INFORMATION TO USERS

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

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

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

Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps.

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

Bell & Howell Information and Learning 300 North Zaeb Road. Ann Arbor, Ml 48106-1346 USA 600-521-0600 UMT

EVOLUTION AND BIOGEOGRAPHY OF MALESHERBIACEAE, AN ENDEMIC FAMILY OF ARID WESTERN SOUTH AMERICA

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of the Ohio State University

By

Karla Marie Gengler, B.A.

*****

The Ohio State University 2000

Dissertation Committee: Approved by

Professor Daniel J. Crawford, Co-adviser Professor Andrea Wolfe, Co-adviser Daniel J. CrawfoE Professor John Wenzel Department of Evolution, Ecology, and Organismal Biology K , L L '^/ Andrea Wolfe, Department of Evolution, Ecology, and Organismal Biology UMI Number 9982566

Copyright 2000 by Gengler, Karla Marie

Ail rights reserved.

UMI*

UMI Microform9982566 Copyright 2000 by Bell & Howell Information and teaming Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code.

Bell & Howell Information and teaming Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 Copyright by Karla Marie Gengler-Nowak 2000 ABSTRACT

Maiesherbiaceae, which are endemic to arid Peru, Chile, and adjacent Argentina, provide an interesting research opportunity, as the family’s small size and distribution in a virtual continental island make it ideal for phylogenetic and biogeographic studies. The family’s distribution in both Chile and Peru is unusual since the adjacent countries share very few desert taxa, so Maiesherbiaceae are an ideal group for the study of the biogeography of the entire dry region west of the Andes. To study the biogeography of the family, a phylogeny based on ITS sequence data was reconstructed. This analysis confirmed the monophyly of the family and showed it to consist of five major clades.

Four of the five major clades contain morphologically very similar species; the two species comprising the Hfth share few such traits. The five major clades are herein recognized as subgenera.

These subgenera correspond closely to the distributions of species in the family, suggesting that the physical history of the region strongly influenced the evolution of

Maiesherbiaceae. The morphologically distinct Peruvian species form a subgenus and constitute a terminal clade in the family, indicating the family originated in Chile and subsequently migrated to Peru. By examining the geological history of the region, it was

u inferred that the family split from Tumeraceae in the late Miocene. A rapid radiation giving rise to the fîve subgenera occurred in the Pliocene, with a more recent radiation event affecting much of the family in the Pleistocene.

Allozyme diversities of four Peruvian inter-Andean species were examined. The data suggest that at least two of the species are the result of founder events; morphological similarities among the four closely related taxa indicate that they may be recently diverged. Founder events may be relatively common in the flora of the arid inter-Andean valleys because these valleys are quite isolated by intervening regions of high elevation.

The M. humilis Poepp. species complex was also studied in detail because the four species and three varieties previously described are difficult to distinguish. A phenetic

analysis of morphological data gathered from herbarium specimens. The further emphasized the similarities among groups of organisms. Five clusters are recognized as

varieties of a single species, M. humilis. One former species, M. multiflora Ric., is not

recognized as separate from M. humilis var. parviflora, and two others, M. gabrielae Ric.

and Af. taltalina Ric., are recognized as varieties of M. humilis.

Ill Dedicated to my husband, Joseph Nowak

IV ACKNOWLEDGMENTS

I thank my primary adviser, Dr. Daniel J. Crawford, for his thoughtful critiques and advice, which encouraged me greatly during the course of my studies. I also thank my former adviser. Dr. Tod F. Stuessy, for introducing me to Maiesherbiaceae, helping me begin my studies, and offering much-needed advice on travel in South America and morphological work. Dr. Andrea Wolfe has also been a source of encouragement and helpful critiques of my work; I further thank her for agreeing to co-ad vise me. I thank Dr.

John Wenzel for use of his copies of the programs NONA, Pee-Wee, and Hennig86, his many hours of assistance with these programs, and especially for spending so much time teaching me cladistic theory. I am grateful to Dr. John Freudenstein for his help with the translation of my diagnoses into Latin and with the rules of botanical nomenclature. My phenetic and genetic diversity studies were improved greatly by the helpful comments of

Dr. Theresa Culley and Lisa Wallace.

My three field seasons in South America would not have been successful without the assistance, advice, and field support of the faculty, staff, and students of a number of

Chilean and Peruvian institutions. I gratefully thank the botany department of the

Universidad de Concepcidn for their hospitality and advice as I studied their large collection of Malesherbia. I also thank the staff of the Museo Nacional de Historia

Natural In Santiago, Chile and the faculty at the Universidad de La Serena’s herbarium.

My Peruvian field studies were made possible through the assistance of the faculty, staff, and students at the Museo Nacional de Historia Natural in Lima, the Universidad

Nacional de San Cristobal de Huamanga in Ayacucho, Museo Contisuyo in Moquegua, the Universidad Nacional de San Agustin in Arequipa, and the Universidad Nacional del

Centro in Huancayo. I extend my thanks to the Institute de Recursos Naturales of Peru for permission to collect and export specimens from that country.

I thank the U.S. Department of Education for granting me a Foreign Language and

Area Studies Fellowship for the study of Spanish, which was necessary for my field work. This study was funded by NSF grant DEB-9623496, the Beatley Herbarium Fund,

Sigma Xi Grants-in-Aid of Research, a Tinker Foundation Grant, and the Flora of Chile

Project.

VI VITA

November 6, 1970 Bom—Adrian, Minnesota

1993 B.A. Biology, College of St. Benedict, St. Joseph, Minnesota

1994, 1996, 1997, 1998, 1999 Graduate Teaching Associate The Ohio State University

1995, 1996 Herbarium Curatorial Assistant The Ohio State University

2000-present Instructor, Department of Biology, Denison University

PUBLICATIONS

I. D M. Spooner, D. De Jong, B.-Y. Sun, T.F. Stuessy, K.M. Gengler, G.L. Nesom, & P £ . Berry. 1995. Chromosome counts of Compositae from Ecuador and Venezuela. Ann. Missouri Hot. Gard. 82:596-602.

2. T.F. Stuessy & K. Gengler. 1995. Maiesherbiaceae. Pp. 297-298. In S. Takahashi et al. (eds.). The World of Plants. Asahi Shimbun, Tokyo. vii FIELDS OF STUDY

Major Field: Evolution, Ecology, and Organismal Biology

vui TABLE OF CONTENTS

Page

Dedication ...... iv

Acknowledgments ...... v

Vita...... vii

List of Figures ...... xi

Chapters;

1. Introduction ...... l

1.1 Literature Cited...... 9

2. Genetic diversities of four little-known species of Malesherbia (Maiesherbiaceae) endemic to the arid inter-Andean valleys of Peru ...... 10

2.1 Introduction ...... 10 2.2 Materials and Methods ...... 12 2.3 Results...... 13 2.4 Discussion...... 15 2.5 Literature Cited...... 21

3. Phenetic analyses of morphological traits in the Malesherbia humilis complex. (Maiesherbiaceae) ...... 27

3.1 Introduction ...... 27 3.2 Materials and Methods ...... 30 3.3 Results...... 32 3.4 Discussion ...... 34 3.4.1 Biogeography of M. humilis complex ...... 41 3.4.2 Conclusion ...... 42 3.5 Taxonomy ...... 43 3.6 Literature Cited...... 47

IX 4. Molecular phylogeny and taxonomy of Maiesherbiaceae ...... 62

4.1 Introduction ...... 62 4.2 Materials and Methods ...... 63 4.2.1 Sampling ...... 63 4.2.2 DNA Isolation and Sequencing ...... 64 4.2.3 Data Analyses ...... 65 4.3 Results...... 68 4.3.1 Sequence data ...... 68 4.3.2 Phylogeny ...... 68 4.4 Discussion ...... 71 4.5 Taxonomy ...... 79 4.6 Key to the Subgenera of Malesherbia...... 80 4.7 Literature Cited...... 85

5. Reconstruction of the biogeographic history of Maiesherbiaceae ...... 97

5.1 Phylogeny and distribution of M^esherbiaceae ...... 99 5.2 Physical history of arid western South America and the Andes ...... 101 5.3 Potential divergence times: perspectives from the molecular clock ...... 109 5.4 Proposed reconstruction of the evolutionary history of Maiesherbiaceae ...... 114 5.5 Conclusions ...... 121 5.6 Literature cited...... 122

List of References ...... 135 LIST OF FIGURES

Figwr? Pag?

2.1 Collection information (vouchers deposited at OS) ...... 24

2.2 Genetic diversity statistics for Malesherbia species and populations studied. Total heterozygosity (Ht), within-population heterozygosity (Hs), proportion of diversity among populations (Gst), total among-population diversity (DsO. proportion of polymorphic loci (P), and the mean number of alleles per locus (A) are shown ...... 25

3.1 Map of northern and central Chile. Political boundaries and the distribution of the M. humilis complex are indicated, with the exception of the Argentinean populations ...... 49

3.2 Populations and specimens used in the analyses. Tentative identifications were made prior to analysis. Abbreviations: G = M. gabrielae, Q = A/, humilis var. propinqua, H = A/, humilis var. humilis, V = Af. humilis var. parviflora, T = A/, humilis var. taltalina, U = undetermined, Prov. = province, Qda = quebrada (dry ravine) ...... 50

3.3 Characters and states used in the phenetic analyses ...... 57

3.4 UPGMA phenogram using average taxonomic distance. “Humilis” contains 33 OTU’s, “propinqua” 10, “parviflora” 62, “gabrielae” 6, and “taltalina” 2 ...... 59

3.5 Eigenvector coefficients for Principal Components Analysis. Character loading values are for the first three Principal Component (PC) axes. The percent variation explained by each axis is also presented ...... 60

3.6 Results of the principal components analysis ...... 61

xi 4.1 Distribution of Maiesherbiaceae ...... 88

4.2 Accessions collected for DNA isolation and sequencing. All vouchers are deposited at OS unless otherwise noted. Starred accessions were included in the maximum likelihood analyses ...... 89

4.3 Consensus tree derived from four most parsimonious trees calculated by NONA. Dashes indicate support values less than 50 ...... 92

4.4 The consensus tree obtained when ambiguous sites were removed and the data matrix re-analyzed in NONA. The arrows indicate clades with topological changes from the trees yielded by parsimony analysis of the complete data matrix. Dashes indicate support values less than 50 ...... 94

4.5 The tree yielded by maximum likelihood analysis of the complete data set using the HKY85 model. The arrow m ar^ the clade that disappears in analysis of the unambiguous data set...... 95

4.6 Comparison of the morphologies of Af. auristipulata, Af. turbinea, and Af. haemantha...... 96

5.1 Distribution of Maiesherbiaceae ...... 127

5.2 Strict consensus cladogram of four most parsimonious trees derived from ITS sequence data. Bremer and Jackknife support values are reported above each branch; values below some of the branches are divergence times (mya) calculated using the molecular clock. Subgenera are labeled in bold-face, and their distributions are indicated in parentheses ...... 129

5.3 Distributions of the subgenera of Maiesherbiaceae ...... 130

5.4 The four high elevation regions in the range of Maiesherbiaceae. The Norte Grande, marginal desert, and Norte Chico are indicated (Paskoff 1977); central Chile (33°-39“S) is not shown ...... 132

5.5 Accessions used to calculate average sequence divergences for the modified Li relative rates test. All collections are deposited at OS except where noted ...... 133

XII 5.6 Average sequence divergence values for terminal clades. The average divergence values between the listed species and a single accession of all the other species in its immediate clade were calculated. Differences in divergence values for the various accessions of a single species were minimal and are not considered here. For those terminal clades containing more than two taxa, the range of divergence values is given ...... 134

Xlll CHAPTER 1

INTRODUCTION

Along the western coast of South America lies one of the driest deserts in the world, the Atacama Desert. The Atacama Desert in northernmost Chile is the hyperarid center of an extremely long, narrow, and arid region extending from 5°S (the Peru-Ecuador border) to about 33°S (dry mediterranean central Chile). Because the Andes Mountains border the entire desert on its eastern edge and wetter climates prevail at its narrow northern and southern ends, the desert is isolated from other dry areas on the continent. This isolation, combined with habitat diversity created by rapid elevation changes from sea level to the nearby Andes and the great latitudinal range of the desert, produced a flora with many endemic taxa.

One endemic family is Maiesherbiaceae. The family has only one genus,

Malesherbia, with 24 species. Maiesherbiaceae, close relatives of Tumeraceae and

Passifloraceae (Fay et al. 1997), are somewhat unusual plants morphologically. The flowers have a corona and an androgynophore. The perianth and the corona perch at the top of an elongated floral tube, which is most likely axial tissue (A. Bernhard, pers. comm.). The entire plant body of all but one species is densely covered in trichomes.

Glandular hairs exuding an oily, slightly sticky liquid usually dot the stem and line the

1 leaf and stipule margins. Within the confines of these traits, however, the family has evolved a wide variety of forms. A group of Peruvian species are few-branched bushes reaching up to 1.5 m in height, whereas some Chilean taxa are less than 30 cm tall.

Flower color ranges from white to purple to blood red and sulfur yellow. Floral tubes may be funnelfbrm, campanulate, or tubular and range in size from less than a centimeter to several centimeters in length.

Maiesherbiaceae range over much of the desert of western South America. From its northernmost populations near the city of Lima, Peru ( 12°S) to those in the in the

Argentinean Andes (36°S), the various species are found in almost every habitat type except the barren, hyperarid Atacama Desert. Some species are native to coastal habitats, but most grow in the foothills of the Andes or in the Andes themselves of Chile, Peru, and western Argentina. The majority of species are narrowly endemic, although they may be very common in their range. A number of species, especially in Peru and northernmost

Chile, are extraordinarily rare: M. turbinea is known from one location, M. weberbaueri from two or three, M. auristipulata one valley, M. tenuifolia a handful of valleys. The rarest may be Af. tocopillana, a species known only from Tocopilla, along the coast of northern, hyperarid Chile. Collection data suggest that Af. tocopillana only appears in the few years after sufficient rainfall, and two decades may pass between such rains (Ricardi

1967).

Maiesherbiaceae provide an interesting research opportunity, as the family’s small size and distribution in a virtual continental island make it ideal for phylogenetic and biogeographic studies. There are no phylogenetic-based studies of the biogeography of any taxon native to the desert of western South America, although there is a comprehensive biogeographic study of the region based on the distributions of taxa and

the region’s geology, topology, and climate (Rundel et al. 1991). In that study, Rundel et

al. (1991) noted a turnover in the constituents of the flora at the northern edge of the

Atacama Desert (which coincides approximately with the political border).

Maiesherbiaceae, which are native to both Peru and Chile, provide an almost unique

opportunity to study the biogeography of the adjacent regions at the same time, exploring

how and if the forces affecting the family’s evolution differed between Peru and Chile

and whether the Atacama Desert has been a biogeographic barrier for this family.

Initial examination of herbarium specimens quickly revealed morphological data

suggesting the Atacama Desert region has played an important role in the evolution of the

family. Species found in Peru and in less arid pockets of the Atacama Desert are

morphologically distinct from species found further to the south. These differences were

recognized taxonomically at one time; the Peruvian group was Malesherbia, whereas the

Chilean species were grouped into Gynopleura. {Gynopleura was later lumped into

Malesherbia, which was a positive change, since I could find no morphological features

uniting the genus.) In addition to this morphological disjunction around the Atacama

Desert, species on both sides of the desert could be grouped into geographically and

morphologically cohesive groups, indicating that, even on this smaller scale, the physical

history of the region influenced the evolutionary history of the family. To investigate

these hypotheses, a phylogenetic reconstruction of the family is needed to test the

monophyly of the morphological groups and, thus, the proposed influence of the region’s

history on the taxon. As I studied the individual species both in the herbarium and in the field, further research problems presented themselves. One arose after my first two rield trips, when it became very clear that those species for which there were few specimens were not just poorly collected, but usually rare. I wondered what this rarity meant for the genetic diversities of the four species in inter-Andean valleys in Peru, since their populations tend to be near small farms and local roads. I conducted a small genetic diversity study of the species in those valleys to understand how much variability exists and to use that information to hypothesize how the species evolved.

In this document, I first deal with the genetic diversity study using allozymes of the four, closely related Peruvian shrubs, one of the few genetic diversity studies of agriculturally unimportant plants native to arid Peru (Sahley 1996, Crawford et al. 1993).

The four species in question are known from few populations, so leaves were collected from much of the known range of each species. Logistically, this was a difficult task: ice was almost never available to keep the leaf material cool until the samples could be refrigerated. Despite these problems, the enzymes survived long enough to determine intraspecific diversities, although they degraded before interspecific comparison gels could be made. Only one of the four species had relatively high isozyme variation, which might be attributed to its large population sizes and the diversity of habitats it occupies.

The other three species had relatively low diversities. The diversities of two were so low and the number of reported populations so small that it seems clear that they evolved recently via founder events. Natives of the arid inter-andean valleys might very often be the descendents of founder events, since these arid valleys are deep and isolated by intervening areas of inhospitable high elevation and more humid lower elevations. The second problem to present itself was what I came to refer to as the Af. humilis species complex, a group of four species which are very difficult to distinguish consistently, especially at the extremes of each species’ variability. It was obvious after a struggle with the dozens of herbarium specimens at my disposal that something needed to be done with the taxonomic tangle. I attempt to unsnarl the taxonomic problems in the Af. humilis complex using a phenetic analysis of morphological data gathered from herbarium specimens. The herbarium specimens were a source of data covering more than one hundred years of collections, the entire geographical and morphological range of the complex, and yearly climatic variation. The phenetic analyses revealed five major clusters of specimens roughly corresponding to some of the previously recognized taxa.

One species, Af. multiflora, could not be distinguished from Af. humilis var. parviflora, so it was lumped together with the latter taxon. The clusters were not easily distinguished by any of the traits used in the phenetic analysis. 1 therefore recommend that only one species be recognized, Af. humilis, with five varieties; var. taltalina, var. parviflora, var. humilis, var. propinqua, and var. gabrielae.

Phylogenetic analysis of Maiesherbiaceae in preparation for the biogeographic study was conducted using molecular data. Although the morphology supported the existence of at least four groups of closely related species, there was nothing in the morphology to suggest how these groups might be related to one another. Parsimony and maximum likelihood analyses of all the species of Maiesherbiaceae and two outgroups from

Tumeraceae were conducted using a data matrix constructed from ITS sequences. The phylogenetic analyses agreed with each other to a large extent: the root fell among the

Chilean species, and four strongly support clades were found. A fifth clade, weakly supported in parsimony, was broken in the maximum likelihood analysis.

The supports for the relationships among these major clades were low. These support

values and the lack of morphological data pertaining to the relationships among major groups of species suggest that the five clades evolved rapidly. Resolution of relationships among species within the clades themselves is also poor, again suggesting that the

modem species evolved rapidly and so recently that they have not yet developed ITS autapomorphies.

The five clades of morphologically similar species are herein taxonomically

recognized at the subgeneric level. Breaking the family into these five subgenera will

accurately reflect the great morphological variability of the genus without the large

number of name changes erection of five separate genera would entail. In addition, one of

these subgenera, subg. Albitomenta, is only weakly supported as monophyletic in

parsimony analysis, suggesting that future research in the family could bring about

changes in this group's taxonomy. Using subgenera rather than genera to classify the

family is prudent in light of the uncertainty surrounding this single clade.

