaiian Islands and Coprosma

Although the overall biogeographic relationships among Pacific Coprosma are not well defined, significant morphological and habitat differences among the Hawaiian species of Coprosma are the cause for two (Wagner et al., 1999) or three (Fosberg, 1948) colonization events hypothesized to have occurred to the archipelago. Twelve of the 13

Hawaiian species share a similar habit (sparingly branched shrubs or small trees), large leaves, tubular stipules, 5(-6)-merous narrowly campanulate to funnelform corollas, axillary inflorescences, red to orange drupes (Oliver, 1935; Wagner et al. 1999), and a

chromosome number (2n) of 44 (Skottsberg, 1953). Species in this group are found in

mesic to wet forests (sometimes in bogs and sub-alpine woodlands) at altitudes ranging

from 300-2290m (Wagner et al., 1999). As these orange-fruited Hawaiian species share

morphological traits with all species of the Marquesas Islands and some New Zealand

species, the 12 Hawaiian species are thought to be a monophyletic lineage and the result

of one colonization event (Oliver, 1935; Heads, 1996; Gardner 2002b). Fosberg (1948)

suggested, however, that the orange-fruited Hawaiian species resulted from two separate

colonizations as C. elliptica and C. montana have solitary axillary flowers, (or

occasionally two flowers per inflorescence in C. montana) which differs from the

numerous flowers per female inflorescence in other orange-fruited Hawaiian species. In contrast to the orange-fruited species, Oliver (1935) and Fosberg (1948) agree that the

thirteenth Hawaiian species, C. ernodeoides, represents an independent colonization from

the orange fruited species, as it is a prostrate shrub with small leaves without apparent

secondary venation, connate stipules, 4-merous long-tubular corollas, a terminal solitary

inflorescence, shiny black drupes and has a chromosome number (2n) of 220 (Skottsberg,

1953). In addition, C. ernodeoides is a pioneer of open canopy habitats on recent lava

flows occurring only at high elevations on east Maui and Hawai`i islands. Based on morphological treatments, C. ernodeoides was allied more closely to C. archboldiana from New Guinea, C. pumila of Tasmania (Oliver, 1935; Merrill & Perry, 1945), C. talbrockiei (New Zealand) and C. moorei of Tasmania (Heads, 1996) rather than to the other twelve Hawaiian species.

Other remote Pacific localities and Coprosma Many high islands of the remote Pacific contain at least two species of Coprosma.

These include Samoa, Rapa Iti Island of the Austral Islands, Norfolk Islands, Kermadec

Islands, and the Juan Fernández Islands (Oliver, 1935; Heads, 1996; Gardner 2002a).

Insular localities with only a single species known include Bougainville Island, Pitcairn

Island, Vanuatu, Fiji, and Cook Islands (Heads, 1996; Gardner 2002a). While not a

diversity center of the genus like the Hawaiian Islands, notable radiations occurred on

three other Pacific localities. The Marquesas Islands in the northern French Polynesia

have six species of Coprosma (Wagner & Lorence, 2011). These species are thought to

be the result of a single dispersal from an unknown locality (Wagner, 1990), but they

share morphological similarities with orange-fruited Hawaiian and some New Zealand

taxa (Heads, 1996; Florence & Lorence, 1997; Gardner, 2002b). Elsewhere in French

Polynesia, an assumed separate radiation to—unrelated to Hawaiian species—are taxa occurring on the Society Islands, Pitcairn Islands, Austral Islands and Cook Islands. The

relationships among species are poorly understood, largely due to their remote localities

and few collections of these species (Merril & Perry, 1945; Heads 1996). The five species

from Lord Howe Island were studied extensively. Papadopulos et al. (2011) indicate at

least two independent colonizations of Coprosma have established the genus on the

island. Furthermore, this study suggests sympatric speciation was highly probable

explanation of speciation mechanism for these species. The speciation of Lord Howe

Island Coprosma—along with Lord Howe endemic Metrosideros (Myrtaceae) and Howea

(Arecaceae)—among the few genera for which sympatric speciation is documented.

The systematic relationships of Coprosma

Tribe Anthospermeae

The genus Coprosma is one of 12 genera of the tribe Anthospermeae (Rubiaceae) as circumscribed by Puff (1982). Tribe Anthospermeae is atypical in distribution in comparison to other Rubiaceae tribes, as it is a southern temperate radiation, whereas the family is generally pan-tropical (Bremekamp 1966, Puff 1982, Robbrecht 1988). Tribe

Anthospermeae is divided into three distinct subtribes: Anthosperminae of Africa,

Opercularinae of Australia, and Coprosminae in Oceania and South America (Puff,

1986). The subtribe Coprosminae, of which Coprosma is the largest member, has primarily diversified in New Zealand, New Guinea, Australia and insular localities of the remote Pacific Ocean (Anderson et al., 2001; Heads, 1996; Puff 1982, 1986). Two genera, Nertera and Leptostigma extend the range of subtribe Coprosminae (and tribe

Anthospermeae) to South America, the Greater Antilles and Tristan da Cunha Islands of the southern Atlantic Ocean (Heads, 1996).

Like the atypical distribution of tribe Anthospermeae, it is also atypical in floral morphology in comparison to other members of the family. Primarily, taxa of Rubiaceae have flowers that are entomophilous and therefore the flowers are brightly colored, tubular, bisexual producing copious amounts of nectar, and often arranged in dense clusters (Puff, 1986; Puff & Robbrecht, 1988). However, the floral morphology of taxa of tribe Anthospermeae is representative of wind pollination and flowers are actinomorphic, not brightly colored, odorless, lacking nectaries and relatively inconspicuous (Puff,

1982). Anthospermeae corollas are generally shortly tubular with styles usually occurring in pairs of two that are deeply divided to the base and covered in receptive papillae. Male

filaments are pendulous, and often several times longer than the corolla, which helps to

facilitate pollen exchange via wind (Puff, 1982, 1986). Fruit of Anthospermeae usually

consist of two pyrenes surrounded by mesocarp and exocarp. Fruits can be dry (subtribe

Anthosperminae and subtribe Opercularinae) or fleshy (subtribe Coprosminae). Fleshy

fruits are well developed in subtribe Coprosminae, especially in Coprosma,

Durringtonia, and Nertera; Normandia and Leptostigma fruits are often described as

semi-fleshy (Puff 1982, 1986; Gardner, 2002).

Subtribe Coprosminae

The subtribe Coprosminae consists of five genera primarily centered in the

Australasian-South Pacific region (Fosberg, 1982). Three genera—Coprosma,

Leptostigma, and Nertera—have numerous species and large distributions that overlap in

many areas. Conversely, Normandia of New Caledonia and Durringtonia of northern

New South Whales and Queensland, Australia are monotypic representing unique distributions for subtribe Coprosminae. Morphologically, members of Coprosminae are separated from other subtribes by their fleshy or semi-fleshy fruit, as opposed to dry fruits. The fleshy mesocarp of Normandia and Leptostigma, which reduces and become dry at maturity is termed semi-fleshy (Fosberg, 1982; Puff 1982). Fleshy fruit of the

Coprosma, Durringtonia, and Nertera is much more well developed. Taxa of subtribe

Coprosminae range from herbaceous perennials, as in Durringtonia, Leptostigma, and

Nertera, to perennial shrubs for Normandia and Coprosma. The greatest diversity of growth forms can be found among the species of Coprosma, which exhibit trailing, mat- forming, shrub, and large tree habits. Many New Zealand species are furthermore highly

divaricate, a condition in which branching occurs at wide angles as to cause an

interlocking mass of vegetation (Gardner 2002; Oliver, 1935).

Closely related genera to Coprosma

The genus Leptostigma

The genus Leptostigma consists of seven species from Australia (2), New Zealand

(1) and South America (4) (Fosberg, 1982; Thompson, 2010). Species of South America

are generally restricted to small disjunct populations in Andean Cordilleras of Bolivia,

Chile, Ecuador and Colombia (Fosberg, 1982). Habitat in Australasia consists of

temperate sclerophyll forest of Southeastern Australia (James 1992) and moist herblands

and grassy turflands of New Zealand (Johnson & Brooke, 1989). Leptostigma is closely

associated with Nertera in which it shares a creeping rhizomatous herbaceous habit,

small leaves and bisexual flowers, and superficially appears very similar (Fosberg, 1982;

Puff 1982, 1986; Gardner 1999). However, Leptostigma is distinguished by the presence

of well-developed and persistent calyx lobes, non-sheathing stipules with glandular

colleters, elongate, tubular corollas, strongly protandrous flowers, far-exerted anthers and semi-fleshy green-brown fruits (Fosberg 1982, Cheeseman 1925, Gardner 1999a). With a chromosome number of 2n=40, Leptostigma is different from other genera of subtribe

Coprosminae where 2n=44 in Nertera and 2n=44-220 in Coprosma (Gardner 1999). The

chromosome number of Normandia and Durringtonia are not known.

The genus Normandia

The genus Normandia is monotypic consisting of only Normandia neocaledonica.

The genus is restricted to New Caledonia and is found on ultramaphic soils on the island

of Grand Terre (Puff, 1982). In its habit, the species is a low shrub to small tree with

coriaceous leaves and hermaphroditic flowers, but unisexual flowers are known (Hooker,

1982; Puff 1982). Flowers are 5-merous, with a longly tubiform corolla with five short

lobes, and a persistent tubular calyx. Inflorescences are large, dichotomously branched

terminal corymbs (Hooker 1872). The fruit of Normandia is semi-fleshy, like that of

Leptostigma, and is crowned by a persistent calyx. Normandia is the only taxon of

subtribe Coprosminae in New Caledonia, despite suitable habitat clearly exists here for

other members. van Balgooy (1966) noted the particular absence of Coprosma from New

Caledonia resulting in a ‘lacuna’ in its otherwise pan-Pacific distribution. Hooker (1873) was enamored with Normandia neocaledonica describing the genus as having the habit of

Psychotria, flowers of Coprosma and the fruit of Spermacoce thus combining characteristics of three distantly related Rubiaceae genera. The relatedness of Normandia to other members of the subtribe Coprosminae was previously disputed. In its initial description, Hooker (1872) indicated it scarcely differed from Coprosma, differing only in the habit and characters of the flower and fruit. Baillon (1880) argued these genera were distinct and later Fosberg (1982) omitted Normandia from the subtribe

Coprosminae altogether. Normandia was moved into subtribe Coprosminae by (Puff,

1982), as he believed it was closely allied with Coprosma.

The genus Durringtonia

Durringtonia paludosa is a monotypic genus like that of Normandia, but was

discovered much later in 1985 in Queensland, Australia. Durringtonia was initially

placed into its own subtribe, the Durringtonieae, but was noted to have close affinities to

both Anthosperminae and Coprosminae subtribes (Henderson & Guymer 1985, Puff

1986). Based on its geographic distribution in Australia and morphology, it is now considered to belong to the Coprosminae (Puff & Robbrect, 1988; Heads, 1996).

Durringtonia is a rare perennial herb endemic to freshwater peat swamps between northern New South Whales and southern Queensland, Australia (Henderson & Guymer

1985). Aerial stems are to 1m in height and are distinguished from other genera in the

Coprosminae by its single pyrene per fruit, unilocular ovary (where the second carpel is aborted), and a single stigma (Henderson & Guymer, 1985; Puff & Robbrecht, 1988).

The inflorescence is a thrysoid panicle with dioecious flowers. Male flowers have five stamens and a vestigial style and ovary. Female flowers have five corolla lobes, four to six persistent calyx lobes subtended by two bracts, and lack vestigial male organs. The fruit is a succulent, orange-yellow drupe with a single pyrene (Henderson & Guymer

1985).

The genus Nertera

The genus Nertera of 13 species has the largest distribution of any genus of the subtribe Coprosminae extending from Australia, New Zealand, through New Guinea and the Philippines, northwards into southern China and Taiwan. The range further includes

numerous islands of the remote Pacific, as well as South America, the Greater Antilles

and Tristan da Cunha Islands in the Atlantic Ocean (Oliver 1935; Fosberg, 1982; Heads,

1996; Wagner et al., 1999, Gardner, 1999b; Andersson, 2000) This impressive range is

due almost entirely from the range one widespread taxon, Nertera granadensis. At least

five taxa are endemic to New Zealand (Heads 1996, Gardner 1999b). Recent molecular

studies suggest the Asian species N. yamashitae and N. sinensis are dubiously placed in

Nertera (G. Jakubowsky & Michael Kiehn, pers. comm.).

Among the species of Nertera, there is a great deal of morphological uniformity.

All species are herbaceous, often forming trailing mats from extensive rhizomes.

Woodiness of older stems is scarcely discernable, with only and extremely limited

secondary thickening (Gardner, 1999). Species occur in wet to mesic environments from

lowland forest understories to subalpine habitats. They are commonly found in humic

soils, such as decaying logs, living trees, sphagnum seeps, snow tussocks, and cushion

bogs (MacMillan, 1995; Wagner et al., 1999; Gardner, 1999b).

Despite Nertera as placed in the tribe Anthospermeae, which is characterized by

anemophillous taxa, Gardner (1999b) argues that Nertera has floral characteristics

indicative of entomophily or perhaps ‘splash’ pollination based on observations flowers

with short, campanulate corollas and shortly exserted anthers which do not shed pollen

freely. Entomophily has yet to be demonstrated in Nertera, and it seems unlikely given

that a nectary disc is conspicuously absent from the genus (Andersson, 2000). However, a

sweet-tasting nectar-like fluid was observed on the stigmas in both New Zealand and the

Hawaiian Islands (A. Markey & J. Cantley, pers. obs.). The argument for splash

pollination is due to a situation in which flowers open upward and could therefore could

interact with water droplets from above. Nertera flowers are bisexual, 4-merous, with

minute-absent calyx lobes, and a relatively truncate, funnel-form corolla (Cheeseman,

1925; Gardner, 1999a, 1999b). They are solitary, axillary or terminal and sessile

(Cheeseman, 1925; Allan 1961; Gardner, 1999a, 1999b). The fleshy fruit is generally

orange-red in color (Gardner, 1999b), but a variant of Nertera depressa in New Guinea is

known to occasionally have black fruit (van Royen, 1983).

The position of Nertera in relation to Coprosma and Leptostigma has been a subject of debate. Heads (1996) sunk Nertera and Leptostigma to sections of Coprosma

arguing that they scarcely differ from Coprosma except for that of their herbaceous habit

and bisexual flowers. Although similar in appearance, Nertera is different from

Leptostigma in its succulent fruit, and weakly protogynous flowers. Leptostigma fruits

are semi-fleshy and flowers are strongly protogynous decreasing the probability of

selfing (Cheeseman, 1925; Gardner 1999a). Nertera is further distinguished from

Coprsoma in its hermaphroditic flowers, possible different pollination mechanism,

herbaceous habit and chromosome number (Oliver, 1935; Gardner 1999a).

Morphological circumscription of Coprosma

Floral and fruit morphology

Coprosma is generally dioecious, however, C. moorei and C. talbrockiei are

strictly bisexual. Nevertheless, at least 28 species are documented to exhibit leaky-

dioecy, where there is an occasional production of bisexual flowers (Oliver, 1935; Moore

& Mason, 1974; van Royen, 1983; Utteridge 2002, pers. obs.). Stamens typically number

four or five, but can be up to ten (Oliver, 1935; van Royen, 1983). The ovary is usually

biloculate with one solitary ovule per locule (Oliver, 1935; Puff, 1982). The calyx is

smaller than the corolla, and can be reduced or even absent in male flowers (Oliver,

1935; Allan, 1961; Markey & de Lange, 2003). Corresponding with anemophily, the

flowers are small and colored dull hues of green, yellow, cream and brown, and often

stippled red or purple. The inflorescences range from large, thyrso-paniculate (thyrsoid)

to considerably reduced inflorescences of solitary flowers (Puff, 1982; Puff & Robbrecht,

1988; Gardner 2002b).

The fruit is a fleshy, globose drupe with two (occasionally three or four) plano-

convex pyrenes (Cheeseman, 1925; Oliver, 1935). The exo- and mesocarp form a

succulent layer around the drupe, which consists of the embryo and testa encased in a

heavily sclerotinized endocarp (Gardner, 2002a). The radical emerges prior to the

cotyledons, and this is achieved by the separation of an apical operculum from the

endocarp (Gardner, 2002a). There are a number of fruit colors (white, blue, black, violet,

yellow, pink, orange, red), but most species are orange-red in color (Oliver, 1935).

Leaves and interpetiolar stipules

Most leaves of Coprosma are simple, entire and coriaceous (Oliver, 1935).

Domatia are present on leaves of many species, a trait that is found in no other genus of

tribe Anthospermeae (Puff, 1982). Not unlike that of other members of the Rubiaceae,

there are generally two leaves per node in Coprosma species, occurring with a pair of

interpetiolar stipules, opposite in phyllotaxy (Oliver, 1935). Occasionally, leaves occur in

whorls of three, which is known for some taxa of the Hawaiian Islands (Wagner et al.,

1999; pers. obs), Marquesas Islands (Wagner & Lorence, 2011), New Guinea (Gardner,

2002b; van Royen, 1983; Utteridge, 2002) and C. talbrockiei of New Zealand (Markey & de Lange 2003). Leaf size is highly variable within the genus and often correlated with habit. For example, micro and leptophyllous leaves are associated with mat forms and

low shrubs whereas mesophyllous leaves are more associated with large shrubs and trees

(Wardle 1991). Variation in leaf characters can also vary within a species, an even within

an individual (pers. obs.). Oliver (1935) notes that leaf variability often is associated with

the growth conditions of individuals of the same species in different microhabitats (ex.

larger leaves tend to occur in area with closed canopies, smaller leaves in exposed

habitat).

The morphology of interpetiolar stipules has received much attention as a reliable

character for the determination of individual species that are otherwise are very similar in

morphology (see Oliver, 1935; Allan 1961; Glenney et al., 2010). Stipules margins range from entire to denticulate, and sometimes form a sheath around the stem. Stipules can have a variety of appendages such as cilia, denticles, or glandular colleters that line the margin and outer surface and are often diagnostic for species recognition (Allan, 1961;

Utteridge, 2002; Markey & de Lange 2003).

Secondary Growth and Habit

Coprosma is distinguished from Nertera and Leptostigma in secondary growth

and woody stems (Oliver, 1935). Within the genus, there are three major growth forms: i)

large leaved, openly branching tall shrubs and trees, ii) small leaved divaricating shrubs

and subshrubs, and iii) prostrate, mat-forming plants. The divaricating habit is restricted

to New Zealand (Oliver, 1935; Heads, 1996). Divarication is a combination of branches opening at wide angles with short internodes that interlace with neighboring branches.

The resulting habit occurs in low shrubs or erect trees in New Zealand and perhaps best described as an extensively interlaced mass of tangled branches (Greenwood & Atkinson,

1977; Tomlinson, 1978; Wardle, 1991). Small leaves less than 25mm in any dimension are notable in species expressing divarication (Kelly, 1994), but many species without divarication also have small leaves (i.e. the mat-forming, sometimes trailing prostrate shrub, C. ernodeoides of the Hawaiian Islands).

Current systematic treatment of Coprosma

Previous treatments on the systematic relationships and taxonomy of Coprosma species were based on morphology (Oliver, 1935; Allan, 1961; Heads, 1996; Gardner,

2002a), karyology (Hair, 1963; Buezenberg, 1983; Dawson, 1995; Dawson &

Buezenberg, 2000), flavonoid chemotaxonomy (Wilson, 1979; 1984), and biogeographic investigations (Heads, 1996). However, as the morphological features of Coprosma are highly plastic and despite a thorough systematic treatment, few reliable characters are able to confidently infer relationships. These treatments subdivided Coprosma into two subgenera based primarily on leaf size, inflorescence structure and habit (Hooker, 1967;

Oliver, 1935; Allan, 1961; Heads, 1996). Oliver (1935) provided the first genus wide treatment, describing seven sections he considered to reflect natural relationships. Oliver

(1935) provided the first hypothesis regarding the morphological evolution of the genus in which he suggested taxa of “simple” morphology were more plesiomorphic in characters than those taxa, which were increased morphological complexity.

Allan (1961) provided the second genus wide treatment for Coprosma, but admittedly followed most of that by Oliver (1935), “for convenience of reference, not from any convictions concerning phylogenetic development.” Allan (1961) deviated from

Oliver (1935) by elevating one of Hooker’s (1867) subdivisions, based on inflorescence

structure and leaf size, to the level of subgenus. The most recent treatment is that by

Heads (1996) who used two subgenera: Coprosma (eight sections and one informal group) and Lucidae (three sections and four informal groups). Based on biogeographical and morphological evidence, the two subgenera of Heads (1996) represented an east to west biogeographic split. The small leaved and solitary flowered species of subgenus

Coprosma represent New Zealand, Australia and Indo-Pacific localities. Conversely, large leaved and complex inflorescences of the subgenus Lucidae are found in New

Zealand and west into the remote Pacific. The small leaved solitary flower inflorescence of C. ernodeioes of Hawaiian Islands was considered to be an outlier for the biogeographical distribution of subgenus Coprosma. Heads (1996) provided some speculations as to the relationships between species in different groups of the subgenera, but did not formally present any sort of phylogenetic hypothesis. However, the species pair C. arborea and C. spathulata was suggested as a clear intermediate morphology between the two subgenera.

Dissertations aims and hypotheses

This dissertation aimed to elucidate the evolution and historical biogeography of the genus Coprosma (Rubiaceae). Phylogenetic reconstructions were performed utilizing chloroplast and nuclear DNA and morphological data from nearly all known species.

Both ancestral state reconstruction and molecular clocking analyses were applied to infer the biogeographical history of the genus. The dissertation research is presented in three chapters, each of which are intended for independent publication in peer reviewed journals. The three broad aims and numerous hypotheses of this dissertation are outlined below.

Aim A. To investigate the origins of Hawaiian Coprosma taxa.

Hypothesis 1: Hawaiian Coprosma are the result of two independent colonization events.

Hypothesis 2: Orange-fruited Hawaiian taxa are monophyletic and closely related to taxa from the Marquesas Islands.

Hypothesis 3: Black fruited Coprosma ernodeoides represents a second colonization to the Hawaiian Islands and is not closely related to the orange-fruited Hawaiian taxa.

Significant morphological and habitat differences among the Hawaiian species of

Coprosma are the cause for two (Wagner et al. 1990, 1999) or three (Fosberg, 1948)

colonization events hypothesized to have occurred to the archipelago. Twelve of the 13

Hawaiian species share a similar habit, large leaves, tubular stipules, 5(-6)-merous

narrowly campanulate to funnelform corollas, axillary inflorescences, red to orange

drupes (Oliver, 1935; Wagner et al. 1999), and a chromosome number (2n) of 44

(Skottsberg, 1953) with Marquesas Islands and some New Zealand species. The

thirteenth Hawaiian species, C. ernodeoides, is believed to represent an independent

colonization as it is a prostrate shrub with small leaves without apparent secondary venation, connate stipules, 4-merous long-tubular corollas, a terminal solitary inflorescence, shiny black drupes and has a chromosome number (2n) of 220 (Skottsberg,

1953). Based on morphological treatments, C. ernodeoides was allied more closely to C.

archboldiana Merill & Perry from New Guinea, C. pumila Hook. f. of Tasmania (Oliver,

1935; Merrill & Perry, 1945), C. talbrockiei Moore & Mason (New Zealand) and C.

moorei Rodway of Tasmania (Heads, 1996) rather than to the other twelve Hawaiian

species.

Aim B. To understand the historical biogeography of Pacific Coprosma

Hypothesis 4: Coprosma is not Gondwanan in temporal origin, but rather is much more recently evolved.

Hypothesis 5: Patterns of Coprosma across the Pacific are similar to patterns expected via bird-mediated dispersal.

Hypothesis 6: Leaky-dioecy, woodiness and fleshy fruit are among the innovations aiding the dispersal and establishment of Coprosma across the Pacific.

It is thought to be unlikely that the distribution of Coprosma across the remote

Pacific is Gondwanan in origin as most islands are volcanic and origin and were not emergent for extended millions of years. Furthermore, the extant diversity of Coprosma in New Zealand is likely no older than ca. 30 Ma as the Oligocene Marie Transgression

(Stevens et al., 1988; Cooper & Millener, 1993; Cooper & Cooper, 1995) of the continent would have significantly reduced the biodiversity of the genus if developed prior. Aiding in dispersal and establishment, fleshy-fruit that are ingested internally by birds, leaky- dioecy and woodiness are thought to have aided Coprosma in achieving its extant distribution as these character states are among the most abundant among other successful Pacific lineages (Carlquist, 1967).

Aim C. To understand the morphological relationships among Coprosma taxa and their evolutionary implications

Hypothesis 7: The evolution of morphological characters shifted from ‘simple’ then increased in complexity.

This hypothesis was first proposed by Oliver (1935), but was never tested using molecular phylogenetic techniques. Studies investigating the morphological diversity of

Coprosma pointedly did not make many specific morphological hypotheses following

Oliver (1935). Furthermore, as no phylogeny was produced previously, understanding the morphological characters that have aided in the establishment of one of the most successful genera across the Pacific was not possible and their evolutionary implications unclear.

CHAPTER 2

BIOGEOGRAPHIC INSIGHTS ON PACIFIC COPROSMA (RUBIACEAE) INDICATE TWO COLONIZATIONS TO THE HAWAIIAN ISLANDS

Abstract

Most archipelagoes of the Pacific Ocean, including the Hawaiian Islands, are volcanic in origin and were never connected to continental landmasses. The derivation of the Hawaiian flora is entirely the result of long distance dispersal and in situ speciation from various source areas including the Americas, Asia, and islands of Oceania. To assess the origins of Hawaiian Coprosma (Rubiaceae), one of the largest and most widely distributed genera across the Pacific, molecular phylogenetic analyses were performed utilizing sequences from internal and external transcribed spacer regions (ITS and ETS) of nuclear ribosomal DNA and the rps16 plastid DNA intron, from which phylogeographic patterns within the genus were then assessed. Our analyses suggest two independent colonization events of Coprosma to the Hawaiian Islands. Twelve of the thirteen Hawaiian Coprosma species are monophyletic and closely related to Marquesas

Islands species plus one species from the Austral Islands. The analyses indicate that

Coprosma ernodeoides is representative of a separate colonization to the Hawaiian

Islands from an uncertain origin, and is closely associated with C. atropurpurea of New

Zealand and C. pumila of Tasmania. Similar to the Hawaiian Islands, the pattern of multiple independent colonization events to a single Pacific locality was also found for six South Pacific localities, as well as Australia. Understanding the origins of Hawaiian

Coprosma adds a new pattern of plant dispersal to our understanding of Pacific biogeography, particularly in reference to multiple independent colonizations to single insular localities.

Introduction

The Pacific Ocean covers nearly half of the surface of Earth and encompasses

over 20,000 islands distributed over millions of km2 of open water (Keast, 1996).