A second reason for describing the clades as subgenera rather than genera lies in the

relationship of Maiesherbiaceae to Tumeraceae. Although the scope of this project did

not include an interfamilial study, such research may reveal Maiesherbiaceae to be

embedded within Tumeraceae, which was shown to be the closest relative of

Maiesherbiaceae by rbcL data (Fay et al. 1997). In her work with Piriqueta

(Tumeraceae), a genus suggested to be related to Maiesherbiaceae (Ricardi 1967), Arbo (1995) maintains that it is closely related to Tumera, and that these genera are derived taxa in Tumeraceae. Tumera and Piriqueta are the only two genera of Tumeraceae in

South America (the remainder are Old World taxa or in Cuba and Central America), so they are the most likely sister taxa of Maiesherbiaceae. If future studies show

Tumeraceae to be paraphyletic, Maiesherbiaceae may be subsumed into Tumeraceae as a genus. Name changes would be minimal in this scenario if I recognize only subgenera in

Malesherbia at this point.

My last chapter relies upon this phylogeny to reconstruct the biogeography of

Maiesherbiaceae. Because the species comprising the subgenera are also native to geographically discrete regions, I conclude that the history of the region heavily influenced the evolution of the family. To understand which physical events may have affected Maiesherbiaceae (and presumably other taxa in the region), I reconstruct the physical history of the desert and the adjacent Andes Mountains. There are numerous studies of the geology and climatic history of Chile especially, largely due to interest in vast mineral deposits in the Atacama Desert. By comparing this history of the region with the phylogeny of the family, I determined that the family most likely evolved in the late

Miocene, when westem South America became permanently arid. The fîve subgenera probably rapidly radiated in the Pliocene. It was during this time that the family migrated from Chile to Peru across the biogeographic barrier in northem Chile. The fluctuating climates in the Pleistocene and very early Holocene likely led to the evolution of the modem species. Late Pleistocene changes appear to have most heavily influenced the evolution of the Peruvian clade, subg. Malesherbia, and the montane Chilean and

Argentinean species, subg. Xeromontana. These subgenera are the most speciose, with most of their species species sharing almost identical ITS sequences with at least one sister species. Evidence from the molecular clock, which was not rejected in most cases, supports these interpretations of the biogeography of Maiesherbiaceae.

These data provide further evidence that the Peruvian and Chilean deserts’ floras evolved separately and can be studied in isolation. Phylogenetic analysis of other endemics to the desert of westem South America will lead to a fuller understanding of the relationship between these floras and the physical history of the region. Inclusion of the phylogenies of other taxa also will test the general applicability of the biogeographic scenarios presented here for Maiesherbiaceae. Literature Cited

Arbo, M M 1995. Turaeraceae Parte I. Piriqueta. Flora Neotropica. Monograph 67. The New York Botanical Garden, Bronx, New York.

Crawford, D.J., A. Segàstegui A., T.F. Stuessy, & I. Sânchez V. 1993. Variaciôn aloenzimàtica en la rara especie endémica peruana Chuquiraga oblongifolia (Asteraceae). Amaldoa 1:73-76.

Fay, M F., S.M. Swensen & M.W. Chase. 1997. Taxonomie affinities of Medusagyne oppositifolia (Medusagynaceae). Kew Bulletin 52: 111-120.

Ricardi, M. 1967. Revision taxonômica de las Malesherbiâceas. Gayana, Botânica no. 16: 3-139.

Rundel, P.W., M O. Dillon, B. Palma, H.A. Mooney, S.L. Gulmon, & J.R. Ehleringer. 1991. The phytogeography and ecology of the coastal Atacama and Peruvian deserts. Aliso 13: 1-49.

Sahley, C.T. 1996. Bat and hummingbird pollination of an autotetraploid columnar cactus, Weberbauerocereus weberbaueri (Cactaceae). American Journal of Botany 83: 1329-1336. CHAPTER 2

GENETIC DIVERSITIES OF FOUR LITTLE-KNOWN SPECIES OF MALESHERBIA (MALESHERBIACEAE) ENDEMIC TO THE ARID INTER-ANDEAN VALLEYS OF PERU

Introduction

Studies of genetic variation have become increasingly important for understanding the evolutionary lability of a species (Godt et al. 1997, Richards & Leberg 1996, Hamrick &

Godt 1990) and postulating historical causes for the levels of diversity observed (e.g., Arft

& Ranker 1998, Qiu & Parks 1994). Conservation status, consequently, is often determined in part by these studies. Of special interest are narrowly endemic species because of their generally low levels of allozyme variation (Hamrick & Godt 1990). In South America, the

Pacific coastal desert and dry Andes of Peru harbor many endemics and distinctive assemblages of species in a mosaic of habitats created by rapid elevational changes and a network of deep river canyons with arid climates (Rauh 1985).

Although unique and often rare taxa populate this desert and the adjacent arid mountains, few genetic diversity studies have been conducted on these plants. The few studies including Andean and desert Peruvian plants have almost exclusively been in agriculturally important genera, such as Solanum (Quiros et al. 1990), Lycopersicon (Bretô et al. 1993, Rick et al. 1976, Rick & Fobes 1975), and Phaseolus (Tohme et al. 1996, Singh et al. 1991). Agriculturally unimportant components of the native flora of these regions have been studied only twice, to our knowledge. Crawford et al. (1993) determined the genetic diversity of Chuquiraga oblongifolia SegdsL & Sânchez, a rare and endangered native of

10 the Andes of northern Peru; Sahley (1996) analyzed allozyme diversity as part of a reproductive biology study of a population of the cactus Weberbauerocereus weberbaueri

(Schumann ex Vaupel) Backeberg from the southern foothills of the Andes.

One of the distinctive taxa of the Pacific coastal desert is Malesherbiaceae; it is one of only two endemic families of the area. Nine of the 24 species in the single genus

Malesherbia are Peruvian (Ricardi 1967, K. Gengler, unpubl. data). These nine species are found from central to southern coastal and Andean Peru, largely on the arid slopes of river gorges or road cuts. Only two species have relatively widespread distributions; the remaining seven species are restricted to small areas and a handful of populations.

Regardless of the size of the range, the Peruvian species exist in few, widely scattered, remote populations separated by mountain ranges or high foothills.

Malesherbia in south-central coastal and Andean Peru is composed of four species of small to moderately sized shrubs with large, showy inflorescences of red, orange, or yellow flowers. The small diameter of their woody stems suggests they are short-lived. These species, M. scarlatiflora Gilg, M. splendens Ric., M. tubulosa (Cav.) J. St. Hil., and Af. weberbaueri Gilg, constitute a monophyletic group (K. Gengler, unpubl. data). Although closely related, they are easily distinguished by floral and foliage color, pubescence, leaf shape, and plant size. Malesherbia scarlatiflora is the most widespread of the group; populations are known from the department (or province) of Lima and from Huancavélica and Apurimac in the central Andes. Malesherbia tubulosa (Cav.) J. St. Hil. is an endemic of the river valleys of Lima, where it maintains large populations. Malesherbia splendens and

Af. weberbaueri are known from perhaps three locations each. Malesherbia splendens is endemic to the Lurin valley of Lima; the two known populations of Af. weberbaueri var. weberbaueri are found only in the Rio Mantaro valley of the central Peruvian department of

Huancavélica, and the single population of Af. weberbaueri var. galjttfii (J.F. Macbr.) Ric. was collected in the department of Junfn. Because of their scattered distributions, rarity,

11 endemic status, and the lack of such studies of any plants of this region, an initial allozyme study was conducted to determine levels of genetic diversity within each species and to examine the probable effects of biogeography, phylogeny, habitat, and human use on this diversity.

Materials and Methods

Fresh leaf material was collected from each population encountered of M. scarlatiflora,

Af. splendens, M. tubulosa, and M. weberbaueri var. weberbaueri (Figure 2.1).

Malesherbia weberbaueri var. galjufii could not be located for this study. Two populations of Af. scarlatiflora and two populations of Af. tubulosa were sampled, while one population each of Af. splendens and Af. weberbaueri was collected. Leaves were gathered from all six individuals of the small Matucana population of Af. scarlatiflora (Af. scarlatiflora 1), whereas 14-24 individuals were sampled from across all others. Leaves were refrigerated at

4°C within 24 hours of collection. Samples of Af. weberbaueri were refrigerated for approximately one month, but samples for the other populations were gathered within two weeks of travel in an insulated container from Lima to the United States, where they were immediately refrigerated. Leaf samples were ground in approximately 1 ml of grinding buffer adjusted to a pH of 7.5. This buffer consisted of 10% glycerol, 0.10 M tris, 1.0 mM tetrasodium salt of EDTA, 10 mM MgCl2,11 mM KCl, 74 mM 2-mercaptoethanol, and

5-10 mg polyvinylpyrrolidone (modified from Gottlieb 1981). Grindates were frozen at

-80°C until electrophoresis.

Seven enzyme systems (malate dehydrogenase [MDH], isocitrate dehydrogenase

[IDH], glucose-6-phosphate dehydrogenase [PGD], leucine aminopeptidase [LAP], phosphoglucomutase [PGM], phosphoglucoisomerase [GPI], and triosephosphate isomerase [TPI]) were assayed in 12.2% (w/v) starch gels using the methods of Crawford et al. ( 1997), which were slightly modified (the pH of the buffer used to separate LAP, PGM,

12 and GPI was 8.1). Staining protocols and nomenclature follow those of Wendel and

Weeden (1989). Individual populations wgre initially resolved on separate gels. Summary gels of the populations of M. scarlatiflora and Af. tubulosa included samples representing all alleles found in the population gels to determine homologies among alleles and loci.

Three additional enzymes for Af. splendens and Af. weberbaueri were resolved in discontinuous vertical slab polyacrylamide gels following Davis (1964). Alcohol dehydrogenase (ADH), aspartate aminotransferase (AAT), and superoxide dismutase

(SOD) were resolved for five samples each of Af. splendens and Af. weberbaueri chosen for their strong activity in starch gel electrophoresis.

Genetic interpretation of banding patterns was based on the minimum number of loci expected for each enzyme in diploid plants and the known active subunit composition of the enzymes (Weeden & Wendel 1989). Faintly staining or poorly resolved bands were not included in the fînal analysis. GeneStat (version 3.3) (Lewis & Whitkus 1989) was used to calculate Nei’s (1973) gene diversity statistics (Ht, Hs, and Dst), Hamrick and Godt’s

(1990) Gst, the mean proportion of polymorphic loci (P), and the mean number of alleles per locus (A).

Results

A total of 15 presumed loci, Mdh-l, Mdh-2, Gpi-I, Gpi-2, Idh-I, ldh-2, Lap-l, Lap-2,

Pgd-l, Pgd-2, Pgd-3, Pgm-l, Pgm-2, Tpi-1, and Tpi-2, was examined for all populations.

Seven additional presumptive loci, Aat-1, Aat-2, Adh-1, Adh-2, Sod-I, Sod-2, and Sod-3, were available for Af. splendens and Af. weberbaueri. Not all loci or enzyme systems stained strongly or clearly enough to interpret for every species or individual. At least eight purported loci, however, were scored for each population. Starch summary gels of the two

13 populations of M. scarlatiflora and M. tubulosa clearly indicated that bands had the same

mobilities, permitting the calculation of diversity statistics between populations within each of these species.

All individuals of all populations showed a three-banded, balanced staining pattern for

IDH, a dimeric enzyme (Weeden & Wendel 1989), suggesting that these species are fîxed

heterozygotes and thus have two copies of the IDH gene, either in the cytosol or plastid.

Banding patterns for PGD in M. splendens also indicate that there has been a gene

duplication because the majority of individuals are identical, triple-banded heterozygotes.

Diversity statistics, proportions of polymorphic loci, and mean numbers of alleles per

locus are shown in Figure 2.2. Of all the populations studied, M. tubulosa 2 has the greatest

within-population genetic diversity (Hs = 0.182), proportion of polymorphic loci (P =

0.364), and mean number of alleles per locus (A = 1.818). Malesherbia scarlatiflora 1, M

weberbaueri, M. scarlatiflora 2, and M. splendens have successively lower values for Hg:

M. tubulosa 1 has the least within-population allozyme diversity (Hs = 0.056). Malesherbia

tubulosa 1 also has the lowest proportion of polymorphic loci (P = 0.125), whereas M.

weberbaueri, M. scarlatiflora I, M. splendens, and A/, scarlatiflora 2,respectively, have

increasingly greater proportions of polymorphic loci. The smallest mean number of alleles

per locus (A = 1.214) was found in Af. splendens. This average is Increasingly greater in Af.

weberbaueri, Af. tubulosa 1, Af. scarlatiflora 2,and Af. scarlatiflora 1, respectively.

Malesherbia scarlatiflora and Af. tubulosa are the only species for which more than

one population was available. With Ht = 0.206, Af. tubulosa is more genetically diverse than

Af. scarlatiflora (Ht = 0.083). Mote of the total diversity is apportioned between

populations of Af. tubulosa (Gst = 0.461) than between those of Af. scarlatiflora (Gst =

0.060). Malesherbia tubulosa also has a greater proportion of polymorphic loci (P = 0.364)

and more alleles per locus (A = 1.818) than Af. scarlatiflora (P = 0.273, A = 1.364).

14 Discussion

Allozyme diversities of the species in this study, with the exception Af. tubulosa, are lower than or comparable to the levels observed in other endemic and short-lived species

(Hamrick & Godt 1990). These low levels of diversity may result from bottlenecks, low variation in the progenitor, inbreeding, and/or little variation carried over from progenitor to founding populations (Godt et al. 1997, Leberg 1992, Gottlieb 1973). Determining what effects these phenomena may have had on species requires information other than allozyme data. Sources of information useful in determining the cause(s) of low levels of variation include phylogeny (Godt et al. 1995), historical and current human intervention (Percy &

Cronk 1997, Rieseberg et al. 1989), biogeography (Godt et al. 1996, Crawford et al. 1992,

Schwaegerle & Schaal 1979), and habitat and range data (Godt et al. 1996, Purdy et al.

1994).

Phylogenetic analysis of the family using the ITS regions of nuclear ribosomal DNA places Chilean species of Malesherbia basal to all the Peruvian species (K. Gengler. unpubl. data). Malesherbia scarlatiflora, Af. splendens, Af. tubulosa, and Af. weberbaueri constitute a monophyletic clade with no resolution of relationships among species, a fînding further supported by the morphological similarities of the four species studied here. Unlike other Malesherbia, these four species are upright, generally densely pubescent bushes with large racemes of long, tubular flowers. Their similar morphologies and sequences indicate that they are the result of a relatively recent and rapid radiation from a common ancestor.

Low levels of allozyme variation may be expected for diploid species of recent origin, since they have not had enough time to accrue allozyme variation since their origins (Gottlieb

1973).

The most allozymically variable of the four species, Af. tubulosa, is endemic to four valleys in the department of Lima. The two large populations from neighboring valleys sampled for this study differ in allozyme diversity. Malesherbia tubidosa I, found near

15 Suico in the Rimac valley, has lower diversity than Af tubulosa 2, located east of Yaso in the

Chilldn valley (Figure 2.2). In comparison with populations of other endemic or woody, short-lived species, intrapopulational diversity in Af. tubulosa is2 high. Malesherbia tubulosa I, in contrast, has somewhat lower diversity than populations in these two categories (Hamrick & Godt 1990: Figure 2.2). Both populations are large (Af. tubulosa I has 75-100 plants; Af. tubulosa probably2 represents a portion of a large metapopulation, as hundreds of plants were seen on steep cliffs and inaccessible ledges), but Af. tubulosa I lacks some alleles found in Af. tubulosa 2. The absence of these alleles accounts for most of the differences in allelic diversity between the populations. These alleles may have been lost from Af. tubulosa 1 during a bottleneck event, or they may have arisen in and are specific to the large Af. tubulosa population.2 It is also plausible that the differences are due to sampling error. Maintenance of large population size for Af. tubulosa 2, and thus greater allozyme variation, may be facilitated by its location in the remote Chillôn valley, where human disturbances from traffic and agriculture are low. The Surco population, in contrast, is found close to farms along a major highway in Peru. The disturbance may cause a reduction in seedling survival rates or may affect pollinator behavior, resulting in lowered seed set and perhaps loss of alleles.

Total diversity in Af. tubulosa is much higher than reported for the average endemic, short-lived perennial, as is the amount of diversity apportioned among populations

(Hamrick & Godt 1997: Figure 2.2). Interpopulational differentiation could arise from a lack of gene flow between populations brought about by geographical isolation. The two populations are separated by at least 70 km if elevations suitable to Af. tubulosa (which occurs only below 2500 m) are followed. Although it is possible that they are genetically connected by intervening populations, the geographical distance between the two makes it unlikely.

16 Allozyme diversity in Af tubulosa is the greatest found among the species studied, possibly because it has more general habitat requirements and is under little pressure from collectors. The species grows in disturbed soils, out of rock faces, and along cliffs. Both populations of Af tubulosa studied cover relatively large areas and diverse habitats, which may lead to preservation of greater diversities than those found in the other three species.

Also found in the department of Lima is Af scarlatiflora, a large, attractive, orange- flowered plant reported not only from three valleys in Lima, from which two populations were collected, but also from central Andean Peru. Malesherbia scarlatiflora 1 of Matucana consisted of six individuals, and, not surprisingly, diversity is similar to or slightly lower than that of populations of other endemic or woody, short-lived species (Hamrick & Godt

1990; Figure 2.2). Collection and historical information suggest that this population is experiencing a bottleneck. The population was found on an almost inaccessible hillside, and even the local herbalists were unfamiliar with this visually striking and regionally prized medicin^ plant, suggesting that the species has been rare in the area for many years. It also seems unlikely that the large number of herbarium collections dating back to 1909 would have been made if the population were historically this small and unreachable.

The Huachupampa population, Af. scarlatiflora 2, is much larger, and the diversity is higher than in the smaller population, Af. scarlatiflora 1. Diversity is similar to values reported for populations of endemic species and somewhat lower than those of woody, short-lived species (Hamrick & Godt 1990: Figure 2.2). This population is heavily harvested as an herbal medicine (K. Gengler, pers. obs.), so it may be at risk of loss of diversity due to overcollection. The proportion of total diversity allocated between the two populations of Af. scarlatiflora is low (Hamrick & Godt 1997: Figure 2.2). This could result from interpopulational gene flow, or, alternatively, these populations may be part of a

17 larger, single population. Suitable habitat likely exists between them, but this area is remote and little-known botanically. Collecting efforts should be concentrated in this region to determine if the species is more widespread than it now appears to be in Lima.

Allozyme diversities of two of the rarest species in Malesherbiaceae, M. splendens and

Af. weberbaueri, were also examined. One of the three known populations of each was sampled. Although the population sampled forAf. splendens is relatively large with 50-100, mostly adult, plants, it is restricted largely to the gullies of a single, dry slope in the Lurin valley and the road cut passing through it. The lack of genetic diversity in this large population suggests that Af. splendens may be suffering the effects of a past bottleneck

(Hamrick & Godt 1990,1997; Figure 2.2). The lack of divergence among Af. scarlatiflora,

Af. splendens, Af. tubulosa, and Af. weberbaueri in the ITS regions and the restricted distribution of Af. splendens further suggest that it evolved recently. The data therefore indicate that Af. splendens may be recovering from a founder event-induced bottleneck.

Spéciation by founder effect is probably common in the patchy landscape of the department of Lima, in which coastal plains lead quickly to foothills and eventually to the high altitude puna.

Because of this mosaic environment, it is improbable that Af. splendens present today represents the rermiants of a once widespread species. The establishment of numerous populations after the evolution of this species was likely hindered by its environment.