Prominent among these are the Hawaiian Islands, one of the world’s most isolated

archipelagoes where the nearest island archipelago is the Marquesas Islands and nearest

continental landmass is North America, both >3500 km away (Wagner, 1991). These

islands were never connected to another landmass, as they arose de novo from volcanoes

that emerged from the ocean floor (Wagner, Herbst & Sohmer, 1999; Clague et al.,

2010). As a result of this isolation, the derivation of the Hawaii flora is entirely the

product of long distance dispersal and in situ speciation (Carlquist, 1980; Wagner & Funk

1995). Molecular studies over the past ~20 years confirm that the highly endemic

Hawaiian flora is derived from multiple geographic source areas. These include tropical,

temperate and boreal regions of the Americas, Asia, Australasia, other Pacific Islands,

and even Africa (see Baldwin & Wagner, 2010; Keeley & Funk 2011). The Hawaiian

flora also includes many examples of extensive speciation, among them the Hawaiian

silversword alliance (see Carlquist et al., 2003 for a review), Hawaiian mints (Lindqvist

& Albert, 2002; Lindqvist et al., 2003), and Hawaiian lobeliads (Givnish et al., 2004;

Givnish et al., 2009). Each of these radiations is the result of speciation from a single

ancestral colonizing event, but more than one colonizing lineage within numerous genera

has also been reported for Hawaiian endemics (Keeley & Funk, 2011). For example, two

independent colonization events are known for a number of taxa including Euphorbia L.

(Yang & Berry, 2011; Yang 2012), and Santalum L. (Harbaugh & Baldwin 2007;

Harbaugh, 2008). In the case of Hawaiian Rubus L., two colonizations were unexpectedly

inferred as the species are quite similar morphologically (Howarth et al., 1997; Alice &

Campbell, 1999). More than two colonizations have been rarely detected in molecular

phylogenetic analyses where it is known for only Scaevola Lindl. (three: Howarth et al.,

2003; Howarth & Baum 2005) and the fern genus Dryopteris Adans. (five: Geiger &

Ranker, 2005).

The genus Coprosma J.R. Forst. & G. Forst. is one of the largest and most widely

distributed in the Pacific (Fig. 2.1), with 110+ species found east to west from Borneo

across the Pacific to the Juan-Fernández Islands of Chile, and north to south from the

Hawaiian Islands to the subantarctic Islands south of New Zealand (Fig 2.1; Oliver, 1935;

Heads, 1996). Species of the genus are predominately dioecious and wind-pollinated; the fruit are fleshy drupes with colors ranging from red and orange, to white, blue, and black

(Oliver, 1935) and species are typically dispersed by frugivorous birds or lizards (Lord &

Marshall, 2001; Lord et al., 2002). Coprosma has a primary center of diversity in New

Zealand, with over 55 species, and a secondary center in the Hawaiian Islands, with 13 species. New Guinea may also represent a center of diversity; however, Utteridge (2002) and Gardner (2002a) disagree on the number of species, five or c. 15, respectively. The relationships among the Pacific island species are unclear and few hypotheses have been proposed (Oliver, 1935; Heads, 1996). As a consequence, the geographic site of origin of the Hawaiian progenitor(s) and relationships to species elsewhere in the Pacific remain unknown.

Significant morphological and habitat differences among the Hawaiian species of

Coprosma are the cause for two (Wagner et al. 1990, 1999) or three (Fosberg, 1948) hypothesized colonization events occurring to the archipelago (Fig. 2.2). Twelve of the

13 Hawaiian species share a similar habit (sparingly branched shrubs or small trees),

large leaves, tubular stipules, 5(-6)-merous narrowly campanulate to funnelform corollas, axillary inflorescences, red to orange drupes (Oliver, 1935; Wagner et al. 1999), and a

Figure 2.1. Map of the Pacific Ocean indicating selected island archipelagos and continental landmasses discussed in the text. chromosome number (2n) of 44 (Skottsberg, 1953). Species in this group are found in mesic to wet forests (sometimes in bogs and sub-alpine woodlands) at altitudes ranging

from 300-2290m (Wagner et al., 1999). As these orange-fruited Hawaiian species share

morphological traits with all species of the Marquesas Islands and some New Zealand

species, (Fig. 2.2) the 12 Hawaiian species are thought to be a monophyletic lineage and

the result of one colonization event (Oliver, 1935; Heads, 1996; Gardner 2002b). Fosberg

(1948) suggested, however, that the orange-fruited Hawaiian species resulted from two

separate colonizations as C. elliptica W.R.B. Oliv. and C. montana Hillebr. have solitary

axillary inflorescences, (or occasionally two flowers per inflorescence in C. montana)

differing from the numerous flowers per inflorescence in the other orange-fruited

Hawaiian species. In contrast to the orange-fruited species, Oliver (1935) and Fosberg

(1948) agree that the thirteenth Hawaiian species, C. ernodeoides A. Gray, represents an

independent colonization from the orange fruited species as it is a prostrate shrub with

small leaves without apparent secondary venation, connate stipules, 4-merous long-

Figure 2.2. Examples of reproductive and vegetative characters in Hawaiian and closely related Coprosma species. (a) Coprosma ochracea var. kaalae of Oàhu Island, Hawaiian Islands, (b) C. longifolia of Oàhu Island, Hawaiian Islands, (c) C. esulcata of Nuku Hiva Island, Marquesas, and (d) C. cymosa of Hawaiì Island, Hawaiian Islands. All with orange drupes and large leaves indicating some of their shared morphological characteristics. (e) Black-fruited C. ernodeoides of East Mauì and Hawaiì Island, Hawaiian Islands is closely related to (f) dark-red/purple fruited C. atropurpurea of New Zealand (notice female flowers and immature fruits). Photographs by Jason T. Cantley (a- b, d-f) and Steve Perlman (c).

tubular corollas, a terminal solitary inflorescence, shiny black drupes and has a

chromosome number (2n) of 220 (Skottsberg, 1953). In addition, C. ernodeoides is a

pioneer of open canopy habitats on recent lava flows occurring only at high elevations on

east Mau`i and Hawai`i island. Based on morphological treatments, C. ernodeoides was

allied more closely to C. archboldiana Merill & Perry from New Guinea, C. pumila

Hook. f. of Tasmania (Oliver, 1935; Merrill & Perry, 1945), C. talbrockiei Moore &

Mason (New Zealand) and C. moorei Rodway of Tasmania (Heads, 1996) rather than to the other twelve Hawaiian species.

Despite speculation that more than one colonization event has occurred for

Hawaiian Coprosma, no study had previously tested these hypotheses. The present study investigates the origins and relationships of Hawaiian Coprosma by addressing the following questions: 1) How many colonization events of Coprosma were there to the

Hawaiian Islands? and 2) What are the biogeographical relationships of Hawaiian

Coprosma to other species throughout the Pacific? We used DNA sequence data from two nuclear (ITS and ETS) and one plastid region (rps16 intron) to construct the first molecular phylogeny for Coprosma species across the Pacific including the primary center of diversity in New Zealand. Additionally, the size of the genus Coprosma (110+ species) and its extensive geographical range allowed for a broader comparison of biogeographical patterns with those known for other Pacific plant groups.

Materials and Methods

Sampling

Eighty-five taxa representing four genera were sampled: Coprosma, Durringtonia

R. Henderson & Guymer, Nertera Banks & Sol. ex Gaertn. and Normandia Hook. f. Of these, the 80 Coprosma taxa selected represent species from across the entire geographic distribution of the genus except three Pacific localities (Cook Is., Pitcairn Is., and

Bougainville Is.; see Fig. 2.1 for a map of Pacific Island localities), but nuclear DNA of

the internal transcribed spacer (ITS) was obtained for C. laevigata Cheeseman of Cook

Is.

and C. benefica W.R.B. Oliv. of Pitcairn Island. Therefore, separate ‘ITS-only’

phylogenetic analyses were completed including these two species plus an additional

seventeen species for which ITS sequences were available on GenBank (see, Table 2.1

for a complete list of taxa, collection localities, voucher information and GenBank

accession numbers). Durringtonia, Normandia, and Nertera were selected as outgroups based on broader phylogenetic studies of the Rubiaceae (Anderson, Rova & Andersson,

2001; Markey et al., 2004).

DNA extraction, amplification and sequencing

DNA was obtained from herbarium and silica-dried material, and from the

Hawaiian Plant DNA Library (Randell & Morden, 1999). Samples were prepared by

hand grinding leaves with liquid nitrogen in a mortar and pestle followed by extraction

using a DNeasy Plant Mini Kit (Qiagen, Valencia, California, USA). Samples older than

1990 and specimens for which DNA preservation was poor at time of collection due to

difficult field conditions in remote Pacific islands), were difficult to extract and the

QiaAmp DNA Stool Mini Kit was used in these cases, following the manufacturer’s

protocols with modified volumes and with a lengthened initial incubation period of 2 hours at 70˚C.

Table 2.1. Voucher information and GenBank accession numbers. C. = Coprosma; D. = Durringtonia; N. = Nertera; No. = Normandia; Species Collector and Herb- Collection ITS Genbank ETS rps16 Collection Number arium Locality No. Genbank Genbank No. No. C. acerosa A. Markey & J. Cantley HAW Dunedin, New KF688388 KF688305 KF688223 NZ242 Zealand C. acutifolia J. Cantley 227 HAW (Kermadec Is.) KF688389 KF688306 KF688224 cult. New Zealand C.arborea J. Cantley 198 HAW Auckland, New EU169116.1 Zealand C. areolata A. Markey & J. Cantley HAW Auckland, New KF688390 KF688307 KF688225 NZ241 Zealand C. atropurpurea A. Markey & J. Cantley HAW South Island, New KF688424 KF688340 KF688256 NZ206 Zealand C. barbata T. Utteridge 34327 KEW Mt. Jaya, KF688467 KF688382 KF688299 Indonesia C. baueri R. O. Gardner 5901 BISH Anson Bay, KF688464 KF688379 KF688296 Norfolk Is. C. benefica G. McCormack 72/17 CHR Pitcairn Is. KF688473 C. brunnea A. Markey & J. Cantley HAW Arthur's Pass, KF688391 KF688308 KF688226 NZ221 New Zealand C. chathamica J. Cantley 219 HAW (Chatham Is.) KF688392 KF688309 KF688227 cult. Wellington New Zealand C. cheesmanii A. Markey & J. Cantley HAW Dunedin, New KF688393 KF688310 KF688228 NZ215 Zealand C. ciliata J. Cantley s.n. HAW cult. Wellington, KF688425 KF688341 KF688257 New Zealand C. colensoi A. Markey & J. Cantley HAW Catlins NP, New AY357777.1 NZ237 Zealand C. cookei T.J. Motley s.n. PTBG Rapa Iti, Austral KF688456 KF688371 KF688288 Is. C. crassicaulis Skottsberg s.n. GB Mt. Kinabalu, KF688394 KF688311 Borneo C. crassifolia A. Markey & J. Cantley HAW New Zealand KF688395 KF688312 KF688229 NZ240 C. crenulata A. Markey & J. Cantley HAW Arthur's Pass, KF688396 KF688313 KF688230 NZ214 New Zealand C. cuneata A. Markey NZ414 CHR Blue Mtns, New JF950726.1 KF688356 Zealand C. cymosa S. Perlman NTBG Hawaii, Hawaii KF688462 KF688377 KF688294 C. decipiens J. Smith-Dodsworth s.n. AK New Zealand KF688474 C. decurva A. Markey & J. Cantley HAW Mt. Cargill, New KF688397 KF688314 KF688231 NZ222 Zealand C. depressa A. Markey & J. Cantley HAW New Zealand KF688475 KF688303 NZ205 C. discolor not vouchered KF688398 KF688315 C. obconica var. P.J. de Lange 9763 CHR Surville Cliffs, KF688441 KF688273 distantia New Zealand C. divergens M. Prebble s.n. CHR New Guinea KF688399 KF688316 KF688232 C. dodonaeifolia D. Glenny 10594 CHR Auckland, New KF688431 KF688347 KF688263 Zealand C. dumosa A. Markey & J. Cantley HAW KF688426 KF688342 KF688258 NZ234 C. elatirioides A. Markey & J. Cantley HAW New Zealand KF688400 KF688317 KF688233 NZ226 C. elliptica T. Flynn & J. Cantley NTBG Kauai, Hawaii KF688455 KF688370 KF688287 7475 C. ernodeoides T. Gallaher 330 HAW Hawaii, Hawaii KF688436 KF688351 KF688268 C. esulcata J-Y. Meyer s.n. PTBG Ua Pou, KF688458 KF688373 KF688290 Marquesas

Table 2.1. (Continued) Voucher information and GenBank accession numbers.

Species Collector and Collection Herb- Collection Locality ITS ETS rps16 Number arium Genbank No. Genbank No. Genbank No. C. K. Wood 10137 PTBG Fatu Hiva, KF688457 KF688372 KF688289 fatuhivaens Marquesas is C. J. Cantley 219 HAW Wellington, New KF688469 KF688384 KF688301 foetidissima Zealand C. A. Markey & J. Cantley HAW Arthur's Pass, New KF688401 KF688318 KF688234 fowerakeri NZ212 Zealand C. foliosa J. Cantley JC-207 HAW Oahu, Hawaii KF688447 KF688362 KF688279 C. J. Cantley JC-221 HAW Wellington, New KF688402 KF688319 KF688235 grandifolia Zealand C. hirtella Thompson & Hewat 883 BISH NSW, Australia JF950730.1 C. C. Beard s.n. AK Lord Howe Is. JF950734.1 huttoniana C. Papadopulos s.n. KEW Lord Howe Is. JF950736.1 inopinata C. A. Markey & J. Cantley HAW Cass, New Zealand KF688403 KF688320 KF688236 intertexta NZ224 C. kauensis S. Perlman 18648 BISH Kauai, Hawaii KF688453 KF688368 KF688285 C. laevigata G. Mccormack, not Rarotonga, Cook KF688476 vouchered Islands C. D. Mueller-Dumbois BISH Lord Howe Island JF950740.1 lanceolaris 92091015 C. J. Cantley 11-250 HAW Oahu, Hawaii KF688445 KF688360 KF688277 longifolia C. lucida J. Cantley 203 HAW Auckland, New KF688404 KF688321 KF688237 Zealand C. A. Markey & J. Cantley HAW Dunedin, New KF688405 KF688322 KF688238 macrocarpa NZ208 Zealand C. menziesii J. Cantley 274 HAW Hawaii, Hawaii KF688406 KF688323 KF688239 C. meyeri D. Lorence 19659 NTBG Hiva Oa, Marquesas KF688460 KF688375 KF688292 C. A. Markey & Cantley HAW South Island, New KF688407 KF688324 KF688240 microcarpa NZ203 Zealand C. montana J. Cantley 271 HAW Maui, Hawaii KF688450 KF688365 KF688282 C. neglecta J. Cantley 190 HAW Auckland, New KF688470 KF688385 KF688302 Zealand C. J. Florence 4342 PTBG Nuku Hiva, KF688477 nephelephil Marquesas a C. A. Markey NZ426 HAW Coronet Peak, New KF688437 KF688352 KF688269 niphophila Zealand C. nivalis A. Markey NZ167 CHR Australia KF688408 KF688325 KF688241 C. obconica J. Cantley s.n. HAW Wellington, New KF688409 KF688326 KF688242 Zealand C. ochracea J. Cantley JC-56 HAW Oahu, Hawaii KF688465 KF688380 KF688297 C. Gagne & Mont. s.n. BISH Mt. Orohena, KF688463 KF688378 KF688295 orohenensis Society Is. C. J. Cantley 180 HAW New Zealand KF688429 KF688345 KF688261 parviflora C. B. Clarkson 1023 AK North Island, New KF688438 KF688353 KF688270 pedicellata Zealand C. A. Markey & J. Cantley HAW Arthur's Pass, New JF950746.1 perpusilla NZ216 Zealand C. Rova & Gustavsson 2470 GB Fiji AF257909.1 persicifolia C. petiolata P.J. de Lange K849 AK Raoul Island, KF688439 KF688354 KF688271 Kermadec Is. C. petriei A. Markey & J. Cantley HAW New Zealand KF688410 KF688327 KF688243 NZ200 C. pilosa D. Mueller-Dumbois BISH Norfolk Is. KF688442 KF688357 KF688274 92091609

Table 2.1. (Continued) Voucher information and GenBank accession numbers.

Species Collector and Collection Herb- Collection Locality ITS ETS rps16 Number arium Genbank Genbank Genbank No. No. No. C. propinqua A. Markey s.n. CHR Chatham Is. KF688411 KF688328 KF688244 var. martinii C. propinqua E. Cameron 15619 AK New Zealand KF688440 KF688355 KF688272 var. propinqua C. pseudociliata D. Glenny 10647 CHR New Zealand KF688432 KF688264 C. A. Markey & J. Cantley HAW Arthur's Pass, New KF688412 KF688329 KF688245 pseudocuneata NZ217 Zealand C. pubens J. Cantley 297 HAW Hawaii, Hawaii KF688448 KF688363 KF688280 C. pumila A. Markey TAS06 WA Tasmania, Australia KF688478 C. putida D. Mueller-Dumbois BISH Lord Howe Is. JF950755.1 92090730 C. pyrifolia M. Pilna s.n. HAW (Juan-Fernandez Is.) KF688443 KF688358 KF688275 cult. Chile C. quadrifida Beesley 640 BISH NSW, Australia JF950756.1 C. rapensis J-Y. Meyer & Perlman BISH Rapa Iti, Austral Is. KF688461 KF688376 KF688293 1021 C. repens Poor J. Cantley NZ206 HAW cult. Auckland, New KF688430 KF688346 KF688262 Knights Form Zealand C. repens J. Cantley 110 HAW (New Zealand) cult. KF688413 KF688330 KF688246 Hawaii C. rhamnoides A. Markey & J. Cantley HAW New Zealand KF688414 KF688331 KF688247 NZ224 C. rhynchocarpa J. Cantley 11-262 HAW Hawaii, Hawaii KF688449 KF688364 KF688281 C. rigida A. Markey & J. Cantley HAW Catlins NP, New KF688415 KF688332 KF688248 NZ235 Zealand C. robusta J. Cantley 217 HAW New Zealand KF688416 KF688333 KF688249 C. rotundifolia A. Markey & J. Cantley HAW New Zealand KF688417 KF688334 KF688250 NZ228 C. rubra A. Markey & J. Cantley HAW New Zealand KF688418 KF688335 KF688251 NZ227 C. rugosa J. Cantley 213 HAW Wellington, New KF688419 KF688336 KF688252 Zealand C. savaiiensis A. Whistler 9999 HAW Upolu, Samoa KF688479 C. serrulata A. Markey & J. Cantley HAW Arthur's Pass, New KF688420 KF688337 KF688253 NZ213 Zealand C. spathulata D. Glenny 10456 CHR North Island, New KF688433 KF688348 KF688265 Zealand C. sp. nov. Papadopulos s.n. KEW Lord Howe Island JF950762.1 C. strigulosa A. Whistler 9899 HAW Upolu, Samoa KF688446 KF688361 KF688278 C. taitensis var. J-Y. Meyer PAP Society Is. KF688444 KF688359 KF688276 taitensis C. tayloriae A. Markey & J. Cantley HAW New Zealand KF688427 KF688343 KF688259 NZ236 C. temetiuensis D. Lorence 8931 PTBG Marquesas KF688459 KF688374 KF688291 C. tenuicaulis A. Markey NZ507 WA New Zealand KF688434 KF688349 KF688266 C. tenuifolia J. Cantley 220 HAW New Zealand KF688428 KF688344 KF688260 C. ternata Hank & Oppenheimer s.n. BISH Molokai, Hawaii KF688452 KF688367 KF688284 C. virescens A. Markey & J. Cantley HAW Otago Penninsula, KF688421 KF688338 KF688254 NZ239 New Zealand C. waima J. Cantley 196 HAW cult. Auckland, New KF688468 KF688383 KF688300 Zealand

Table 2.1. (Continued) Voucher information and GenBank accession numbers.

Species Collector and Collection Herb- Collection Locality ITS ETS rps16 Number arium Genbank Genbank Genbank No. No. No. C. wallii J. Cantley 212 HAW Wellington, New KF688422 KF688339 KF688255 Zealand D. paludosa NSW 154507 NSW Australia KF688423 N. dichondrifolia A. Markey CHR New Zealand KF688472 KF688387 KF688304 N. depressa A. Markey & J. Cantley HAW Dunedin, New KF688471 KF688386 NZ230 Zealand N. granadensis J. Cantley 38 HAW Oahu, Hawaii KF688451 KF688366 KF688283 No. Munzinger & McPherson MO New Caledonia KF688454 KF688369 KF688286 neocaledonica 532

The full internal transcribed spacers 1 and 2 and the 5.8s nuclear ribosomal DNA

gene region (ITS) was amplified using the primers ITS4 (5’-TCC TCC GCT TAT TGA

TAT GC-3’) and ITS5 (5’-GGA AGT AAA AGT CGT AAC AAG G-3’) after (Wagstaff

& Garnock-Jones, 1998). The primers ETS-IGSchaos (5’-GCC TGT TCA CGC ACT

ACC C-3’) and ETS77R (5’-GAG CCA TTC GCA GTT TCA CAG-3’) were specifically

designed for the nuclear external transcribed spacer (ETS) region of Coprosma by

Wichman (2000). Primers for the plastid rps16 intron, rps16F (5’-GTG GTA GAA AGC

AAC GTG CGA CTT-3’) and rps16R2 (5’-TCG GGA TCG AAC ATC AAT TGC

AAC-3’) are those of Oxelman et al. (1997). All PCR reactions were performed with 25

µl of reaction cocktail containing 12.75 µl of sterilized H2O, 2.0 µl of 20 mM dNTPs

(Pharmacia) in an equimolar solution, 2.5 µl of 10x PCR reaction Buffer A (Promega),

1.25 µl of 25 mM MgCl2, 0.5 µl 10 mg/ml Bovine Serum Albumin (Sigma), 1 µl of 10

mM of each of the two primers, 0.5 µl Biolase Red Taq DNA polymerase enzyme

(Bioline) and 4 µl of DNA template. The final amount of DNA template and PCR

reaction cocktail and Taq was adjusted as necessary to generate sufficient PCR products

for DNA sequencing. ITS amplifications were performed on a Bio-Rad thermal cycler

c1000 using an initial denaturing step of 94˚C, 2m; followed by 35 cycles of (94˚C, 1

min; 50˚C, 1 min; 72˚C, 2 min), then a final 72˚C, 7 min elongation step. Amplification

of the ETS region was achieved using the same PCR cycling program as the ITS region

with a raised annealing temperature of 62˚C. For the plastid rps16 intron the

amplification program was an initial denaturing step of 95˚C, 2 min, followed by 35

cycles of (95˚C, 0.5 min; 57˚C, 1 min; 72˚C, 2 min) and a final elongation step of 72˚C, 7

min. Samples were purified prior to sequencing using an Exo-Sap enzymatic PCR

product pre-sequencing protocol (USB) for 45 minutes.

A final volume of 8.2 µl was used for sequencing reactions. This consisted of 2.0

µl of sterilized H2O, 3.2 µl of 1 mM primer and 3.0 µl of purified DNA template.

Sequencing was conducted at the Advanced Studies of Genomics, Proteomics and

Bioinformatics facility at the University of Hawai‘i at Manoa. Sequences were edited

using Sequencher 3.1.10 (Gene Codes Corp., Ann Arbor Michigan, USA) and aligned by

MUSCLE (Edgar, 2004) using default parameters as implemented in MEGA 5 (Tamura

et al., 2011). Alignments were then manually inspected by eye and adjusted where

necessary.

Phylogenetic and biogeographic analyses

To test for congruence among gene regions and between plastid and nuclear

genomes, all gene regions were subjected to the partition homogeneity test (ILD) as

implemented in PAUP* ver. 4.0b10 (Swofford, 2003) using 1000 replicates, with TBR

(tree-bisection reconnection) branch swapping and the MulTrees option turned on. No

significant differences in topology were found between the separate partitions (p=0.53) so

a concatenated data set was used for Maximum Likelihood (ML) and Bayesian Inference

(BI) analyses. Prior to running ML and BI analyses, the same best-fit model (GTR + I +

Γ) was selected for all individual gene regions as well as the combined nuclear data set

using the Akaike Information Criterion (AIC) in jModelTest 2 (Posada & Buckley,

2004). ML analyses were implemented in RaxML v7.0.4 (Stamatakis, 2006) for

individual and combined datasets. Nonparametric bootstrap replicates (1000) for all ML

analyses were calculated with the thorough bootstrap replicate option selected and

allowing all free model parameters to be estimated.

BI analyses were implemented in MrBayes (Hueselenbeck & Ronquist, 2001) for

individual regions and the concatenated datasets. The combined dataset was divided into five partitions (ITS1, 5.8s, ITS2, ETS, and rps16) and Markov Chain Monte Carlo

(MCMC) sampling was performed with two replicates of four chains (one hot and three cold) each with a heating temperature of 0.2. Five million generations were completed

with sampling occurring every 1000 generations. A discarded burn-in period was

estimated by visually inspecting plotted log likelihood values versus generation time to

determine the point at which convergence had been reached using Tracer v1.5 (Rambaut

& Drummond, 2007). The remaining trees were combined to construct a consensus tree

where Bayesian posterior probabilities (PP) were calculated for internal node support of

the resulting phylogenetic reconstruction. A Separate ML and BI analysis of only the ITS

region was run with an additional 17 Coprosma species obtained from GenBank for a

total of 97 taxa. Parameters for each analysis were the same as the above analyses. This

separate analysis was completed to infer the relationships of the additional sixteen species

for which only the ITS nrDNA region was available on GenBank.

Geographical areas were mapped onto the BI tree (=ML topology) using the parsimony criterion to elucidate major dispersal patterns among Pacific species using

Mesquite (Maddison, 2001). Geographical distributions were divided into five unordered area states based on the presence of one or more endemic species in respective localities: i) Hawaiian Islands, ii) New Zealand, iii) Australia, iv) Indo-Pacific Islands (i.e. New

Guinea and islands to the west), and v) other Pacific Islands.

Results

Maximum Likelihood and Bayesian Inference Analyses

The total sequence lengths and the number of parsimoniously informative characters (PIC) (both with and without outgroup taxa) are listed in Table 2 for each

DNA region and for the combined data set. The aligned matrix contained some indels, but none were phylogenetically informative as they were either autapomorphic or outgroup specific. The individual nuclear (ITS and ETS) ML and BI phylogenetic reconstructions were similar in their topologies and support for nodes and resolution was increased when combined. The individual rps16 plastid phylogenetic reconstructions were fairly unresolved, however, the addition of rps16 to the total combined dataset (ITS

+ ETS + rps16) resulted in a phylogenetic reconstruction with better support and resolution than nuclear alone (Fig. 2.3). Furthermore, the resulting combined phylogenetic reconstruction was consistent in topology for the individual ITS and ETS cladograms. Results from the Partition Homogeneity Test showed no significant differences between individual DNA regions, or the nuclear and plastid datasets (p=0.53) suggesting that the regions share a common ancestry. Topologies of the combined dataset

were congruent between ML (Tree Length=0.407026) and BI (-lnL=7277.861) analyses,

although support values varied slightly (Fig. 2.3 & 2.4).

Table 2.2. Number of characters and parsimoniously informative characters (PIC) for each gene region and the combined data.

DNA Total No. of PIC No. of PIC

sequence Number of including without

Region Characters outgroups outgroups

ITS 574 117 53

ETS 494 75 60

rps16 870 40 16

combined 1939 232 129

Coprosma was supported as monophyletic consisting of three well-supported clades, each with many taxa and with strong resolution between them (Fig. 2.3 & 2.4).