Suitable patches of eroded, arid, rocky soils found between 2400 m and 30(X) m seem widely scattered. Dispersal of seeds or pollen to distant sites is likely a rare event; the species lacks obvious long-distance dispersal mechanisms, and it is improbable that pollinators would travel the long distances between populations in the rugged landscape.

Malesherbia weberbaueri var. weberbaueri, an endemic known only from the central

Peruvian departments of Huancavélica, has low to average diversity for endemic, short-lived species (Hamrick & Godt 1990,1997: Figure 2.2). This population of Af. weberbaueri is

18 one of only two reported for the variety and three for the species; the only population of the other variety, Af. weberbaueri var. g a lji^i, has almost certainly been extirpated, probably as a result of extensive irrigation at its only known location in Junm. Thus the population examined in this study very likely represents one of two extant for this species. The Mayocc population of Af. weberbaueri var. weberbaueri is found exclusively on one hill in the driest portion of the entire Rio Mantaro valley of central Andean Peru (O. Tovar, pers. comm.). The soil of this hill, unlike all the others in the area, is light yellow. Few other plant species common on neighboring hills grow at this site, suggesting that the soil is quite different and that Af. weberbaueri is an edaphic endemic. Unfortunately, the soil’s composition remains unknown.

The rarity of appropriate habitat, lack of divergence in ITS sequences, and the low number of alleles per locus suggest that Af. weberbaueri, like Af. splendens, is a neospecies, which may have arisen after a founder event resulting from seed dispersal. These seeds may have been from Af. scarlatiflora, which is geographically the closest species of

Malesherbia, or from a common ancestor of the two species. Low levels of allozyme variation may persist because of strong selection for plants tolerant of the habitat and the high levels of ovary herbivory by Heliconia butterfly larvae (K. Gengler, pers. obs.).

Further work with this species should include detailed cytological, ecological, and interspecific allozyme and molecular studies to examine the origin of this alleged edaphic endemic.

The allozyme diversities of the species of Malesherbia studied here range from the depauperate to high for the average endemic or short-lived, woody perennial species.

Continued intensive harvesting of Af. scarlatiflora for medicinal use in Lima may lead to loss of evolutionary potential due to lowered allelic diversity (Godt et al. 1997, Godt et al.

1996, Richards & Leberg 1996, Hamrick & Godt 1990, Nei et al. 1975), and eventually to local extinction. Malesherbia weberbaueri and Af. splendens, both very rare species, are

19 also in precarious situations. Although the populations found of these species are large, clearing and irrigation of these sites would be devastating for both species. For the time being, the aridity and poor soil in which they thrive shelter them. Nevertheless, even these conditions were not enough to safeguard the only known population of Af. weberbaueri var. galjufii. Hopefully the remoteness of the localities of Af. weberbaueri var. weberbaueri and

Af. splendens will afford them added protection.

Like other areas of Peru better known tor their diversities, the Pacific coastal desert is under increasing pressure from humans. The abundance of irrigated fields and orchards in the river valleys suggests that the plants and animals of these valleys may be subject to population crashes and extirpation. Fortunately, the species of Malesherbia studied here are found in the valleys' driest and poorest soils, which have yet to be exploited. Other endemic species to these areas, however, may be under more immediate pressures. Further studies of the distributions and genetic diversities of endemics to the Pacific coastal desert and the adjacent arid Andes may aid in the preservation of its unique flora.

20 Literature Cited

Arft, A ^ . & T.A. Ranker. 1998. Allopolyploid origin and population genetics of the rare orchid Spiranthes diluvialis. American Journal of Botany 85: 110-122.

Bretô, M.P., M.J. Asins, & E.A. Carbonell. 1993. Genetic variability in Lycopersicon species and their genetic relationships. Theoretical and Applied Genetics 86: 113-120.

Crawford, DJ., E. Landolt, D.H. Les, & E. Tepe. 1997. Allozyme variation and the taxonomy of Wolffiella. Aquatic Botany 58:43-54.

, A. Segàstegm A., T f. Stuessy, & I. Sànchez V. 1993. Variaciôn aloenzimàtica en la rara especie endémica peruana Chuquiraga oblongifolia (Asteraceae). Amaldoa l: 73-76.

T.F. Stuessy, D.W. Haines, M.B. Cosner, M. Silva O., & P. Lopez. 1992. Allozyme diversity within and divergence among four species of Robinsonia (Asteraceae: Senecioneae), a genus endemic to the Juan Fernandez Islands, Chile. American Journal of Botany 79: 962-966.

Davis, B.J. 1964. Dis-electrophoresis Q. Methods and application to human serum proteins. Annals of the New York Academy of Science 21:404-427.

Godt, M.J. W., B.R. Johnson, & J.L. Hamrick. 1996. Genetic diversity and population size in four rare southern Appalachian plant species. Conservation Biology 1 0 :7 9 ^0 5 .

, J. Walker, & JJ-. Hamrick. 1995. Genetic diversity in a threatened wetland species, Helonias bullata (Liliaceae). Conservation Biology 9:596-604.

. 1997. Genetic diversity in the endangered lily Harperocallis flava and a close relative, Tofieldia racemosa. Conservation Biology 11:361-366.

Gottlieb, L.D. 1973. Genetic differentiation, sympatric spéciation, and the origin of a diploid species of Stephanomeria. American Journal of Botany 60:545-553.

1981. Electrophoretic evidence and plant populations. Progress in Phytochemistry 7: 1^6.

Hanuick, J.L. & M.J.W. Godt. 1990. Allozyme diversity in plant species. Pp. 43-63. In Plant population genetics, breeding, and genetic resources, A.H.D. Brown , M.T. Clegg, A.L. Kahler, and B.S. Weir, eds. Sinauer Associates, Sunderland, Massachusetts.

& -----. 1997. Effects of life history traits on genetic diversity in plant species. Pp. 103-118. In Plant life histories: ecology, phylogeny, and evolution, J. Silvertown, M. Franco, & JI.. Harper, eds. Cambridge University Press, Cambridge, UK.

Leberg, P i,. 1992. Effects of population bottlenecks on genetic diversity as measured by allozyme electrophoresis. Evolution 46:477-494.

Lewis, P.O. & R. Whitkus. 1989. GeneStat (version 3.3) for microcomputers. American Society Plant Taxonomy Newsletter 2:15-16.

21 Nei, M. 1973. Analysis of gene diversity in subdivided populations. Proceedings of the National Academy of Sciences USA 70:3321-3323.

, T. Maruyama, & R. Chakraborty. 1975. The bottleneck effect and genetic variability in populations. Evolution 29: I-10.

Percy, D.M. & Q.C.B. Cronk. 1997. Conservation in relation to mating system in Nesohedyotis arborea (Rubiaceae), a rare endemic tree from St. Helena. Biological Conservation 80:135-145.

Purdy, B.C., RJ. Bayer. & S.E. MacDonald. 1994. Genetic variation, breeding system evolution, and conservation of the narrow sand dune endemic Stellaria arenicola and the widespread S. longipes (Caryophyllaceae). American Journal of Botany 81:904-911.

Qiu, Y.-L. & C.R. Parks. 1994. Disparity of allozyme variation levels in three Magnolia (Magnoliaceae) species from the southeastern United States. American Journal of Botany 81: 1300-1308.

Quiros, C.F., S B. Brush, D.S. Douches, K.S. Zimmerer, & G. Huestis. 1990. Biochemical and folic assessment of variability of Andean cultivated potatoes. Economic Botany 44: 254-266.

Rauh, W. 1985. The Peruvian-Chilean deserts. Pp. 239-267. In Hot deserts and arid shrublands, M. Evenari, I. Noy-Meir, & D.W. Goodall, eds. Elsevier, Amsterdam.

Ricardi, M. 1967. Revisiôn taxonômica de las Malesherbiâceas. Gayana, Botânica no. 16: 3-139.

Richards, C. & P.L. Leberg. 1996. Temporal changes in allele frequencies and a population's history of severe bottlenecks. Conservation Biology 10:823-839.

Rick, C M. & J.F. Fobes. 1975. Allozyme variation in the cultivated tomato and closely related species. Bulletin of the Torrey Botanical Club 102:375-384.

, E. Kesicki, J.F. Fobes, & M. Holle. 1976. Genetic and biosystematic studies on two new sibling species of Lycopersicon from interandean Peru. Theoretical and Applied Genetics 47:55-68.

Rieseberg, L.H., S. Zona, L. Aberbom, & T.D. Martin. 1989. Hybridization in the island endemic, Catalina mahogany. Conservation Biology 3:1-7.

Sahley, C.T. 1996. Bat and hummingbird pollination of an autotetraploid columnar cactus, Weberbauerocereus weberbaueri (Cactaceae). American Journal of Botany 83: 1329-1336.

Schwaegerle, K.E. & B.A. Schaal. 1979. Genetic variation and founder effect in the pitcher plant Sarracema purpurea L. Evolution 33: 1210-1218.

Singh, S.P., R. Nodari, & P. Gepts. 1991. Genetic diversity in cultivated common bean: I. Allozymes. Crop Science 31: 1^23.

22 Tohme, J., D. Orlando G., S. Beebe, & M.C. Duque. 1996. AFLP analysis of gene pools of a wild bean core collection. Crop Science 36: 1375-1384.

Weeden, N.F. & J.F. Wendel. 1989. Genetics of plant isozymes. Pp. 46-72. In Isozymes in plant biology, D.E. Soltis & P S. Soltis, eds. Dioscorides Press, Portland, Oregon.

Wendel, J.F. & N f. Weeden. 1989. Visualization and interpretation of plant isozymes. Pp. 5-45. In Isozymes in plant biology, D.E. Soltis & P S. Soltis, eds. Dioscorides Press, Portland, Oregon.

23 Taxon Collector & Number Collection Information

A/, scarlatiflora I KMG & Roque 350 Slopes of Qda. Paihua, near Matucana, 2410m; Prov. Huarochiii; Lima, Peru ll°50.25rS 76°22.82rW May 17, 1997

Af. scarlatiflora 2 KMG & Salvador P. 369 On road from San Juan de Iris to Huachupampa, 2950m; Prov. Huarochirf, Distr. Huachupampa; Lima, Peru 11°42.308’S 76“33.695’W May 23, 1997

Af. splendens KMG & Roque 356 1km before San José de los Chorrillos on road to Huarochirf, 2590m; Prov. Huarochirf; Lima, Peru 12°08.085’S 76°26.473’W May 18, 1997

Af. tubulosa KMG & Roque 354 km 62-64 on road from Lima to Huancayo, below Surco, along roadside, 1800m; Prov. Huarochirf; Lima, Peru 11°50’S 76“22’W May 17, 1997

Af. tubulosa 2 KMG & Bedoya 362 5km east of Yaso, 1770m; Prov. Canta; Lima, Peru 11“32.499’S 76“42.522’W May 19, 1997 Af. weberbaueri var. KMG 288 Hill east of Mayocc, left side of Rfo weberbaueri Mantaro, 2280m; Prov. Churcampa, Distr. La Merced de Haser; Huancavélica, Peru 12“48.122’S 74°22.661’W May 2, 1997

Figure 2.1. Collection information (vouchers deposited at OS).

24 Taxon N H t" Hs" Gst Dst P A

Malesherbia scarlatiflora 0.083 0.078 0.059 0.005 0.273 1.364

M. scarlatiflora 1 6 - 0.094 - - 0.200 1.300

M. scarlatiflora 2 14 - 0.070 - - 0.273 1.273

Malesherbia splendens 21 - 0.057 — - 0.214 1.214 Malesherbia tubulosa 0.206 0.1 II 0.483 0.095 0.364 1.818

M. tubulosa 1 24 - 0.056 - - 0.125 1.250

« M. tubulosa 2 20 - 0.182 - - 0.364 1.818 Malesherbia weberbaueri 19 0.079 0.154 1.231

Figure 2.2. Genetic diversity statistics for Malesherbia species and populations studied. Total heterozygosity (Ht), within-

population heterozygosity (Hs), proportion of diversity among populations (Gst), total among-population diversity (Dst),

proportion of polymorphic loci (F), and the mean number of alleles per locus (A) are shown, (continued) Figure 2.2. (continued)

Population-level diversities for 100 - 0.063 - 0.263 1.39 average endemic species^ Species-level diversities for 52-100 0.096 0.063 0.248 0.400 1.80 average endemic species^

Population-level diversities for II - 0.094 0.313 1.55 average woody, short-lived species^

Species-level diversities for 8-17 0.097 0.094 0.088 0.418 1.54 average woody, short-lived species^

Species- level diversities for 30 0.083 - 0.325 0.321 average endemic, short-lived perennial^

Ht and H$ are equivalent to Hamrick & Godt’s (1990, 1997) Hes and Hep, respectively.

^Diversity statistics compiled from Hamrick & Godt ( 1990).

^ Diversity statistics compiled from Hamrick & Godt (1997). CHAPTERS

PHENETIC ANALYSES OF MORPHOLOGICAL TRAITS IN THE MALESHERBIA HUMIUS COMPLEX (MALESHERBIACEAE)

Introduction

A small, little-known group of xerophytic plants endemic to the Peruvian and Chilean deserts and adjacent Argentina comprise Malesherbiaceae. They are most closely related to

Tumeraceae, a largely neotropical family (Fay et al. 1997). Malesherbiaceae are distinguished from their closest relatives by the presence of an androgynophore (a stalk elevating the ovary above the floral tube base), a well developed corona, and a floral tube of axial origin bearing the perianth on its upper margin. In addition, most Malesherbiaceae have glandular hairs covering much of the plant body. The 27 species of the family’s single genus, Malesherbia, colonize a wide variety of habitats, including the mediterranean foothills of central Chile, the inter-Andean valleys of Peru, and even the dry ravines of the northern Chilean Atacama Desert, a section of the driest coastal desert in the world

(Trewartha 1966). A diverse range of floral and vegetative morphologies has arisen in the family, probably in response to the patchy environment in which it lives.

One of the most distinctive groups of species in the family is the M. humilis complex.

The taxa treated at the species level by Ricardi (1967) are M. gabrieiae Ric., M humilis

Poepp., Af. multiflora Ric., and Af. taltalina Ric.. They are characterized by the production of multitudes of small flowers (floral tubes are 1.5-12 mm), bractless petioles, leaves that are often very small and/or withered upon flowering, and a very short (under 30 cm), bushy

27 habit. This combination of features is unique in the family, and there is no doubt that the taxa form a monophyletic group. The four taxa are difficult to distinguish because most features seem to intergrade from taxon to taxon. Moreover, most leaf traits are not useful characters for all specimens, since leaves are often missing or withered.

Members of the M. humilis complex, which are largely Chilean, are found from

Santiago (33°20’S) to Guatacondo (20°S6’S) in the north (Figure 3.1). (A few collections have been made in Prov. Neuquén, Argentina, but these will not be considered here.)

Collections have been made up to 2900 m, but the majority are from under 1000 m. The preferred habitats are rocky, arid soils characteristic of dry ravines, washes, and road cuts.

Although the climate of this entire range is dry, there is a gradient of increasing aridity from the mediterranean south to the hyperarid north. An anticyclone stationed over the Facifîc

Ocean off the coast of Chile blocks most of the moisture-laden polar air from moving north; the anticyclone’s influence extends as far south as Huasco (28°30’S) in southern Region m (Chilean political divisions; Figure 3.1). In this area, occasional precipitation supports more permanent vegetation than found further to the north (Rundel et al. 1991 ; Gajardo

1993). At about 27“S, the climate becomes so dry that most populations of the Af. humilis complex are restricted to the coast. Occasionally, Af. humilis is found in dry ravines in the far north where moisture probably is available from the mountain runoff or fossil water deposits.

Ricardi (1967) last revised the family, and he was the first to treat the Af. humilis complex in its entirety. He expanded the concept of the first species described in the group,

Af. humilis, to encompass two other species as varieties [Af. humilis Poepp. var. parviflora

(Phil.) Ric. and Af. humilis Poepp. var. propinqua (Gay) Ric.]. Ricardi’s three varieties of

Malesherbia humilis are distributed over almost the entire range of the complex, and the species is the only taxon in the complex with linear stipules. The type variety is found in the southern part of the range, from Santiago to Huasco. It has floral tubes 4 J-7 .0 mm in

28 length. Malesherbia humilis var. propinqua, in contrast, is endemic to the Elqui Valley

(30°S) and the surrounding area in Region IV. It is distinguished by its 8-11 nun long floral tube. Malesherbia humilis var. parviflora has minute, sometimes obsolete, stipules and small floral tubes (2.5-3.S mm). This last variety is found along the coast and in the pre-Andean foothills of Regions 0 and HI; its distribution overlaps that of M. humilis var. humilis in the Huasco region. The distribution of a second species, M. multiflora, largely overlaps that of M. humilis var. parviflora. Ricardi (1967) considered M. multiflora to be a pre-Andean species and M. humilis var. parviflora a coastal species. Malesherbia multiflora is distinguished from M. humilis var. parviflora by the presence of a mucron on the capsule and the absence of stipules.

The remaining two taxa are very restricted in distribution. Malesherbia taltalina is endemic to the dry ravines around the coastal town Taltal (25°26’S) and is distinguished largely by tiny flowers having orbicular petals which are much smaller than the sepals. The petals have a single, glandular hair at the tip. This species lacks stipules, like M. multiflora, and has a dense, corymbiform inflorescence. The last species, M. gabrieiae, is in many ways at the other extreme of the complex’s morphological spectrum from M. taltalina. It is distributed in the Elqui Valley, like M. humilis var. propinqua. The flowers of this species, which is currently only known from the type, are the largest in the complex (floral tubes are over 12 mm long), and the stipules are bifid.

Although the descriptions of the taxa in Ricardi’s treatment should allow their separation based upon stipule and floral features, in practice this is often not the case.

Malesherbia humilis var. propinqua and M. gabrieiae are almost indistinguishable, as are

M. multiflora and M. humilis var. parviflora. Malesherbia humilis var. humilis grades into

M. humilis var. propinqua at one extreme of its morphological variation; specimens at the other extreme of its variation grade into M. multiflora and M. humilis var. parviflora.

Because there seems to be a gradation of traits from M. taltalina through M. multiflora and

29 into Af humilis var. humilis, which in turn appears to intergrade with Af humilis var. propinqua and Af gabrieiae, the taxonomic utility of these species and varieties is questionable.

A detailed analysis of the group is necessary to determine whether the Af. humilis complex is best treated taxonomically as a large, polymorphic species or rather as a series of very similar yet distinguishable taxa. If more than one taxon can be recognized, it would be further necessary to determine which morphological characters should be utilized in the construction of keys to allow identification of individuals in the field and herbarium.

Therefore, a phenetic study of morphological data was conducted to determine if Ricardi’s

(1967) revision accurately reflects the distribution of morphological characters in these taxa.

Materials and Methods

In total, 113 operational taxonomic units (OTU’s) were examined for this study (Figure

3.2). Multiple herbarium sheets from a single collection were grouped into a single OTU;

16 such OTU’s were used. Of the 97 collections for which only one sheet was available, 69 consisted of more than one individual plant One collection was split into two separate

OTU’s because initial examination suggested two separate collections may be mounted on the same sheet (Johnston 4746; F, US, GH). Each OTU was tentatively assigned to a taxon based on Ricardi’s (1967) descriptions (Figure 3.2). The character states of both quantitative and qualitative floral characters were recorded (Figure 3.3); only qualitative vegetative characters were used, because sufficient leaf material for measurements was not available on many specimens and because it was not possible in most cases to use comparable leaves. At least five separate measurements were taken (from different plants and herbarium sheets, if possible) to determine the averages of the quantitative characters for

30 floral characters (the androgynophore length, however, is the average of three measurements from different plants). Stipules were considered to be present if even one was found on one of the plants comprising an OTU.