Resolution within each clade was variable, as both strongly supported and comb topologies were recovered. Clade 1 consisted of species from Borneo, New Guinea, and

New Zealand. Clade 2 consisted of species from Australia, Fiji, Hawaiian Is., Lord Howe

Island, New Zealand and Samoa. Clade 3 was the largest clade reconstructed, which included species from Austral Is., Australia, Chatham Is., Cook Is., Hawaiian Is., Juan-

Fernandez Is., Kermadec Is., Lord Howe Is., Marquesas Is., New Zealand, Norfolk Is.,

Samoa, and Society Is. (Fig. 2.3 & 2.4). Hawaiian Coprosma species were found in two separate clades. Of the species sampled in this study, the Hawaiian taxon C. ernodeoides of clade 2 was most closely related to C. atropurpurea (Cockayne & Allan) L.B. Moore of New Zealand. All other Hawaiian taxa formed an independent, monophyletic lineage

in clade 3, which was sister to the Marquesas Islands species and C. cookei Fosberg from closely related to C. tenuifolia of New Zealand rather than any other Pacific species samples in this study (Fig. 3).

The ITS-only phylogenetic analyses—which included ITS data for an additional

17 Coprosma species—produced a similar topology to the three region analyses (Fig.

2.4). The BI and ML ITS-only topologies were similar and the same phylogenetic relationships were supported as in the three region analyses (i.e. six Pacific Island localities colonized at least twice—including the Hawaiian Islands—from independent dispersal events (see below), Coprosma is monophyletic and is the sister taxon to

Nertera; Fig. 2.4). The addition of 17 taxa in the ITS-only phylogenetic reconstructions supported the same findings of the three region analyses. Sixteen of the additional species added were not directly related to either of the two Hawaiian lineages. The seventeenth additional species, Coprosma pumila of Tasmania, was reconstructed in a polytomy with

C. atropurpurea of New Zealand and C. ernodeoides of the Hawaiian Islands.

Biogeographic Relationships

The phylogenetic reconstructions with mapped biogeographic areas suggested an

origin of Coprosma in New Guinea or New Zealand. All lineages, but perhaps excluding

New Guinea, could have originated from New Zealand (Figs. 2.3 & 2.4). Two

colonization events to the Hawaiian Islands were inferred. The twelve Hawaiian species

of the HMR clade may have origins from the Marquesas or Austral Islands, but it was not

possible to determine the patterns of dispersal among these three archipelagoes given

their unresolved relationship. The second colonization event inferred to the Hawaiian

Figure 2.3. Phylogenetic reconstruction of Bayesian Inference showing a 50% consensus topology based on two nuclear (ITS and ETS) and plastid (rps16) DNA regions. Numbers above branches represent Bayesian posterior probabilities and Maximum Likelihood

bootstrap values, respectively. Abbreviations of geographic localities following taxa are as follows: AUST=Australia (including Tasmania), BORNEO=Borneo Island, CHAT=Chatham Islands, HAWAII=Hawaiian Islands, JF=Juan-Fernandez Islands, KERM=Kermadec Islands, LHI=Lord Howe Island, MARQ=Marquesas Islands, NCAL=New Caledonia, NG=New Guinea, NORF=Norfolk Islands, NZ=New Zealand, RAPA=Rapa Iti Island (of the Austral Islands), SAMOA= Samoan Islands, SOCI=Society Islands, and WIDE=widespread taxon. Hawaiian taxa indicated in bold. Branch colors representing geographical distribution are indicated and were coded as following (1. Indo-Pacific [New Guinea, Borneo, Sulawesi], 2. Australia, 3. New Zealand, 4. South Pacific [Austral Is., Chatham Is. Juan-Fernandez Is., Kermadec Is., Lord Howe Is., Marquesas Is., Norfolk Is., Samoan Is., Society Is.], and 5. the Hawaiian Is.). Inset is the phylogram of the Bayesian Inference.

Islands was represented by C. ernodeoides. Clade 2, of which C. ernodeoides was a member, consists of eight geographical locations, which were incompletely resolved. It was therefore not possible to determine the biogeographic origin for C. ernodeoides

(Figs. 2.3 & 2.4).

Similar to the two colonization events to the Hawaiian Islands, we inferred that two independent colonizations occurred each to the Austral Is. (C. rapensis F. Br. and C. cookei), Chatham Islands (C. chathamica Cockayne and C. propinqua var. martini

W.R.B. Oliv.), Kermadec Is. (C. acutifolia F. Muell. ex Benth. and C. petiolata Hook. f.),

Lord Howe Is. (C. prisca W.R.B. Oliv. and a monophyletic lineage of five endemic species [see Fig. 4]) Norfolk Is. (C. baueri Endl. and C. pilosa Endl.) and Samoan Is. (C. savaiiensis Rech. and C. strigulosa). Each of these species pairs was located in different clades or subclades of the phylogenetic reconstructions, which suggested multiple independent colonization events (Figs. 3 & 4).

Figure 2.4. Phylogenetic reconstruction of Bayesian Inference showing a 50% consensus topology based the nuclear ITS region only and including the 18 taxa not included in the previous analysis (indicated with *). Hawaiian taxa indicated in bold. Numbers above branches represent Bayesian posterior probabilities and Maximum Likelihood bootstrap values, respectively. Geographic distributions following taxa names are as in Fig 3. Branch colors, representing geographical distribution, are as indicated and as coded in Fig. 2.3.

Discussion

Results from this study indicate that the biogeography of Coprosma across the

Pacific was complex and involved many dispersal events in order to explain its extant distribution. We inferred two successful colonizations of Coprosma to the Hawaiian

Islands, as previously hypothesized based on morphology (Oliver, 1935; Wagner, 1991).

However, the hypothesis of three colonizations (Fosberg, 1948) was not supported.

Additionally, the Hawaiian Islands were not the only Pacific archipelago with two

independent colonization events of Coprosma. Our results suggested multiple

independent colonization events of Coprosma occurred to each the Austral, Chatham,

Kermadec, Lord Howe, Norfolk and Samoan Islands and Australia, highlighting the

biogeographic complexity of the successful establishment of this genus across the Pacific.

Multiple colonizations to a single insular locality were previously documented for many

other Pacific genera. For example, molecular studies support the idea of three

colonizations of Scaevola to the Hawaiian Islands presumably via internal bird ingestion

and water dispersal (Howarth et al., 2003; Howarth & Baum, 2005), two colonizations of

internally bird dispersed Melicope J.R. Forst. & G. Forst. to the Marquesas Islands

(Harbaugh, 2009), and two colonizations of wind dispersed Metrosideros to Lord Howe

Island (Papadopulos, 2011). The emerging biogeographic pattern of Coprosma differs

from these lineages as it has colonized at least seven insular localities (plus Australia) a

minimum of two times representing the most known cases of this for any genus

investigated to date.

In the Hawaiian Islands, the largest radiation of Coprosma consisted of twelve

orange-fruited species that are sister to Marquesan species plus C. cookei of Rapa Iti

Island of the Austral Islands (HMR clade; Fig. 2.3). The close relationship between

Hawaiian and Marquesan species was postulated to exist in previous morphological

treatments due to a shared morphology. Common features include ciliated stipules with a

central emarginate appendage, large leaf scars and buttresses, pale wrinkled bark,

conspicuous primary venation, and a tendency towards the multiplication of floral parts

(Heads, 1996; Florence & Lorence, 1997; Gardner, 2002b). Nevertheless, orange-fruited

Hawaiian Coprosma are immediately distinguished by an inflorescence in which

peduncles, located in the axils of leaves, are not closely grouped together (Florence &

Lorence, 1997). The close relationship of plant taxa in the Hawaiian and Marquesan

islands is well known for many other genera as well, including, Bidens L. (Ganders et al.,

2000), Kadua Cham. & Schltdl. (Motley, 2003), Plantago L. (Dunbar-Co et al., 2008), and many others. Furthermore, Coprosma shares an intriguing biogeographic similarity with Astelia Banks & Sol. Ex R. Br. (Birch et al., 2012) and Metrosideros Banks ex

Gaertn. (Wright et al., 2001; Percy et al., 2008) as New Zealand taxa from all of these genera form a close relationship with taxa from the Marquesas Islands and the Hawaiian

Islands. Further investigation of this unexpected biogeographic relationship is needed, but it is clear that both internal bird ingestion and wind borne seeds are two dispersal syndromes aiding their establishment.

While it is apparent that the twelve orange-fruited Hawaiian Coprosma are monophyletic, very little genetic variation is detectable among them based on nuclear ITS and ETS and plastid rps16 intron (Fig 2.3 & 2.4) despite the significant amount of morphological diversity occurring among these species. A situation in which there are great morphological differences among species within a radiation, accompanied by few detectable genetic differences of the molecular markers investigated is documented for oceanic island taxa in the Hawaiian Islands (Keeley & Funk, 2011), Galápagos (Tye &

Francisco-Ortega, 2011) and Macaronesian Islands (Lledó et al., 2011). In the Hawaiian

Islands, this situation is documented in Schiedea (Weller et al., 1995; Soltis et al. 1996),

Tetramolopium Nees (Okada et al., 1997; Lowery & Crawford, 1985; Okada et al.,

2000), and Psychotria L. (Nepokroeff et al., 2003), among other genera. The lack of

genetic differences in this type of molecular marker is often hypothesized to be the result

of a rapid and recent radiation into multiple ecologically distinct habitats after initial

establishment of one or few colonizing ancestors that would allow little time for genetic

mutations to accumulate (Keeley & Funk, 2011). This could explain the pattern observed

among the radiation of orange-fruited Hawaiian Coprosma if they happen to represent a

recent arrival to the archipelago. However, the results obtained from the ITS-only

phylogenetic analyses must be considered cautiously as there are known limitations

utilizing these data (Alvarez & Wendel, 2003; Nieto-Feliner & Rosello, 2007).

Unfortunately, it was not possible to explore these known limitations (e.g. the presence

paralogs or orthologs, often resulting from polyploidization or hybridization) as the

additional sequences were collected from GenBank. However, the inferred result of two

colonization events to the Hawaiian Islands is likely not affected by these known

limitations, as both molecular and morphological data independently support our

interpretation.

The most recent common ancestor of the HMR clade is shared with C. tenuifolia from New Zealand rather than to any other Pacific island locality. Within clade 3, all species from New Zealand and Kermadec Island (with the exception of C. propinqua) belong to two well-defined morphological sections: Australes and Lucidae (see Allan

1961; Heads, 1996; Gardner, 2002b). However, the phylogenetic reconstructions suggest

these classifications are artificial, as they were recovered as polyphyletic. Sections

Australes (C. grandifolia Hook. f., C. macrocarpa Cheeseman, C. acutifolia, C. tenuifolia, C. robusta Raoul, and C. waima A.P. Druce), Lucidae (C. lucida J.R. Forst. &

G. Forst. and C. dodonaeifolia W.R.B. Oliv.) and the HMR lineage are similar in

morphology with red to orange fruits, large-leaves, ternate phyllotaxy and long-sheathing deciduous stipules. Ternate phyllotaxy is uncommon of species in sections Australes,

Lucidae and the HMR lineage, and within Coprosma in general, but it is known from species within these three related groups and a few distantly related New Guinea species.

Furthermore, the most remotely located Coprosma species, C. pyrifolia of Juan-

Fernández Islands, is nested among the sections Australes, Lucidae and the HMR lineage

and is not directly related to the HMR lineage (Fig. 2.3 & 2.4) further supporting these

morphological groups as polyphyletic.

The second inferred colonization of Coprosma to the Hawaiian Islands includes

only black-fruited C. ernodeoides. Morphologically, C. ernodeoides is distinct from the

HMR lineage as it has small leaves without conspicuous secondary venation, 4-merous

long-tubular corollas, a terminal solitary inflorescence, shiny black fruits and is a low

shrub with a prostrate to mat-forming habit (Fig. 2.2). Results from the phylogenetic analyses indicate that C. ernodeoides is closely related to C. atropurpurea of New

Zealand and C. pumila of Tasmania (Fig. 2.3 & 2.4). Moreover, C. atropurpurea and C. pumila share the above mentioned morphological characters with C. ernodeoides (Merrill

& Perry, 1945; Orchard, 1986). All three species—C. pumila, C. ernodeoides and C.

atropurpurea—are polyploid (2n=220; Dawson, 1995; Dawson & Beuzenberg, 2000)

whereas most species in the genus, including those known of the HMR lineage, are

diploid (2n=44) (Beuzenberg, 1983; Wagner et al., 1999; Dawson & Beuzenberg 2000).

Although these morphological and cytological characters appear to support the close

relationships among these species, it should be noted that C. archboldiana from New

Guinea (Merrill & Perry, 1945), C. talbrockiei of New Zealand, and C. moorei of

Tasmania (Heads, 1996) are also similar in their morphology to C. ernodeoides.

Unfortunately, we were unable to obtain material of these species for this investigation.

For these reasons, the origin of Hawaiian C. ernodeoides currently remains unclear as

New Zealand, Australia, or New Guinea are all possible source areas. Additionally, the

relationship of C. ernodeoides to C. persicifolia of Fiji and C. strigulosa of Samoa,

should be investigated in more detail (clade 2, Fig. 2.4). These three species are pioneers

on relatively recent lava flows, are some of the few species in the genus to have black

fruits (Oliver, 1935) and share other floral and vegetative characters (Oliver, 1935;

Whistler, 1992; Wagner et al., 1999). Due to affinities shared among these species, it is

possible that the arrival of C. ernodeoides to the Hawaiian Islands was via stepping-stone

dispersal from New Zealand and/or Tasmania through New Guinea, Fiji and Samoa in

some fashion or as a direct dispersal from a locality farther afield (i.e., New Zealand or

Tasmania without intervening archipelagos acting as stepping stones).

In our attempt to understand the origins of Coprosma in the Hawaiian Islands, it is clear that the biogeographic patterns of the genus across the Pacific are complex. It is also clear that dispersal from New Zealand to other locations in the Pacific was fundamental to the biogeography of Pacific Coprosma. Multiple dispersals to the same area, like the

Hawaiian Islands, are characteristic of the genus, as is the development of centers of speciation and radiations on many archipelagos. Among the biological factors that likely aided establishment across the Pacific is internal bird mediated dispersal. All Pacific species outside of New Zealand have bird dispersed orange or red fruits, with three notable black-fruited exceptions: C. ernodeoides of the Hawaiian Islands, C. persicifolia of Fiji and C. strigulosa of Samoa. Coprosma ernodeoides is commonly dispersed by the endemic Hawaiian Nēnē goose (Branta sandvicensis Vigors) (Baldwin, 1947), but the dispersers of C. persicifolia and C. strigulosa are not known. Although, Lee et al. (1994) found black fruits to be highly conspicuous in the diets of New Zealand birds, but it is unclear if this also true for birds Pacific wide. Coprosma species occupy a broad range of habitats, have considerable morphological plasticity in habit, commonly hybridize and some species are polyploids (Oliver, 1935; Wagner et al., 1999; Beuzenberg, 2000).

These factors, particularly hybridization and polyploidy, are known to be important in the establishment of the species-rich silversword alliance in the Hawaiian Islands (Carlquist et al., 2003), and may have similarly benefited Coprosma establishment by allowing high levels of genetic diversity to be maintained, thus providing a store of variability for adaptive radiation following arrival to the Hawaiian Islands. Whatever the reasons,

Coprosma is among the most successful genera across the Pacific, dispersing thousands of kilometers, and establishing in nearly all high islands and archipelagos from New

Zealand to Juan-Fernández Islands to the Hawaiian Islands, and to Borneo. Furthermore, most Pacific localities were colonized by Coprosma on multiple occasions by separate lineages, and have subsequently radiated into a wide variety of habitats. Understanding the origins of Hawaiian Coprosma, and indeed the biogeographical patterns of the genus across the Pacific greatly furthers our understanding of the evolution and relationships of

Pacific floras.

CHAPTER 3 BIOGEOGRAPHY AND EVOLUTIONARY DIVERSIFICATION OF INSULAR COPROSMA (RUBIACEAE): ONE OF THE LARGEST AND MOST WIDELY DISTRIBUTED GENERA IN THE PACIFIC

Abstract

The aim of this research to investigate the historical biogeography and associated factors of the Pacific genus Coprosma (Rubiaceae) through phylogenetic analyses utilizing chloroplast and nuclear DNA sequence data occurring across the Pacific. Sequence data were analyzed using Bayesian and Maximum likelihood approaches to construct a phylogeny of Coprosma including 105 of 110 taxa from across the geographic range of

the genus plus six outgroups. Divergence times were calibrated from fossil data using a

Bayesian log-normal relaxed molecular clock. Ancestral areas were reconstructed using

the statistical dispersal-vicariance (s-DIVA) and Bayesian binary method (BBM) models

and morphological characters were traced onto the phylogeny using Mesquite.

Coprosma diverged from Nertera in New Zealand in the Oligocene. Subsequent

diversification of New Zealand taxa is correlated with tectonics of the Miocene. Species

with orange fruit, large leaves, leaky-dioecy and non-divaricating habit are the common shared features of species occurring elsewhere from New Zealand and at least 30 dispersals are required in order to explain the distribution across the Pacific. The divergence of Coprosma from Nertera in the Oligocene was likely facilitated by innovations of woodiness and dioecy, which allowed Coprosma a greater ability to diversify in novel environments. The dynamic landscape of New Zealand during the

Miocene is correlated to the development of all major lineages and morphological diversity of Coprosma. Dispersal out from New Zealand and among Pacific island

localities is correlated with the morphological combination of fleshy orange fruit, large leaves, leaky-dioecy and a non-divaricating habitat. These characters suggest bird- mediated dispersal and increased establishment likelihood for insular Coprosma taxa compared to novel character states of New Zealand taxa.

Introduction

Biogeographic patterns of plants of the Pacific remain poorly understood, but are slowly elucidated in recent studies (Winkworth et al., 2002; Gillespie et al., 2008,

Muellner et al, 2008). Perhaps this is a consequence of the difficulties of working in the most isolated localities on Earth, resulting in a poor understanding of species relationships to each other and their area of origin (Price & Wagner, 2004; Keeley &

Funk, 2011). However, a strong rationale for conducting research in the Pacific was recognized as early as the 19th century by Hillebrand (1888, p. xxix) who noted, “the evolution theory could hardly find a more favorable field for observation than an isolated island-group in mid-ocean.” Islands represent simple systems, in comparison to continental floras, in which specific evolutionary factors of a lineage can be more easily isolated for focused research (Carlquist, 1977). It is also possible to deduce a wider understanding of important evolutionary responses by examining common biogeographic patterns of lineages across the Pacific and how they are influenced by their biological constraints (i.e. dispersal abilities, niche requirements) within a context of known island geological history (Price & Wagner, 2004).

The biogeographic patterns of Coprosma (Rubiaceae) are among those that remain poorly understood, despite its place as one of the most widespread and species rich genera in the Pacific (Cantley et al., in press). The ~110 species of Coprosma are located on nearly every major island archipelago of the Pacific, east to west from the Juan

Fernández Islands (near Chile) to Borneo, and south to north from the subantarctic islands of New Zealand to the Hawaiian Islands (Fig. 3.1). The centre of diversity and the hypothesized origin of the genus is New Zealand, where there are more than 55 species

Figure 3.1. Map of landmasses of the Pacific Ocean. Labeled localities indicate presence of one or more Coprsoma species.

(Oliver, 1935; Gardner 2002b; Cantley et al., in press). A secondary centre of diversity is recognized in the Hawaiian Islands where there are 13 species (Wagner et al., 1999), and possibly in New Guinea where there is a disputed number of taxa (5-15 species; Gardner

2002a; Utteridge, 2002). Smaller concentrations of species occur in the Marquesas

Islands (6 species; Wagner & Lorence, 2011), Lord Howe Island (5 species;

Papadopoulos et al., 2011), and Australia (8 species; Thompson, 2010) and numerous

other insular locations. Few species of Coprosma are widespread, and many are endemic

to individual islands within archipelagos.

Despite the remarkable distribution of Coprosma in the Pacific, our current

understanding of the biogeographic patterns and associated dispersal and establishment

factors are limited at best. Heads (1996) was the first to propose a genus wide

biogeographic hypothesis based upon morphological observations. The hypothesis

recognized an east to west biogeographic split of two subgenera, which overlapped

significantly in New Zealand, and to a minor degree in the Hawaiian Islands. Coprosma

species distributed in the west (plus one Hawaiian taxon) were classified into subgenus

Coprosma having small leaves and solitary female flowers. Conversely, species in the

eastern distribution were classified into subgenus Lucidae having large leaves and

multiple female flowers. Heads (1996) further suggested that the east-west subgeneric split of Coprosma was the vicariant result of a widespread Pacific ancestor of

Gonwanaland. Gardner (2002) refuted this assumption suggesting that the observed distribution of Coprosma could represent a biological reality rather than a vicariant explanation. Gardner (2002) notes that the small leaved taxa of Head’s subgenus

Coprosma, which occupy cold high elevation habitats, are found in all colonisable habitats, including the outlying distribution of the Hawaiian Islands. Beyond these subgeneric hypotheses, no other relationships among Pacific taxa were proposed.

Following the biogeographic hypotheses of Heads (1996) and Gardner (2002), it is not clear if the distribution of Coprosma is the result of an ancient Gondwanan split, recent dispersals throughout the Pacific, or both. The distribution of Coprosma is highly disjunct and almost certainly involved long distance dispersals for its establishment, as most localities with Coprosma are volcanic in origin and were never connected to any other landmass (Cantley et al., in press; Price and Wagner, 2004). Furthermore, these

localities are often separated from each other by thousands of kilometres (ex. >3500 km

between the Hawaiian Islands and Marquesas Islands). Coprosma taxa in New Zealand,

Australia and Malesia, however, could represent a vicariant distribution of Gondwanaic

times (see Fig. 3.1). But given the complex geologic history of New Zealand and unclear

relationships among Coprosma taxa, it is difficult to speculate a precise biogeographic

history.

Whatever the timing of origin for Coprosma, many taxa have traits are suspected

to have aided Coprosma in achieving its highly disjunct distribution and successful

establishment across the Pacific. Long distance dispersals must have played an important

role, at least for a number of non-New Zealand taxa. It is believed that birds might have mediated these dispersals as they are shown to be highly attracted to the fleshy two- seeded drupes of many Coprosma taxa (Fig. 3.2). Fleshy fruits are known as one of the most successful mechanisms in the dispersal of Pacific lineages (Carlquist, 1967). Birds eating these fruits may carry seeds for long distances internally, as well as by external adhesion to feathers and feet. Almost all insular Coprosma fruit are orange, but a number of other colors (blue, white, yellow, purple) exist primarily in New Zealand. Lee et al.

(1988) identified reptiles and birds as the primary dispersers of Coprosma fruits and indicated orange fruits were more conspicuous to New Zealand birds than other colours.

Orange fruits of Coprosma are also a favoured food of some Hawaiian bird species (Pers. obs.).

In addition to the ability to disperse well, colonizing taxa must also be able to overcome stresses of establishment in novel insular habitats (Carlquist, 1967). Among the most important aspects of becoming established is successful reproduction, especially in the face of the low numbers of initial colonizers. Coprosma maintains a number of traits that are cited as important in the establishment of insular species. These include apomixis

(Heenan et al., 2002; Heenan et al., 2003), hybridization (Oliver 1935; Allan, 1961;

Figure 3.2. Vegetative and reproductive morphology of Coprosma and Nertera. (a-c) Flowers, (a) female flowers of C. colensoi of New Zealand, (b) male flowers of C. ochracea of the Hawaiian Islands, and (c) two male and one bisexual flower of C. serrulata of New Zealand. (d) C. montana of the Hawaiian Islands as a tall tree pictured with author Jason Cantley, (e) C. menziesii of the Hawaiian Islands as an understory shrub, note orange fruit, (f) C. wallii of New Zealand indicating a strong divaricating habit. (g-i) Fruit colors, (g) black fruit of C. ernodeoides of the Hawaiian Islands, (h) C. pumila of Tasmania with red-burgundy fruit, (i) C. brunnea of New Zealand with blue fruit. (j) Nertera granadensis indicating trailing herbaceous habit and orange fruit. Scale bars: (a) 0.75 cm, (b-c) 1 cm, (d) 1 m, (e) 2 cm, (f) 0.25 m, (g-i) 0.5 cm, (j) 2cm. Photos a-g and j by Jason T. Cantley and h-i by Adrienne Markey.

Gardner, 2002), depressed levels of concerted evolution (Wichman et al., 2002), polyploidy (Beuzenberg, 1983), and leaky-dioecy (Fig 3.2). Leaky dioecy is more common in insular taxa than continental counterparts and is often cited as important for the establishment of dioecious taxa, such as Coprosma. This breeding strategy of

Coprosma may have significantly aided in reproductive assurance of species establishment after long distance dispersal events across the Pacific as leaky dioecy could aid the establishment of taxa from a small founding population, but also promote outcrossing, which has advantages for adaptation (Vamosi et al., 2003; Barrett et al., 2003).

It is evident that Coprosma is well adapted to dispersal and radiation among the Pacific islands, but the factors contributing to this success remain to be elucidated.

It was the purpose of this study to establish the relationships among Coprosma species by constructing a molecular phylogeny and using this information to gain an understanding of the patterns of radiation and the factors contributing to the success of

Coprosma in the Pacific. Molecular clocking techniques and ancestral area reconstruction were applied to provide an estimate of dates of dispersal and sites of probable origin.

Morphological characters were also mapped onto the phylogenetic reconstructions. We specifically aimed to answer the following questions: a) what are the biogeographic relationships among all Pacific species of Coprosma? b) are there specific adaptations that have aided Coprosma in achieving its current distribution? and, c) how does the biogeographic pattern of Coprosma compare to other Pacific lineages?

Materials and Methods

Sampling

Of approximately 110 Coprosma species, 105 taxa were selected representing the

entire geographic distribution of the genus, including all but Bougainville Is. (C.

bougainvilleensis) and (see Fig. 1 for a map of Pacific Island localities). Five genera

from tribe Anthospermeae, Durringtonia, Leptostigma, Nertera, Normandia and

Opercularia, were used as outgroups (Anderson et al., 2001; Markey et al., 2004).

DNA extraction, amplification and sequencing

DNA was obtained from herbarium and silica dried material, and from the

Hawaiian Plant DNA Library (Randell and Morden, 1999). Samples were prepared by

hand grinding leaves with liquid Nitrogen in a mortar and pestle followed by extraction using DNeasy Plant Mini Kit (Qiagen, Valencia, California, USA). For difficult-to-

extract samples, the QiaAmp DNA Stool Mini Kit was used following the manufacturer’s

protocols, with a lengthened initial incubation period of 2 hours at 70˚C. All PCR

reactions were performed with 25 µl of reaction cocktail containing 12.75 µl of sterilized

H2O, 2.0 µl of 20 mM dNTPs (Pharmacia) in an equimolar solution, 2.5 µl of 10x PCR

reaction Buffer A (Promega), 1.25 µl of 25 mM MgCl2, 0.5 µl 10 mg/ml Bovine Serum

Albumin (Sigma), 1 µl of 10 mM of each of the two primers, 0.5 µl Biolase Red Taq

DNA polymerase enzyme (Bioline) and 4 µl of DNA template. The final amount of DNA template and PCR reaction cocktail and Taq was adjusted as necessary to generate sufficient PCR products for DNA sequencing. Samples were purified prior to sequencing using an Exo-Sap enzymatic PCR product pre-sequencing protocol (USB) for 45 minutes.

A final volume of 8.2 µl was used for sequencing reactions. This consisted of 2.0

µl of sterilized H2O, 3.2 µl of 1 mM primer and 3.0 µl of purified DNA template.

Sequencing was conducted at the Advanced Studies of Genomics, Proteomics and

Bioinformatics facility at the University of Hawai‘i at Manoa. Sequences were edited using Sequencher 3.1.10 (Gene Codes Corp., Ann Arbor Michigan, USA) and aligned by

MUSCLE (Edgar, 2004) using default parameters as implemented in MEGA 5 (Tamura

et al., 2011). Alignments were then manually inspected and adjusted as needed.

Phylogenetic and morphological analyses

To test for congruence among gene regions and between chloroplast and nuclear

genomes, all gene regions were subjected to the partition homogeneity test (ILD) as

implemented in PAUP* ver. 4.0b10 (Swofford, 2003) using 1000 replicates, with TBR

(tree-bisection reconnection) branch swapping and the MulTrees option turned on. No

significant incongruities were found among any of the individual partitions or between

the chloroplast genome regions (rps16 intron and trnQ-rps16 intergenic spacer) when compared to nuclear (ITS and ETS) genome regions (p=0.60), so a concatenated data set was used for Maximum Likelihood (ML) and Bayesian Inference (BI) analyses. Prior to running ML and BI analyses, the same best-fit model (GTR + I + Γ) was selected for all individual gene regions as well as the combined nuclear data set using the Akaike

Information Criterion (AIC) in jModelTest 2 (Posada and Buckley, 2004). ML analyses were implemented in RaxML v7.0.4 (Stamatakis, 2006) for individual and combined datasets. Nonparametric bootstrap replicates (1000) for all ML analyses were calculated

with the thorough bootstrap replicate option selected and allowing all free model

parameters to be estimated.

BI analyses were implemented in MrBayes 3.1.2 (Hueselenbeck and Ronquist,

2001) for individual regions and the concatenated datasets. The combined dataset was

divided into six partitions (ITS1, 5.8s, ITS2, ETS, rps16 intron and trnQ-rps16) and

Markov Chain Monte Carlo (MCMC) sampling was performed with two replicates of

four chains (one hot and three cold) each with a heating temperature of 0.2. Five million

generations were completed with sampling every 1000 generations. Burn-in was

estimated using Tracer v1.5 (Rambaut and Drummond, 2007) with visual inspection of

the plotted log likelihood values. Trees generated before convergence was reached were

discarded. The remaining trees were combined into a consensus tree with Bayesian

posterior probabilities (PP) calculated for internal node support.

Morphological traits of fruit colour, reproductive mode and habit were overlaid on the BI (=ML topology) tree using Mesquite (Madison et al., 2001). Accelerated transformation was used to minimize the number of convergent gains of character states.

Fruit colour was divided into seven categories: black, blue, burgundy, orange, red, white

(unpigmented) and polymorphic. Coprosma cheesemanii, often polymorphic in colour

was coded as orange as this was the fruit colour in our collection. Reproductive syndrome

was scored as hermaphroditic, dioecious with occasional bisexual flowers (leaky dioecy),

and strictly dioecious following the field observations of and reported by Gardner

(2000a). Habit was divided into four categories: herbaceous, suffruticose (weakly woody,

especially at the base), woody with open growth habit, and woody with branches strongly

divaricating.

Ancestral state reconstructions

Based on the presence of one or more endemic species in each respective locality

(see Fig. 1) the geographical range of Coprosma was divided into five unordered areas for the reconstruction of ancestral distributions (1. Malesia [New Guinea, Borneo,

Sulawesi], 2. Australia, 3. New Zealand, 4. Southwest Pacific [Austral Is., Chatham Is.

Juan Fernandez Is., Kermadec Is., Lord Howe Is., Norfolk Is., Samoan Is., Society Is.], and 5. Northeast Pacific [Hawaiian Is., Marquesas Is.]) Per the recommendation of the

RASP 2.1b manual, the smallest number of initial ancestral area states is preferred and preliminary tests using alternative combinations of ancestral areas showed no significant effect on our results, which is in support of this recommendation. Both Statistical

Dispersal Vicariance Analysis (s-DIVA) and Bayesian Binary methods (BBM) analyses were implemented in RASP 2.1b (Yu et al., 2010; Yu et al., 2011) to reconstruct ancestral distribution states. These methods average frequencies of ancestral range distribution across all trees generated from Monte Carlo Metropolis Chain (MCMC) output. For s-

DIVA analyses, the maximum number of areas was kept at three, as no species in the genus occurs in more than two areas (i.e. there are few widespread taxa). Possible ancestral states at each node were then obtained from s-DIVA output. BBM analyses were implemented with four million MCMC generations with four chains (one hot, three cold) run simultaneously with trees sampled every 1000 generations. Given that the best- fit model selected from jModelTest is not implemented natively in RASP, the model F81

+ G was used. A burn in period equivalent to 25% was discarded ensuring convergence of the sampled trees before combining into a consensus tree.

Molecular dating analyses

Prior to implementing the dating analyses in BEAST 1.7.5 (Drummond et al.,

2012), the best-fit model of molecular evolution was determined for each of the six partitions in jModelTest (Posada & Crandall, 1998) using the Bayesian Information

Criterion (BIC). The GTR+G model was the selected model for most of the partitions.

Alternate models were selected for trnQ-rps16 and ETS; however GTR+G also received a high BIC value. As GTR+G is used natively within BEAST, these models were chosen.

Analyses were then run on the CIPRES Science Gateway version 3.2 portal (Miller et al.,

2012).

To produce dated phylogenetic reconstructions, three separate analyses were performed in BEAST each with independent constraints. A fourth analysis was completed as a total evidence approach in which all constraints were applied to the analysis concurrently. The first analysis constrained the root height of the entire phylogenetic reconstruction between 0-50 Ma, as this is a reasonable time estimation of evolution based on previous studies of the Rubiaceae (Bremer & Eriksson, 2009). The second analysis included the 0-50 Ma root height and an additional fossil pollen calibration point for the monophyletic lineage of Coprosma and Nertera. This node was constrained to 5.3-31 Ma with fossil pollen evidence in New Zealand (Graham, 2009).

The third constraint independently assessed the ITS rate of molecular evolution and the rate range was set to 0.38 x 10-9 to 8.34 x 10-9 with an initial starting value of 1.99 x 10-9 after Kay et al. (2006). This range represents estimated ITS rates known for most angiosperms and the initial start value was chosen as the rate found for Gaertnera

(Rubiaceae), as the most reasonable estimate given the caveat of Kay et al. (2006) that phylogenetic relatedness is not necessarily correlated with the rate of molecular change

For all dated phylogenetic reconstructions, the complete concatenated dataset

(four DNA regions; 112 taxa) was used. All species of the genus Coprosma and Nertera

(except C. talbrockiei and C. moorei) were set as monophyletic as indicated from BI and

ML analyses. A starting tree was generated without date or rate prior information for ten million generations, which circumvented initial trees starting with likelihoods of negative infinity. Each analysis was run for 300 million generations with sampling occurring every 10,000 generations. Sampled generations were then downsampled in LogCombiner

1.7.5 (Drummond et al., 2012) and a consensus tree was generated using TreeAnnotator

1.7.5 (Drummond et al., 2012) with maximum clade credibility, mean node heights and discarded burn in. Trees were visualized in FigTree 1.3.1 (Rambaut & Drummond,

2010).

Results

Phylogenetic Analyses

BI and ML phylogenetic analyses reconstruct Coprosma s.s. with five major lineages (Clades A-E; Fig. 3.3) and two monotypic lineages (C. pseudocuneata and C. linariifolia). The first diverging lineage of Coprosma s.s. (Clade A) is composed of eleven taxa; five taxa are from New Zealand and six taxa from New Guinea. Three New

Zealand taxa (C. crenulata, C. foetidissima, and C. serrulata) are nested among the New

Guinea taxa. Sister to Clade A is C. pseudocuneata plus a polytomy of Clade B, Clade C and (Clade D + Clade E + C. linariifolia). Clade B is composed of 17 New Zealand taxa

and C. pilosa of Norfolk Island. Clade C consists of taxa from six different locations

(Australia, Fiji, Hawaii, New Zealand, Lord Howe Island, Samoa) despite only having

eight taxa. The 26 taxa of Clade D are primarily from New Zealand (16 taxa), but are also

from the Austral Islands, Australia, Chatham Islands, Cook Islands, Kermadec Islands,

Lord Howe Island, Norfolk Island, Society Islands, and Pitcairn Island. Clade E is the

largest clade with 38 taxa. Within Clade E, 12 taxa are from Hawaii, 11 from New

Zealand, and six from Marquesas. The remaining Clade E taxa are species found in

Australia, Austral Islands, Juan Fernandez Islands, Kermadec Islands, and Samoa.

Nertera is supported as the sister genus to Coprosma supporting the findings of Cantley

et al. (in press). Additionally, C. moorei and C. talbrockiei are most closely related to

Durringtonia paludosa of Australia and are therefore not part of the monophyletic

Coprosma s.s. lineage.

Geographic origin(s) of Coprosma s.s. and non-New Zealand Pacific Taxa

The three BEAST permutations (root height constraint, Coprosma + Nertera fossil constraint, and combined root height and Coprosma + Nertera fossil constraint) all

inferred similar age estimates at nodes. However, the 95% HPD confidence intervals

were tighter in the combined analyses (the total evidence approach) and are used here

(Fig. 5). The combination of molecular clocking and ancestral state analyses indicate that

the most recent common ancestor (MRCA) of Coprosma s.s. and its sister genus Nertera

occurred in New Zealand during the Oligocene c. 25 Ma (Fig. 3.3 & 3.4). All clades of

Figure 3.3. Phylogenetic reconstruction of Bayesian Inference showing a 50% consensus topology based on two nuclear (ITS and ETS) and two plastid (rps16 and trnQ) DNA regions. Numbers above branches represent Bayesian posterior probabilities and

Maximum Likelihood bootstrap values, respectively. Clades A-E and Coprosma s.s. are as referred to in the text. Branch colors representing geographical distribution are indicated and were coded as following (1. Indo-Pacific [New Guinea, Borneo, Sulawesi], 2. Australia, 3. New Zealand, 4. South Pacific [Austral Is., Chatham Is. Juan-Fernandez Is., Kermadec Is., Lord Howe Is., Marquesas Is., Norfolk Is., Samoan Is., Society Is.], and 5. the Hawaiian Is.). Inset is the phylogram of the Bayesian Inference.

Coprosma s.s. are inferred to have a New Zealand origin where most of the diversification has occurred less than 7 Ma (with the exception of taxa in Clade C

(discussed below). Taxa of Clade A suggest a single dispersal from New Zealand to New

Guinea followed by a back dispersal to New Zealand. Clade B is almost entirely of New

Zealand origin albeit for a single dispersal event inferred to Norfolk Island (C. pilosa).

The biogeographic origins for taxa of Clade C are unclear as the eight taxa are from six different Pacific localities (Australia, Fiji, Hawaii, New Zealand, Lord Howe Island,

Samoa). Clade D taxa are of New Zealand origin and there are four dispersals from New

Zealand to other localities inferred. Inter-Pacific dispersal events, i.e. those not including

New Zealand as a dispersal sink or source, are also inferred (e.g. between C. laevigata of

Cook Islands and C. taitensis plus C. orohenensis of the Society Islands). Clade E taxa share a New Zealand MRCA, but the majority of taxa are found in Hawaii and Marquesas

Islands. Hawaiian taxa of Clade E are infereed to have arrived less than 3 Ma from the

Marquesas Islands or Rapa Iti of the Austral Islands. All Marquesan plus Hawaiian species are sister to C. cookei of the Austral Islands.

Morphological trait evolution in dispersal and establishment

Reproductive Strategy

Dioecy is inferred to have only evolved once for the MRCA of Coprosma s.s.

(Fig. 3.4a) and no reversals to bisexuality were found. Dioecy is not always fully expressed, however. For example, all species in Clade A are known to have a leaky dioecious reproductive strategy while species of Clades B, D, and E have both ‘leaky’ and ‘purely’ dioecious taxa. Clade C is the only strictly dioecious lineage recovered. All outgroup members (including C. moorei and C. talbrockiei) have a bisexual reproductive

strategy without any indications of dioecy. However, a strict division of leaky dioecy and

‘perfect’ dioecy is not found.

Habit

The evolution of woodiness is inferred to have occurred for the MRCA of

Coprosma s.s. (Fig. 3.4c). All outgroup taxa, including its sister genus, Nertera, as well

as the two segregate species C. moorei and C. talbrockiei, are herbaceous. While woodiness emerged only once, the woody subcategories were not monophyletic (Fig.

3.4b). Taxa with strongly divaricating branches, a variation of woodiness are found in

Clades B, D, and E. Taxa of Clade A include trailing liana/ mat forming species C. niphophila and C. perpusilla, suffruticose New Guinea taxa plus C. crenulata of New

Zealand (‘weakly woody’ after Gardner, 2002), as well as the two woody New Zealand taxa C. serrulata and C. foetidissma. Mat-forming or trailing liana taxa occur in Clades

A, C, and D.

Figure 3.4. Character states of Coprosma and outgroups traced onto the Bayesian 50% phylogenetic reconstructions. (a) reproductive strategy, (b) habit, (c) fruit color.

Fruit Color

Reconstruction of the evolution of fruit color (Fig. 3.4c) suggests orange as the initial character state for the MRCA of all clades except for Clade D, which is inferred as unpigmented (=white, clear). Sublineages of Clade D include the development of blue, yellow, orange and polymorphic lineages. Black-fruited taxa occur only in Clades B and

C and taxa with polymorphic fruit colors are only present in Clades B and D. Deep violet-fruited taxa are found in Clades B, C and D. Of the 38 taxa in Clade E, 36 are orange fruited. The remaining two taxa are unpigmented. Orange fruit were independently evolved in the outgroup taxon Durringtonia paludosa of Australia.

Discussion

Character evolution aiding dispersal and establishment of Coprosma s.s.

Our analyses suggest that two morphological innovations of Coprosma s.s. were key in shaping its evolutionary trajectory: a shift to dioecy and a woodiness; neither of which occur in Nertera or outgroup taxa (Fig. 3.4a, b). The shift to dioecy in Coprosma s.s. from a bisexual one draws parallels to that of the well-studied Hawaiian genus

Schiedea. For example, the colonizing ancestor of the 35 extant species of Hawaiian

Schiedea was hypothesized to be hermaphroditic. During its radiation within the

Hawaiian Islands, dioecy evolved twice and is correlated with anemophily and more open drier habitats. In contrast, hermaphroditic Schiedea taxa are associated with moist forest habitats and are not wind pollinated (Sakai et al., 2006). Given these environmental differences and other data, it was hypothesized that the shifts towards drier more open habitats were associated with a selective pressure for outcrossing via wind pollination in

Schiedea in order to cope with a lack of pollinators in novel open environments. A

similar logic can be applied for Coprosma s.s., whereas new habitats in New Zealand were created continuously over the past ca. 25 Ma (Galloway & Kemp, 1981; Smith,

1982), Coprosma s.s. similarly would have been faced with changing environments, particularly with regard to direct exposure to drier, windier, more open habitats over large areas and elevational gradients. Selective pressures could then be expected to promote the evolution of dioecy in a similar manner as is hypothesized for Hawaiian Schiedea.

Determining the applicability of this explanation to the evolution of dioecy in Coprosma will require further study, however.

Insular dioecious taxa, such as that of Coprosma s.s., are often considered disadvantaged over hermaphroditic lineages, yet dioecious taxa are disproportionately more abundant on islands than in continental floras (Crawford et al., 2011). The disadvantages of dioecy were outlined by Baker (1955) who proposed that self- compatible species would be more likely to successfully establish from a single propagule than obligate out-crossing species, which inherently requite at least two individuals to reproduce. Despite this apparently logical reasoning, a paradox exists in which many insular floras—such as those of the Hawaiian, Reunion, Ogasawara and Juan

Fernández Islands—have a high proportion of dioecious taxa represented among their floras in comparison to continental floras. In explanation of this paradox, several authors

(see Barrett, 1996; Baker & Cox, 1984) argue some insular taxa are not completely dioecious in their reproductive strategy, but rather, some individuals exhibit leaky dioecy, which is the ability to occasionally produce hermaphroditic flowers allowing asexual reproduction. Correlating leaky dioecy to establishment of lineages on islands, few initial colonists would be required for establishment in leaky dioecious individuals, despite the

outcrossing nature of a dioecious reproductive strategy. Leaky dioecy is documented for many Coprosma s.s. taxa across the Pacific (Fig. 3.4a), and as Coprosma s.s. is one of the most widespread and species rich genera in the Pacific this suggests that leaky dioecy is a highly successful strategy, at least for many Coprosma taxa.

Woodiness, a character found in Coprosma s.s., but not in its sister group, is also correlated with dioecy in many floras including those of Hawaiian Islands (Sakai et al.

1995), Panama (Croat, 1979), and Puerto Rico (Flores & Schemske, 1984). The correlation of woodiness and dioecy is often attributed to heightened selective pressures for outcrossing in long-lived plants, which may otherwise self (geitonogamy) reducing/plateauing the genetic variation in its population (Bawa 1980; Steiner 1988).

The innovation of woodiness in Coprosma s.s. increases the longevity of individuals, promoting a temporally lengthened ability to contribute genetic resources to subsequent generations in comparison to a short-lived herbaceous species like Nertera.

All but six non-New Zealand taxa are small trees with orange fruit and a non- divaricating habit (Fig. 5). The six exceptions are Hawaiian Coprosma ernodeoides

(trailing shrub with black fruits), C. archboldiana of New Guinea (trailing shrub and burgundy fruits; not sampled), Australian C. nivalis (trailing/mat forming with bluish- purple fruits), C. pilosa of Norfolk Island (small tree with pale violet-grey fruits),

Samoan C. strigulosa (trailing shrub with black fruits) and Fijian C. persicifolia (small tree with blue-black fruits). Clearly, orange fruit color was important in the dispersal of

Coprosma s.s. out of New Zealand to the rest of the Pacific (Fig 3.4c). Lee et al. (1994) noted that orange and black fleshy fruits of Coprosma taxa were highly conspicuous to birds against green foliage and are commonly found in the diets of many New Zealand

forest birds. In relation to dispersal ability, fleshy fruit that is internally ingested by birds—like those of orange and black-fruited Coprosma s.s. taxa—is the most common dispersal mechanism known for plant colonists to the Hawaiian Islands (Carlquist, 1967),

Juan Fernández Islands (Bernardello et al., 2001, 2006), Azores (Schafer, 2002),

Galápagos (McMullen, 1999) and Canary Islands (Bramwell & Bramwell, 2001).

Furthermore, studies of Hawaiian, Canary and the Juan Fernández Islands indicate fleshy fruits are in strong correlation with a dimorphic breeding systems, like dioecy, further solidifying these character states as successful for insular taxa (Sakai et al, 1995;

McMullen, 1999; Bramwell & Bramwell, 2001).

Historical biogeography of Coprosma s.s.

Time, origin, and extensive radiation in New Zealand

Our analyses indicate the MRCA of Coprosma s.s. and Nertera originated in New

Zealand during the late Oligocene ca. 25.6 Ma (Fig. 3.5). The timing of this divergence corresponds with major subduction of the New Zealand landmass ca. 25-22 Ma—referred to as the Oligocene Marine Transgression (OMT)—in which at least a significant portion of New Zealand is thought to have become submerged (Stevens et al., 1988; Cooper &

Millener, 1993; Cooper & Cooper, 1995). Controversy exists around the OMT of New

Zealand, primarily concerning the degree of submergence (i.e. complete or incomplete;

Pole, 1994; Le Masurier, 1996). This argument affects the inference of the age of the

New Zealand flora as it is either considered ‘recent of primarily oceanic origins’ or

Gondwanan. The temporal evolution of Coprosma s.s. and Nertera inferred in our analyses neither confirms nor rejects a total submergence of the New Zealand. If

complete submergence of New Zealand occurred, their MCRA could have colonized

New Zealand from Australia or some other nearby area, which was not submerged during

the Oligocene. Alternatively, in the scenario of an incomplete drowning of New Zealand

during the OMT, the MRCA of Coprosma s.s. and Nertera would have been able to persist on the low-laying emergent islands of New Zealand during the Oligocene, albeit likely having experienced a bottleneck in diversity as a result of a diminished size and habitat remaining on the continent. An incomplete drowning of New Zealand is supported by Onychophora (Allwood et al., 2010) Astelia (Birch et al., 2012), Agathis

(Knapp et al., 2007) and giant weta taxa (Anostostomatidae; Trewick and Morgan-

Richards, 2005), but at this point is difficult to say which is the most accurate description regarding the evolution of Coprosma s.s. and Nertera. Furthermore, the location of

Coprosma/Nertera-like microfossil records from mid Miocene and mid Oligocene fossil deposits of Murray Basin of southeastern Australia and New Zealand (Couper, 1960;

Graham, 2009; MacPhail, 1999) could support both scenarios. Moreover, Sharma &

Wheeler (2013) point out that the New Zealand geology predisposes endemic lineages to root age shifts when utilizing molecular clocking techniques affecting the temporal inference of lineages and could certainly be the case in our dated analyses here.

Following events of the OMT, the extant diversity of Coprosma s.s. in New

Zealand diverged primarily in the mid Miocene between ca. 10-8.5 Ma (Fig. 3.5) and can be correlated generally with New Zealand’s history of rapid change in of tectonic activity, often refered to as the Kaiokoura Orogeny. These changes included slow uplift commencing ca. 25 Ma with the activation of the modern Pacific-Australian plate boundary and climaxed around 5 Ma (Cooper & Cooper, 1995; Kamp, 1992). The newly

formed land resulting from this uplift was rugged in topography and much larger than

previously in the Oligocene (Ollier, 1986). Drastic changes occurred to the New Zealand

landscape during the Pliocene onward, where a markedly increased tempo of tectonic

activity, volcanism (Ollier, 1986; Cox and Findlay, 1995; Batt et al., 2000), and over 20

glaciation cycles are documented (Suggate, 1990; Pillans et al., 1992). The geoclimatic

events of New Zealand shaped the contemporary landscapes of the continent creating

habitat heterogeneity, which in turn promoted the opportunity for niche differentiation

and adaptive changes to accumulate among the New Zealand biota (Winkworth et al.,

2005). Our analyses indicate Coprosma s.s. lineages underwent adaptive radiations into

multiple habitats at high and low elevations and it is during this time that nearly all extant

New Zealand taxa of Coprosma s.s are indicated to have diverged.

The pattern of Coprosma s.s. diversification in New Zealand is not always clear,

but it appears a number insights can be made. Firstly, it appears that niche partitioning

can explain observed diversity, at least for some taxa. In the most simple of cases, this is

exemplified in three taxa resultant from a back dispersal from New Guinea to New

Zealand less than 6 Ma (Clade A; Fig. 3.3). From a shared ancestor adapted to the cold

highlands of New Guinea, these three New Zealand species diverged into a non-

divaricating subalpine shrub of boulderfields and tussock habitat with large coriaceous

leaves (C. serrulata), a divaricating subalpine shrub of boulderfields and tussock habitat

with small coriaceous leaves (C. crenulata) and a beach forest shrub found from near the

coast to subalpine habitat without coriaceous leaves (C. foetidissima). Alternatively, a

second pattern is exemplified by taxa of clade B, which have very similar morphological

characters and often hybridize with one another. Given their young age, these species

could be the product of Pliocene glaciation cycles of New Zealand in which there is

incomplete lineage sorting or the beginning stages of niche partitioning among these taxa.

Another inference regarding the patterns of Coprosma diversification is to note that

dispersal away from New Zealand, other Pacific localities often results in a

‘homogenized’ Pacific morphology (ex. non-New Zealand taxa of Clades D & E; Fig.

3.3). More specifically, species across the Pacific, although not all directly related, evolved morphologies that are orange-fruited, large leaved and non-divaricating. This is perhaps due to insular selective pressures favoring these character states resulting in convergent evolution. For example, the largest non-New Zealand radiation of Clade D includes seven species with orange fruits, large leaves and a non-divaricating habit, which is closest in morphology with Marquesan and Hawaiian taxa, but their closest relatives are from New Zealand having small leaves, blue fruits and are divaricating. The precise factors promoting a homogenized non-New Zealand Coprosma are not clear from this study, but warrant further research.

Figure 3.5. Cladogram from BEAST 1.7.5 analyses. Branch node position indicate mean prior distributions whereas number on nodes indicate the estimated 95% HPD range.

Dispersals from New Zealand to elsewhere

Establishment of Coprosma s.s. outside of New Zealand occurred in a rather

complex manner. Our results indicate that a minimum of 30 long distance dispersal

events are required in order to explain the extant distribution Coprosma s.s. across the

Pacific (Fig. 3.5). Inferred dispersals occurred directly to distant localities, in a stepping-

stone fashion, as back dispersals and as repeated dispersals to the same geographic

locality. The number of dispersals documented is remarkably high, suggesting frequent

movement of seeds from source areas to places around the Pacific. Furthermore, the

frequency of movement suggests that the dispersals are not governed by the slim statistics

of a sweepstakes chance arrival to novel environment, but rather that more directed

dispersal was likely occurring. The distribution of Coprosma s.s. includes nearly every

high island archipelago in the Pacific for which suitable habitat exists. Even pollen

records near the summit of Rapa Nui indicate that Coprosma once occupied this island

before human degradation of that environment (Azizi & Flenley, 2008; Rull et al., 2010).

Notably, Coprosma s.s. is not present in Southeast Asia, the Galápagos Islands or New

Caledonia. However, these localities likely are unsuitable for the establishment of

Coprosma s.s. as it is too dry in the Galápagos and hot/humid in most of South East Asia.

Coprosma s.s. is curiously absent from New Caledonia, but the monotypic outgroup

taxon Normandia neocaledonica does occur here, which perhaps excludes the

colonization of Coprosma s.s. through competitive forces, as it already occupies its

habitat niche or these taxa are now extinct.

Two large Pacific lineages are the result of stepping-stone dispersal between proximally close islands/archipelagoes. The first of these lineages was established from

New Zealand ca. 5.6 Ma involving Austral, Cook, Lord Howe, Norfolk, Pitcairn, and

Society Islands (Fig. 3.3, clade D). The precise biogeographical movement between these

localities is not clear from the results. However, from a parsimonious standpoint, it is less

likely that direct dispersals from New Zealand established taxa in each of these far-flung localities. Alternatively—and perhaps more likely—the extant distribution is the result of stepping-stone dispersals through multiple archipelagos/islands (ex. from New Zealand to the Cook Islands then to the Society Islands then to Austral Islands and finally to Pitcairn

Island). Despite our inability to infer the precise biogeographical relationships between these taxa, an early division separates taxa on islands to the northwest of New Zealand

(Lord Howe Islands and Norfolk Island) from northeastern taxa (Austral, Cook, Pitcairn, and Society Islands). The largest radiation of Coprosma s.s. outside of New Zealand is the result of stepping-stone dispersals between Austral, the Hawaiian Islands and

Marquesas Islands ca. 6.3 Ma; correlative with the ages of the currently emergent islands.

The precise biogeographical relationship among these three localities is tenuous, but it is suggested that the MRCA arrived to Marquesas Islands first and then two dispersals are inferred: from the Marquesas to the Hawaiian Islands and from the Marquesas to the

Austral Islands. It is also possible that the MRCA arrived first to the Austral Islands and reached the Hawaiian Islands via a stepping stone dispersal via the Marquesas Islands.