NTSYSpc version 2.0 (Rohlf 1998) was used to analyze the resulting data matrix. The data were standardized, then used to compute Manhattan distance, average taxonomic distance, and Euclidian distance. Phenograms were constructed using the unweighted pair- group method using arithmetic averages (UPGMA). To test for consistency between the resulting phenogram and the original data, cophenetic correlation coefflcients were calculated for each phenogram-data matrix pair. The phenogram with the highest cophenetic correlation coefficient is reported here. Large clusters were then isolated from the data set and analyzed separately to eliminate any influence outside populations may have on intracluster structuring (Whitkus & Packer 1984).

A principal components analysis was also performed. The standardized data were used to compute a correlation matrix, from which the eigenvectors were extracted and OTU’s plotted. The first three eigenvectors are presented.

Discriminant analyses were performed on the quantitative characters using SAS Release

6.12 (SAS Institute, Cary, NC) to determine which characters, if any, are most useful in distinguishing among the clusters of OTU’s found in UPGMA and ordination analyses.

The PROG DISCRIM command (SAS 1989) was employed in this analysis. The populations were assigned to a group for this analysis based on their position in the cluster analysis using average taxonomic distance; discriminant analysis was not used to identify any groups.

31 Results

Initial herbarium studies carried out prior to numerical analysis led to tentative identirications of the 113 OTU’s in the study. It was most diffîcult to distinguish M. multiflora from Af. humilis var. parviflora, Af. humilis var. propinqua from Af. gabrieiae, and Af. humilis var. humilis from Af. humilis var. propinqua. Morphological variation, especially in flower sizes, was essentially continuous from Af. gabrieiae, with its relatively few, large flowers, through Af. humilis var. propinqua, Af. humilis var. humilis, Af. humilis var. parviflora, Af. multiflora, and Af. taltalina, which has multitudes of extremely small flowers.

Of the three coefficients used to construct phenograms, average taxonomic distance yielded the phenogram with the least distortion (r = 0.87). There are two large and two small clusters in this tree (Figure 3.4). For convenience, the clusters will be referred to by the speciflc epithet or varietal name of the predominant taxon (as identified before the computer analysis) composing the cluster; these names will be unitalicized and in quotes to prevent confusion with the proper taxonomic name. One of the large clusters, cluster A, is composed of those populations identified as Af. humilis var. humilis and Af. humilis var. propinqua. A subcluster B (“propinqua”) within this “humilis” cluster contains Af. humilis var. propinqua populations, two Af. humilis var. humilis collections, and one Af. gabrieiae specimen. Separate UPGMA analysis of this cluster did not reveal any further substructuring; this separate analysis will therefore not be further considered.

The other large cluster, C (“parviflora”), contains those populations tentatively identifled as Af. multiflora and Af. humilis var. parviflora, including the type specimen of

Af. multiflora. The small subclusters in the total data analysis correspond neither to a taxon nor to a geographical region. Although many Af. multiflora populations clustered together.

32 others are mixed with specimens which usually could not be assigned to either taxon with certainty. Separate UPGMA analysis of this cluster failed to cluster these taxa together as well.

Cluster D (“gabrieiae”) contains only those specimens initially identified as Af. gabrieiae or Af. humilis var. propinqua. The type specimen for Af. gabrieiae is a member of this group. The Hnal cluster, E (“taltalina”), contains only two populations, both identified as Af. taltalina. One of these specimens is the type for that species. No other specimens were tentatively identified as Af. taltalina.

In ordination analysis, most of the variation was explained by principal component axis

I (40.97%), to which petal length, floral tube shape, and sepal length contributed most of the variation (Figure 3.5). Principal component axis U explained 16.23% of the variation. The presence or absence of anastomosing sepal veins, floral tube length, and androgynophore length contributed most to this axis. The third axis explained only 8.17% of the variation.

Sepal pubescence, petal apex vestiture, and the pedicel length composed principal component axis HI.

Ordination (Figure 3.6) separates the OTU’s into two or possibly three groups corresponding closely to those in the cluster analysis. The cluster “taltalina” is differentiated from “parviflora” only by principal component axis 3. The “parviflora” +

“taltalina” cluster is well differentiated from the remaining OTU’s, which fall into a loose group. The “humilis” populations lie between the “parviflora” + “taltalina” cluster and those populations assigned by cluster analysis to “propinqua” and “gabrieiae. ”

“Propinqua” and “gabrieiae” are only very weakly separated from each other and from

“humilis.”

Discriminant analyses were performed to determine if any of the characters used in this study deflne the clusters found in the UPGMA and ordination analyses. The five clusters

from the UPGMA analysis, “parviflora,” “humilis,” “propinqua,” “taltalina,” and

33 “gabrielæ,” were used to group the OTU’s. Only two of the clusters had diagnostic characters. "Taltalina" could be accurately distinguished by the presence of an apical, glandular petal hair. “Gabrielae” was distinguished by its pilose androgynophore.

“Propinqua” + “gabrielae” could be identifîed by these plants’ large funneiform floral tubes, which were longer than 7 mm.

With the exception of the shape of the pressed floral tube, there was considerable overlap in the character states of “parviflora” and “humilis.” The shape of the floral tube did separate the two, although there was a single anomalous population of “parviflora” with funneiform tubes like those of “humilis.”

Discussion

The results of the ordination and UPGMA analyses are largely consistent with

Ricardi’s (1967) concept of the morphological divisions in the M. humilis complex, although his concepts of some taxonomic boundaries require revision. Clusters roughly corresponding to all four of his species were found. Ricardi’s broad concept of M. humilis, however, is inconsistent with these results. Malesherbia humilis var. parviflora and Af. multiflora are indistinguishable from each other and together they are distinct from the type variety of M. humilis.

Ricardi (1967) based his distinction between Af. humilis var. parviflora and Af. multiflora largely upon Af. multiflora's lack of stipules and the presence of a mucron at the capsule’s apex. Malesherbia humilis var. parviflora has stipules and lacks the capsular mucron. In the preliminary herbarium studies, however, most of the populations belonging to “parviflora” having stipules were also found to have a mucron. One population lacking stipules also lacked a capsular mucron. Moreover, populations with flowers lacking the

34 mucron fail to cluster together in both the total data UPGMA analysis and the separate analysis of this cluster. The presence or absence of a mucron on the capsule’s apex is

therefore an unreliable taxonomic character.

In the UPGMA analysis of the total data set, there is also no cohesive pattern for the

absence or presence of leaf stipules (as opposed to stipules found on the bracts) in

“parviflora.” Leaf stipules are present in three subclusters scattered among OTU’s lacking

them, indicating that stipule presence does not group those populations as much as other characters tie them to populations lacking stipules. In separate analysis of the cluster, all

populations having leaf stipules group together. However, this stipulate subcluster is well embedded in the middle of the cluster, suggesting that this small subcluster is so strongly

associated with exstipulate populations that continued recognition of these populations as a

separate taxon would be biologically artificial.

Populations in the “parviflora” cluster—with or without stipules or a capsular

mucron—tend to have wineglass-shaped floral tubes (these floral tubes do not flare open in

a funneiform fashion), withered lower leaves, corymbose panicles, no bract stipules, and a

trend toward smaller floral parts than members of “humilis, ” “propinqua,” or

“gabrielae. ”

In addition to sharing these morphological traits, M. multiflora and Af. humilis var. parviflora also share geographic distributions. Ricardi (1967) described the distribution of

M. humilis var. parviflora as coastal, with some populations approaching the interior,

whereas Af. multiflora was described as a native of the pre-Andes and upper reaches of the coastal desert However, specimens originally identified as Af. multiflora (as well as those

listed in Ricardi’s revision as Af. multiflora) were often collected near the coast.

Additionally, a number of collections of Af. humilis var. parviflora are from the interior

35 desert and higher elevations. Coastal specimens may be stipulate or exstipulate, further evidence that the presence or absence of leaf stipules is not a taxonomically useful character in this cluster.

Their overlapping distributions and morphological similarity indicate that there is little evidence supporting the continued recognition of M. multiflora as separate from M. humilis var. parviflora. Malesherbia humilis var. parviflora has nomenclatural priority, leaving M. multiflora as a synonym.

Very similar to M. humilis var. parviflora is “taltalina,” which is endemic to the ravines and hills around Taltal (25°26’S). Its distribution falls within that of “parviflora,” which is also found in Taltal’s ravines. “Taltalina” forms its own cluster in UPGMA analysis, but it is only marginally distinct from “parviflora” in ordination. In its description, M. taltalina is distinguished by its obconical floral tube, orbicular and small petals with an apical glandular hair, violet flowers, and anastomosing sepal veins (Ricardi 1967). “Taltalina” is not as clearly separated from “parviflora” as the description suggests. Although “taltalina” does differ from “parviflora” by the presence of an apical glandular petal hair, the two groups have similar floral tube shapes. Violet flowers and anastomosing sepal veins are also found among members of “parviflora.” The type specimen of M. taltalina does have some flowers with orbicular petals, but others have much narrower petals like those of the second specimen identified as “taltalina” and those of “parviflora.” In general, the petal length to sepal length ratio of “taltalina” is smaller than that of members of “parviflora,” although one collection of “parviflora” does have a smaller ratio than both collections of “taltalina.”

Because this tendency toward small petals and the presence of the apical glandular petal hair do permit recognition of the cluster, it should continue to be seen as a distinct taxon.

The second major group in the M. humilis complex is the “humilis” cluster (Figures

3.4 & 3.6). Ordination analysis indicates that “humilis” is the nexus of the entire complex.

It is morphologically intermediate to “parviflora” and “propinqua,” which were both

36 considered to be varieties of M. humilis in Ricardi’s (1967) work. “Humilis” is distinct from “parviflora” in ordination analysis, but it is only weakly separated from

“propinqua.”

Morphologically, “parviflora” is best distinguished from “humilis” by the shape of the flattened floral tube. The floral tubes of “parviflora” are, with the exception of one population, wineglass-shaped. In “humilis,” on the other hand, the floral tubes are always funneiform (the floral tube flares open at the throat). Although it was not available for the statistical analysis, examination of the type specimen of M. humilis var. parviflora at SCO showed that it, too, has the wineglass-shaped floral tube characteristic of “parviflora.” The single population clustering with “parviflora” despite its funneiform floral tube was collected well north of southwestern, coastal Region m near Huasco (28°30’S), where the two clusters, as well as other elements of the northern and southem coastal desert floras, overlap in distribution (Gajardo 1993). The appearance of funneiform flowers in a

“parviflora” collection may be indicative of introgression or an ancestral polymorphism.

No other character used in this study separates the clusters as well as the floral tube shape. There are, however, some general trends. “Parviflora” tends to have smaller petals, sepals, and floral tubes, although there is extensive overlap with “humilis” in these features.

When the floral tube length to width ratio is used in conjunction with locality data, the populations studied here can be separated. Collections made north of the coastal Huasco region or in the pre-Andean Huasco region belong to either “parviflora” or “taltalina,” whereas those made south of that region are “humilis,” “propinqua,” or “gabrielae.” In the coastal Huasco region, all of the populations of “parviflora” used in this study have floral tube length to width ratios greater than 2.0, whereas those of “humilis” have ratios less than 1.6.

Further studies of “humilis” and “parviflora” should include more extensive field observations and sampling, especially where the two overlap in the region just southwest of

37 Huasco. Available data are inadequate to address possible interbreeding, ecological differentiation, and population sdecture in this region. These data are critical to understand the relationship between these very similar varieties. Although the relationship between the two taxa is not well understood, the morphological data do support the recognition of two taxa.

"Parviflora" and “humilis” are difficult to distinguish when dealing with small- proportioned specimens of the latter group. When individuals of "humilis” have larger flowers, they become difficult to distinguish from members of “propinqua.” Recognition of “propinqua” taxonomically was supported by UPGMA analysis, in which “propinqua” clustered with “humilis” at a relatively high average taxonomic distance (approximately

1.06; Figure 3.4). Ordination analysis (Figure 3.6) reaffirmed the affînities between the two clusters, but it also exposed the close ties between the morphologically similar “propinqua” and “gabrielae,” the two clusters with the largest flowers in the complex. The close relationship between “gabrielae” and “propinqua” may have been lost in the UPGMA analysis because this method of analysis can distort basal relationships (Sneath & Sokal

1973).

The distinction is somewhat tenuous between “humilis” and “propinqua,” which is intermediate to “humilis” and “gabrielae.” “Propinqua” and “gabrielae” have floral tubes 7 mm long or more. In “humilis,” floral tubes are almost always less than 7 mm in length, although some specimens may have a few flowers with floral tube lengths slightly exceeding 7 mm. These specimens, however, are few, and the vast majority of flowers on each specimen have floral tubes under 7 mm in length. Many of the specimens assigned to

“propinqua” or “gabrielae” also have bifîd or lobed leaf stipules. In contrast, all specimens of “humilis” have linear, unlobed stipules.

“Gabrielae” is perfectly distinguished from “propinqua” by an unusual trait in the Af. humilis complex, its pilose androgynophore. The presence of a diagnostic character and the

38 distinctiveness of “gabrielae” in ordination and cluster analyses support the recognition of that cluster taxonomically. “Propinqua” should also continue to be recognized because it is distinguishable from “humilis” by its longer floral tubes and often bifid stipules. Its glabrous androgynophore distinguishes it from “gabrielae.”

“Gabrielae,” “propinqua,” and “humilis” have more or less separate distributions.

Although a few collections of “humilis” were made in the Elqui Valley (30°S), the majority were found closer to the coast and around the cities La Serena (29°S4’S) and Ovalle, which have coastal and interior matorral steppe vegetation types, respectively (Gajardo 1993). In contrast, “gabrielae” is endemic to the eastern Elqui Valley and the eastern tributaries of the

Rio Elqui, where the vegetation borders on shrubby pre-Andean steppe (Gajardo 1993).

Most specimens of “propinqua” were collected in the Elqui Valley, although a few were found 40-50 km to the north and to the south near Ovalle. It is associated with montane flowering desert and interior matorral steppe (Gajardo 1993).

Within the Elqui Valley and its eastern tributaries, the locality data for “propinqua” and

“gabrielae” indicate that the two have adjacent, yet separate, distributions. Members of

“gabrielae” were all collected east of Diaguitas (65 km from the ocean), whereas those of

“propinqua” were collected further to the west, south, or north. This distributional difference roughly corresponds to changes in vegetation type from the flowering desert of the hills to the shrubby pre-Andean steppe (Gajardo 1993). “Humilis” has been collected in both eastern and southem areas of the valley.

Both phenetic and distribution data support recognition of “propinqua” and

“gabrielae.” However, confusion existing in the typification of “propinqua” could affect

“gabrielae.” Ricardi (1967) named three specimens collected by Claudio Gay (number

345) deposited in F, NY, and GH as isotypes of Af. humilis var. propinqua Gay (= Af. propinqua). Gay did not refer to any specimens in his description, and Ricardi did not designate which of the three specimens should be the neotype. I have examined all of the

39 specimens of the Af. humilis complex belonging to these institutions, and only one herbarium sheet of plants collected by Gay exists in each herbarium. There is little evidence, however, that the three specimens belong to the same collection other than the fact that they all were originally deposited in the Natural History Museum of France, Gay’s home country.

The specimen deposited at F Is the only one labeled “Gay 345.” This specimen, incidentally, is a member of the cluster “propinqua.” The herbarium sheet at NY has two labels: on the first is written “Hennegart” and “ 1876,” presumably the date of collection, and on the second label, which is handwritten on different paper, is found the name “Gay,” but no number. This specimen has much smaller flowers than the one at F and is a member of “humilis.” In his description. Gay (1846) describes Af. propinqua as having a floral tube 7 mm in length. This description does not match the NY specimen, which has smaller flowers. Moreover, Gay could not have collected this specimen if 1876 is the collection date—he died in 1873 (Brummitt & Powell 1992).

The third specimen is deposited at GH, and this specimen is labeled only with the name

“Gay” and the country of origin, “Chili” (.sic). A later worker added information that appears to be taken directly from the distribution information given by Gay (1846) in his description of the taxon. There is no indication in the flora that this distribution was to be taken as exact locality data for his collections, which were not cited. There is also no clue to the relationship between this specimen, which clusters with “gabrielae,” and the F specimen.

Because these three specimens belong to three different clusters (“humilis,”

“propinqua,” and “gabrielae”), which will each be given taxonomic status, it is imperative that the relationships among them be established. The F specimen is the only one that unambiguously is “Gay 345.” It is quite clear from the morphology and the labels that the

NY herbarium sheet does not belong to the same collection (or even the same collector) as

40 the F specimen, so the NY specimen was mistakenly included in the list of isotypes. The

GH specimen is so ambiguously labeled that it is impossible to determine if it belongs to the Gay 345 collection. Since both the label and the phenetic analysis do not support inclusion with the F specimen, the GH specimen should be considered to be simply Gay s.n. and not tied to the F specimen of M. humilis var. propinqua. Designation of the F herbarium specimen as the neotype of Af. humilis var. propinqua would rid the need for nomenclatural novelties. The phenetic cluster “propinqua” would thus be associated with

Af. humilis var. propinqua. “Gabrielae” is associated with Af. gabrielae, since its type is included in that cluster.

Because the differences among the five clusters that are to be recognized taxonomically are minor, they are best recognized as five varieties of a single species, Af. humilis. Each variety contains the type of a taxon that Ricardi or previous workers described, so no new names are needed, although two new combinations, made here, are necessary. The varieties recognized are Af. humilis Poepp. var. gabrielae (Ric.) Gengler (“gabrielae”), Af. humilis

Poepp. var. humilis (“humilis”), Af. humilis Poepp. var. parviflora (Phil.) Ric.

(“parviflora”), Af. humilis Poepp. var. propinqua (Gay) Ric. (“propinqua”), and Af. humilis Poepp. var. taltalina (Ric.) Gengler (“taltalina”).

BiogeographyofM. humilis complex—Hypotheses of the biogeography and evolution of the Af. humilis complex may be derived from the distinctive distributions of the five varieties and their morphological affinities. The combined distributions of Af. humilis var. humilis and Af. humilis var. parviflora cover almost the entire range of the complex. Judging from their broad distributions and their great morphological diversity, they are most likely the oldest varieties in the complex. The fact that these taxa meet only in southwestern Region HI near Huasco, where the vegetation type and climate is in transition hrom semi-aridity to desert, suggests that establishment of this climatic pattern split the distribution of the

41 ancestor to Af. humilis var. humilis and Af. humilis var. parviflora. Alternatively, dispersal over this climatic transition zone with secondary contact between the resulting varieties may have created their largely allopatric distribution.

The other varieties are restricted in distribution, and they may be the results of recent range expansions to the north and into the pre-cordillera. In the north, Af. humilis var. taltalina is probably derived from Af. humilis var. parviflora, with which it is sympatric. The flora of Taltal is unusually rich due to the heavy ocean fogs which coat the steep slopes of the coastal mountains (Johnston 1929; Rundel et al. 1991), and it is not surprising that a variety of Af. humilis has evolved in that unusual habitat.

One of the other two endemics in the complex, Af. humilis var. propinqua, is probably derived from the southem Af. humilis var. humilis, judging from the morphological similarities between the two varieties. The morphological affinity between Af. humilis var. propinqua and Af. humilis var. gabrielae and their distinct but adjacent distributions further suggests that these varieties are phylogenetically closely related. The northern part of

Region IV, which includes the Elqui Valley, is noted for its endemism (Ricardi 1967), so the concentration of three varieties of Af. humilis in this region reflects the overall diversity of the area. Although the flora is known to be diverse, little is known about community structure or distributions of plants within the area (Gajardo 1993), except for the Doha Ana range in the east (Squeo 1994). Such information would be very useful in understanding the evolutionary path of Af. humilis in this part of its distribution.

Conclusion—The similarities among clusters formed in ordination and UPGMA analyses suggest that the clusters are best treated as five varieties of a single, diverse species, Af. humilis. These varieties correspond in large part to Ricardi’s concepts of taxonomic

42 boundaries in the M. humilis complex. The results of this study highlight the need for more extensive studies to test the hypotheses presented here of the circumscriptions and origins of the varieties of Af. humilis.

Taxonomy

1. Floral tube funneiform when pressed flat, flaring from the base of the ovary to the

throat. Distribution; 29°S-33°20’S

2. Floral tube length, on average, less than 7 mm. Leaf stipules linear and lacking

lobes ...... 1. Af. humilis var. humilis

2’. Floral tube length greater than or equal to 7 mm. Leaf stipules often bifid or lobed,

sometimes linear and/or unlobed.

3. Androgynophore pilose over entire length 2. Af. humilis var. gabrielae

3 \ Androgynophore glabrous, or pilose only at the base of the ovary

3. Af. humilis var. propinqua r . Floral tube wineglass-shaped when pressed flat, remaining narrow after flaring at the

base of the ovary. Distribution: 29“S - 21°S

4. Petals lacking a single, apical glandular hair ...... 4. Af. humilis var. parviflora

4’. Petals with a single, apical glandular hair S. Af. humilis var. taltalina

1. Malesherbia humilis Poepp. var. hum ilis = Malesherbia humilis D. Don, Edinb. New

Philos. Joum. 1II. 1831 = Af. humilis D. Don var. delta Gay, Flora Chilena 2:425.

1846. = Af. humilis D. Don var. b Reiche, Anal. Univ. Chile, 98:734. 1897. =

Gynopleura humilis (D. Don) Roemer, Syn. Pepon. 211. 1846. = G. dilatata Walpers,

Rep. Bot. Syst. 2:223. 1843. =Af. dilatata (Walp.) F. Philippi, Anal. Univ. Chile 59:

125.1881. — Proposed holotype: Region V, Dpto. Los Andes, between Las Vizcachas

& Rio Colorado, Poeppig 134 (FI; isotype - MO)

43 This variety is distinguished from var. parviflora and var. taltalina by the shape of

its floral tube, which is funneiform, flaring at the throat of the tube. In the Huasco

region, where var. humilis and var. parviflora overlap and this character is sometimes

not useful, the ratio of the floral tube length to width can be used to separate the two

varieties. Malesherbia humilis var. humilis has a ratio less than 1.6. This variety has

floral tubes less than 7 nun in length and has linear, unlobed stipules. The type variety

has a wide distribution from Huasco (28°30’S) south to Santiago. Some collections

have been made in the Elqui Valley, but the majority are from nearer the coast. The

Argentinean collections are probably of this variety.

Ricardi (1967) named isotypes for this name, but did not designate any one

specimen as the type. Poeppig (1829) described M. humilis in the context of an article

describing specimens he collected in the Rfo Colorado valley of Region V. The two

specimens used in this study labeled Poeppig 134 were collected in this valley and fit

the description he gives. They are almost certainly the material upon which he based his

description, and therefore the type should be chosen from them. The F specimen is the

more intact of the two examined and would be the best choice.

2. Malesherbia humilis Poepp. var. gabrielae (Ric.) Gengler, comb. nov. = Malesherbia

gabrielae Ricardi, Gayana 16: 120.1967. — Holotype: Region IV, Depto. Elqui, km 9

camino Vicuna-Guanta, crece entre cantos rodados, Ricardi, Marticorena, & Matthei

7/26 (CONG!)

This variety produces floral tubes greater than 7 mm in length, and its stipules are

almost always bifid. It is distinguished from var. propinqua by its pilose

androgynophore. It is endemic to Region IV’s eastern Elqui Valley and the Rio Elqui’s

eastern tributaries.

44 3. Malesherbia humilis Poepp. var. propinqua (Gay) Ric. = Malesherbia propinqua

Gay, Flora Chilena 2:424.1846. = Af. humilis D. Don var. propinqua (Gay) Reiche,

Anal. Univ. Chile, 98:734.1897. — Proposed neotype: Chile, Gay 345 (F!)

This variety has flowers with longer floral tubes than var. humilis (greater than 7

mm), and its leaf stipules are usually lobed, although sometimes linear, unlobed stipules

appear. It differs from var. gabrielae in that its androgynophore is not pilose along its

entire length. The variety is distributed in Region IV in the western portion of the Elqui

Valley, the region 40-50 km north of the valley, and area surrounding Ovalle.

4. Malesherbia humilis Poepp. var. parviflora (Phil.) Ric. = Malesherbia parviflora

Philippi, Anal. Univ. Chile 84:977. 1893. = Af. humilis D. Don sensu Philippi, Viage

Des. Atacama 192, 1860. = Af. humilis Poepp. sensu Johnston, Contr. Gray Herb. 85:

79.1929. =Af. multiflora Ricardi, Gayana 16: 126. 1967.— Holotype: Region m,

Dpto. Chanaral, Chanarcillo, P. Ortega X-1876 (SGO!)

Members of this variety possess floral tubes that do not flare open at the tubes’

throats, but rather they remain constricted to approximately the width of the ovary. The

petal apices of this variety have no glandular hair, as var. taltalina has. The morphology

of this taxon varies widely: some specimens are highly branched and bear hundreds of

tiny flowers, whereas others have only a few branches with a handful of flowers.

Stipules and a mucron at the capsule’s apex may or may not be present. In the Huasco

region, var. parviflora is distinguished from var. humilis by a floral tube length to width

ratio greater than 2.0.

This northern variety extends from southem Region m (Huasco) north to

Guatacondo (20°56’S) and from the coast into the Andean foothills. In the north, the

distribution is strictly coastal, save the single population found in a dry ravine near

Guatacondo.

45 s. Malesherbia humilis Poepp. var. taltalina (Ric.) Gengler, comb. nov. - Malesherbia

taltalina Ric., Gayana Botànical6:123.1967. — Holotype: Region II, Dpto. Taltal,

Quebrada Anchuna, ca. 1 km de la entrada, flores violadas, crece entre cantos rodados,

Ricardi 2530(CONCl)

This variety is distinguished by the glandular apical hair on each petal. The petals

tend to be very small relative to the sepals, although this feature cannot be used to

separate the variety from var. parviflora with great accuracy. Like var. parviflora, var.

taltalina produces many, very small flowers with floral tubes that do not flare open at

their apices.

This variety is endemic to the lomas around Taltal in Region U. Although it hasn’t

been collected in the nearby lomas of Paposo, the variety’s distribution may well extend

there.

46 Literature Cited

Brummitt, R.K. & CÆ. Powell. 1992. Authors of Plant Names. Royal Botanic Gardens, Kew.

Fay, M f ., SAi. Swensen, & M.W. Chase. 1997. Taxonomic affinities of Medusagyne oppositifolia (Medusagynaceae). Kew Bulletin 52: 111-120.

Gajardo, R. 1993. La vegetaciôn natural de Chile: clasificaciôn y distribuciôn geografica. Editorial Universitaria, Santiago, Chile.

Gay, C. 1846. Historia fisica y politica de Chile. Tomo segundo. Paris.

Johnston, I.M. 1929. Papers on the flora of northern Chile. Contributions from the Gray Herbarium of Harvard University 85:1-171.

Poeppig, E. 1829. Schreiben des jezt in Chile reisenden Hm. Dr. Poppig. Notizen aus dem Gebiete der Natur- und Heilkunde 23:290-293.

Ricardi, M. 1967. Revision taxonômica de las Malesherbiaceas. Gayana Botànica No. 16: 1-139.

Rohlf, FJ. 1998. NTSYSpc version 2.0. Exeter Software, Setauket, NY.

Rundel, P.W., M.O. Dillon, B. Palma, H.A. Mooney, S.L. Gulmon, & J R. Ehleringer. 1991. The phytogeography and ecology of the coastal Atacama and Peruvian deserts. Aliso 13: W 9 .

SAS. 1989. SAS/STAT user’s guide, version 6. Fourth edition, volumes 1 & 2. SAS Institute Inc., Cary, NC.

Sneath, P.H.A. & R.R. Sokal. 1973. Numerical Taxonomy. W.H. Freeman and Company, San Francisco, California.

Squeo, F.A., R. Osorio, & G. Arancio. 1994. Flor de Los Andes de Coquimbo: Cordillera de Dona Ana. Editorial Universitaria, Chile.

Trewartha, G.T. 1966. The Earth’s Problem Climates. University of Wisconsin Press, Madison, WI.

Whitkus, R. & J.G. Packer. 1984. A contribution to the taxonomy of the Carex macloviana aggregate (Cyperaceae) in western Canada and Alaska. Canadian Journal of Botany 62: 1592-1607.

47 Figure 3.1. Map of northern and centrai Chile. Political boundaries and the distribution of the M. humilis complex are indicated, with the exception of the Argentinean populations.

48 18 - —18

Region I ) - f

20 - -2 0

«P Guatacondo

22 - -2 2

Region II Antofagasta 24 - - 2 4

26 - Region III - 2 6

28 - - 2 8 Huasco i____

Ij, s,mm. ^iSllL/'JrB,'"Sv 30 - — 30 alia Region IV

32 - - 3 2 Region V

-'S ^Santiago 34 - - 3 4 Region VI

Figure 3.1 49 Tentative Collector Locality Holding herb identification G Montero 11686 Region IV, Prov. Elqui, Elqui Valley OS

0 Puin s.n. Region IV, Prov. Elqui, Elqui Valley ULS

0 Ricardi, Marticorena, & Matthei 1126 Region IV, Prov. Elqui, Elqui Valley CONG (Type) Q/P Gajardo 3002 Region IV, Prov. Elqui, Elqui Valley CONG Q Arancio 91764 Region IV, Prov. Elqui, Elqui Valley ULS Q Behn 1310 Region IV, Prov. Elqui, Elqui Valley CONG Q Gajardo 2935 Region IV, Prov. Elqui, Elqui Valley CONG Q Gay 345 unknown F (Type) Q Gay s.n. Region IV, Prov. Elqui GH Q Jiles 5149 Region IV, Prov. Elqui, Condoriaco OS Q Johnston 6235 Region IV, Prov. Elqui, Estero Guanta GH Q Looser 4287 Region IV, Prov. Elqui, Elqui Valley GH Q Wagenknecht 18447 Region IV, Prov. Elqui, Elqui Valley F, GH, MO, UG Q Worth & Morrison 16354 Region IV, Prov. Elqui, Elqui Valley GH, UG, MO Q/H Jiles 4748a Region IV, Prov. Limarf, El Altar OS

Figure 3.2. Populations and specimens used in the analyses. Tentative identifications were made prior to analysis. Abbreviations:

G = M. gabrielae, Q = M. humilis var. propinqua, H = M. humilis var. humilis, V = Af. humilis var. parviflora, T = Af. humilis var. taltalina, U = undetermined, Frov. = province, Qda = quebrada (dry ravine), (continued) Figure 3.2. (continued)

H Ban'os 5524 Region IV, Prov. Limarf, Ovaile US

H Barros 3778 Region IV, Prov. Choapa, East of Illapel GH

H Claude Joseph 5420 Region IV, Prov. Elqui, Andacollo US

H Claude-Joseph 722 Region Metropoliiana US

H Philippe s.n. unknown us

H Fraga 1332 Region IV, Prov. Choapa, Tilama GH

H Gay or Hennegart 1876 unknown NY

H Jiles 3050 Region IV, Prov. Limarf, Ovalle OS

H Jiles 3267 Region IV, Prov. Limarf, Ovalle OS U) H Jiles 4771 Region IV, Prov. Limarf, Ovalle OS H Jiles 5675 Region IV, Prov. Limarf, Combarbala OS

H Looser 4244 Region IV, Prov. Elqui, Elqui Valley GH

H Looser 4288 Region IV, Prov. Elqui, Elqui Valley GH

H Looser 5555 Region Meuopolitana, Qda. Ramôn GH

H Marticorena, Rodriguez, & Wendt 1776 Region III, Prov. Huasco, Qda. El Morado F

H Marticorena, Rodriguez, & Wendt 1830 Region III, Prov. Huasco, Chanaral de F Aceituna H Montero 1839 Region IV, Prov. Elqui, Coquimbo GH

H Philippi s.n. Region MeUopolitana F

H Poeppig 134 Region V, between Las Vizcachas & Rfo F, MO (Type) Colorado Figure 3.2. (continued)

H Ricardi & Marticorena 4904 Region III, Prov. Huasco, Incahuasi OS

H Rose 19216 Region IV Prov. Choapa, Choapa NY

H Skottsberg 772 Region IV, Prov. Limarf, Fray Jorge F, NY

H Solbrig 3068 Region IV, Prov. Elqui, between La Serenu GH & Condoriaco H Taylor, Bohien, & Marticorena 10672 Region IV, Prov. Elqui, El Tofo MO H Wagenknecht 18480 Region IV, Prov. Elqui, Elqui Valley MO, UC, GH, F H Wcrdermann 119 Region IV, Prov. Elqui, Elqui Valley GH, MO, F, UC H unknown Region IV, Prov. Elqui, Coquimbo US

M/H Barros 10127 Region IV, Prov. Elqui, Coquimbo CONC » M/H Billiet & Jadin 5289 Region IV, Prov. Limarf, Fray Jorge MO M/H Dillon & Tiellier 4981 Region IV, Prov. Elqui, Totoralillo OS

M/H Lammers, Bæza, Penaillo 7587 Region IV, Prov. Limari, Corral Quemado F

M/H Zôilner 8270 Region III, Prov. Elqui, Guanaqueros MO

M/V Barros 5538 Region II, Prov. Antofagasta, Taltal UC

M/V Billiet & Jadin 5496 Region II, Prov. Antofagasta, Paposo MO

M/V Dillon, Dillon, & Pobleto 5557 Region II, Prov. Antofagasta, Paposo F

M/V Dillon &Teillier 5122 Région II, Prov. Antofagasta, Paposo OS

M/V Gengler 106 Region II, Prov. Antofagasta, Qda. Anchuna OS

M/V Gengler 190 Region I, Prov. Iquiqui, Qda. Guatacondo OS Figure 3.2. (continued)

M/V Hoffmann 212 Region II, Prov. Antofagasta, Taltal CONC

M/V Johnston 3627 Region II, Prov. Antofagasta. Qda. La US Chimba M/V Johnston 4962 Region III, Prov. Copiapô, Cerrillos GH

WW Johnston 5096 Region II, Prov. Antofagasta, Prov. Taltal GH

M/V Maiticorena s.n. Region 111, Prov. Huasco, Cerro Colorado CONC

MA/ Marticorena, Matthei, & Quezada 469 Region IV, Prov. Elqui, Cuesta Pajonales OS

M/V Montero 11020 Region II, Prov. Huasco OS

M/V Quezada & Ruiz 88 Region II, Prov. Antofagasta. Qda. La CONC Chimba MA/ Quezada & Ruiz 130 Region II, Prov. Antofagasta, Caleta El CONC Cobie 52 MA/ Ricardi 2268 Region III, Prov. Copiapô, between Copiapô OS & Vallenar M/V Ricardi 2316 Region III, Prov. Huasco, Agua Amarga OS

M/V Ricardi 2633 Region II, Prov. Antofagasta, Qda. Paposo OS

MA/ Ricardi & Marticorena 3740 Region III, Prov. Huasco, Estancia Manilas CONC

M/V Taylor, Bohien, & Marticorena 10759 Region II, Prov. Antofagasta, Taltal MO

M/V Werdermann 778 Region II, Prov. Antofagasta, Taltal US, NY, MO, F. GH, UC MA/ Worth & Morrison 15797 Region II, Prov. Antofagasta, Qda. Taltal UC

V Garaventa 4737 Region III, Prov. Huasco, Canto del Agua CONC

V Jafftiel 1058 Region II, Prov. Tocopilla, Tocopilla GH

V Johnston 5095 Region II, Prov. Antofagasta, Qda. Taltal GH Figure 3.2. (continued)

V Morong 1288 Desert Atacama MO, NY, US, GH, US V Ricardi, Marticorena, & Matthei 1313 Region III, Prov. Copiapd, Caleta Obispito OS

M Dillon & Dillon 5686 Region II, Prov. Antofagasta, Taltal OS

M Dillon & Teillier 5023 Region III, Prov. Huasco, Domeyko OS

M Dillon & Teillier 5298 Region II, Prov. Antofagasta, Miguel Diaz F

M Dillon & Teillier 5324 Region II, Prov. Antofagasta F

M Gigoux s.n. Region III, Prov. Copiapô, Qda. GH Chanchoquin M Greninger 26 Region III, Prov. Copiapô, Potrerillos GH

M Johnston 3693 Region III, Prov, Copiapô, Pouerillos GH, F M Johnston 4746 Region III, Prov. Copiapô, Potrerillos F, US, GH

M Johnston 4759 Region III, Prov. Copiapô, El Barquito GH

M Johnston 5012 Region III, Prov. Copiapô, Copiapô GH

M Marticorena, Mattei, & Quezada 466 Region IV, Prov. Elqui, Cuesta Pajonales OS

M Marticorena, Stuessy, & Baeza 9813 Region III, Prov. Copiapô, Qda Puquios OS

M Marticorena, Stuessy, & Baeza 9883 Region III, Prov. Copiapô, El Salvador OS

M Marticorena, Stuessy, & Baeza 9892 Region III, Prov. Copiapô, 42.2 km west of OS Diego de Almagro M Morrison 17091 Region II, Prov. Antofagasta, Qda. Taltal UC, GH

M MuAoz, Teillier, & Meza 2698 Region III, Prov. Copiapô, Copiapô SGO

M Munoz, Teillier, & Meza 2699 Region III, Prov. Copiapô, Copiapô OS, MO Figure 3.2. (continued)

M Pennell 13033 Region II, Prov. Antofagasta US, F, NY, GH

M Ricardi 5503 Region II, Prov. Antofagasta, Antofagasta F

M Ricardi 5530 Region III, Prov. Copiapô, Las Bombas F

M Ricardi & Marticorena 4422 Region III, Prov. Huasco, Qda. Carrizal Bajo OS M Ricardi, Marticorena. & Matthei 518 Region III, Prov. Copiapô, Puquios OS

M Ricardi, Marticorena, & Matthei 520 Region III, Prov. Copiapô, Puquios OS

M Ricardi, Marticorena, & Matthei 664 Region III, Prov. Copiapô, south of Copiapô OS

M Ricardi, Marticorena, & Matthei 1245 Region III, Prov. Huasco, 30 km south of OS Vallenar M Ricardi, Marticorena, & Matthei 1461 Region III, Prov. Copiapô, Paipote OS

M Ricardi, Marticorena, & Matthei 1641 Region III, Prov. Copiapô, Puquios OS

M Ricardi, Marticorena, & Matthei 1689 Region III, Prov. Huasco, Estaciôn Romero F, CONC (Type)

M Rosas 398 Region III, Prov. Huasco, 20 km north of OS Vallenar M Schlegel 7706 Region II, Prov. Tocopilla, Tocopilla CONC

M Taylor, von Bohien, & Marticorena 10699 Region III, Prov. Copiapô, Caldera MO

M Teillier 932 Region III, Prov. Huasco, 10 km east of NY Huasco M Werdermann 1523 Region III, Prov. Copiapô, Caldera NY

M Worth & Morrison 16210 Region III, Prov. Huasco, 60 km cast of UC, GH Vallenar M Zôllner 14760 Region II, Prov. Antofagasta, Paposo MO

T Ricardi 2530 Region II, Prov. Antofagasta, Taltal CONC (Type) Figure 3.2. (continued)

T Beetle 26176 Region II, Prov. Antofagasta, Taltal GH, UC, MO

U Behn s.n. Region IV, Prov, Elqui, Elqui Valley CONC 1. Lower leaf (0 = leaves green, unwithered; 1 = leaves brown, withered) 2. Pedicel length (mm) 3. Leaf stipules (0 = absent; 1 = present) 4. Bract stipules (0 = absent; 1 = present) 5. Inflorescence (0 = loose panicle; 1 = corymbose panicle) 6. Floral tube length (mm) 7. Floral tube length / width ratio 8. Floral tube shape when pressed (0 = wineglass-shaped; 1 = funnel-shaped) 9. Sepal length (mm) 10. Sepal veins (0 = not anastomose; I = anastomose) 11. Adaxial face of sepals pubescence (0 = glabrous; I = pubescent with scattered hairs) 12. Petal length (mm) 13. Petal tip vestiture (0 = lacking a glandular hair at apex; 1 = having a glandular hair at apex) 14. Androgynophore length (mm) 15. Androgynophore vestiture (0 = glabrous; 1 = pilose) 16. Capsule tip (0 = amucronate; 1 = mucronate)

Figure 3.3. Characters and states used in the phenetic analyses.