Direct dispersals—i.e. those from New Zealand directly to another locality— appear to have played as important of a role as stepping-stone dispersals have for the establishment of Coprosma s.s. across the Pacific. In fact, the greatest range expansion of

Coprosma s.s. occurred ca. 4 Ma from a single direct dispersal event of 9,000 km from

New Zealand to the Juan Fernández Islands. This distance is roughly equivalent to the

distance between Berlin, Germany and Tokyo, Japan. The majority of inferred direct

dispersal events occurred to Australia and archipelagoes proximal to New Zealand

(Chatham, Kermadec, Lord Howe and Norfolk Islands). However, despite the inference

of these as direct dispersal events, extinct taxa could represent intermediate localities

confounding our interpretation.

Repeated dispersals to the same geographic locality is important in discussing the

overall biogeographic pattern of Coprosma s.s. across the Pacific because they indicate

frequent movement among Pacific localities. Repeated colonizations refers to two

colonizing lineages establishing at a single insular locality independently from each

other. For example, the Samoan Islands were colonized twice, once by an ancestor of C.

strigulosa and then separately by an ancestor of C. savaiiensis (Fig. 3.3). Repeated

dispersals occurred to at least seven different Pacific localities. This pattern is not novel

to Coprosma s.s., but Coprosma s.s. represents the highest documented cases from a

single angiosperm genus thusfar. For example, three known colonizations of Scaevola are

documented for the Hawaiian Islands via internal bird ingestion and water dispersal

(Howarth et al., 2003; Howarth & Baum, 2005). In addition, there were two colonizations of internally bird dispersed Melicope J.R. Forst. & G. Forst. to the Marquesas Islands

(Harbaugh, 2009), and two colonizations of wind dispersed Metrosideros to Lord Howe

Island (Papadopulos, 2011). Only Santalum compares with Coprosma s.s. in total number

of repeatedly colonized locations which occurred three times: to the Bonin Islands, the

Hawaiian Islands and Cook Islands (Harbaugh et al 2008).

Conclusions

Assessing the overall pattern of movement that is inferred for Coprosma s.s.

across the Pacific indicates no clear progression in size, age or direction of movement for

the entire lineage. Rather, the pattern seems stochastic in process in which movements

have occurred in all directions, at all times and have either radiated into large numbers of

new taxa upon arrival to a new locality, or have not. Coprosma s.s. achieved its current

distribution though greater than 30 bird-mediated dispersal events over a relatively short

period, geologically speaking. The unpredictability of its establishment and reproductive

success upon arrival makes it difficult to postulate the precise character or combinations

of characters that were biogeographically important, or may highlight that the

combination of traits increases its potential successful establishment. The genus shares various similarities with some other widespread lineages such as Astelia (Asteliaceae),

Metrosideros (Myrtaceae), but also differs, which has resulted in a unique biogeographic

history for Coprosma. Coprosma and Astelia both share fleshy orange fruits that were

likely dispersed via internal transport by birds, and both are dioecious (Birch et al., 2012).

Both also derive largely from ancestral lineages of New Zealand and have cases of both

direct long distance and stepping stone dispersal. However, extant taxa of Astelia have

reached only a small portion of the areas in which Coprosma is found (Birch et al., 2012;

Birch et al., 2013). Metrosideros has a similar widespread distribution across the Pacific

as Coprosma, but does not share the same dispersal mechanism or breeding system (i.e.

Metrosideros species are bisexual and have wind dispersed seeds; Wright et al., 2001;

Percy et al., 2008). Metrosideros also differs from Coprosma with a considerably smaller

number of species, but this could be because of over-dispersal (frequent) of their small

wind borne seeds (a continuance of gene flow from a source population slowing the rate

of endemicity) or for some other unknown reason. Santalum also shares many

distributional similarities with Coprosma. Santalum is widespread, bisexual and has

fleshy bird dispersed fruits, but is not as abundant as Coprosma.

Assessment of the factors that allow taxa to become widespread in the Pacific

coupled with new insights from Coprosma, we can more readily identify major

evolutionary factors that promote the evolution of multiple endemic species and insular

adaptive radiations in lineages that colonize frequently. For Coprosma, birds likely

carried the seeds frequently to new landmasses, which is perhaps slightly more

successful than passive wind dispersal. However, the sheer number of wind-dispersed

seeds alone could equal the heightened chances of arriving to an island via bird-mediated

dispersal. Coprosma exhibits a great range of environmental tolerance and its variability in breeding system is also in part responsible for the successful establishment of the

Pacific. This study certainly does not identify all of the factors that were historically important for the success of insular plants, but the combination of traits in Coprosma was

clearly powerful in the establishment of this widespread and species rich insular Pacific

radiation.

CHAPTER 4 MORPHOLOGICAL, ECOLOGICAL AND MOLECULAR SUPPORT FOR NEW SUBGENERIC RELATIONSHIPS OF COPROSMA (RUBIACEAE)

Abstract

Coprosma is one of most widely distributed and most species rich genera of the

Pacific with over 110 species north to south from New Zealand to the Hawaiian Islands

and east to west from Borneo to the Juan Fernandez Islands. Morphological complexity

among taxa makes it difficult to establish relationships among the species as there is often

considerable variability. The traditional, morphologically based description that divided

Coprosma into two subgenera was based only on leaf size and number of flowers per inflorescence, but many believed this to be an incomplete morphological assessment. In order to assess relationships among Coprosma species and to evaluate morphological characters applied to understanding the evolution of the genus, molecular phylogenetic analyses for 105 Coprosma taxa plus outgroups from two nuclear (ITS and ETS) and two chloroplast (rps16 and trnQ) DNA regions using Bayesian Inference and Maximum

Likelihood methods were conducted. Morphological characters were traced on the resultant phylogenetic reconstructions using Mesquite 1.75 with the parsimony criterion.

Results indicate that existing subgenera and sections are polyphyletic as delimited by previous morphological treatments. Upon the initial divergence of Coprosma from sister genus Nertera, the evolution of dioecy, woodiness and a switch to predominately more exposed environments is indicated. However, Coprosma retained characters that are plesiomorphic to both genera including: leaves smaller than 25mm in length, 1(-2) flowers on a terminal inflorescence borne on short branches, number of calyx and corolla lobes 4, stamens 4, orange fruit and 2n=44. The evolution of characters of extant clades

trends towards the multiplication of parts (ex. solitary flowers to axillary peduncles with many flowers per inflorescence, corolla lobes from 4 to 5 or more, etc). Morphological trends among lineages do not clearly divide sections with synapomorphies, but rather show a great deal of homoplasy often expressed as multiple independent gains of a character state. Based on results of morphological and molecular data, a taxonomic revision of the genus with six subgenera is outlined.

Introduction

Taxonomic History

Coprosma was first collected in 1769 during a voyage of Captain James Cook to New Zealand. Sir Joseph Banks and Dr. D. C. Solander described seven species under the genus Pelaphia, but this name was never published. It was not until J. R. Forster and

G. Forster examined material from Captain Cook’s second voyage to New Zealand in

1773 that the name Coprosma was applied to the genus and validly published in 1776

(Oliver, 1935). The genus name Coprosma is a derivative of the Greek/Latin words

“kopros” and “osome” roughly translating to ‘smelly corpse,’ and is in reference to the pungent, fetid scent of methanethiol released from the bruising of leaf tissue in the type species, C. foetidissma (Oliver, 1935; Webb, 1996) and many other species of the genus

(pers. obs.).

Since the initial description of Coprosma in 1776, only three comprehensive systematic treatments of the genus were published. The first of these was by Oliver

(1935), which has yet to be superseded as the most comprehensive monograph of the genus. More recently, Heads (1996) produced a controversial revision of the genus, which was then refuted and slightly modified by Gardner (2002a). Coprosma is a taxonomically challenging group as few morphological characters distinguish species. A high level of interspecific morphological variation compounds the difficulty in species identification and delimitation. Furthermore, hybrids are common. For example, C. propinqua of New Zealand is known to hybridize with at least twelve other species

(Oliver, 1935). For these reasons, taxonomists have avoided attempts to hypothesize phylogenetic relationships among Coprosma species. Joseph Hooker (1853) mentioned,

“[Coprosma species] are so variable that I quite despair at reducing them to anything like proper order,” and Cheeseman (1925) described it as, “next to Veronica, the genus

Coprosma is the most puzzling in the New Zealand flora.” More recently, Fosberg (1982) claimed, “to gain an understanding of this genus alone would be a major undertaking.”

Following the above observations, few phylogenetic hypotheses exist for taxonomic ranks below the generic level. Oliver (1935) proposed sections in which

‘simple’ morphology (i.e. few floral parts, solitary flowers, etc.) were more plesiomorphic in character states than those sections that were ‘increased morphological complexity’ (i.e. more derived). By his definition simple morphology included species with small leaves, prostrate habit and solitary flowers in comparison and species with a

‘complex morphology’ had large leaves, clustered inflorescences and a larger growth habit. Modern treatments (Oliver, 1935; Heads, 1996; Gardner, 2002) subdivide

Coprosma based primarily on leaf size (greater or less than 2cm in length), inflorescence structure (solitary or not) and habit (Hooker, 1867), but a great deal of overlap and uncertainty in the validity of these characters in distinguishing species remains. In his combined use of morphological characters and geographic distributions of species, Heads

(1996) proposed a biogeographic hypothesis in which Coprosma was separated into two subgenera divided along an east-west boundary from an ancient vicariant split of a previously widespread Pacific centered taxon. The separation of his two subgenera was based on morphology where species with large leaves and complex inflorescences were of subgenus Lucidae and small leaved species with solitary female flowers were of subgenus Coprosma. Species of the western subgenus Coprosma are from New Zealand,

New Guinea, Australia and high islands of the Indo-Pacific region (Sulawesi, Borneo,

and Java) with the small leaved, solitary flowered C. ernodeoides of Hawaiian Islands considered as an outlier of subgenus Coprosma. Subgenus Lucidae overlaps with

subgenus Coprosma only in New Zealand and the Hawaiian Islands, but is otherwise

found throughout the remote Pacific where subgenus Coprosma is not found. Although

Heads (1996) provided a biogeographic explanation for his subgeneric classification,

formal hypotheses for the relationships between sections was not made.

Gardner (2002a) disagreed with Heads, and argued that the reasoning for the

observed east-west pattern of distribution of Coprosma might reflect a biological reality,

rather than an ancient vicariant event of a widespread ancestor. For example, Gardner

(2002) noted that most species within western subgenus Coprosma are adapted to

temperate environments (e.x. the highlands of New Guinea, the Southern Alps of New

Zealand and subalpine ecosystems of the Hawaiian Islands). As it stands, there is little

available habitat in the eastern distribution of the genus as many of the islands (such as

Samoa, Fiji and the Marquesan Islands) are not high enough currently to support a

montane flora. However, the Hawaiian Islands are high enough, and furthermore, the

species C. ernodeoides is found in the respective montane habitat found these islands.

Gardner (2002a) therefore refuted the subgeneric classification of Coprosma by rejecting the rationale of Heads (1996) that the biogeographic distribution plus floral and leaf characters. However, even though Gardner (2002a) believed the conclusions of Heads

(1996) were based on an incomplete analysis of characters, Gardner (2002) agreed that an east-west division could in fact be a reality at a lower taxonomic level for at least some

sections of the genus.

Coprosma and uncertain position of allied genera

Relating to the difficulty in species identification of Coprosma, the taxonomic identity of the closely related genera Nertera and Leptostigma is also uncertain.

Leptostigma and Nertera are superficially similar morphologically (small leaves, herbaceous creeping habit, bisexual flowers) and previous considerations proposed these three genera possibly represent a single lineage (Hooker, 1867; Cheeseman, 1887). Some

Coprosma species are scarcely different from Nertera and Leptostigma taxa (ex. C. perpusilla and C. niphophila), but primarily differ in having secondary growth and a dioecious breeding system. The three genera remained separated until the controversial revision by Heads (1996) in which Leptostigma and Nertera were sunk into Coprosma as sectional divisions of subgenus Coprosma. Heads (1996) considered the morphological characters of Leptostigma and Nertera to be well within the circumscription of Coprosma sensu strictu. However, Gardner (1999a; 1999b), and other authors previously (Oliver,

1935; Fosberg, 1982) defended the position of Nertera and Leptostigma at the rank of genus based upon a number of reliable morphological characters. Primarily, floral

(stamen & corolla length) and fruit characters (fleshy vs. semi-fleshy) plus ploidy levels differentiate Leptostigma and Nertera from each other. These two genera are in turn differentiated from Coprosma in their herbaceous habit, and bisexual flowers. Within

Coprosma, there is a wide range of morphological diversity among taxa whereas

Leptostigma and Nertera taxa are quite uniform in morphology (Fosberg, 1982; Puff,

1986; Gardner 1999a, 1999b, 2002a). A phylogenetic study of the tribe by Anderson et

al. (2001) supported the retention of the three genera as each their own monophyletic

lineage, although not all species were sampled.

Reproductive and vegetative characters of Coprosma

Floral and Fruit morphology

Coprosma is generally regarded as dioecious, however, this is not strictly the

case. At least 28 species occasionally produce bisexual flowers (leaky-dioecy) and C.

moorei and C. talbrockiei are entirely hermaphroditic (Oliver 1935, Moore & Mason

1974, van Royen 1983, Utteridge 2002, pers. obs.). Stamens typically number four or five, but can be up to ten (Oliver 1935; van Royen 1983). The ovary is usually biloculate with one ovule per locule (Oliver 1935, Puff 1982). The calyx is smaller than the corolla, and can be reduced or even absent in male flowers (Oliver 1935, Allan 1961, Markey &

de Lange 2003). Corresponding with anemophily, the flowers are small and colored dull

hues of green, yellow, cream and brown, and often stippled red or purple. The

inflorescences range from large, thyrso-paniculate (thyrsoid) to considerably reduced inflorescences of solitary flowers (Puff 1982, Puff & Robbrecht 1988, Gardner 2002b).

The fruit is a fleshy, globose drupe with two (occasionally three or four) plano- convex pyrenes (Cheeseman 1925, Oliver 1935). The exo- and mesocarp form a succulent layer around the drupe, which consists of the embryo and testa encased in a heavily sclerotinized endocarp (Gardner, 2002a). The radical emerges prior to the cotyledons, and this is achieved by the separation of an apical operculum from the endocarp (Gardner, 2002a). There are a number of fruit colors (white, blue, black, violet, yellow, pink, orange, red), but most species are orange-red in color (Oliver, 1935).

Leaves and Interpetiolar Stipules

Most leaves of Coprosma species are simple, entire and coriaceous (Oliver,

1935). Domatia are present on the leaves of many species, a trait that is found in no other genus of tribe Anthospermeae (Puff, 1982), although they are absent from many species with small leaves. Like that of other members of the Rubiaceae, most Coprosma species have opposite leaves with a pair of interpetiolar stipules (Oliver, 1935). Occasionally, leaves occur in whorls of three, which is known for some taxa of the Hawaiian Islands

(Wagner et al., 1999; pers. obs), Marquesas Islands (Wagner & Lorence, 2011), New

Guinea (Gardner, 2002b; van Royen, 1983; Utteridge, 2002) and C. talbrockiei of New

Zealand (Markey & de Lange 2003). Leaf size is highly variable within the genus and often correlated with habit. For example, micro and leptophyllous leaves are associated with mat forms and low shrubs whereas mesophyllous leaves are associated with large shrubs and trees (Wardle, 1991). Variation in leaf characters can also vary within a species, an even within and individual (pers. obs.). Oliver (1935) notes that leaf variability is often associated with growth conditions of individuals of the same species in different microhabitats (ex. larger leaves tend to occur in area with closed canopies, smaller leaves in exposed habitat).

Interpetiolar stipules morphology has received much attention as a reliable character for the determination of individual species that are otherwise are very similar in appearance (see Oliver, 1935; Allan 1961; Glenney et al., 2010). Stipule margins range from entire to denticulate, and sometimes form a sheath around the stem. They can have a variety of appendages such as cilia, denticles, glandular colleters that line the margin and

outer surface, which are often diagnostic for species recognition (Allan, 1961; Utteridge,

2002; Markey & de Lange 2003).

Secondary Growth and Habit

The woody perennial habits of Coprosma species can be categorized into three major growth forms: i) large leaved, openly branching tall shrubs and trees, ii) small leaved divaricating shrubs and subshrubs, and iii) prostrate, mat-forming plants. The divaricating habit is restricted to New Zealand (Oliver, 1935; Heads, 1996). Divarication is the combination of branches opening at wide angles with short internodes that interlace with neighboring branches. The resulting habit can occur in low shrubs or erect trees and perhaps is best described as an extensively interlaced mass of tangled branches

(Greenwood & Atkinson, 1977; Tomlinson, 1978; Wardle, 1991). Small leaves less than

25mm in any dimension are notable in species expressing divarication (Kelly, 1994), but many species without divarication also have small leaves (i.e. the mat-forming, sometimes trailing and prostrate shrub, C. ernodeoides of the Hawaiian Islands).

Similar to the plasticity of vegetative and reproductive characters observed at both inter and intraspecific level in Coprosma, it is not surprising that the genus also occupies numerous different habitats. However, Coprosma is particularly abundant in exposed environments such as forest margins, wetlands, tussock lands, grasslands, solifluction terraces and boulder fields. It is particularly represented in shrublands and grasslands in subantarctic, montane and dry coastal situations. (Oliver, 1935; Heads,

1996; A. Markey & J. Cantley, pers. obs.). In New Zealand, the genus is pioneer of areas of primary and secondary succession (Heads, 1996). In the Hawaiian Islands, C.

ernodeoides is a primary successional species on new lava flows (Wagner et al, 1999; J.

Cantley pers. obs.). Some species are commonly found in coastal situation such as in

Kermadec, Lord Howe, New Zealand, and Norfolk Islands (Gardner 2002a).

As a molecular phylogeny did not exist previously and as difficult

morphological identification of species has hindered our understanding, no test of the

relationships below the rank of genus has been thoroughly completed for Coprosma.

This investigation sought to combine molecular phylogenetic data with morphological,

ecological, and karyological data in order to better understand the evolutionary history of

Coprosma species and their relationships to one another. Specifically, we aimed to test

Heads’ (1996) hypothesis that Coprosma can be subdivided into two subgenera. We also examined the statement of Oliver (1935) that species have evolved from a ‘simple’ to a more complex morphology.

Methods

Taxonomic sampling

Coprosma (105/110 species) and five outgroup taxa, Durringtonia,

Leptostigma, Nertera, Normandia and Opercularia were sampled for this study. Samples include the entire geographic range of Coprosma, except for Bougainville Island for which samples were not available. Herbarium material for morphological study was obtained from herbaria and personal collections in Hawaii and New Zealand.

Durringtonia, Leptostigma, Nertera, Normandia and Opercularia, were used as outgroups for phylogenetic purposes (Anderson et al., 2001; Markey et al., 2004).

Molecular data and phylogenetic analyses

DNA was obtained from herbarium and silica dried material, and from the

Hawaiian Plant DNA Library (Randell and Morden, 1999). Samples were prepared by

hand grinding leaves with liquid nitrogen in a mortar and pestle followed by extraction

using DNeasy Plant Mini Kit (Qiagen, Valencia, California, USA). For difficult-to-

extract samples, the QiaAmp DNA Stool Mini Kit was used following the manufacturer’s

protocols, with a lengthened initial incubation period of 2 hours at 70˚C. Primer

sequences used and Polymerase Chain Reaction (PCR) protocols are as listed in Table 2

for each DNA region sequenced. All PCR reactions were performed with 25 µl of

reaction cocktail containing 12.75 µl of sterilized H2O, 2.0 µl of 20 mM dNTPs

(Pharmacia) in an equimolar solution, 2.5 µl of 10x PCR reaction Buffer A (Promega),

1.25 µl of 25 mM MgCl2, 0.5 µl 10 mg/ml Bovine Serum Albumin (Sigma), 1 µl of 10

mM of each of the two primers, 0.5 µl Biolase Red Taq DNA polymerase enzyme

(Bioline) and 4 µl of DNA template. The final amount of DNA template and PCR

reaction cocktail and Taq was adjusted as necessary to generate sufficient PCR products

for DNA sequencing. Samples were purified prior to sequencing using an Exo-Sap

enzymatic PCR product pre-sequencing protocol (USB) for 45 minutes.

A final volume of 8.2 µl was used for sequencing reactions. This consisted of 2.0

µl of sterilized H2O, 3.2 µl of 1 mM primer and 3.0 µl of purified DNA template.

Sequencing was conducted at the Advanced Studies of Genomics, Proteomics and

Bioinformatics facility at the University of Hawai‘i at Manoa. Sequences were edited using Sequencher 3.1.10 (Gene Codes Corp., Ann Arbor Michigan, USA) and aligned by

MUSCLE (Edgar, 2004) using default parameters as implemented in MEGA 5 (Tamura

et al., 2011). Alignments were then manually inspected and adjusted as needed.

To test for congruence between gene regions and between chloroplast and nuclear

genomes, all gene regions were subjected to the partition homogeneity test (ILD) as

implemented in PAUP* ver. 4.0b10 (Swofford, 2003) using 1000 replicates, with TBR

(tree-bisection reconnection) branch swapping and the MulTrees option turned on. No

significant incongruities were found among any of the individual partitions or between

the chloroplast genome regions (rps16 intron and trnQ-rps16 intergenic spacer) when

compared to nuclear (ITS and ETS) genome regions (p=0.60), so a concatenated data set

was used for Maximum Likelihood (ML) and Bayesian Inference (BI) analyses. Prior to

running ML and BI analyses, the same best-fit model (GTR + I + Γ) was selected for all

individual gene regions as well as the combined nuclear data set using the Akaike

Information Criterion (AIC) in jModelTest 2 (Posada and Buckley, 2004). ML analyses

were implemented in RaxML v7.0.4 (Stamatakis, 2006) for individual and combined

datasets. Nonparametric bootstrap replicates (1000) for all ML analyses were calculated

with the thorough bootstrap replicate option selected and allowing all free model

parameters to be estimated.

BI analyses were implemented in MrBayes (Hueselenbeck and Ronquist, 2001) for individual regions and the concatenated datasets. The combined dataset was divided into six partitions (ITS1, 5.8s, ITS2, ETS, rps16 intron and trnQ-rps16) and Markov

Chain Monte Carlo (MCMC) sampling was performed with two replicates of four chains

(one hot and three cold) each with a heating temperature of 0.2. Five million generations were completed with sampling occurring every 1000 generations. A discarded burn-in

period was estimated by visually inspecting plotted log likelihood values versus

generation time to determine the point at which convergence had been reached using

Tracer v1.5 (Rambaut and Drummond, 2007). The remaining trees were combined to construct a consensus tree where Bayesian posterior probabilities (PP) were calculated for internal node support of the resulting phylogenetic reconstruction.

Morphological and taxonomic data

Morphological and taxonomic data were collected from multiple sources including: herbarium material from AK, BISH, CANB, CHR, HAW, MEL, PAP, PTBG and WELT; from personal observations of living material in New Zealand (2010) and

Hawaii (2008-2013); and extensively from taxonomic literature sources and personal communications. The primary sources of Coprosma taxonomic information were obtained from Oliver (1935), Allan (1961), Heads (1996), Gardner (2002) and (Glenny et al., 2010). However, as many species of Coprosma were only treated in their initial descriptions on remote Pacific islands, numerous other sources were included—but not limited to—Wagner et al (1999) for Hawaiian taxa, Florence (1986) and Wagner &

Lorence (2011) for Marquesan taxa, Fosberg (1937) for taxa of southeastern Polynesia,

Gardner (2002b) and Utteridge (2002) for New Guinea/ Indo-Pacific taxa, and Thompson

(2010) for Australian taxa.

To assess the phylogenetic utility of characters previously used to delineate species and subsections, data was divided into vegetative, reproductive, habitat and chromosome number traits. Two vegetative traits that have been regarded as the most important vegetative characters in past taxonomic treatments were assessed: leaf size and

habit. Leaf sizes were subdivided into three categories: taxa in which all leaves are i) small <25mm in length, ii) ±25mm (including those species in which at least some leaves are recorded both larger and smaller than 25mm in length), and iii) >25mm. Habit was divided into four categories: herbaceous, woody & non-divaricating, woody & strongly divaricating, and mat-forming/ prostrate trailing.

A total of eight reproductive traits were assessed: fruit color, inflorescence type, number of male flowers per inflorescence, number of female flowers per inflorescence, number of stamens per male (or bisexual) flower, number of male corolla lobes, number of male calyx lobes (if present), and breeding system. Fruit colour was divided into seven categories: black, blue, burgundy, orange, red, white (unpigmented) and polymorphic.

Six types of inflorescence were categorized: i) 1(-2) terminal flower(s) on short branches, ii) stout axillary peduncles with many flowers, iii) fascicle-like brachyblast, iv) axillary but non-pedunclate (usually 1(-2) flowers), v) either terminal or axillary flowers (usually

1(-2) flowers), and vi) panicles. Many species of Coprosma maintain a degree of variability in floristic traits (ex. 4 or 5 stamens per flower of a taxon), so the most commonly noted condition was used where possible. Breeding system was divided into three categories: hermaphroditic, dioecious with occasional bisexual flowers (=leaky dioecy), and strictly dioecious. Chromosome counts were divided into four categories: i)

2n=44, ii) 2n=88, iii) 2n=132, iv) 2n=220. Habitat was assessed for each species and was divided into five categories: exposed, not exposed, both exposed and not exposed, coastal, and bogs.

Figure 4.1. Phylogenetic reconstruction of a 50% Bayesian Inference topology indicating taxonomic affiliations of Heads (1996) and the new sectional classification.

Evolution of Morphological Characters

Ancestral morphological characters were traced onto the Bayesian Inference tree generated from the molecular data using the parsimony criterion in Mesquite version

2.75 (Maddison and Maddison 2010).

Results

Polyphyly of subgeneric and sectional divisions of Heads (1996)

BI and ML analyses were similar in topology as no major differences occurred between them. ML bootstrap values were generally lower than Bayesian posterior probabilities (Fig. 4.1). Furthermore, findings support Andersson et al. (2001) and (see

Chapters 2 & 3) that Coprosma is sister to Nertera while Leptostigma is less closely related. Two species of Coprosma, C. moorei and C. talbrockiei, are more closely related to Durringtonia paludosa of Australia.

The assessment of the bipartite subgeneric relationships of Heads (1996) (i.e. assessment of the monophyly of subgenus Coprosma and subgenus Lucidae) indicates that both subgenera are polyphyletic. Subgenus Coprosma is first diverging and subgenus

Lucidae is interspersed many times throughout (Fig. 1; blue and green branches).

Additionally, all sections as described by Heads (1996) were polyphyletic (Fig. 4.1). In particular, section Microcoprosma occurs in nine different lineages. Two species, C. talbrockei and C. moorei of section Moorei were found to be outside the Coprosma sensu stricto. The informal group “Hawaii” after Heads (1996) only occurs in two lineages where only the placement of C. serrulata differed from Heads (1996) prediction.

Vegetative and chromosome number evolution (Fig. 4.2)

Outgroup genera (Nertera, Leptostigma, Durringtonia) are all herbaceous and relatively small plants, which appears to represent the ancestral condition. Coprosma, on the other hand is woody with shrubs and trees among its members. Additionally, while most Coprosma species share small leaves and a chromosome count of 2n=44 with its sister genus Nertera, there are also counts of 2n=88, 132 and 220. The primary vegetative difference separating Nertera and Coprosma is the development of secondary growth

resulting in a woody habit of Coprosma. There is also no evidence of polyploidy in the

outgroup genera.