57 Figure 3.4. UPGMA phenogram using average taxonomie distance. “Humilis” contains 33

OTU’s, “propinqua” 10, “parviflora” 62, “gabrielae” 6, and “taltalina” 2.

58 humilis

c |-CZj propinqua

parviflora

gabrielae □ taltalina I I I I I I I 1-----1 I I I I I I I ' T - T - I 1 2 .4 8 1.89 130 0.70 0 .1 1 Average taxonomic distance Figure 3.4 59 Character PC 1 (40.97%) PC 2 (16.23%) PC 3 (8.17%)

Floral tube length 0.6985 0.6314 0.0849 Floral tube lengthrwidth -0.6950 0.5114 0.0570 Sepal veins 0.3111 -0.6644 0.1791 Pedicel length -0.1346 0.3704 0.3159 Leaf stipule 0.6938 -0.2949 0.1460 Bract stipule 0.8620 -0.2426 0.0152 Capsular mucron -0.7632 0.3244 -0.2602 Floral tube shape 0.8824 -0.3466 -0.0190 Sepal length 0.8823 0.3649 0.0251 Sepal vestiture -0.2138 -0.0344 0.7283 Petal length 0.8976 0.3537 -0.0396 Petal apex vestiture -0.1977 -0.0765 0.6514 Androgynophore length 0.6233 0.6258 -0.0330 Androgynophore vestiture 0.5285 0.5118 0.1761 Inflorescence type -0.5068 0.2246 0.2830 Lower leaf state -0.5396 0.1307 -0.0735

Figure 3.5. Eigenvector coefficients for Principal Components Analysis. Character loading values are for the fîrst three Principal Component (PC) axes. The percent variation explained by each axis is also presented.

60 "gabrielae" Y "hum ilis" if "parviflora" ■ "propinqua" $ "taltalina** #

ON

PCI

Figure 3.6. Results of the principal components analysis. CHAPTER 4

MOLECULAR PHYLOGENY AND TAXONOMY OF MALESHERBIACEAE

Introduction

Malesherbiaceae are an angiosperm family closely related to Tumeraceae and

Passifloraceae (Fay et al. 1997). The distributions of the three are centered in the neotropics, but, unlike their relatives, the members of Malesherbiaceae are xerophytes restricted to the

Andes mountains and the coastal desert on their western flank in Chile, western Argentina, and Peru (Figure 4.1 ). This desert region covers more degrees of latitude than any other intensely arid desert (Trewartha 1966), and Malesherbiaceae are one of only a few taxa occurring over a large portion of it. The family is therefore ideal for the study of the biogeography of the entire region, which until recently has not been investigated in any detail (Rundel et al. 1991). No studies of the biogeography have been conducted using phytogenies of the taxa.

Malesherbiaceae exhibit a suite of morphological characters reminiscent of both

Tumeraceae and Passifloraceae. The variously colored flowers of Malesherbiaceae are characterized by a long floral tube (probably of axial origin [A. Bernhard, pers. comm.]) from which the perianth and a corona arise, a prominent androgynophore, and a persistent perianth. Unlike Passifloraceae, the three styles are free and widely spaced upon the ovary apex. The seeds of Malesherbiaceae lack the aril characteristic of their relatives. The leaves of the often densely pubescent Malesherbiaceae are almost always ciliate with glandular hairs exuding an oily, sticky fluid. Most species have stipules having up to ten lobes. Recent

62 work on the nature of the unusual floral structures in Passifloraceae (Bernhard 1999) may be augmented with studies of the morphology of Malesherbiaceae, which may lead to a better understanding of floral evolution in these families. In addition to these morphological traits, these largely malodorous perennials produce cyanogenic glycosides; they are unique in that tetraphyllin A and tetraphyllin B are the dominant compounds (Spencer & Seigler

1978).

Malesherbiaceae are little known outside of Chile, where most species occur. The first comprehensive treatment of Malesherbiaceae was a traditional revision by Ricardi (1967). In his treatment, Ricardi recognized 27 species in a single genus, Malesherbia. Recent phenetic analysis of a species complex in the family indicates that three of Ricardi’s species are best treated as varieties (Gengler, chapter 3); in this study, therefore, 24 species will be recognized. Although there appear to be several morphologically cohesive groups of species

In the genus, Ricardi erected no sections or subgenera to categorize this diversity.

The focus of this study is the construction of a robust hypothesis of the phylogeny of

Malesherbiaceae using the internal transcribed spacer regions (ITS) of nrDNA. This phylogeny will form the basis for the recognition of subgeneric categories. It will also be used in future studies of the biogeography of the Pacific and Andean desert region of central South America.

Materials and Methods

Sampling. Leaf material was collected for 39 accessions of Malesherbia representing all 24 species and dried in silica gel (Figure 4.2). When available, more than one population of each species was sampled; more extensive sampling, however, was generally not feasible due to the rarity of many of the species and/or their inaccessibility. Four species could not be located in the field; DNA was extracted from herbarium material of these species as well as additional accessions of six other species previously collected in the field. Fresh leaves of

63 the outgroup taxa, Piriqueta caroliniana and Tumera scabra of Tumeraceae, were collected from greenhouse-grown plants. This family was chosen because Tumeraceae have been implicated as the sister taxon of Malesherbiaceae (Fay et al. 1997) and because Ricardi

(1967) noted that Piriqueta and Malesherbiaceae share some morphological characters, such as a corona and floral tube. Tumera is a neotropical genus closely related to Piriqueta

(Arbo 1995). The only other New World genera in Tumeraceae are Cuban and central

American and are therefore unlikely sister taxa.

DNA Isolation and Sequencing. Depending upon the amount of material available, total

DNA extractions were carried out using a modified large-scale CTAB extraction protocol

(Doyle & Doyle 1987) with subsequent puriHcation on CsCI/ethidium bromide gradients

(Palmer 1982) or a CTAB miniprep protocol after Doyle & Doyle (1987) and Cullings

(1992). Miniprep-isolated DNA was not purifîed by CsCI gradient.

Direct, symmetrical amplifîcation of the ITS regions and intervening 5.8S region was carried out using primers ITS-4 and ITS-5 of White et al. (1990), with ITS-S modified according to Sang et al. (1995). In a few cases, amplifîcation of DNA from herbarium material was particularly difficult, so ITS 1 was amplifîed using primers ITS-2 (White et al.

1990) and the modifîed ITS-5, while ITS 2 was amplifîed using primers ITS-3 (White et al.

1990) and ITS-4. The PCR reaction followed a hot start protocol: the initial cycle was 5 min. at 95C followed by 6 min. at 72C, during which time Taq polymerase was added to each reaction volume and the reactions capped with two drops of mineral oil. The next 30 cycles consisted of 1 min. at 95C, I min. at 50C, and 45 sec. at 72C. The extension segment

(at 72C) was extended by 4 sec. each cycle after the fîrst segment of 45 sec. A final extension of 5 min. at 72C completed the reaction. Amplification products were purified either in agarose gels (Ix TAE) by electrophoresis and subsequent separation from the

64 agarose using glass milk (U.S. Bioclean, Amersham Corp., Arlington Heights, IL) or by centrifugal filtration using Ultrafree-MC filter tubes (Millipore Corp., Bedford, MA).

Recovered, clean amplification products were concentrated for sequencing.

Manual sequencing of the purified, double-stranded ITS amplification products was performed using the Sequenase version 2.0 (Amersham Corp., Arlington Heights, IL) dideoxy chain termination method with forward primers ITS-5 and ITS-3 and reverse primers ITS-2 and ITS-4. The reaction protocol followed that of Sang et al. (1994).

Electrophoresis of the sequences was performed in 6% acrylamide gels using wedge spacers. Gels were run at 1500 mA until bromphenol stain migrated to the end of the gel

(approx. 2.5 hours) and subsequently fixed in 10% acetic acid. They were then dried to 3-

MM Whatman filter paper for 2 hours under vacuum before exposure to Kodak XAR x-ray film for one to seven days. The ITS regions of two species, M. weberbaueri and M. turbinea, were sequenced using an automated sequencer due to manual sequencing difficulties (Molecular Genetics Facility of the University of Georgia Research Services,

Athens, GA). All sequences were submitted to GenBank under accession numbers

AF276315-AF276420.

Data Analyses. Boundaries of the ITS regions were determined using sequences previously published for the coding regions of ISS, 5.8S, and 26S adjacent to the ITS regions (compiled in Torres et al. 1990). Sequences were aligned using Clustal W

(Thompson et al. 1994) with a 10-point gap penalty and a transition preference. Minor

manual adjustments to the computer alignment were made.

Parsimony analysis of the data set utilized NONA 1.6 (Goloboff 1993), Pee-Wee 2.15

(Goloboff 1993), Hennig86 (Farris 1988), and PAUP*4.0b2 for Macintosh (Swofford

1999). In NONA, 100 replicates were performed, holding 60 trees per replicate and using tree bisection-reconnection. Further swapping was not necessary. These commands were

65 also used in Pee-Wee, which optimizes the fît of the characters on the tree by their implied weights. In Hennig86, successive approximations character weighting was performed after conversion to the required numerical format. lUPAC symbols for polymorphisms were converted to unknowns by default for this process. Gaps are treated as missing data in all three programs, and polymorphic sites are accepted by NONA and Pee-Wee.

In PAUP*, a heuristic search was conducted using the tree bisection-reconnection

(TBR) branch-swapping algorithm with MULPARS on. A second analysis was also performed to explore multiple islands. 2500 trees were generated by saving five minimum-

length trees resulting from each of 500 replicates utilizing nearest-neighbor interchanges

(NNI) branch-swapping. This pool of trees was filtered to eliminate all but the shortest trees

(301 steps, excluding uninformative characters). These trees then formed the basis fora

second analysis using TBR branch-swapping. Two separate analyses were performed,

saving 2500 trees the fîrst time and S(X)0 trees the second.

Jackknife, Bremer, and bootstrap support values were computed using Xac (Farris

1996), NONA, and PAUP*, respectively. In the calculation of the jackknife values, 1(XX)0

replicates were performed with branch swapping and five random addition sequences per

replicate. Bremer supports were calculated for the consensus of the most parsimonious trees

from NONA. Because Davis (1995) found that Bremer support values may be inflated if all

the topologies in suboptimal trees have not been searched, and since it is computationally diffîcult to calculate all of these trees, three separate runs were performed to determine if the

results vary as more trees are searched. Unchanging support values across runs would

suggest that most of the topologies have been explored and that the support values are not egregiously overestimated. In these three runs, 2502,5502, and 10502 trees up to five steps

longer were calculated. Bootstraps were calculated from 5000 replicates using “fast” step­

wise addition, which Mort et al. (2000) found to yield bootstraps comparable to a heuristic

search.

66 Maximum likelihood was also employed using PAUP*. To decrease computer run times, I removed duplicate accessions (with the exception of the two sequences of Af. deserticola and M. lanceolata and ) sequences which were identical to others included in the matrix (Figure 4.2). (A parsimony analysis using NONA was conducted on the reduced data set to ensure that the topology obtained from parsimony remained stable.) The data were analyzed using the HKY85 model (Hasegawa et al. 1985) with an empirically determined transition to transversion ratio of 1.525. The gamma distribution of the shape parameter was set to 1.375. Ten random addition sequences were performed.

A second data set was constructed in which all sites within the ingroup with ambiguous alignments (21 total) were deleted to determine what their effects are on the complete data set. Parsimony analyses using NONA and Pee-Wee were performed and Xac and Bremer support values calculated as above. This data set was also used to search for a maximum likelihood tree. The transidon to transversion rado was empirically determined to be 1.579 and the gamma distribution was set to 1.252.

Since the branch between the ingroup and the outgroup is very long ( 104 mutations using ACCTRAN opdmizadons on the complete data set in PAUP*) and the root falls on one of the longer branches in the ingroup, long branch attracdon could be a source of potendal phylogenetic problems (Felsenstein 1978, Hendy & Penny 1989, Huelsenbeck

1997). Siddall & Whiting (1999) noted that the eliminadon of either of the long branches should have no effect on the placement of the other long branch if these branches are not attracdng each other. Therefore, sequences of the outgroup or the (A/, linearifolia - Af. paniculata) lineage from the complete and the unambiguous data sets were removed, with subsequent data analysis using NONA and Pee-Wee.

67 Results

Sequence Data. The ITS regions of the 24 species recognized by Gengler (unpublished data) were successfully amplified and sequenced. The aligned sequences were 466 bp long with 175 informative characters. Absolute lengths of the combined ITS 1 and ITS 2 regions ranged from 420 bp in the outgroups, Tumera scabra and Piriqueta caroliniana, to 450 bp in Af. ardens (KMG 185), Af. campanulata (Ricardi, Marticorena, & Matthei 1748), Af. densiflora, Af. lirana var. subglabrifolia, and Af. rugosa (KMG 37). The shortest sequence within Malesherbiaceae was that of Af. ardens (KMG 184), which had 444 bp. Within the ingroup, indels of one to three bases were scattered. A few indels in the outgroups were longer; in addition to scattered indels of one to four bases, there was one gap of 11 bases in the outgroup not found in the ingroup. G+C content ranged from 52.2% in Af. ardens

(KMG 185) to 61.9% in Tumera scabra.

Sequence divergences in Malesherbiaceae calculated with MEGA 1.0 (Kumer et al.

1993) using the Kimura 2-parameter model were greatest between all accessions of Af. tenuifolia and Af. paniculata ( 10.00%). Divergences were zero between Af. humilis var. parviflora (Stuessy 9892) and three other accessions of that variety (KMG 109, 110, &

190); Af. tocopillana and Af. ardens (KMG 187), Af. angustisecta and Af. arequipensis

(KMG 58, Dillon et al. 4793); and among Af. scarlatiflora (all accessions), Af. splendens, and Af. weberbaueri. Divergences between the outgroup and ingroup ranged from 41.99% to 54.02%; divergence between T. scabra and P. caroliniana was 29.01%. Despite these differences, the outgroup and ingroup sequences could be easily aligned by eye for approximately half the sequence.

Phylogeny. Malesherbiaceae were strongly supported as a monophyletic family in all phylogenetic analyses. Parsimony analysis of informative characters of the equally weighted data matrix using NONA yielded four shortest trees of 301 steps with a consistency index

68 of 76 and retention index of 92 (Figure 4.3) These trees differ in two respects only: the (M. ardens - M. tocopillana) clade is sister to the (Af. auristipulata Af. - turbinea) clade in two trees, whereas in other two trees, the (Af. ardens - Af. tocopillana) clade is sister to the (Af. arequipensis Af - tenuifolia) clade. The other inconsistency lies in (Af. scarlatiflora Af.- weberbaueri) clade. The two accessions of Af. tubulosa are either most closely related and sister to the remaining taxa in the clade, or one of the accessions falls within the polytomy containing Af. scarlatiflora, Af. splendens, and Af. weberbaueri.

The heuristic search in PAUP* yielded 50000 most parsimonious trees of 301 steps.

The great discrepancy between the results of the two programs is likely due to differences in

treatment of polymorphisms, clades appearing in only one optimization, and methodologies

regarding the reporting of unsupported clades (Goloboff 1993-1997 demonstration

program documentation). The consensus of the 50000 trees was identical to the consensus

tree of the four found by NONA (Figure 4.3). The consensus trees from the island analyses

using 2500 and 5000 minimum-Iength trees were also the same as the NONA consensus.

Therefore, only the four parsimony trees resulting from the NONA analyses will be further

considered.

Two analyses were performed using weighted data. The program Pee-Wee generated

two trees of greatest fit (total fit = 1828.9). One (301 steps) was identical to the most

parsimonious trees in which the (Af. ardens - Af. tocopillana) clade is sister to the (Af.

arequipensis - Af. tenuifolia) clade (Figure 4.3). The other Pee-Wee tree (302 steps)

differed from the first in that the two accessions of Af. lactea were basal to those of Af. fasciculata.

Successive approximations character weighting was also employed. This method

generated 384 trees. Like the Pee-Wee trees, the strict consensus tree also placed the (Af.

69 ardens - M. tocopillana) clade sister to the (Af. arequipensis Af. - tenuifolia) clade. The (Af. linearifolia - Af. paniculata) clade collapsed in this tree. The remainder of the tree was identical to the parsimony trees (Figure 4.3).

Bremer support values for the trees were constant across the three runs, although the number of trees investigated was almost quadrupled from 2502 to 10502. This suggests that the number of different kinds of topologies in the set of trees up to five steps longer is relatively small, indicating that the Bremer support values were not grossly overestimated.

Parsimony analysis of the data set in which ambiguous sites were removed resulted in some loss of resolution and changed supports for some branches in comparison to the analyses of the complete data set (Figure 4.4). The tree resulting from the maximum likelihood analysis differs little from the parsimony trees (Figure 4.5).

In the consensus tree of the parsimony analyses, Malesherbiaceae were broken into several well supported clades (clades A , B, D, and E; Figure 4.3) and one clade with weak support (clade C, Figure 4.3). Relationships among species were poorly resolved. The terminal clade (clade E), composed exclusively of Peruvian and extreme northern Chilean species, was further subdivided into two clades. Clade El, with four species, was strongly supported as monophyletic, but its sister clade, E2, was weakly supported. This second clade was composed of Af. haemantha and three small clades; the relationships among these components remain unresolved. This clade disappeared in the parsimony and maximum likelihood analyses of the completely unambiguous data (Figures 4.3 & 4.4). In the maximum likelihood analyses of the unambiguous and complete data sets, the (Af. ardens -

Af. tocopillana) clade was sister to the (Af. auristipulata - M. turbinea) clade (Figure 4.5).

The remainder of the species are strictly Chilean. Weakly supported in a sister position to the Peruvian clade is a relatively large, strongly supported clade of seven species (clade D,

Figure 4.3). The larger of the two lineages in clade D (clade D2) showed Af. campanulata and Af. lanceolata to be sister taxa, but the relationships among this lineage, Af. lirana, Af.

70 obtusa, and M. rugosa were unresolved. Some loss of resolution in this clade resulted when the ambiguous sites were removed (Figure 4.4). The second lineage (clade Dl) contains only M. deserticola and M. densiflora.

The sister clade (clade C) to the preceding taxa contains only M. fasciculata and Af. lactea, which are very weakly supported as a monophyletic group by two homoplasious characters in the NONA analysis and in one tree of the Pee-Wee analysis (Figure 4.3). In the maximum likelihood analyses, M. lactea broke from M. fasciculata and migrated to the base of the tree, attaching to the rest of the ingroup between M. linearifolia, M. paniculata, and M. humilis and the remaining taxa (Figure 4.5).

All analyses agree that Af. humilis and the clade containing Af. linearifolia and Af. paniculata are located near the base of the tree (Figures 4.3,4.4, & 4.5). In the parsimony analyses, Af. humilis was sister to clade C+D+E and the (Af. linearifolia Af. - paniculata) lineage was basal to the rest of the family.

NONA and Pee-Wee analyses of the complete and unambiguous data sets with the elimination of the outgroups each yielded four trees identical to those recovered by the original parsimony analysis of the complete data set. When the (Af. linearifolia - Af. paniculata) lineage was removed, however, topological rearrangements resulted: using the complete data set, M. fasciculata moved to the base of the ingroup, and the remaining major lineages and Af. lactea formed various relationships with each other. Using the unambiguous data set, Af. fasciculata was again basal for the ingroup, with Af. lactea falling as the sister to the remaining taxa.

Discussion

As implied by the family’s unique combination of morphological characteristics and in accordance with other workers’ conclusions (Ricardi 1967, Takhtajan 1980, Cronquist

1981), Malesherbiaceae are monophyletic. Fay et al. (1997), using rbcL sequences, showed

71 Malesherbiaceae to be most closely related to Tumeraceae, with Passifloraceae as their sister. However, neither that study nor this FTS-based study address the relationships among the genera of Tumeraceae and Malesherbiaceae. It is unknown if Malesherbiaceae and Tumeraceae split before Tumeraceae radiated (i.e. Tumeraceae are monophyletic) or if

Malesherbiaceae constitute a highly modified branch of Tumeraceae. Given Tumeraceae s distribution in South America and Africa, however, it seems biogeographically unlikely that both the families are monophyletic. Raven & Axelrod (1974) contend that Tumeraceae became established on both continents during the Paleocene (65-45 mya), and

Malesherbiaceae are almost certainly not that ancient. Further investigation using a gene evolving at a rate intermediate to rbcL and ITS, such as ndh¥ or matK, and sampling of all the genera in Tumeraceae and Malesherbiaceae may better elucidate the precise nature of the relationship between the two families.

The placement of the root of Malesherbiaceae impacts the formation of biogeographical inferences for the family. The parsimony analyses show the root to fall between the (M. linearÿôlia - M. paniculata) lineage and the rest of the family. The outgroup, however, is defined by a large number of autapomorphies, and it could be argued that the placement of the root is the result of long branch attraction (Felsenstein 1978, Hendy & Penny 1989,

Huelsenbeck 1997). Siddall & Whiting ( 1999) noted that a pair of long branches cannot attract each other in a spurious topology if one of those branches were absent. The stability of the ingroup topology to the removal of the outgroup indicates that the placement of the

(Af. linearifolia - Af. paniculata) branch relative to other ingroup taxa is not misleading.

Removal of the (Af. linearifolia Af. - paniculata) branch, on the other hand, destabilized the relationships among the major lineages enjoying good support in the parsimony analysis, broke up the (Af. fasciculata - Af. lactea) lineage, and placed the root between Af. fasciculata and the other taxa. This topological rearrangement may be the result of the outgroup branch attracting to the rather long M. fasciculata branch and/or the lack of data

72 concerning the relationships among the major lineages. Alternatively, the results of sampling

error (by ignoring one of the major family lineages) may be influencing the new topology;

Lecointre et al. (1993) investigated the effects of taxon sampling and found incomplete

samples often yield misleading phylogenetic hypotheses. To investigate these possibilities

more fully using parsimony analysis, a data set which resolves the relationships among the

major lineages of Malesherbiaceae with greater confidence is needed.

Maximum likelihood has been suggested to be less susceptible to long branch attraction

(Huelsenbeck 1997,1998; but see Siddall & Whiting 1999). Analyses of the reduced data

set with representatives of all the major lineages and the outgroup also suggest that the (Af. fasciculata Af.- lactea ) clade is a source of topological instability. In these analysis, Af.

lactea (rather than M. fasciculata, as in the parsimony analysis performed without the

purported basal lineage of Malesherbia) was pulled to the base of the ingroup (Figure 4.5).

Maximum likelihood and parsimony methods, then, are in disagreement as to the exact

placement of the root. All analyses agree, however, that the root lies among the Chilean taxa

(clades A, B, C, and D) and involves either the (Af. linearifolia Af. - paniculata) or (Af.

fasciculata Af.- lactea) clade.

Although there are few morphological characters supporting the relationships among the

clades at the base of the ITS phylogeny, the monophyly of four of the Hve major clades

found in the parsimony analysis is strongly supported by morphology (Figure 4.3). These

four clades, clades A, B, D, and E, are also well supported by the ITS data. The Hrst of these

clades (clade A) contains Af. linearifolia and Af. paniculata. These species have intensely

blue or purple petals and sepals that are very large compared to the relatively inconspicuous

floral tube. The flowers are in panicles atop long, slender branches bearing shortly

pubescent leaves. These species are native to central mediterranean and semi-arid Chile.

Clade B (Figure 4.3) contains only Af. humilis. This species is composed of fîve

varieties (Gengler, unpublished data) distributed over 1T in latitude along the coast and in

73 the pre-Andes of Chile, with some disjunct populations reported from Province Neuquén,

Argentina (37°30’S) (Hunziker & Espinar 1967). These plants, which do not grow taller than 30 cm, produce multitudes of tiny flowers with white or pale lavender or blue perianth parts and often greenish floral tubes. The pedicels are bractless. Stipules may be obsolete, and the leaves and flowers are pilose. Maximum likelihood analyses indicate that this species forms a monophyletic clade with Af. linearifolia and Af. paniculata.

Morphologically, Af. humilis is so divergent from the other species that there is no evidence from morphology that the two lineages are very closely related.

Species adapted to montane and very dry desert habitats comprise the well supported clade D (Figure 4.3). This lineage is differentiated from others by white flowers tinted with clear yellow in the throat (with the exception of Af. obtusa, which has pale blue flowers) and a pilose androgynophore (the plesiomorphic condition for the family is glabrous). The seven species of this clade are separated into two lineages. Malesherbia deserticola and Af. densiflora (clade Dl) are distinguished by having unequal, asymmetrically shaped stipules or stipules largely fused with the leaf blade and petiole. In clade D2, the stipules of the flve species are obsolete or reduced to small flaps of tissue. The coronas of this clade’s species are very short or reduced to a thick band of tissue at the base of the perianth.

The last clade of Chilean species in the parsimony tree contains M. fasciculata and Af. lactea (clade C, Figure 4.3). Molecular data do not strongly support their close relationship or their placement in midst of the phylogeny. The morphologies of these species are very different from other species of Malesherbia. Unlike other Malesherbia, they have entire- margined leaves lacking stipules, and usually these leaves lack prominent glandular hairs. A purple hue often tints their white flowers. Both species have long, slender branches relative to plant size, although those of Af. lactea are sinuous and largely underground. These species are heavily covered with white, tomentose hairs. Outside of these similarities, Af. lactea and M. fasciculata share few features. The intemodes of the slender branches of Af.

74 fasciculata are very long, and the rather small leaves are elliptic to narrowly lanceolate. The ends of the branches bear small flowers arranged in racemes or dichasia which are then compressed into compact globes. The anthers and pollen are blue. This species is native to the pre-Andean semi-arid region of Chile. Malesherbia lactea, in contrast, has very short intemodes and spatulate leaves, and the flowers are either solitary or in few-flowered racemes. The anthers and pollen are yellow. Its native habitat lies above 3500 m in the

Andes of Chile and Argentina, running from 24°12’S to 28°30’S. If these species are sister species, the lack of morphological and molecular evidence is probably due to rapid divergence shortly after the evolution of the lineage. Habitat differentiation may have led to this rapid divergence.

The family’s final major lineage (clade E, Figure 4.3) is largely Peruvian, with three species (M. tenuifolia, M. tocopillana, and M. auristipulata) in three separate clades native to extreme northern Chile. Innovations in this Peruvian clade are the presence of a thickened band of androecial tissue where the filaments become free at the apex of the androgynophore and the appearance of red and orange pigmentation. Their ovaries are cylindrical rather than globose, and the valves of their capsules, which extend beyond and often tear open the perianth and corona at maturity, are expanded. The long, tubular shape, reddish hues, and exerted stamens and styles of the flowers suggest that many are hummingbird pollinated.

Two clades comprise the Peruvian lineage. Clade El (Figure 4.3) is composed of species having long, tubular flowers with constricted throats and petals with pilose abaxial and adaxial surfaces. The flowers are arranged into dense racemes held erect on sturdy stems. Three of the four species are tall shrubs. The ITS data failed to resolve the relationships among the four species, although one parsimony tree showed Af. tubulosa to be basal to the other three species.

75 The original parsimony analyses support monophyly of clade E2 (Figure 4.3) weakly supported in the original parsimony analyses, but analyses of the unambiguous data set and the lack of morphological features disagree. Low confidence for monophyly of this clade is best viewed as strong support for inference of a rapid radiation culminating in a diversity of

Peruvian lineages united by a suite of unique morphological characters.

Malesherbia auristipulata and M. turbinea comprise a clade of plants with blood-red flowers having black anthers and black glandular hairs lining the apices of the petals and sepals. Their flowers are easily dislodged from the racemes. The dark green leaves, which have broad, lobed stipules, are also lined with black glandular hairs. Unlike most Peruvian species, Af. auristipulata and Af. turbinea have floral tubes which are wider at the throat than in the middle of the tube. Access to the throat of the flower, however, is still restricted by the corona, which is narrower than the floral tube throat and which forms a glossy sheath exceeding the length of the perianth. This constriction of the throat of the flower by a narrow corona in this lineage and by a narrow perianth in other Pemvian species suggests that it may be functionally important, perhaps by promoting of pollinator speciflcity or restricting access to the flower by some nectar robbers.

Of unknown affinities is Af. haemantha, which closely resembles Af. auristipulata and

Af. turbinea with its blood-red floral tube, black pollen and anthers, and glandular hairs lining the leaves and perianth (Figure 4.6). Its corona is also relatively long, although approximately equal to the perianth, rather than greatly exceeding it. Unlike Af. auristipulata and Af. turbinea, Af. haemantha's leaves are deeply pinnatifid and subtended by lanceolate stipules. The unusual shared traits suggest that the three species are closely related, but the lack of molecular evidence for such a relationship may indicate that Af haemantha split from the ancestor of Af. auristipulata and Af. turbinea very early in the

76 evolutionary history of the lineage. Malesherbia auristipulata and M. turbinea are found near each other on opposite sides of the Chilean-Peruvian political border; Af. haemantha is a narrow endemic living about 500 km to the northwest.

The most molecularly isolated clade in the family is composed of Af. angustisecta, Af. arequipensis, and Af. tenuifolia. These species have highly dissected, revolute-margined leaves and are covered with white, tomentose hairs. When present, the stipules are also highly dissected. These plants are small bushes bearing their flowers in panicles, the plesiomorphic condition of the family. The three species are also unusual in that they lack the intense floral colors of other Peruvian species. Malesherbia arequipensis and Af. angustisecta have greenish-white flowers, although the upper portion of the style and the stigma of Af. angustisecta are pink. The flowers of Af. tenuifolia are a pale pink-red when exposed to sunlight; parts of flowers lying on the ground or in the shade lack pigmentation.

Malesherbia arequipensis, in addition, has lost the ring of androecial tissue at the apex of the androgynophore and the long, tubular floral tube characteristic of all the other members of the Peruvian clade.

The last lineage in clade E2 contains Af. ardens and Af. tocopillana, which bear their flowers in dense, long racemes held on sturdy, semidecumbent stems. The corona is deeply and sharply toothed, and the leaves are lobed and covered with glandular hairs. The characteristics shared by these species led Rundel et al. (1991) to postulate that they are closely related. Ricardi (1967) suggested that Af. tocopillana, a very rare endemic to

Tocopilla, Chile, is more closely allied to the shrubby species in clade El. The ITS and morphological data clearly support Rundel et al.’s (1991) assessment of the affinities of Af. tocopillana.

There is little molecular resolution of relationships among the lineages in clade El.

Morphological data are also equivocal. The Af. ardens clade shares with the Af. auristipulata clade a reddish corona (although the color is blood red in the Af. auristipulata

77 clade and more orangish-red in the Af. ardens clade) and the presence of a glandular hair at the apex of the petal. Supporting the closer relationship of the Af. ardens clade to the Af. angustisecta clade is the presence of short sepals relative to the floral tube, a floral tube more than twice as long as wide, and the absence of a glandular hair at the apex of the sepal.

Further evidence is necessary to resolve the relationships among the clades in this diverse lineage.

The phylogeny presented here of Malesherbiaceae suggests that the family has undergone several periods of rapid radiation, some which have been recent. The earliest radiation appears to have given rise to the major lineages of species, judging from the lack of strong support for the relationships along the backbone of the tree. The absence of morphological characters supporting these relationships also indicates that those ancestors split soon after their evolution.

Low levels of sequence divergence have been cited as evidence of recent and rapid

radiations (Carlquist 1974, Sang et al. 1994, Francisco-Ortega et al. 1996, Hodges 1997,

Baldwin 1997). Clade D in Malesherbiaceae appears to have experienced such a recent radiation. The lack of molecular support for the relationships among Af. rugosa, Af. lirana,

Af. obtusa, and the (Af. lanceolata Af. - campanulata) lineage indicates that spéciation

occurred very rapidly. This clade of montane and very arid desert species almost certainly

radiated relatively recently as well. Few or no autapomorphies define the species, indicating

that there has been insufflcient time for these species to accumulate changes in their ITS

sequences.

Similar trends are evident in the Peruvian clade, which apparently speciated rapidly twice

in its history. The very low support for the monophyly of clade E2 indicates that the

Peruvian clade rapidly evolved into today’s flve diverse lineages. Because 12 mutations and

distinctive morphological traits support the clade, this initial radiation must have occurred

quite some time after arrival in Peru. Modem species in both subclades appear to be young;

78 of the twelve species In the Peruvian clade, ten are defîned by only one autapomorphy or none at all. As in Chile, the paucity of autapomorphies defining species which are morphologically distinct indicates that they have not been separated long enough to develop rrs autapomorphies. Subclade El has been most affected by the conditions promoting spéciation; spéciation has left few clues to the relationships among its four morphologically distinct species.

Taxonomy

Although the relationships among the five major lineages are not strongly supported, the lineages, with the exception of the (A/, fasciculata - M. lactea) lineage, are themselves well defined by both morphological and molecular data. The array of morphological diversity in

Malesherbia should be recognized taxonomically. To give the well supported lineages

(clades A, B, D, and E) formal taxonomic names forces the naming of the poorly supported

{M. fasciculata - M. lactea) lineage, which could later prove to be paraphyletic. However, the benefit of creating taxonomic categories—nomenclatural recognition of the morphological diversity in the family—outweighs the disadvantage of potential future taxonomic changes for two species. Recognition at the level of subgenus is appropriate in this situation. Not only does this avoid the creation of many small genera (and therefore new binomials), but also the possibility of further binomial changes if future work establishes the phylogenetic positions of M. fasciculata and M . lactea.

79 Key to the Subgenera of Malesherbia 1. Petals dark blue or purple, petals and sepals as long or longer than floral tube; plants

erect 1. suhg. Cyanpetala r . Petals white, yellow, red, orange, or pale blue to violet; floral tube longer than perianth

parts; plants erect or semidecumbent

2. Leaf margins completely entire; leaves exstipulate; leaves and stem white-tomentose;

flowers solitary or in few-flowered or many-tlowered and globose racemes

2. sxxbg. Albitomenta

2.' Leaf margins other; leaf and stem pubescence hispid, pilose, or velutinous; if

tomentose, leaves deeply pinnatisect or flowers tubular, campanulate, or obconical

3. Shrublets ( ^ 0 cm); floral tube funnelform and tightly constricted around the

androgynophore ..... 3. subg. Parvistella

3’. Shrubs; floral tube campanulate, tubular, obconical, or broadly funnelform and

expanded around androgynophore

4. Floral tube red, orange, or yellow; if white or greenish, leaves also finely

pinnatisect; androgynophore ± glabrous, usually bearing a ring of thickened

tissue at the base of the ovary; ovary cylindrical or conical

4. subg. Malesherbia

4’. Floral tube white, greenish, or sky blue; androgynophore pilose and lacking

ring of thickened tissue; ovary globose ..... 5. subg. Xeromontana

80 1. Malesherbia subg. CyanpeUüa Gengler, subg. nov. Frutices 25-150 cm alta ramis longis gracilibus erectis caule brevi exorientibus; foliis lanceolatis vel pinnatisectis, inferioribus 23-120 mm longis, stipulis saepe lobatis; tube floris late infundibuliformi colore inconspicuo; perianthio tubum floris aequantia vel longiote, atrocyaneo vel atropurpureo. — Typus: Af. linearifolia (Cav.) Pers.

Shrubs 25-150 cm tall with long, slender, erect branches arising from a short stem.

Leaves lanceolate or pinnatisect, 23-120 mm long, stipules often lobed. Floral tube broadly funnelform, inconspicuously colored; perianth equal to or longer than the floral tube, dark blue or dark purple.

Native to Chile (Regions ID, IV, V, VI, VII, and Metropolitana); in the north of its range, it is largely found in the Andean foothills. In the south, the subgenus ranges from the coast to the foothills. Ricardi (1967) notes with some doubt the claim thatAf. linearifolia is native to Argentina; Hunziker & Espinar (1967) found no conflrmation for its presence in that country.

Species included. — Malesherbia linearifolia (Cav.) Pers., Af. paniculata D. Don.

2. Malesherbia subg. Albitomenta Gengler, subg. nov. Perennes tectus trichomatibus albis tomentosis ramis graciles; foliorum margine integro trichomatibus glandulosis vulgo deflcientibus; tubo floris minus quam 1.2 cm, albo interdum tincto violaceo. —

Typus: Af. lactea Phil.

Perennials covered with white tomentose hairs with branches slender; leaf margins entire, glandular hairs usually lacking; floral tubes less than 1.2 cm long, white

sometimes violet-tinted.

81 Native to above 3500 m in the Andes of Chile (Regions Q, m, and IV) and adjacent

Argentina (Provinces La Rioja and San Juan). Also found in the semi-arid, pre-Andean foothills of Chilean Regions IV, V, VI, and Metropolitana.

Species included. — Malesherbia lactea Phil., M. fasciculata D. Don.