Leaf size and habit, characters used in past treatments to separate taxa do not

show clear phylogenetically based patterns. Both Nertera and Coprosma have small

leaves (<25 mm) but within Coprosma, where both large and small leaves are found, leaf

size (greater or less than 25mm) has shifted at least 10 times, making the character highly

homoplasious in nature. The woody habit of Coprosma also varies greatly over the

phylogenetic reconstructions. Similarly, there are strongly divaricating shrubs or trees,

trailing or mat-forming species and typical non-divaricating open-branched shrub or tree

species in multiple locations of the phylogenetic reconstructions. Coprosma shares an

ancestral chromosome number with Nertera of 2n=44. However, 2n=44, 88, 132, or 220

are found sporadically found throughout Coprosma. The fixation of 2n=220 for three

species (C. ernodeoides, C. atropurpurea and C. pumila) is the only specific instance in

which chromosome number is phylogenetically informative.

Figrue 4.2. Vegetative morphological characters reconstructed on the 50% Bayesian Inference topology. A) phylogeny representing the new sectional arrangement from the taxonomic treatment of this study. B) Habit characteristics. D) Leaf sizes of species

Reproductive character evolution (Fig. 4.3 & 4.4)

Dioecy (Fig. 4.3)

The evolution of dioecy appears first in the most recent common ancestor of

Coprosma separating it from the bisexual outgroups. Many Coprosma species produce some bisexual flowers, but these leaky-dioecious taxa do not form a monophyletic

100

lineage. There is further homoplasy in this character as there reversals to leaky dioecy

occur after the initial divergence of taxa to dioecy. Leaky-dioecy is indicated to have

evolved independently at least eight times (Fig. 4.1).

Inflorescence structure (Fig. 4.3)

The earliest diverging clades of Coprosma share a solitary terminal inflorescence

produced on arrested lateral branches. This condition is shared with sister genus Nertera

and outgroup genus Leptostigma. Fascicle-like inflorescences have evolved only once,

but axillary pedunculate inflorescences have evolved multiple times. The vast majority of

species have axillary inflorescences with well developed peduncles, especially those

species that occur outside of New Zealand.

Fruit Color (Fig. 4.3)

Reconstruction of the evolution of fruit suggests the ancestor of Coprosma and

Nertera was orange in fruit color. Independently, orange fruit were evolved in the outgroup taxa Durringtonia paludosa and C. moorei of Australia. Orange fruit is inferred

as the color for the most recent common ancestor of all lineages. Unpigmented fruit

appeared early in the clade that is now defined as Subgenus Acerosae (see new

taxonomic implication discussion below). Yellow, blue and orange fruit evolved

independently in several species of subgenus Acerosae. Black fruit evolved

independently as it is found in six unrelated species.

101

Figuer 4.3. Reproductive and karyological characters reconstructed on the 50% Bayesian Inference topology. A). evolution of breeding system. B) reconstruction of inflorescence type. C). Fruit color and D). chromosome evolution.

102

Male floral characters (Fig. 4.4)

As indicated in section, 3.3.1, the development of dioecy evolved in Coprosma

in which male and female flowers were separate. Early diverging lineages of Coprosma,

which are closely related to the sister genus Nertera, share the same number of floral

parts. For example, Nertera and the early diverging lineages of Coprosma share 4-merous corolla and calyx lobes, and four stamens. The plesiomorphic condition for the number of flowers per inflorescence of Coprosma is 1(-2). An early loss of the male calyx occurs in all lineages after the grade of monotypic C. pseudocuneata. Derived lineages, predominately on islands outside of New Zealand, have a much higher number of floral parts.

Increase in flower part number

The highly derived clades experienced an increase in the number of floral reproductive parts. The geographic locality of these species represents the secondary center of diversity for the genus and has diverged the most in the Hawaiian Islands. The increase of floral parts is clearly a fixation in insular habitats has significant implications regarding the successful establishment and colonization of many Pacific archipelagos.

For example, the duplication of stamens to 8 in number, from the plesiomorphic condition of 4, potentially doubles the amount of pollen available for fertilization of a female flower. For those individuals with 8 stamens, a higher rate of successful fertilization events could have selected for the retention of this duplicative trait.

103

Figure 4.4. Reproductive characteristic of the male inflorescence reconstructed on the 50% Bayesian Inference topology. A) number of male flowers per inflorescence. B). Number of lobes of the male calyx. C.) number of lobes in the male corolla. D.) number of stamens per flower.

104

Discussion The morphological evolution of Coprosma inferred from this study supports

the phylogenetic hypothesis put forth by Oliver (1935) in which taxa have evolved from a

more ‘simple’ morphology to more complex states. However, the subgeneric and

sectional relationships of the most current taxonomic assessment of the genus by Heads

(1996) were not supported by the phylogenetic reconstructions of this study. For

example, the evolution, some taxa with reversals to the plesiomorphic condition in the

most derived lineages (Fig. 4.2 & 4.3). Additionally, the evolution of calyx lobes is

interesting, the most recent ancestor of the genus is hypothesized to have had four lobes,

but an early loss/ reduction of the calyx is also indicated. This is then followed by

regaining the calyx and an elaboration of the number of lobes in some lineages.

The characters associated with the divergence of Coprosma from Nertera are marked by the development of dioecy and woodiness in Coprosma. Hypothesized characters for the most recent common ancestor (MRCA) of all Coprosma species (not including C. moorei or C. talbrokei) include: 1(-2) flowers on terminal a inflorescence

having four calyx lobes, four corolla lobes, four stamens or (two stigmas if female) and a

small trailing woody habit with small leaves. The retention of vestigial floristic sex

organs of the opposite sex may have occurred in these early lineages, but we are unable

to verify this. The clade that is sister to the remaining species of the genus contains C. niphophila and C. perpusilla, which share the most character states with the MRCA of

Coprosma. However, these two species have a higher chromosome number signifying the possibility of multiple ancestors in their evolutionary history.

The evolution of dioecy in Copromsa is correlated with the evolution of wind pollination, as it is speculated that Nertera is possibly splash or insect pollinated

105

(Gardner, 1999). This hypothesis needs field observations as wind-pollination has not

been ruled out for Nertera, but ‘small species of flies’ have been seen feeding on the

pollen of Nertera species (Gardner, 1999) and the orientation of Nertera flowers could

indeed represent the pollination via water splashes. Gardner (1999; A. Markey & J.

Cantley, pers. obs.) also mentions stigmas of at least one species of Nertera produces an

exudate that small flies were seen feeding on, but the identity and function of this exudate

is unknown.

Variations in vegetative characters such as leaf size and habit were likely

correlated to selective pressures of various habitats and other environmental/ecological

characters acting on individual lineages of Coprosma. For instance, species that occur in

the remote Pacific predominately share large leaves, orange fruit and an understory shrub

habit. These characters would have promoted the successful establishment of taxa across

the Pacific, representing convergent morphologies in response to insular selective

pressures within Coprosma.

Taxonomic implications

The subgeneric and sectional classifications by Heads (1996) were not

supported in the phylogenetic reconstructions of this study. None of the characters used

by Heads (1996) to delimited the subgeneric ranks, leaf size (small or large),

inflorescence structure (solitary flowers or not solitary) and biogeography (east versus west vicariant biogeographic split overlapping in New Zealand) was shown to be phylogenetically informative. Our analyses indicate that the morphological evolution of

leaf size and inflorescence structure include a great deal of homoplasy in which many

reversals between the character states have occurred which created a complex

106

evolutionary history between taxa. Therefore, our findings support the speculation of

Gardner (2002) that delimitation of species based upon leaf size and inflorescence

structure alone is not sufficient enough to delimit subgeneric/sectional divisions. Also,

the vicariant split of Coprosma as suggested by Heads (1996) would have had to occur much longer ago than the estimated age of the genus by Cantley et al., (Chapter 3) during the Oligocene. Furthermore, for many of the traits assessed in this study, morphological homoplasy compounds the inference of relationships between groups of species within the genus. However with the new phylogenetic analyses, we were able to develop a new subgeneric and sectional classification for the genus Coprosma and further support the retention of Nertera and Leptostigma as independent monophyletic genera.

All sections and subgenera variously defined by Oliver (1935), Allan (1961),

Heads (1996), were found to be non-monophyletic (Fig 4.1). From the characters

assessed in this study, it is clear that morphological characters of Coprosma maintain

high levels of homoplasy, which had hindered the understanding of natural sublineages

without the aid of molecular phylogenetic techniques.

In the taxonomic treatment below, Coprosma is divided into six monophyletic

subgenera and several sections. In almost all cases, no there is no single synapomorphic

character that defines a natural section, but rather various combinations of characters are

needed. Brief explanations are given when needed to clarify their morphological,

geographic and molecular relationships. To note, this revision represents the baseline

investigation into characters that separate individual lineages. The treatment does not

specifically take into account the effects of hybridization of taxa. Furthermore, the

implications of chromosome number evolution (perhaps resulting in taxa of allopolyploid

107

origins) were not considered, as we were not able to collect relevant data to address these

situations. Further work is clearly needed, but will require extensive fieldwork across the

Pacific or directed investigations at the sectional or lower divisions within the genus.

Section Microcoprosma of Heads (1996) is polyphyletic among nine separate lineages in the phylogenetic reconstruction. Similarly, section Moorei is not a natural section of Copromsa (see notes on section Pumilae below), and therefore neither names are applied to the following systematic account. The transfer of C. talbrockei and C. moorei to a new genus is beyond the scope of this paper and is to be discussed by Markey et al. (in prep).

A key for Netera, Leptostigma and subdivisions of Coprosma

1a. Plants herbaceous, flowers bisexual ……………………………………………..……2 1b. Plants with secondary growth (shrubs, trailing or erect trees), dioecious or leaky- dioecious………………………………….……………………………...... …Coprosma, 3 2a. Flowers strongly protandrous, male stamens exerted far beyond corolla at maturity……………………………………………………………………..…Leptostigma 2b. Flowers not protandrous, male stamens not exerted far beyond corolla……….Nertera 3a. Leaky dioecy not known………………………………………………………………4 3b. Leaky dioecy known for at least one species………………………………………….6 4a. Plants with large united stipules, cuneate-oblong coriaceous leaves and long female calyx lobes……………………………………..…...………..…subgenus Pseudocuneatae 4b. Plants otherwise……………………………………………………………………….5 5a. Habit highly divaricating of New Zealand or large leaved plants with deep violet drupes from Norfolk Island…………………………………………subgenus Parviflorae 5b. Habit not divaricating (mat-forming, trailing, shrubs or trees)……………………………………………………………………subgenus Pumilae 6a. Inflorescence terminal or occasionally axillary (but not pedunculate)……………..…7 6b. Plants not divaricating or mat-forming, leaves >25mm in length, inflorescence axillary peduncle, male calyx lobs 4-6, male corolla lobes 6-8, drupes orange…………………………………………………………...... ……subgenus Lucidae 7a. Plants having leaves >25mm in length with orange fruit or leaves <25mm in length with blue, translucent (white) or polymorphic………………..……….subgenus Acerosae 7b. Plants with leaves >25mm in length, orange to red fruit………...subgenus Coprosma

108

Taxonomic circumscription of subgenus Coprosma Heads

13 species: Borneo (1), Java (1), New Guinea (5), New Zealand (3), Sulawesi (1) and widespread in New Zealand and Australia (2). =sect. Coprosma Heads; sect. Cuneatae Allan (incl. sect. Foetidae Allan). Type species: C. foetidissma J.R. Forst. & G. Forst., Char. Gen. Pl. 69. 1775.

Erect shrubs or small trees in temperate forest or alpine habitat, leaves smaller than 25mm in length, leaf margin entire or toothed, reproductive strategy leaky dioecious, terminal inflorescences usually with 1(-2) male or female flowers, male calyx absent or lobes 4, male corolla lobes 4(-6), stamens 4, stigmas 2(-4), fruit orange to red, 2n=44 or 132.

C. barbata Utteridge, NG C. foetidissma .R. Forst. & G. Forst., NZ C. brassii Merr. & L.M. Perry, NG C. niphophila Orchard, NZ/Australia C. celebica W.R.B. Oliv., Sulawesi C. papuensis W.R.B. Oliv., NG C. crassicaulis Stapf., Borneo C. perpusilla Colenso, NZ/Australia C. crenulata W.R.B. Oliv., NZ C. serrulata Hook. f. ex Buchanan, NZ C. divergens W.R.B. Oliv., NG C. sundana Miq., Java C. elegans Utteridge, NG C. wollastonii Wernham, NG

Subgenus Coprosma of 13 species is the first diverging lineage of Copromsa from the sister genus Nertera (Fig. 4.1). Species exhibit a range of habit from trailing/mat forming (C. niphophila and C. perpusilla), erect shrubs (Indo-Pacific species and C. crenulata), and small trees (C. foetidissma and C. serrulata). Chromosomal number is

2n=44 except in C. foetidissma and C. perpusilla where 2n=132. All species occasionally produce bisexual flowers, but their functionality is unknown. Consistently,

C. niphophila, and occasionally C. perpusilla, vestigial anthers/carpels are retained in the hermaphroditic state (Orchard, 1986). Stigmas and carpels of C. perpusilla are commonly four (J. Cantley, pers. obs.). Coprosma as a whole has entire leaf margins, but three nested New Zealand species (C. crenulata, C. foetidissma, and C. serrulata) and New

109

Guinean C. papuensis share serrated leaf margins, which are produced by a ‘saw-toothed

flattening of the upper epidermis’ (Gardner, 2002a). Despite a disputed number of taxa

from New Guinea between Utteridge (2002) and Gardner (2002b), these New Guinea

species are the result of a single long distance dispersal event from New Zealand (Cantley

et al., Chapter 3). Conversely, the three nested New Zealand species are the result of a

single back dispersal from New Guinea to New Zealand (Cantley et al., Chapter 3). New

Guinea species (not C. archboldiana [Subgenus Pumilae section Pumilae])—as well as

C. crassicaulis of Borneo, C. celebica of Sulawesi and C. sundana of Java—are shrubs

that are widely branching. All species of subgenus Coprosma have terminal flowers with

orange fruit.

Taxonomic circumscription of subgenus Pseudocuneatae (Allan) J. Cantley stat. nov.

1 species: New Zealand. =sect. Pesudocuneatae Allan. Type species: C. pseudocuneata W.R.B. Oliv. ex Garn.-Jones & Elder., New Zealand J. Botany 34: 140. 1996.

Erect shrubs to 3m tall in forest and 5cm in exposed alpine habitats with dense foliage, leaves narrow-ovate or cuneate-oblong and coriaceous, reproductive strategy dioecious, leaky dioecy not known, inflorescence terminal with solitary male or female flowers on short lateral branches, male calyx absent, male corolla lobes 4, stamens 4, stigmas 2, fruit orange to red or less commonly pale purple, pale yellow or unpigmented, 2n=132.

Oliver (1935) states that “[Coprosma pseudocuneata] is an easily recognized species, not very closely related to any other…” It is placed in its own monotypic section due to its position in the grade between section Coprosma and the four other subgenera of the genus in the molecular phylogeny (Fig 4.1). The morphological assessment failed to recover any unique characters, but Oliver (1935) notes large united stipules, cuneate- oblong coriaceous leaves and long female calyx lobes as autapomorphic.

110

Taxonomic circumscription of subgenus Parviflorae (Allan) stat. nov. J. Cantley

17 species: New Zealand (16), Norfolk Island (1). =sect. Parviflorae Allan. Type species: C. parviflora Hook. f., Bot. Antarct. Voy. II. (Fl. Nov.-Zel.). 1: 106. 1852

Highly divaricating erect shrubs to small trees or non-divaricating small trees, open or closed habitats, leaves usually less than 25mm in length or much larger, reproductive strategy dioecious, leaky dioecy not known, inflorescence fascicle-like or terminal and solitary, male flowers usually 1(-2) per inflorescence, male calyx absent, male corolla lobes 4, stamens 4, stigmas 2, fruit black or dark violet or orange to red or unpigmented, 2n=44, 88, or 132.

C. areolata Cheeseman, NZ C. pedicellata Molloy, de Lange & B.D. C. cheesmanii W.R.B. Oliv., NZ Clarkson, NZ C. ciliata Hook. f., NZ C. pilosa Endl., Norfolk Is. C. decurva Heads, NZ C. pseudociliata Hook. f., NZ C. dumosa (Cheeseman) G.T.Jane, NZ C. rhamnoides A. Cunn., NZ C. fowerakeri D.A. Norton & de Lange, C. rotundifolia A. Cunn., NZ NZ C. tayloriae A.P. Druce ex G.T.Jane, NZ C. microcarpa Hook. f., NZ C. tenuicaulis Hook. f., NZ C. neglecta Cheeseman, NZ C. wallii Petrie, NZ C. parviflora Hook. f., NZ

The relationship of subgenus Parviflorae to other sections is not clear based on

the phylogenetic analyses. Furthermore, the 17 New Zealand taxa in this clade are often

hard to distinguish, as all are highly divaricating shrubs to small trees with microphyllous

leaves smaller than 25mm in length (with the exception of C. pilosa of Norfolk Island).

Coprosma pilosa has leaves much larger than 25mm and is not divaricating, characters

that likely developed after colonization of Norfolk Island with new selective pressures.

All species of subgenus Parviflorae have a fascicle-like inflorescence, which is otherwise unknown in the genus. Other reproductive characters of subgenus Parviflorae are male flowers predominately without a calyx, four corolla lobes and four stamens. Polyploid species in this section are common where 2n=44, 88, or 132. Fruit color of subgenus

111

Parviflorae is mostly orange to red, but some are black (C. areolata and C. tenuicaulis), unpigmented (C. ciliata and C. microcarpa) and deep violet (C. pilosa and C. parviflora) fruits. Three species have highly polymorphic fruit colors (C. dumosa, C. pedicellata, and

C. tayloriae).

Taxonomic circumscription of subgenus Pumilae (Allan) J.T. Cantley stat. nov.

2 sections: 14 species total . =sect. Pumilae Allan. Type species: C. pumila Hook f., Bot. Antarct. Voy. II. (Fl. Nov.-Zel.). 1: 111. 1852

Erect shrubs or trees, often in exposed or forested habitats, leaves >25mm in length or much larger, reproductive strategy dioecious, leaky dioecy not known, inflorescence axillary peduncle or terminal on short lateral branches, male flowers 1(-2) or 6 per inflorescence, male calyx absent or lobes 4 or 5, male corolla lobes 4 or 5, stamens 4 or 5, female flowers 1(-2) or numerous per inflorescence, stigmas (1-)2(-3), fruit color orange to red or black, or burgundy 2n=44 or 220.

Subgenus Pumilae is newly described here containing 14 species in two sections. The relationships of subgenus Pumilae to other subgenera is not clear given the topology of the phylogenetic reconstructions. The two sections are very different in morphology including chromosome number, inflorescence type, leaf size and habit.

Smaller differences that may be phylogenetically uninformative include a variable number of stamens and stigmas.

112

Section Pumilae (Allan) J.T. Cantley comb. nov

5 species: Australia (1), Fiji (1), Hawaiian Islands (1), New Guinea (1), New Zealand (1). Type species: C. pumila Hook f., Bot. Antarct. Voy. II. (Fl. Nov.-Zel.). 1: 111. 1852

Trailing and mat-forming or erect non-divaricating shrubs to trees often in rocky or exposed habitats such as braided rivers beds or bogs or forest clearings or recent lava flows, leaves less than 25mm in length or much larger, reproductive strategy dioecious, leaky dioecy not known, inflorescence terminal on short lateral branches with 1(-2) male flowers and 1(-2) to numerous female flowers, male calyx absent, male corolla lobes 4, stamens 4, stigmas (1-)2(-3), fruit color orange or black or red or burgundy, 2n=220.

C. archboldiana Merr. & L.M.Perry, NG C. ernodeoides A.Gray, Hawaii C. atropurpurea (Cockayne & Allan) C. persicifolia A.Gray, Fiji L.B.Moore, NZ C. pumila Hook. f., Australia

Section Pumilae, of five species is markedly disjunct in geographic distribution with one species each in New Zealand, New Guinea, Australia, the Hawaiian Islands and

Fijian Islands. Species most often have male flowers with absent male calyxes, four corolla lobes and four stamens. All species further share 2n=220, although chromosome counts are unknown for C. archboldiana and C. persicifolia. Excluding C. persicifolia, leaves are smaller than 25mm in length on terminal inflorescences usually with 1(-2) flowers. Species are trailing/mat-forming in habit and often occupy specialized habitats such recent lava flows (C. ernodeoides), braided rivers (C. atropurpurea) and bogs (C. pumila), but are also found in open or closed forests, scrublands or tussocks. Coprosma persicifolia is clearly an anomalous taxon in section Pumilae as it has solitary flowers in the axils of leaves and is a erect shrub with leaves larger than 25mm. Little has been published on this Fijian species, but it is an early colonizer of lava flows and forest clearings (Oliver, 1935). Variability is known in floral characters of C. ernodeoides including individuals with one, two or, three stigmas, and variable stamen and corolla

113

lobe numbers that usually occur in multiples 4s, but multiples of 3s and 5s are also

observed (J. Cantley, pers. obs.).

Coprosma ernodeoides and C. archboldiana were previously allied closely by

Merrill and Perry (1945). Both were grouped with C. talbrockei of New Zealand and C. moorei of Tasmania by Heads (1996) due to the presence of acuminate leaves, non- sheathing stipules and the absence of floral bracts in all species. However, from our analyses, leaf and stipule characters are highly plastic and of little use for the identification of monophyletic lineages. Even though these species superficially seem similar, they are clearly unrelated differing primarily in reproductive strategy, chromosome number and large genetic distance.

Section Putidae Allan

9 species: Australia (1), Bougainville Is. (1), Lord Howe Island (5), Samoa (1), and Vanuatu (1). Type species: C. putida C. Moore & F. Muell., Fragm. (Mueller) 7(52): 45. 1869

Erect shrubs or trees, maritime insular forests often exposed habitats, leaves >25mm in length, reproductive strategy dioecious, leaky dioecious not known, inflorescence axillary peduncle, male flowers 1(-2) or 6 per inflorescence, male calyx absent or lobes 4 or 5, male corolla lobes 4 or 5, stamens 4 or 5, female flowers 1(-2) or numerous per inflorescence, stigmas 2, fruit color orange to red or black, 2n=44.

C. bougainvilleensis Gideon, C. novaehebridae W.R.B. Oliv., Bougainville Is. Vanuatu Is. C. hirtella Labill., Australia C. putida C. Moore & F. Muell., Lord C. huttoniana P.S. Green, Lord Howe Is. Howe Is. C. inopinata I. Hutton & P.S. Green, C. sp. nov., Lord Howe Is. Lord Howe Is. C. strigulosa Lauterb., Samoa C. lanceolaris F. Muell., Lord Howe Is.

114

Section Putidae is closely related to section Pumilae and a more detailed molecular investigation could result in a better understanding of their morphological and biogeographic relationships. All species are shrubs or trees with leaves larger than 25mm in length, 2n=44, axillary peduncles, and male flowers with an absent calyx, four corolla lobes and four stamens. Lord Howe Island species share two additional synapomorphies, as there is a gain of a male calyx with four lobes and the number of male flowers per inflorescence increases from 1(-2) to six. Coprosma inopinata, C. landeolaris and C. sp. nov. were not included in the molecular analysis of this study, but Papadopulos et al.

(2011) indicate that these species plus C. huttoniana and C. putida are monophyletic and the result of a single dispersal event. Fruit color in section Puditdae is orange, except for

C. strigulosa of Samoa, which has black fruits and C. hirtella, which are strongly red.

Corpsoma hirtella is unique in section Putidae as it has five lobes on the male calyx and corolla with five stamens and many flowers per inflorescence. The historical placement of C. hirtella has been a matter of contention as Oliver (1935) and Gardner

(2002a) did not name any species of close relations. Oliver (1935) and Heads (1996) tentatively suggested it grouped with C. sundana of Java and New Guinean C. papuensis based on leaf characters, but these relationships are not supported by our phylogenetic analyses.

Taxonomic circumscription subgenus Acerosae J.T. Cantley stat. nov.

3 sections: 27 species total. Type species: C. acerosa A. Cunn., Ann. Nat. Hist. 2(9): 207. 1838.

Prostrate or erect shrubs divaricating or mat-forming or open branching in habit, leaves smaller or larger than 25mm in length, reproductive strategy dioecious or leaky dioecy,

115

inflorescence axillary or terminal on arrested branchlets, male flowers 1(-2) or 3-10 inflorescence, male calyx absent or 4, male corolla lobes 4(-5), stamens 4 or 6, female flowers 1(-2) or 3-12 per inflorescence, stigmas 2, fruit blue or translucent or orange or polymorphic, 2n=44, 2n=44 or 132.

Uncertain sectional placement

C. crassifolia Colenso, NZ C. rubra Petrie, NZ C. rigida Cheeseman, NZ

Subgenus Acerosae is most closely related to subgenus Lucidae.

Morphologically, subgenus Acerosae is the most variable subgenus of Coprosma,

spanning, or nearly so, the entire morphological diversity of the genus. The three sections

are distinct from each other. Section Acerosae is the most easily distinguished in which

all species have blue fruits. Coprosma crassifolia, C. rigida and C. rubra of New Zealand

are of uncertain placement as they are on a grade between section Spathulatae and

section Petiolatae + section Acerosae. Their morphology agrees closely with sections

Spathulatae and Acerosae.

Ssection Acerosae (Allan) J.T. Cantley comb. nov.

9 species: Australia (1), New Zealand (8). =sect. Acerosae Allan (incl. Linariifoliae Allan & sect. Antipodae Allan). Type species: C. acerosa A. Cunn., Ann. Nat. Hist. 2(9): 207. 1838.

Prostrate or erect shrubs strongly to weakly divaricating or mat-forming in habit from coastal to subalpine open habitats, leaves smaller than 25mm in length, reproductive strategy dioecious, leaky dioecy not known, inflorescence axillary, male flowers 1(-2) per inflorescence, male calyx absent, male corolla lobes 4, stamens 4, female flowers 1(-2) per inflorescence, stigmas 2, fruit blue or translucent with or without blue epidermal pigmentation, 2n=44.

C. acerosa A. Cunn., NZ C. elatirioides de Lange & A. S. C. brunnea Endl., NZ Markey, NZ C. intertexta G. Simpson, NZ

116

C. nivalis W.R.B. Oliv., Australia C. rugosa Cheeseman, NZ C. petriei Cheeseman, NZ C. virescens Petrie, NZ C. propinqua A. Cunn., NZ

Section Acerosae is closely related to section Petiolatae, but differs primarily in a divaricating or trailing/mat-forming habit, small leaves, non-pedunculate inflorescence, blue fruits and their restriction to New Zealand and Australia. This group was historically difficult to relate to any other in the genus, despite that its monophyly was long speculated by Allan (1961) Heads (1996) and Gardner (2002), citing their blue fruits of significant phylogenetic importance. Indeed, blue fruits appear in no other section of the genus except C. moorei and C. talbrockei, which are now demonstrated as

outside the circumscription of the genus Coprosma sensu strictu.

Our phylogenetic reconstructions support C. propinqua (var. propinqua and

var. martini) within subgenus Lucidae, but we place it here in subgenus Acerosae section

Acerosae based on morphological similarities. Its status is a matter of contention, as it is

known to hybridize with at least 12 other Coprosma taxa from many sections. Oliver

(1935) allied it with C. microcarpa, C. parviflora and C. ciliata (=subgenus Parviflorae),

but was well aware of its known variability and hybridization. Heads (1996) suggested

the species related to section Acerosae, which was retained by Gardner (2002). The

molecular phylogeny groups C. propinqua closely to C. robusta. The well know hybrid

C. x cunninghamii is a well-known hybrid of C. propinqua and C. robusta in which

nearly every leaf intermediate can been found between the two species (Oliver, 1935).

Section Petiolatae (Allan) J. Cantley comb. nov.

12 species: Austral Islands (1), Chatham Islands (1), Cook Islands (1), Kermadec Islands (1), Lord Howe Island (1), New Zealand (1), Norfolk Island (1),

117

Pitcairn Islands (1), Society Islands (4). Type species: C. petiolata Hook. f., J. Proc. Linn. Soc., Bot. 1: 128. 1856.

Erect non-divaricating shrubs or trees in maritime insular forests or coastal habitats, leaves >25mm in length, reproductive strategy dioecious or leaky dioecious, inflorescence axillary peduncle, male flowers 3-10 per inflorescence, male calyx absent or lobes 4, male corolla 4(-5), stamens 4 or 6, female flowers 3-9 per inflorescence, stigmas 2, fruit orange, 2n=44 or 132.

C. baueri Endl., Norfolk Is. C. prisca W.R.B. Oliv., Lord Howe Is. C. benefica W.R.B. Oliv., Pitcairn Is. C. rapensis F. Br., Austral Is. C. chathamica Cockayne, Chatham Is. C. repens A. Rich., NZ C. laevigata Cheeseman, Cook Is. C. setosa J.W. Moore, Society Is. C. orohenensis J.W. Moore, Society Is. C. taitensis A. Gray, Society Is. C. petiolata Hook. f., Kermadec Is. C. velutina Fosberg, Society Is.

Section Petiolatae is closely related to section Acerosae. In the phylogenetic reconstructions, C. intertexta and C. rugosa are reconstructed as within section Peitolatae with low support, but given morphological evidence we do not include either within section Petiolatae. Future genetic studies would be useful to elucidate the precise position of C. intertexta and C. rugosa between both sections. All species of section

Petiolatae are perhaps the result of a single dispersal from New Zealand except for the coastal New Zealand species of C. repens. They are large leaved species commonly found in forest understories to exposed coastal habitats. Five species (C. chathamica, C. repens, C. petiolata, C. baueri and C. prisca) were recognized by Oliver (1935) as sharing coastal habitats with oblong sometimes slightly fleshy leaves.

Coprosma benefica, C. esulcata, C. laevigata, C. orohenensis, C. taitensis and

C. velutina are closely related representing a group widely distributed throughout islands of French Polynesia, Pitcairn and Cook Islands (Fosberg, 1937; Brown 1935; Moore,

1933). These species comprise a species complex that has not been investigated with

118

extensive detail. Coprosma taitensis occupies many islands (Tahiti and Moorea [Society

Islands], Mangareva [Gilbert Islands], Tubuai and Raivavae [Austral Islands]), of which

DNA was sampled from only Tahiti Islands due to the in ability to obtain material from

these remote locations. But, as each Island represents a different variety of C. taitensis, it

would be interesting to investigate their relationships. As nearly no species are

widespread in locations larger than a single archipelago, its seems dubious to not call

these unique species, but an extensive morphological investigation coupled with

molecular phylogenetic techniques is needed to elucidate their precise relationships.

Section Spathulatae (Allan) J. Cantley stat. cons.

3 species: New Zealand. =Sect. Microcoprosma Heads. Type species: C. spathulata A. Cunn., Ann. Nat. Hist. 2(9): 207. 1838.

Erect shrubs to trees, divaricating or non-divaricating habit in open or closed habitats, leaves smaller than 25mm in length or larger, leaves spathulate or with a notably thickened margin of fibers, reproductive strategy dioecious, leaky dioecy not known, inflorescence terminal on arrested branchlets or axillary and non-pedunculate, male flowers 1(-2) per inflorescence or in dense glomerules with numerous flowers, male calyx absent or lobes 4, male corolla lobes 4, stamens 4, female flowers 1(-2) or in clusters of 4-12 flowers, stigmas 2, fruit polymorphic or unpigmented with or without violet flecks or pale yellow, 2n=44.

C. arborea Kirk, NZ C. obconica Kirk ssp. distantia de Lange C. spathulata A. Cunn., NZ & R.O. Gardner, NZ C. obconica ssp. obconica, NZ

Section Spathulatae is sister to section Acerosae plus section Petiolatae. All

authors agree that C. arborea and C. spathulata are each other’s closest relatives

primarily differing in leaf size, numbers of flowers per inflorescence, fruit color and

habit. Coprosma spathulata is generally smaller leaved and is trailing or a small shrub

while C. arborea is larger leaved with more flowers per inflorescence and is one of the

119

larger species in height of the genus (Heads, 1996). Fruits of C. arborea are pale yellow

or unpigmented. Perceived color should be investigated in more detail as pale yellow

fruits might simply be the result of seed visibility through translucent fleshy fruit tissue.

Both varieties of Coprosma obconica differ from C. arborea and C. spathulata

in their divaricating habit and several characters that are unique to the genus including a

notably thickened leaf margin composed of fibers with an encircling marginal vein and

swollen pyrenes with a deep ventral indentation (de Lange & Gardner, 2002). Coprosma

arborea agrees with C. obconica ssp. distantia with its pale yellow fruits and to C.

obconica ssp. obconica in their tree like habits.

Taxonomic circumscription of subgenus Lucidae (C. J. Webb) J.T. Cantley comb. nov.

4 sections: 54 species total. Type species: C. lucida G. Forst., Char. Gen. Pl. 69. 1775

Erect shrubs or trees of open or closed habitats, leaves greater than 25mm in length, reproductive strategy dioecy or leaky dioecy, inflorescence axillary peduncle, male flowers 1-(2) or 4-12, male calyx lobes 4-6 or 6-8, male corolla lobes 4-6 or 6-8, stamens 1(-2) or 4(-6) or 8, female flowers 1(-2) or 3-9 per inflorescence, stigmas 2, fruit orange, 2n=44.

Uncertain sectional placement: 6 species

C. colensoi Hook. f., NZ C. linariifolia Hook. f., NZ C. cuneata Hook. f., NZ C. quadrifida (Labill.) B.L. Rob., C. “decipiens”, NZ Australia C. depressa Colenso ex Hook. f., NZ

The new combination of species found in subgenus Lucidae is similar to the

previous circumscription of Heads (1996). However, it now includes fewer species in

total, and a few species previously under the placement of subgenus Coprosma having

120

small leaves and solitary female flowers. Subgenus Lucidae is most closely related to

subgenus Acerosae, of which both have a high degree of morphological diversity.

Subgenus Lucidae here is divided into three sections, and includes six species of

uncertain sectional affiliations. Subgenus Lucidae encompasses the secondary center of

diversity in the Hawaiian Islands and is the most species rich subgenus containing nearly

half of the described species.

The six species receiving an uncertain status here are closely related in some

manner, to each other and to section Australes and section Grandiflorae. These are C.

colensoi, C. cuneata, C. decipiens, C. depressa and C. quadrifida. All but C. quadrifida

are in a weakly supported clade on a polytomy with section Australes and section

Grandiflorae. Coprosma quadrifida is directly sister to the section Australes (Fig. 1).

However, it is likely the result of an independent dispersal from New Zealand to

Australia. No other Australian species are closely related. Thompson (2010) indicates

that C. nitida of Australia is similarly confused with C. quadrifida sharing similar

stipules, branchlets that end in spiny-tips. Coprosma quadrifida likely shares an ancestor

with section Australes, but differs as it has not experienced an increase of floral parts and a difference of floral parts and the leaves are smaller than 25mm. For this reason, the

morphology of C. quadrifida recalls that more of the plesiomorphic condition. The

molecular phylogeny suggests the position of C. linariifolia is somewhere between

subgenus Acerosae and subgenus Lucidae.

121

Section Australes J. Cantley nom. nov

6 species: Juan Fernández Islands (2), Kermadec Islands (1), New Zealand (3). Type species: C. robusta Raoul, Ann. Sci. Nat., Bot. sér. 3, 2: 121. 1844

Erect shrubs or trees of open or closed forests sometimes coastal, leaves greater than 25mm in length, reproductive strategy dioecy or leaky dioecy, inflorescence axillary peduncle, male flowers 1-(2) or 4-12, male calyx lobes 4-5, male corolla lobes 4-5, stamens 4, female flowers 3-9 per inflorescence, stigmas 2, fruit orange, 2n=44.

C. acutifolia F. Muell. ex Benth., C. oliveri Fosberg, Juan Fernández Is. Kermadec Is. C. pyrifolia Skottsb., Juan Fernández Is. C. macrocarpa Cheeseman, NZ C. robusta Raoul, NZ

The distribution of section Australes is the largest of any section spanning from

New Zealand east to the Juan Fernández Islands. It is perhaps only rivaled only by section Coprosma, which reaches from New Zealand westward to Borneo. It is most closely related to section Polynesica.

Section Polynesica J.T. Cantley nom. nov.

19 species: Austral Islands (1), Hawaii (12), Marquesas Islands (6). Type species: C. longifolia A. Gray, Proc. Amer. Acad. Arts iv. 1860.

Erect shrubs or trees of open or closed forests as well as bogs and subalpine habitat, leaves greater than or around 25mm in length, reproductive strategy dioecy or leaky dioecy, inflorescence axillary peduncle, male flowers 1(-2) or 2-15 per inflorescence, male calyx lobes absent or 4-6 or 8, male corolla lobes 6-8, stamens 1(-2) or 8 or about 10, female flowers 1(-2) or 6-9 per inflorescence, stigmas 2(-4), fruit orange, 2n=44 where known.

122

C. cookei Fosberg, Austral Is. C. montana Hillebr., Hawaiian Is. C. cymosa Hillebr., Hawaiian Is. C. nephelephila J. Florence, Marquesas C. elliptica W.R.B. Oliv., Hawaiian Is. Is. C. esulcata (F. Br.) Fosberg, Marquesas C. ochracea W.R.B. Oliv., Hawaiian Is. Is. C. pubens A. Gray, Hawaiian Is. C. fatuhivaensis W.L. Wagner & C. reticulata J. Florence, Marquesas Is. Lorence, Marquesas Is. C. rhynchocarpa A. Gray, Hawaiian Is. C. folisa A. Gray, Hawaiian Is. C. temetiuensis W.L. Wagner & C. kauensis A. Heller, Hawaiian Is. Lorence, Marquesas Is. C. longifolia A. Gray, Hawaiian Is. C. ternata W.R.B. Oliv., Hawaiian Is. C. menziesii A. Gray, Hawaiian Is. C. waimeae Wawra, Hawaiian Is. C. meyeri W.L. Wagner & Lorence, Marquesas Is.

The name of section Polynesica is dual purposed as it refers to the obvious geographical distribution within three Polynesian archipelagos of the Pacific and to the greater number of floral parts that this subsection exhibits than other sections/subgenera.

Many morphological parts are clearly a simple duplication from the plesiomorphic condition (ex. stamens 8 rather than 4 stamens in Hawaiian C. cymosa, C. pubens, C. waimeae, C. menziesii and C. foliosa), but this is not always the case (for example, C. cookei regularly has 10 stamens). However, the solitary (or sometimes two) female flowers and small leaves of Hawaiian C. elliptica and C. montana were anomalous and called into question the monophyly of the Hawaiian orange fruited taxa (Fosberg, 1948;

Wagner, 1999). Morphologically speaking, Hawaiian and Marquesan taxa are separated on a number of characters and C. cookei shares a seemingly hybrid of characters between them suggesting the plasticity of these characters (as it is almost certainly a dispersal from either Hawaii or Marquesas, not both).

The 12 Hawaiian orange fruited species represent an adaptive radiation throughout the archipelago. Although it has not been studied as extensively as other

123

charismatic Hawaiian lineages such as the Silversword Alliance (Compositae) or

Lobeliads (Campanulaceae), many Coprosma species are in exclusive habitats from one

another (C. elliptica in bogs of Kaua‘i, C. cymosa in mesic forest clearings of the Big

Island only, C. montana in subalpine scrub) or have separated along elevational gradients. For instance, C. waimeae of Kaua‘i Island, C. foliosa of O‘ahu and Maui Nui islands and C. menziesii of the Big Island are each restricted to the lowest elevations of the islands they occupy whereas purportedly sister taxa occur higher in elevation. The relationships between these Hawaiian taxa are unclear as our molecular phylogenetic analyses were unable to provide resolution. Perhaps the only natural pair recognizable based on only morphology is C. longifolia of O‘ahu Island and C. ternata of Moloka‘i

Island, which share long-sheathing deciduous stipules and occasionally ternate phyllotaxy. Further complicating the unclear relationships among the orange-fruited

Hawaiian taxa is the occurrence of hybridization and the difficulty of their identification in distinguishing them from parental species. For instance, morphological characters clearly intergrade into one another where they occur in sympatry. An in depth investigation of this adaptive radiation, perhaps with new molecular techniques such as

Next Generation Sequencing, could elucidate their evolutionary relationships for a new evolutionary understanding.

Section Grandiflorae J.T. Cantley nom nov.

5 species: Australia (1), New Zealand (3), Samoa (1). Type species: C. grandiflora Hook. f. Bot. Antarct. Voy. II. (Fl. Nov.-Zel.). 1: 104. 1852.

Erect shrubs or trees from open and closed habitats, leaves greater than or near 25mm in length, reproductive strategy dioecy and leaky dioecy, inflorescence axillary peduncle, male flowers 4-6 or >12 per inflorescence, male calyx lobes 4-6, male corolla lobes 4-6,

124

stamens 4-6, female flowers 1(-2) or 8-9 or >12 per inflorescence, stigmas 2, fruit orange, 2n=44.

C. dodonaeifolia W.R.B. Oliv., NZ C. savaiiensis Rech., Samoa C. grandiflora Hook. f., NZ C. waima A.P. Druce, NZ C. nitida Hook. f., Australia

The evolution of floral parts of in section Grandiflorae is especially increased.

This is especially apparent in C. waima, where there are more than 12 flowers per male or

female inflorescence. All New Zealand species are restricted to the North Island.

Section Lucidae J.T. Cantley nom. nov.

1 species: New Zealand. Type species: C. lucida G. Forst., Char. Gen. Pl. 69. 1775.

Erect shrubs or trees from open and closed habitats, leaves greater 25mm in length, reproductive strategy dioecy and sometimes leaky, inflorescence axillary peduncle, male flowers in dense clusters, male calyx lobes 4-5, male corolla lobes 5-6, stamens 5-6, female flowers 8-9 or >12 per inflorescence, stigmas 2, fruit orange, 2n=44.

The morphological characters of Coprosma lucida overlap entirely with section

Australes, but it is included here as a monotypic section given its placement among the

molecular phylogenetic reconstructions.

Conclusions

This is the first phylogenetically based study of the genus Coprosma

extensively investigating the morphological evolution of the genus with a taxonomically

robust dataset. Coprosma demonstrates high levels of morphological plasticity, and this

has clearly shaped its evolutionary trajectory. The divergence of Coprosma from its sister

genus Nertera is primarily associated with the evolution of dioecy and secondary growth.

125

Plesiomorphic characters include simple terminal inflorescences, orange fruit, and 4- merous flower parts. Many characters are highly plastic, such as number of flowers per inflorescence, but the general trend agrees with that of Oliver (1935) for the ‘tendency of the multiplication of floral parts.’ Derived lineages exemplifying the largest change from the plesiomorphic conditions have many more flowers with higher floral part number and compound inflorescences. The new taxonomic treatment outlined here reflects natural lineages determined on the basis of molecular relatedness, additional work at both the molecular and morphological level will provide the refinement and better ensure our understanding of the evolution of the genus throughout the Pacific.

126

CHAPTER 5 CONCLUSION

This research was directed at elucidating the evolutionary and biogeographical

histories of the genus Coprosma (Rubiaceae) in the Pacific. There were three broad aims

within this study: (A) to investigate the origins of Hawaiian Coprosma taxa, (B) to

understand the historical biogeography of Pacific Coprosma, and (C) to understand the

morphological evolution of Coprosma and its evolutionary implications. Individual

hypotheses were posed within each of these topics and are discussed below. A synthesis

of the findings and future directions are also presented.

Aim A. To investigate the origins of Hawaiian Coprosma taxa.

Hypothesis 1: Hawaiian Coprosma are the result of two independent colonization events.

Results indicate that Hawaiian Coprosma taxa are the result of two independent

colonization events to the archipelago, supporting this hypothesis, originally articulated

by Oliver (1935) and Wagner (1991). An alternative proposal of three colonizations of

the Hawaiian Islands (Fosberg, 1948), however received no support.

Hypothesis 2: Orange-fruited Hawaiian taxa are monophyletic and closely related to

taxa from the Marquesas Islands.

127

Twelve of the 13 Hawaiian species have red to orange drupes and also share similarities of habit, vegetative and floral morphologies, as well as a chromosome number of 2n=44 (Oliver, 1935; Wagner et al. 1999, Skottsberg, 1953) and therefore presumed to be of common descent. This finding was confirmed here, and also included an additional taxon, C. cookei from Rapa Iti. The latter species was previously thought to be most closely allied to taxa from the Society Islands and Cook Islands taxa based on their morphological similarities and geographical proximity (Florence & Lorence, 1997) and it was somewhat surprising to discover the relationship to Hawaiian taxa. The sister group to the Hawaii + Rapa Iti clade is composed of the orange fruited Marquesan species. The close relationship of plant taxa in the Hawaiian and Marquesan Islands is known for other Pacific genera as well, including Bidens (Ganders et al., 2000), Kadua

(Motley, 2003), among others. In addition, a previously undocumented pattern of relationships was shown for taxa in the Marquesas and Hawaiian Islands and New

Zealand. The most recent common ancestor (MRCA) of the orange-fruited Hawaiian

Coprosma, Marquesan and C. cookei of Rapa Iti is shared with C. tenuifolia from New

Zealand rather than to any other Pacific island locality. These same three areas also have recently discovered relationships within Astelia (Asteliaceae) (Birch et al., 2012) and

Metrosideros (Myrtaceae) (Wright et al., 2001; Percy et al., 2008). Astelia and Coprosma both have fleshy bird-dispersed fruits, however, Metrosideros has small, dry seeds thought to be primarily wind dispersed. Further study of this previously unknown set of biogeographic relationships is needed to determine if it is a more general phenomenon or if Coprosma, Astelia and Metrosideros have now similar distributions but for unrelated reasons.

128

Hypothesis 3: Black fruited Coprosma ernodeoides represents a second colonization to

the Hawaiian Islands and is not closely related to the orange-fruited Hawaiian taxa.

Two (Wagner et al. 1990, 1999) or three (Fosberg, 1948) colonization events for

Coprosma have been hypothesized based on differences in fruit color and morphology.

The black-fruited C. ernodeoides, is distinct from other 12 Hawaiian taxa and here was

shown to be most closely related to other black fruited species, C. atropurpurea of New

Zealand and C. pumila of Tasmania. All three species—C. pumila, C. ernodeoides and C.

atropurpurea—are polyploid (2n=220; Dawson, 1995; Dawson & Beuzenberg, 2000), as

well, whereas most species in the genus, are diploid (2n=44) (Beuzenberg, 1983; Wagner

et al., 1999; Dawson & Beuzenberg 2000). Although these morphological and cytological

characters appear to support the close relationships among these species, it should also be

noted that C. archboldiana from New Guinea (Merrill & Perry, 1945) is also similar in their morphology to C. ernodeoides. Unfortunately, no material of this species was available during this investigation so their relationship remains unknown. As a result,

New Zealand, Australia, or New Guinea are all possible source areas and it is impossible to confirm which of these gave rise to the initial Hawaiian colonist of this lineage.

Additionally, the relationship of C. ernodeoides to C. persicifolia of Fiji and C. strigulosa of Samoa, should be investigated in more detail. These three species are all pioneers on young lava flows, have black fruits and share similar floral and vegetative characters

(Oliver, 1935; Whistler, 1992; Wagner et al., 1999). It is possible that the arrival of C.

ernodeoides to the Hawaiian Islands was via stepping-stone dispersal from New Zealand

129

and/or Tasmania through New Guinea, Fiji and Samoa in some fashion or as a direct

dispersal from a locality farther afield (i.e., New Zealand or Tasmania without

intervening archipelagos acting as stepping stones), but this cannot yet be determined.

Aim B. To understand the historical biogeography of Pacific Coprosma

Hypothesis 4: Coprosma is not Gondwanan in temporal origin, but rather is much more

recently evolved.

Dating analyses indicate that Coprosma originated in the late Oligocene, ca. 25.6

Ma, most likely in New Zealand. At this time New Zealand, Australia and Antarctica were already separated from one another and Gondwana per se no longer existed.

Similarly, Coprosma was inferred to have diverged from its sister genus, Nertera, in the

Oligocene, around the time of the Oligocene Marine Transgression (OMT) of New

Zealand. During the OMT all or a significant portion of the New Zealand landmass was submerged (Stevens et al., 1988; Cooper & Millener, 1993; Cooper & Cooper, 1995). If total submergence occurred then then MCRA of Coprosma and Nertera could have colonized New Zealand from Australia or some other nearby area, which was not submerged during the Oligocene. Alternatively, in the scenario of an incomplete drowning of New Zealand during the OMT, the MRCA of Coprosma and Nertera may have been able to persist on low-laying emergent islands that made up New Zealand during the Oligocene. It is not possible to resolve this for Coprosma, however, as the location of Coprosma/Nertera-like microfossil records from mid-Miocene and mid-

130

Oligocene fossil deposits of Murray Basin of southeastern Australia and New Zealand

(Couper, 1960; Graham, 2009; MacPhail, 1999) could support either scenario.

Hypothesis 5: Patterns of Coprosma across the Pacific are similar to patterns expected

for bird-mediated dispersal.

Coprosma fruits are fleshy and have been shown to be attractive to and eaten by birds in

New Zealand (Lord & Markey, 2004). Our results indicate that a minimum of 30 long

distance dispersal events are required to explain the extant distribution Coprosma across

the Pacific and that these recurring dispersals are to the same islands not to randomly

determined locations as would occur with wind dispersed fruits and oceanic drift.

Multiple visits to the same island or land mass would also be expected during trans-

Pacific and regional journeys in the Pacific as flyways are well documented in the region,

and even though they may change over time they are still regularly used routes for

sustained periods. Additionally, bird dispersal is also the most common explanation for

plant colonizations to the Hawaiian (Carlquist, 1967), Juan Fernández (Bernardello et al.,

2001, 2006), Azores (Schafer, 2002), Galápagos (McMullen, 1999) and Canary Islands

(Bramwell & Bramwell, 2001) and there is no reason to assume that it is not important here.

131

Hypothesis 6: Leaky-dioecy, woodiness and fleshy fruit are among the innovations aiding

the dispersal and establishment of Coprosma across the Pacific.

Studies of Hawaiian, Canary and Juan Fernández Island plants indicate that fleshy fruit is often strongly correlated with a dimorphic breeding system such as dioecy, widely present in Coprosma taxa (Sakai et al, 1995; McMullen, 1999; Bramwell & Bramwell,

2001). Besides the orange and black-fruited species, Coprosma taxa include those with blue, white, and purple drupes as well and all are apparently animal dispersed (by lizards as well as birds in New Zealand). The principal disadvantage of dioecy as outlined by

Baker (1955) is that this condition inherently requires at least two individuals for sexual reproduction. However, despite this logic, a paradox exists in that many insular floras, including those of Hawaii, Reunion, Ogasawara and Juan Fernández Islands, which have a higher proportion of dioecious taxa in their floras than do similar sized continental areas. One explanation for this apparent paradox, (see Barrett, 1996; Baker & Cox, 1984) is that there is a reasonable frequency of incomplete or “leaky” dioecy whereby some individuals occasionally produce hermaphroditic flowers, thus allowing for selfing to occur. With leaky dioecy only a few initial colonists would be required for establishment despite the otherwise obligate outcrossing nature of a dioecious reproductive strategy.

Leaky dioecy is documented among Coprosma species in many locations in the Pacific and appears to be at the very least a good back up strategy for colonizing new areas.

Woodiness, also characteristic of Coprosma (but not its sister genus, Nertera) is

also correlated with dioecy in insular floras including those of Hawaiian Islands (Sakai et

al. 1995), Panama (Croat, 1979), and Puerto Rico (Flores & Schemske, 1984). The

132

positive correlation of woodiness and dioecy has been attributed to the heightened

selective pressures for outcrossing in long-lived plants that might otherwise self too often thus reducing genetic variation within isolated and presumably small populations (Bawa

1980; Steiner 1988). As woodiness increases the longevity of individuals also increases thereby providing more opportunities over time in which to contribute genetic resources to the next generation. Nertera, the sister genus is herbaceous and short lived and is strictly hermaphroditic. The advantages gained by being both woody and dioecious and bird dispersed would be significant in allowing Coprosma and other plants with similar syndromes to successfully colonize Pacific islands.

Aim C. To understand the morphological relationships among Coprosma taxa and their evolutionary implications

Hypothesis 7: The evolution of morphological characters shifted from ‘simple’ to complex.

Coprosma is morphologically plastic with a great range of variation in character states such as number of flowers per inflorescence, presence of simple or compound inflorescences, number of floral parts, leaf size and shape, growth habit, and fruit color, for example. Tracing character states along the branches of the trees generated show that there is no consistent set of character states, only that overall there is a general

‘tendency towards the multiplication of floral parts,’ Oliver (1935) or more broadly interpreted increasing complexity, albeit not strictly so. This study also indicates that the

133

evolution of leaf size and inflorescence structure is highly homoplasic and that there have

been many reversals in character states. Therefore, our findings support Gardner’s (2002)

conclusion that delimitation of species based upon leaf size and inflorescence structure is

not sufficient to delimit subgeneric divisions. Further, the characters used by Heads

(1996) to delimit his subgenera—leaf size (small or large), and inflorescence structure

(solitary flowers or not solitary) (along with and east versus west vicariant split overlapping in New Zealand) could not be used to provide robust phylogenetic or taxonomic assignments. Taxonomic changes suggested on the basis of the current work rely on a combination of relationships established using molecular phylogenetics and morphological characters that are consistent within clades.

Synthesis

This study has resulted in a better understanding of the evolution and biogeography of Coprosma in the Pacific. Among the outstanding features of the genus that have come to light here is the success of the strategy that derives from the combination of morphological and reproductive plasticity coupled with bird dispersal that has given Coprosma an edge over most Pacific island taxa. Additionally, once established on an island or archipelago there is often extensive radiation. If a new colonization is successful, it too may give rise to internal radiations, but into slightly different habitats, or in a different way so that each succeeding colonizing lineage continues to maintain itself. There have been long distance and direct colonizations from New Zealand to the

Hawaiian Islands and other distant archipelagoes, for example, and there have also been shorter “stepping stone” movements from the Marquesas to Hawaii or from Australia to

134

New Zealand and back, and also between other more closely situated archipelagoes

elsewhere in the Pacific. That all of this is relatively recent (primarily >10 Ma) as well is

also notable. As habitats have become available through volcanic island creation

Coprosma has become established (repeatedly) and re-radiated and recolonized. Further

work will reveal more details about the role of chromosomal evolution, frequency and

importance of hybridization in the development and persistence of new lineages. Relating

the biology of Coprosma to the ecology of the islands can also provide insight into some of the larger phenomena associated with the creation of the rich, and different, floras of

the Pacific islands.

135

LITERATURE CITED

Alice LA, Campbell CS. 1999. Phylogeny of Rubus (Rosaceae) based on nuclear ribosomal DNA internal transcribed spacer region sequences. American Journal of Botany 86: 81- 97. Allan H. 1961. Flora of New Zealand: indigenous tracheophyta. Wellington. Allwood, J., Gleeson, D., Mayer, G., Daniels, S., Beggs, J.R. & Buckley, T.R. 2010. Support for vicariant origins of the New Zealand Onychophora. Journal of Biogeography: 37, 669-681. Alvarez I, Wendel JF. 2003. Ribosomal ITS sequences and plant phylogenetic inference. Molecular Phylogenetics and Evolution 29: 417-434. Anderson CL, Rova JHE, Andersson L. 2001. Molecular phylogeny of the tribe Anthospermeae (Rubiaceae): systematic and biogeographic implications. Australian systematic Botany 14: 231-244. Azizi, G. & Flenley, J. 2008. The last glacial maximum climatic conditions on Easter Island. Quaternary International 184: 166-176. Baker, H.G. & Cox, P.A. 1984. Further thoughts on dioecism and islands. Annals of the Missouri Botanical Garden: 244-253. Baker, H.G. 1955. Self-compatibility and establishment after'long-distance'dispersal. Evolution, 9, 347-349. Baldwin BG, Wagner WL. 2010. Hawaiian angiosperm radiations of North American origin. Annals of Botany 105: 849-879. Baldwin PH. 1947. Foods of the Hawaiian Goose. The Condor 49: 108-120. Barrett, S.C. 2003. Mating strategies in flowering plants: the outcrossing–selfing paradigm and beyond. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences 358: 991-1004. Batt, G.E., Braun, J., Kohn, B.P. & McDougall, I. 2000. Thermochronological analysis of the dynamics of the Southern Alps, New Zealand. Geological Society of America Bulletin 112: 250-266. Bawa, K.S. 1980. Evolution of dioecy in flowering plants. Annual Review of Ecology and Systematics 11: 15-39. Bernardello, G., Anderson, G.J., Stuessy, T.F. & Crawford, D.J. 2001. A survey of floral traits, breeding systems, floral visitors, and pollination systems of the angiosperms of the Juan Fernandez Islands (Chile). The Botanical Review 67: 255-308. Bernardello, G., Anderson, G.J., Stuessy, T.F. & Crawford, D.J. 2006. The angiosperm flora of the Archipelago Juan Fernandez (Chile): origin and dispersal. Botany 84: 1266-1281. Beuzenberg EJ. 1983. Contributions to a chromosome atlas of the New Zealand Flora-25 Miscellaneous species. New Zealand Journal of Botany 21: 13-20. Birch J.L., Keeley SC, Morden CW. 2012. Molecular phylogeny and dating of Asteliaceae (Asparagales): Astelia s.l. evolution provides insight into the Oligocene history of New Zealand. Molecular Phylogenetics and Evolution. Bramwell, Z.I. & Bramwell, D. 2001. Wild flowers of the Canary Islands. In. Editorial Rueda, Madrid, Spain Bremer, B. & Eriksson, T. 2009. Time tree of Rubiaceae: phylogeny and dating the family, subfamilies, and tribes. International Journal of Plant Sciences 170: 766-793.

136

Brown FB. 1935. Flora of southeastern Polynesia. III. Dicotyledons. Bernice P. Bishop Museum Bulletin 130: 1-386. Carlquist S. 1980. Hawaii: a natural history. Geology, climate, native flora and fauna above the shoreline. SB Printers, Inc. for Pacific Tropical Botanical Garden: Honolulu. Carlquist SJ, Baldwin BG, Carr GD. 2003. Tarweeds & silverswords: evolution of the Madiinae (Asteraceae). Missouri Botanical Garden Press. Carlquist, S. 1967. The biota of long distance dispersal. I. Principles of dispersal and evolution. Quarterly Reviews in Biology 41: 247-270. Cheeseman T. 1925. Manual of the New Zealand. Flora. Cheeseman TF. 1887. On the New Zealand species of Coprosma. Cheeseman TF. 1909. On the systematic botany of the islands to the south of New Zealand. Philosophical Institute of Canterbury. Clague DA, J. C. Braga, D. Bassi, P. D. Fullagar, W. Renema, Webster. JM. 2010. Early Oligocene to early Miocene carbonate bank deposits on Kōkō seamount, Emperor spelling Seamount Chain: implications for the biogeography of the Hawaiian Islands. Journal of Biogeography 37: 1022-1033. Cooper, A. & Cooper, R.A. 1995. The Oligocene bottleneck and New Zealand biota: genetic record of a past environmental crisis. Proceedings of the Royal Society of London. Series B: Biological Sciences 261: 293-302. Cooper, R.A. & Millener, P.R. 1993. The New Zealand biota: historical background and new research. Trends in Ecology & Evolution 8: 429-433. Couper, R. & Harris, W. 1960. Pliocene and Pleistocene plant microfossils from drillholes near Frankton, New Zealand. New Zealand Journal of Geology and Geophysics 3: 15- 22. Cox, S. & Findlay, R. (1995) The Main Divide fault zone and its role in formation of the Southern Alps, New Zealand. New Zealand Journal of Geology and Geophysics 38: 489-499. Crawford, D., Anderson, G. & Bernardello, G. (2011) The reproductive biology of island plants. The biology of island floras 5: 11-36. Croat, T.B. (1979) The sexuality of the Barro Colorado Island flora (Panama). Phytologia 42 Dawson M. 1995. Contributions to a chromosome atlas of the New Zealand flora - 34 Miscellaneous species. New Zealand Journal of Botany 33: 477-487. Dawson MI, Beuzenberg EJ. 2000. Index of chromosome numbers of indigenous New Zealand spermatophytes. New Zealand Journal of Botany 38: 47-150. de Lange PJ, Gardner RO. 2002. A taxonomic reappraisal of Coprosma obconica Kirk (Rubiaceae: Anthospermeae). New Zealand Journal of Botany 40: 25-38. Drummond, A.J., Suchard, M.A., Xie, D. & Rambaut, A. 2012. Bayesian phylogenetics with BEAUti and the BEAST 1.7. Molecular Biology and Evolution 29: 1969-1973. Dunbar-Co S, Wieczorek AM, Morden CW. 2008. Molecular phylogeny and adaptive radiation of the endemic Hawaiian Plantago species (Plantaginaceae). American Journal of Botany 95: 1177-1188. Edgar RC. 2004. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Research 32: 1792-1797. Feliner GN, Rossello JA. 2007. Better the devil know you? Guidelines for insightful utilization of nrDNA ITS in species-level evolutionary studies in plants. Molecular Phylogenetics and Evolution 44: 911-919.

137

Florence J, D. H. Lorence DH. 1997. Introduction to the flora and vegetation of the Marquesas Islands. Allertonia 7: 226-237. Florence J. 1986. Sertum polynesicum II: RUBIACEAE nouvelles des îles Marquises (Polynésie Française). Bulletin du Muséum National d'Histoire Naturelle. Section B: Adansonia 8: 3-11. Flores, S. & Schemske, D.W. 1984. Dioecy and monoecy in the flora of Puerto Rico and the Virgin Islands: ecological correlates. Biotropica 34: 132-139. Fosberg FR. 1937. Some Rubiaceae of Southeastern Polynesia. Bernice P. Bishop Museum. Fosberg FR. 1948. Derivation of the Hawaiian Flora. In: Zimmerman EC, ed. Insects of Hawaii. Honolulu: University of Hawaii Press. 107-119. Fosberg FR. 1982. A preliminary conspectus of the genus Leptostigma (Rubiaceae). Acta Phytotax. Geobot 33: 73-83. Galloway, R. & Kemp, E. 1981. Late Cainozoic environments in Australia: 51-80. Ecological Biogeography of Australia: i-xix, 1-2142. Gardner R. 1999. 1999. Systematic notes on Leptostigma setulosum (Rubiaceae). New Zealand Natural Sciences 24: 1-8. Gardner R. 2002a. The genus Coprosma (Rubiaceae) in New Guinea. Candollea 57: 97-130. Gardner R. 2002b. Systematic notes on Coprosma (Rubiaceae:Anthospermeae). New Zealand Natural Sciences 27: 41-57. Geiger J, Ranker T. 2005. Molecular phylogenetics and historical biogeography of Hawaiian Dryopteris (Dryopteridaceae). Molecular Phylogenetics and Evolution 34: 392-407. Gillespie, R.G., Claridge, E.M. & Roderick, G.K. 2008. Biodiversity dynamics in isolated island communities: interaction between natural and human‐mediated processes. Molecular Ecology 17: 45-57. Givnish TJ, Millam KC, Mast AR, Paterson TB, Theim TJ, Hipp AL, Henss JM, Smith JF, Wood KR, Sytsma KJ. 2009. Origin, adaptive radiation and diversification of the Hawaiian lobeliads (Asterales: Campanulaceae). Proceedings of the Royal Society B: Biological Sciences 276: 407-416. Givnish TJ, Montgomery RA, Goldstein G. 2004. Adaptive radiation of photosynthetic physiology in the Hawaiian lobeliads: light regimes, static light responses, and whole- plant compensation points. American Journal of Botany 91: 228-246. Glenny D, Cruickshank J, Rolfe J. 2010. Key to Coprosma species of New Zealand. http://www.landcareresearch.co.nz/resources/identification/plants/coprosma-key. Accessed on September 7, 2013. Graham, A. 2009. Fossil record of the Rubiaceae. Annals of the Missouri Botanical Garden 96: 90-108. Greenwood R, Atkinson I. 1977. Evolution of divaricating plants in New Zealand in relation to moa browsing. Proc. NZ Ecol. Soc., 21-33. Hair J. 1963. Cytogeographical relationships of the southern podocarps. Pacific basin biogeography 401: 414. Harbaugh DT, Baldwin BG. 2007. Phylogeny and biogeography of the sandalwoods (Santalum, Santalaceae): repeated dispersals throughout the Pacific. American Journal of Botany 94: 1028-1040. Harbaugh DT, Wagner WL, Allan GJ, Zimmer EA. 2009. The Hawaiian Archipelago is a stepping stone for dispersal in the Pacific: an example from the plant genus Melicope (Rutaceae). Journal of Biogeography 36: 230-241.

138

Heads M. 1996. Biogeography, taxonomy and evolution in the Pacific genus Coprosma (Rubiaceae). Candollea 51: 381-404. Heenan, P., Dawson, M. & Bicknell, R. 2002. Evidence for apomictic seed formation in Coprosma waima (Rubiaceae). New Zealand Journal of Botany 40: 347-355. Heenan, P., Molloy, B., Bicknell, R. & Luo, C. 2003. Levels of apomictic and amphimictic seed formation in a natural population of Coprosma robusta (Rubiaceae) in Riccarton Bush, Christchurch, New Zealand. New Zealand Journal of Botany 41: 287-291. Hillebrand, W. 1888. Flora of the Hawaiian Islands: a description of their phanerogams and vascular cryptogams. Williams & Norgate eds. Hooker J. 1867. On insular floras: a lecture. Journal of Botany 5: 23-31. Hooker JD. 1853. Introductory Essay to the Flora of New Zealand: Published Under the Authority of the Lords Commissioners of the Admirality. Lovell Reeve. Howarth DG, Baum DA. 2005. Genealogical evidence of homoploid hybrid speciation in an adaptive radiation of Scaevola (Goodeniaceae) in the Hawaiian Islands. Evolution 59: 948-961. Howarth DG, Gardner DE, Morden C. 1997. Phylogeny of Rubus subgenus Idaeobatus (Rosaceae) and its implications toward colonization of the Hawaiian Islands. Systematic Botany 22: 433-441. Howarth, D.G., Gustafsson, M.H., Baum, D.A. & Motley, T.J. 2003. Phylogenetics of the genus Scaevola (Goodeniaceae): implication for dispersal patterns across the Pacific Basin and colonization of the Hawaiian Islands. American Journal of Botany 90: 915- 923. Hueselenbeck JP, Ronquist F. 2001. MrBayes: Bayesian inference of phylogeny. Biometrics 17: 754-755. Kamp, P.J., Green, P.F. & Tippett, J.M. 1992. Tectonic architecture of the mountain front‐ foreland basin transition, South Island, New Zealand, assessed by fission track analysis. Tectonics 11: 98-113. Keast A. 1996. Pacific biogeography: patterns and processes. The origin and evolution of Pacific island biotas, New Guinea to eastern Polynesia: patterns and processes: 477-512. Keeley SC, Funk VA. 2011. Origin and evolution of Hawaiian endemics: new patterns revealed by molecular studies. In: Bramwell D and Caujape-Castells J, eds. The biology of island floras: Cambridge University Press. Kelly D. 1994. Towards a numerical definition for divaricate (interlaced small-leaved) shrubs. New Zealand Journal of Botany 32: 509-518. Knapp, M., Mudaliar, R., Havell, D., Wagstaff, S.J. & Lockhart, P.J. 2007. The drowning of New Zealand and the problem of Agathis. Systematic Biology 56: 862-870. Lee WG, Weatherall I, Wilson J. 1994. Fruit conspicuousness in some New Zealand Coprosma (Rubiaceae) species. Oikos 69: 87-94. LeMasurier, W.E. & Landis, C.A. 1996. Mantle-plume activity recorded by low-relief erosion surfaces in West Antarctica and New Zealand. Geological Society of America Bulletin 108: 1450-1466. Lindqvist C, Albert VA. 2002. Origin of the Hawaiian endemic mints within North American Stachys (Lamiaceae). American Journal of Botany 89: 1709-1724. Lindqvist C, Motley TJ, Jeffrey JJ, Albert VA. 2003. Cladogenesis and reticulation in the Hawaiian endemic mints (Lamiaceae). Cladistics 19: 480-495.

139

Lledó MD, Karis PO, Crespo MB, Fay MF, Chase MW. 2011. Endemism and evolution in Macaronesian and Mediterranean Limonium taxa. The biology of island floras. Cambridge University Press, London: 325-337. Lord JM, Markey AS, Marshall J. 2002. Have frugivores influenced the evolution of fruit traits in New Zealand? CAB International Wallingford, UK. Lord JM, Marshall J. 2001. Correlations between growth form, habitat, and fruit colour in the New Zealand flora, with reference to frugivory by lizards. New Zealand Journal of Botany 39: 567-576. Lowrey TK, Crawford DJ. 1985. Allozyme divergence and evolution in Tetramolopium (Compositae: Astereae) on the Hawaiian Islands. Systematic Botany 10: 64-72. Macphail, M. (1999) Palynostratigraphy of the Murray Basin, inland southeastern Australia. Palynology, 23, 197-240. Maddison W, Maddison D. 2010. Mesquite: a modular system for evolutionary analysis. 2010. Version 2.74. Available at: mesquiteproject.org/mesquite/download/ download. html. Maddison W. 2001. Parsimony package for Mesquite 0.95 ed. Markey AS, de Lange PJ. 2003. A new species of Coprosma Sect. Acerosae (Rubiaceae) endemic to the South and Stewart Islands of New Zealand. New Zealand Journal of Botany 41: 459-473. Markey AS, JM L, DA O. 2004. Coprosma talbrockiei: an oddball sheds light on the Coprosminae SYSTANZ meeting. Whakapapa Village. McMullen, C.K. 1999. Flowering plants of the Galápagos. Cornell University Press. Merrill ED, Perry LM. 1945. Plantae Papuanae Archbolianae XVI. Journal of the Arnold Arboretum 26: 229-266. Merrill ED, Perry LM. 1945. Plantae Papuanae Archbolianae XVI. Journal of the Arnold Arboretum 26: 229-266. Miller, M.A., Pfeiffer, W. & Schwartz, T. 2012. The CIPRES science gateway: enabling high-impact science for phylogenetics researchers with limited resources. Proceedings of the 1st Conference of the Extreme Science and Engineering Discovery Environment: Bridging from the eXtreme to the campus and beyond p. 39. Moore J.W. 1933. New and critical plants from Raiatea. The Bishop Museum. Moore L, Mason R. 1974. Coprosma talbrockiei sp. nov. and allied creeping species (Rubiaceae). New Zealand Journal of Botany 12: 137-148. Motley TJ. 2003. Phylogeny of the Hawaiian and Pacific Hedyotis (Rubiaceae); fruit evolution and the implications for conservation and genomics. Muellner, A.N., Pannell, C.M., Coleman, A. & Chase, M.W. 2008. The origin and evolution of Indomalesian, Australasian and Pacific island biotas: insights from Aglaieae (Meliaceae, Sapindales). Journal of Biogeography 35: 1769-1789. Nepokroeff M, Sytsma KJ, Wagner WL, Zimmer EA. 2003. Reconstructing ancestral patterns of colonization and dispersal in the Hawaiian understory tree genus Psychotria (Rubiaceae): a comparison of parsimony and likelihood approaches. Systematic Biology 52: 820-838. Okada M, Lowrey T, Whitkus R. 2000. Quantitative morphological variation in Tetramolopium (Asteraceae) in Hawaii and the Cook Islands. Plant Systematics and Evolution 221: 1-13.

140

Okada M, Whitkus R, Lowrey T. 1997. Genetics of adaptive radiation in Hawaiian and Cook Islands species of Tetramolopium (Asteraceae; Astereae). I. Nuclear RFLP marker diversity. American Journal of Botany 84: 1236-1236. Oliver WRB. 1935. The genus Coprosma. Bulletin of the Bernice P. Bishop Museum: Honolulu. Ollier, C. 1986. The origin of alpine landforms in Australasia. Flora and fauna of alpine Australasia Barlow BA. 3-26. Orchard AE. 1986. A revision of the Coprosma pumila (Rubiaceae) complex in Australia, New Zealand and the Subantarctic Islands. Brunonia 9: 119-138. Oxelman B, Linden M, Berglund D. 1997. Chloroplast rps16 intron phylogeny of the tribe Sileneae (Caryophyllaceae). Plant Systematics and Evolution 206: 393-410. Papadopulos A, Baker W, Crayn D, Butlin R, Kynast R, Hutton I. 2011. Speciation with gene flow on Lord Howe Island. Proceedings of the National Academy of Science of the United States of America 108: 13188-13193. Percy DM, Garver AM, Wagner WL, James HF, Cunningham CW, Miller SE, Fleischer RC. 2008. Progressive island colonization and ancient origin of Hawaiian Metrosideros (Myrtaceae). Proceedings of the Royal Society B: Biological Sciences 275: 1479-1490. Pillans, B. & Wright, I. 1992. Late Quaternary tephrostratigraphy from the southern Havre Trough Bay of Plenty, northeast New Zealand. New Zealand Journal of Geology and Geophysics 35: 129-143. Pole, M. (1994) The New Zealand flora-entirely long-distance dispersal? Journal of Biogeography 625-635. Posada D, Buckley T. 2004. Model selection and model averaging in phylogenetics: advantages of Akaike Information Criterion and Bayesian approaches over Likelihood Ratio Tests. Systematic Biology 53: 793-808. Posada, D. & Crandall, K.A. 1998. Modeltest: testing the model of DNA substitution. Bioinformatics 14: 817-818. Price, J.P. & Wagner, W.L. 2004 Speciation in Hawaiian angiosperm lineages: cause, consequence, and mode. Evolution 58: 2185-2200. Puff C, Robbrecht E. 1988. The taxonomic position of the Australian endemic Durringtonia (Rubiaceae). Australian systematic botany 1: 191-197. Puff C. 1986. Phylogenetic Relationships and Evolutionary Aspects: A Biosystematic Study of the African and Madagascan Rubiaceae-Anthospermeae: Springer. 467-520. Rambaut, A. & Drummond, A. 2009. FigTree v1. 3.1. Computer program and documentation distributed by the author at http://tree.bio.ed.ac.uk/software. Rambaut, A. & Drummond, A.J. 2007. Tracer v1.4. Available at: http://beast.bio.ed.ac.uk/Tracer (accessed 8-8-2013). Randell RA, Morden C. 1999. Hawaiian Plant DNA library II: endemic, indigenous, and introduced species. Pacific Science 53: 401-417. Rull, V., Cañellas-Boltà, N., Sáez, A., Giralt, S., Pla, S. & Margalef, O. 2010. Paleoecology of Easter Island: evidence and uncertainties. Earth-Science Reviews 99: 50-60. Sakai, A.K., Wagner, W.L., Ferguson, D.M. & Herbst, D.R. 1995. Origins of dioecy in the Hawaiian flora. Ecology, 2517-2529. Sakai, A.K., Weller, S.G., Wagner, W.L., Nepokroeff, M. & Culley, T.M. 2006. Adaptive radiation and evolution of breeding systems in Schiedea (Caryophyllaceae), an endemic Hawaiian genus 1. Annals of the Missouri Botanical Garden 93: 49-63.

141

Schafer, H. 2002. Flora of the Azores—A Field Guide. Weikersheim: Margraf vi, 264p.-col. illus.. ISBN 3823613685 En Icones. Geog, 1 Shaw, J., Lickey, E.B., Schilling, E.E. & Small, R.L. 2007. Comparison of whole chloroplast genome sequences to choose noncoding regions for phylogenetic studies in angiosperms: the tortoise and the hare III. American Journal of Botany 94: 275-288. Skottsberg C. 1953. Chromosome numbers in Hawaiian flowering plants. Ark. Bot. 3: 63-70. Smith, J.M. 1982. A history of Australasian vegetation. McGraw-Hill Companies. Soltis PS, Soltis DE, Weller SG, Sakai AK, Wagner WL. 1996. Molecular phylogenetic analysis of the Hawaiian endemics Schiedea and Alsinidendron (Caryophyllaceae). Systematic Botany 21: 365-379. Stamatakis A. 2006. RaxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 22: 26-88. Stamatakis, A. 2006. RaxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 22: 26-88. Steiner, K.E. 1988. Dioecism and its correlates in the Cape flora of South Africa. American Journal of Botany 1742-1754. Stevens, G.R., McGlone, M., McCulloch, B. & Ward, V. 1988. Prehistoric New Zealand. Heinemann Reed Auckland. Suggate, R. 1990. Late Pliocene and quaternary glaciations of New Zealand. Quaternary science reviews 9: 175-197. Swofford DL. 2003. PAUP*: Phylogenetic Analysis Using Parsimony (*and Other Methods). Version 4 beta 10. 4 ed. Sunderland, Massachusetts: Sinauer Associates. Tamura K, Peterson D, N. P, Stecher G, Nei M, Kumar S. 2011. MEGA5: Molecular Evolutionary Genetics Analysis using Maximum Likelihood, Evolutionary Distance, and Maximum Parsimony methods. Molecular Biology and Evolution 28: 2731-2739. Thompson, I.R. 2010. A new species of Leptostigma (Rubiaceae: Coprosminae) and notes on the Coprosminae in Australia. Muelleria 28: 29-39. Tomlinson P. 1978. Some qualitative and quantitative aspects of New Zealand divaricating shrubs. New Zealand Journal of Botany 16: 299-310. Trewick, S.A. & Morgan, Richards, M. 2005. After the deluge: mitochondrial DNA indicates Miocene radiation and Pliocene adaptation of tree and giant weta (Orthoptera: Anostostomatidae). Journal of Biogeography 32: 295-309. Tye A, Francisco-Ortega J. 2011. Origins and evolution of Galapagos endemic vascular plants. The biology of island floras. Cambridge University Press, London: 89-153. Utteridge TMA. 2002. New Species of Coprosma (Rubiaceae) from New Guinea. Kew Bulletin 57: 195-203. Vamosi, J., Otto, S. & Barrett, S. 2003. Phylogenetic analysis of the ecological correlates of dioecy in angiosperms. Journal of Evolutionary Biology 16: 1006-1018. van Royen P. 1983. The alpine flora of New Guinea. In: Cramer J, ed. Vaduz, Germany. Wagner WL, Funk VA. 1995. Hawaiian biogeography: Evolution on a hot spot archipelago. Smithsonian Institution Press: Washington D.C. Wagner WL. 1991. Evolution of waif floras: A comparison of the Hawaiian and Marquesan Archipelagos. In Dudley EC, ed. The unity of evolutionary biology proceedings of the Fourth International Congress of Systematics and Evolutionary Biology. Portland, Oregon Dioscorides Press, 267-284.

142

Wagner, Herbst DR, Sohmer SH. 1990. Manual of the flowering plants of Hawaii. University of Hawaii Press and Bishop Museum Press: Honolulu. Wagner, Herbst DR, Sohmer SH. 1999. Manual of the flowering plants of Hawaii, Revised Edition. University of Hawaii Press and Bishop Museum Press: Honolulu. Wagner, W.L. & Lorence, D.H. 2011. Revision of Coprosma (Rubiaceae, tribe Anthospermeae) in the Marquesas Islands. Botany of the Marquesas Islands: new taxa, combinations, and revisions (ed. by W.L. Wagner and D.H. Lorence), pp. 109-124. PhytoKeys. Wagstaff SJ, Garnock-Jones PJ. 1998. Evolution and biogeography of the Hebe complex (Scrophulariaceae) inferred from ITS sequences. New Zealand Journal of Botany 36: 425-437. Wardle P. 1991. Vegetation of New Zealand. CUP Archive. Webb C. 1996. A rose by any other name: two problems of scent in the naming and typification of New Zealand plants. New Zealand Journal of Botany 34: 281-283. Weller SG, Wagner WL, Sakai AK. 1995. A phylogenetic analysis of Schiedea and Alsinidendron (Caryophyllaceae: Alsinoideae): implications for the evolution of breeding systems. Systematic Botany: 315-337. Whistler WA. 1992. Vegetation of Samoa and Tonga. Pacific Science 46: 159-178. Wichman SR. 2000. Molecular Phylogeny of Coprosma based on rDNA ITS and ETS sequences. Unpublished Masters, University of Auckland. Wichman, S.R., Wright, S.D., Cameron, E., Keeling, K., Jeannette, D. & Gardner, R.C. 2002. Elevated genetic heterogeneity and Pleistocene climatic instability: inferences from nrDNA in New Zealand Coprosma (Rubiaceae) Journal of Biogeography 29: 943-954. Wilson RD. 1979. Chemotaxonomic studies in the Rubiaceae 1. Methods for the identification of hybridisation in the genus Coprosma JR et G. Forst. using flavonoids. New Zealand Journal of Botany 17: 113-116. Wilson RD. 1984. Chemotaxonomic studies in the Rubiaceae 2. Leaf flavonoids of New Zealand coprosmas. New Zealand Journal of Botany 22: 195-200. Winkworth, R.C., Wagstaff, S.J., Glenny, D. & Lockhart, P.J. 2002. Plant Dispersal N.E.W.S. from New Zealand. Trends in Ecology and Evolution 17: 514-520. Winkworth, R.C., Wagstaff, S.J., Glenny, D. & Lockhart, P.J. 2005. Evolution of the New Zealand mountain flora: origins, diversification and dispersal. Organisms Diversity & Evolution 5: 237-247. Wright SD, Yong CG, Wichman SR, Dawson JW, Gardner RC. 2001. Stepping stones to Hawaii: a trans-equatorial dispersal pathway for Metrosideros (Myrtaceae) inferred from nrDNA (ITS + ETS). Journal of Biogeography 28: 769-774. Yang Y. 2012. Phylogenetics and evolution of Euphorbia subgenus Chamaesyce: University of Michigan. Yu, Y., Harris, A. & He, X. (2011) RASP (reconstruct ancestral state in phylogenies) 2.0 beta. WWW program] URL http://mnh.scu.edu.cn/soft/blog/RASP [accessed on 13 June 2011].

143