3. Malesherbia subg. Parvistella Gengler, subg. nov. Fruticuli erecti vulgo ramosissimi ad 30 cm alta; foliis pinnatisectis, lanceolatis vel oblongis, pilosis, margine ciliato trichomatibus glandulosis, inferioribus vulgo marcidis; stipulis interdum diricientibus; pedicellis ebracteatis; tubo floris 4-12 mm longis, 0.5-2.0 mm diametro

(2.0-6.5 mm circumferential, infundibuliformi, virello, dense piloso, fauce nunquam tincta flava; perianthio albo, caelesti vel violaceo; androgynophoro raro piloso. — Typus:

M. humilis Poepp.

Shrublets erect, often highly branched, reaching 30 cm in height; leaves pinnatisect, lanceolate, or oblong, pilose, margins ciliate with glandular hairs, lower leaves often withered, stipules sometimes lacking; pedicels ebracteate; floral tube 4-12 mm long,

0.5-2.0 mm in diameter (2.0-6.5 mm in circumference), funnelform, greenish, densely pilose, throat never tinted with yellow; perianth white, sky blue, or violet; androgynophore rarely pilose.

Native to the coast and the pre-Andes of Chile from Santiago to Guatacondo

(20“56’S) (Regions II, HI, IV, V, and Metropolitana. Also found in Province Neuquén

(37°30’S) of Argentina.

Species included. — Malesherbia humilis Poepp.

4. Malesherbia subg. Malesherbia Gengler, subg. nov. Frutices ramis erectis vel semidecumbentibus; foliis lanceolatis, oblanceolatis, oblongo-ovatis, vel profunde pinnatisectis, dense pubescentibus; tubo floris tubulari vel obconico, auieo, rubello,

82 auriantico vel interdum aibido vel viridi pallido; corona perianthium vulgo superantia vel eo aequantia; androgynophoro vulgo tumido ad basim ovarii, ovario cylindrico vel conico. — Typus: M. tubulosa (Cav.) J. St. Hil.

Shrubs with branches erect or semidecumbent; leaves lanceolate, oblanceolate, oblongo-ovate, or deeply pinnatisect, densely pubescent; floral tube tubular or obconical, yellow, red, orange, or sometimes white or pale green; corona usually exceeding or equaling perianth; androgynophore usually swollen at the base of the ovary, ovary cylindrical or conical.

Native to the arid, deep inter-andean valleys, broad valleys, coast, and Andean foothills of Peru (Departments Lima, Huancavélica, Ayacucho, Junm, Ica, Arequipa,

Moquegua, and Tacna). Also found in the scattered dry river valleys of the Atacama

Desert of northern Chile and along the coast near Tocopilla, Chile (Regions I and II).

Species included. — Malesherbia angustisecta Harms, M. ardens Macbr., M. arequipensis Ric., M. auristipulata Ric., M. haemantha Harms, M. scarlatiflora Gilg,

M. splendens Ric., M. tenuifolia D. Don, M. tocopillana Ric., M. tubulosa (Cav.) J. St.

Hil., M. turbinea Macbr., M. weberbaueri Gilg.

S. Malesherbia subg. Xeromontana Gengler, subg. nov. Frutices ramosissimi erecti vel semidecumbentes; foliis dense pubescentibus; stipulis deficientibus vel redactis vel anisomorphis; tubo floris infundibuliformi vel campanulato, fauce vulgo tincta flava; corona interdum diminuta instar porcae; perianthio albo vel raro caelesti, androgynophoro piloso. — Typus: M. densiflora Phil.

Shrubs, many-branched and erect or semidecumbent; leaves densely pubescent; stipules lacking, reduced, or asymmetrically shaped; floral tube funnelform or campanulate with throat often tinted yellow; corona sometimes reduced to a ridge; perianth white or rarely sky blue; androgynophore pilose.

83 Distributed in montane and very dry desert habitats of Chile (Regions n, HI, IV, V, and Metropolitana) and extreme western Argentina (Provinces La Rioja, Mendoza,

Neuquén, and San Juan).

Species included. — Malesherbia campanulata Ric., M. densiflora Phil., M. deserticola Phil., M. lanceolata Ric., Af. lirana Gay, M. obtusa Phil., M. rugosa Gay.

84 Literature Cited

Arbo, M.M. 1995. Tumeraceae Parte I. Piriqueta. Flora Neotropica. Monograph 67. The New York Botanical Garden, Bronx, New York.

Baldwin, B.C. 1997. Adaptive radiation of the Hawaiian silverswoid alliance: congruence and conflict of phylogenetic evidence from molecular and non-molecular investigations. Pp. 103-128. In Molecular evolution and adaptive radiation, T.J. Givnish & K.J. Sytsma, eds. Cambridge University Press, Cambridge, UK.

Bernhard, A. 1999. Flower structure, development, and systematics in Passifloraceae and in Abatia (Flacourtiaceae). International Journal of Plant Science 160: 135-150.

Carlquist, S. 1974. Island biology. Columbia University Press, New York, NY.

Cronquist, A. 1981. An integrated system of classification of flowering plants. Columbia University Press, New York, NY.

Callings, K.W. 1992. Design and testing of a plant-specific PCR primer for ecological and evolutionary studies. Molecular Ecology 1:233-240

Davis, J.I. 1995. A phylogenetic structure for the Monocotyledons, as inferred from chloroplast DNA restriction site variation, and a comparison of measures of clade support. Systematic Botany 20:503-527.

Doyle, J.J. & J.L. Doyle. 1987. A rapid DNA isolation procedure for small quantities of fresh leaf material. Phytochemical Bulletin 19: 11-15.

Farris, J.S. 1988. Hennig86 version 1.5. MS-DOS program, Port Jefferson, New York.

------1996. Parsimony jackknifing outperforms neighbor-joining. Cladistics 12: 99-124.

Fay, M.F., S.M. Swensen & M.W. Chase. 1997. Taxonomic affinities of Medusagyne oppositifolia (Medusagynaceae). Kew Bulletin 52: 111-120.

Felsenstein, J. 1978. Cases in which parsimony or compatibility methods will be positively misleading. Systematic Zoology 27:401-410.

Francisco-Ortega, J., R.K. Jansen, & A. Santos-Guerra. 1996. Chloroplast DNA evidence of colonization, adaptive radiation, and hybridization in the evolution of the Macaronesian flora. Proceedings of the National Academy of Science USA 93:4085-4090.

Goloboff, P.A. 1993. Pee-Wee version 2.15. MS-DOS program. Published by the author, San Miguel de Tucumân, Argentina.

------. 1993. NONA, version 1.6. MS-DOS program. Published by the author, San Miguel de Tucumân, Argentina.

Hasegawa, M., H. Kishino, & T. Yano. 1985. Dating of the huroan-ape splitting by a molecular clock of mitochondrial DNA. Journal of Molecular Evolution 22: 160-174.

85 Hendy, M.D. & D. Penny. 1989. A framework for the quantitative study of evolutionary trees. Systematic Zoology 38:297-309.

Hodges, S.A. 1997. Rapid radiation due to a key innovation in columbines (Ranunculaceae: Aquilegia) pp. 391-406. In Molecular evolution and adaptive radiation, T.J. Givnish & KJ. Sytsma, eds. Cambridge University Press, Cambridge, UK.

Huelsenbeck, J.P. 1997. Is the Felsenstein zone a fly trap? Systematic Biology 46: 69-74.

------. 1998. Systematic bias in phylogenetic analysis: is the Strepsiptera problem solved? Systematic Biology 47:519-537.

Hunziker, A.T. & L A. Espinar. 1967. Nota aclaratoria sobre las Malesherbiaceae argentinas y una clave para su identificaciôn. Kurtziana 4:83-86.

Kumer, S., K. Tamura, & M. Nei. 1993. MEGA: Molecular evolutionary genetics analysis, version 1.0. The Pennsylvania State University, University Park, PA.

Lecointre, G., H. Philippe, H.L.V. Le, & H. LeGuyader. 1993. Species sampling has a major impact on phylogenetic inference. Molecular Phylogenetics and Evolution 2: 205-224.

Mort, M E., P.S. Soltis, D.E. Soltis & M L. Mabry. 2000. Comparison of three methods for estimating internal support on phylogenetic trees. Systematic Biology 49: 160-170.

Palmer, J.D. 1982. Isolation and structural analysis of chloroplast DNA. Nucleic Acids Research 10: 167-186.

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

Ricardi, M. 1967. Revisidn taxonômica de las Malesherbiaceas. Gayana Botdnica. No. 16: 1-139.

Rundel, P.W., M.O. Dillon, B. Palma, H.A. Mooney, Si,. Gulmon, & J R. Ehleringer. 1991. The phytogeography and ecology of the coastal Atacama and Peruvian deserts. Aliso 13: 1-49.

Sang, T., D.J. Crawford, S.-C. Kim, & T.F. Stuessy. 1994. Radiation of the endemic genus Dendroseris (Asteraceae) on the Juan Fernandez Islands: evidence from sequences of the ITS regions of nuclear ribosomal DNA. American Journal of Botany 81:1494-1501.

------D.J. Crawford, & T.F. Stuessy. 1995. ITS sequences and the phylogeny of the genus Robinsonia (Asteraceae). Systematic Botany 20:55-64.

Siddall, ME. & M.F. Whiting. 1999. Long-branch abstractions. Cladistics 15:9-24.

Spencer, K.C. & D.S. Seigler. 1978. Cyanogenic glycosides of Malesherbia. Biochemical Systematics and Ecology 13:23-24.

Swofford, DX. 1999. PAUP*. Phylogenetic Analysis Using Parsimony (* and Other Methods). Version 4. Sinauer Associates, Sunderland, MA.

86 Takhtajan, A.L. 1980. Outline of the classification of flowering plants (Magnoliophyta). The Botanical Review 46(3): 225-359.

Thompson, J.D., D.G. Higgins, & T.J. Gibson. 1994. Clustal W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position specific gap penalties and weight matrix choice. Nucleic Acids Research 22:4673-4680.

Trewartha, G.T. 1966. The Earth’s problem climates. The University of Wisconsin Press, Madison, Wisconsin.

Torres, R j\., M. Ganal, & V. Hemleben. 1990. GC balance in the internal transcribed spacers ITS I and ITS 2 of nuclear ribosomal RNA genes. Journal of Molecular Evolution 30: 170-181.

White, T.J., T. Bruns, S. Lee, & J. Taylor. 1990. Amplification and direct sequencing of fungal ribosomal RNA genes for phylogenetics. Pp. 315-322. In PCR protocols: A guide to methods and amplifications. M. Innis, D. Gelfand, J. Sninsky, and T. White, eds. Academic Press, San Diego, CA.

87 Lima Peru I

Pacific Ocean Chile/

Santiago

Figure 4 .1. Distribution of Malesherbiaceae. 88 Species Collector

M. angustisecta Harms KMG & Refulio R. 199* M. ardens Macbr. KMG & Arakaki 184* KMG & Arakaki 185 KMG & Arakaki 187 M. arequipensis Ric. Dillon, Sagastegui, & Santisteban 4793 KMG, Salas, & Cuadras 58 KMG & Cuadras 182* M. auristipulata Ric. KMG 61* KMG 65 M. campanulata Ric. Ricardi. Marticorena. & Matthei 1748* M. densiflora Phil. KMG 115* M. deserticola Phil. Dillon & Dillon 6013* KMG 93* Af. fasciculata D. Don KMG 151 Stuessy 9803* Af. haemantha Harms Hutchinson 1278 (UC)* Af. humilis Poepp. KMG 109 var. parviflora (Phil.) Ric. KMG 110 KMG 190 Stuessy 9883* Stuessy 9892 Af. lactea Phil. KMG 44 Stuessy 9876* Af. lanceolata Ric. KMG 54 KMG 55 KMG 133* Stuessy 12809* Af. linearifolia (Cav.) Pers. KMG 23* KMG 36 KMG 155 Af. lirana Gay var. lirana Stuessy 9814* Af. lirana Gay var. subglabrifolia O.K. Hjertling 6312 (A) Af. obtusa Phil. var. obtusa KMG 119a* Af. paniculata D. Don KMG 28* Af. rugosa Gay var. rugosa KMG 37 KMG 114 KMG 118* Af. scarlatiflora Gilg KMG 286 KMG & Romero 295 KMG & Roque 350 KMG & Salvador P. 369

Figure 4.2. Accessions collected for DNA isolation and sequencing. All vouchers are deposited at OS unless otherwise noted. Starred accessions were included in the maximum likelihood analyses, (continued)

89 Figure 4.1. (continued)

Af. splendens Ric. KMG 188* Af. tocopilUma Ric. Dillon & Dillon 5719 Af. tenuifolia D. Don KMG & Refulio 191 KMG & Refulio 192* KMG & Refulio 197 Af. tubulosa (Cav.) J. St. Hil. KMG & Roque 354* KMG & Bedoya 362 Af. turbinea Macbr. KMG & Refulio 198* Af. weberbaueri Gilg var. KMG 288 weberbaueri Piriqueta caroliniana (Walt.) Urban Shore 175* Tumera scabra Millsp. Shore 150*

90 Figure 4.3. Consensus tree derived from four most parsimonious trees calculated by

NONA. Dashes indicate support values less than SO.

91 Piricfueta Tumera M. linearifolia 5/100/100 M. linearifolia 1/63/63 M. linearifolia M. paniculata M. humilis 5/100/ M. humilis 100 4/93/90 M. humilis M. humilis M. humilis 5,100/100 M.fasciculata M.fasciculala 5/98/97 M. lactea M. lactm 5 /9 6 /9 2 M. densiflora 1/63/70 M. deserticola M. deserticola M. obtusa M. lirana v. lirana M. lirana v. subglab. M. rugtm 1 /-/5 3 Af. rugosa Af. rugosa 2/8 4 /8 5 Af. campanulata 1/63/77 Af. lanceolata Af. lanceolata Af. lanceolata Af. lanceolata Af. tubulosa Af. tubulosa Af. scarlatiflora 5/97/97 Af. scarlatiflora Af. scarlatiflora Af. scarlatiflora Af. scarlatiflora Af. splendens 5/99/96 Af. weberbaueri Af. haemantha 2/77/82 Af. auristipulata Af. auristipulata Af. turbinea Af. tocopillana 1/63/60 3/95/96 Af. ardens Af. ardens Af. ardens Bremer / Jackknife / Bootstrap Af. are(ptipensis 5/100/ M. arequipensis # mutations (ACCTRAN) 100 Af. arequipensis Af. angustisecta 1/63/62 Af. tenuifblia Af. tenuifblia Af. tenuifblia Figure 4.3 92 Figure 4.4. The consensus tree obtained when ambiguous sites were removed and the data matrix re-analyzed in NONA. The arrows indicate clades with topological changes from the trees yielded by parsimony analysis of the complete data matrix. Dashes indicate support values less than SO.

93 Piriqueta TUmera M. lirmrifolia 5/100 M. linearifolia m 1/64 M. lineariplia M. paniculata M. humilis M. humilis 5/100 4/93 M. humilis M. humilis M. humilis 5,100 M. fasciculata 1/- M. fasciculata 3/81 m 5/98 M. lactea M. lactea 5/99 M. densiflora 1/64 M. deserticola M. deserticola 3/87 1/- M. obtusa M. lirana v. lirana M. lirana v. subglab. 4/97 M. rugosa -/- M. rugosa M. rugosa 2/85 M. campanulata 1/63 M. lanceolata M. lanceolata 1/- 1/- M. lanceolata M. lanceolata M. tubulosa M. tubulosa 4/94 M. scarlatiflora M. scarlatiflora 1/- M. scarlatiflora M. scarlatiflora M. scarlatiflora M. splendens M. weberbaueri M. haemantha 1/70 M. auristipulata 5/99 ^ M. auristipulata M. turbinea M. tocopilhma 3/95 M. ardens M. ardens M. ardens Bremer / Jackknife M. arequipensis M. arequipensis 5/100 M. arequipensis M. angustisecta 1/64 M. tenuifolia M. tenuifolia M. tenuifblia F igure 4.4 94 Piriqueta Tumera — M. lactea — M. humilis rM . paniculata I M. linearifolia r M. tocopillana ^ M. ardens rM . auristipulata L M. turbinea — 0.01 substitutions/site \ pM. arequipensis M. angustisecta ^M. tenuifolia — M. haemantha M. splendens M. tubulosa M. deserticola M. deserticola M. densiflora M. lanceolata r:M. lanceolata '-M . campanulata - M. —M. rugosa M. firaua 'M.fasciculata

Figure 4.5. The tree yielded by maximum likelihood analysis of the complete data set using the HKY85 model. The arrow marks the clade that disappears in analysis of the unambiguous data set. 95 M. auristipulata M. turbinea M. haemantha

Floral color Red Red Red Anther color Black Black Black Perianth hairs Black & glandular Black & glandular Black & glandular Leaf hairs Black & glandular Black & glandular Black & glandular Corona iength Longer than perianth Longer than perianth Same length as perianth Leaf margins Serrate Serrate Pinnatifid Stipule shape Ovate to auricular Ovate to auricular Lanceolate

Figure 4.6. Comparison of the morphologies of Af. auristipulata, Af. turbinea, and Af. haemantha.

96 CHAPTERS

RECONSTRUCTION OF THE BIOGEOGRAPHIC HISTORY OF MALESHERBIACEAE

The western flank of the Andes Mountains and the narrow coastal area between them and the Pacific Ocean is an arid zone essentially isolated from other dry regions on the

South American continent. This isolation and the environmental diversity of the region likely spurred the development of its unique and varied flora. The arid region of Chile and

Peru is longitudinally very narrow due to the proximity of the Andes to the ocean; elevation consequently increases rapidly from sea level over a very short distance. The degree and nature of the aridity vary greatly with latitude. Capping off the southern extension of the arid zone is mediterranean central Chile. As the latitude decreases, the aridity increases until, in northern Chile, the climate is so severely dry that the region, known as the Atacama

Desert, is one of the driest on the planet. This hyperaridity ends near the political border of

Chile and Peru. In Peru, a broad coastal plain, adjacent dry mountains and foothills, and occasional coastal fog zones, or lomas, characterize the desert. In addition to the arid coast, deep river valleys with arid climates dissect the Andes in Peru. The desert ends at the border of Peru and Ecuador.

Although this region has a wide variety of habitats and a diverse flora, there have been no studies of the biogeography of the area based on the phylogeny of any of the taxa endemic to the region. It is particularly difficult to study the biogeography of the entire

97 region using a phylogeny because there are few examples of monophyletic groups native to both sides of the Atacama Desert. Malesherbiaceae, an angiosperm family shown to be closely related to Tumeraceae by rbcL data (Fay et al. 1997), are one of two families endemic to arid western South America and one of the few taxa found both in arid Peru and

Chile. The family contains one genus, Malesherbia, with 24 species.

This family has a distribution spanning most of the desert, from central Chile to central

Peru and from the coast to the Argentinean Andes, but Malesherbiaceae are notably absent from the hyperarid regions of the Atacama Desert in northem Chile (Figure 5.1). They are native to a variety of arid climates, including mediterranean central Chile, the Peruvian summer-rain deserts, the Andes up to 3700 m, and the edge of the most arid part of the

Atacama Desert. Many species of Malesherbiaceae are endemic to a relatively small area.

Because the family has radiated into such a wide variety of habitats and is native to much of the desert, Malesherbiaceae are an ideal case study of the evolution of the desert’s flora in the context of the geological and climatic history of the entire area.

Previous studies provided a robust estimate of the family’s phylogeny (Gengler, chapter

4) using molecular sequence data, and these data may be further used to calculate divergence times based on a molecular clock, if clock-like evolutionary rates cannot be rejected.

Although use of such clocks warrants caution (Bousquet et al. 1992, Ayala 1997, Sanderson

1998, Doyle & Gaut 2000, Muse 2000), calculating divergence times of lineages for which clock-like evolution cannot be rejected may provide further, albeit somewhat inexact, evidence pointing to paleohistorical events important in the evolution of the family, and of the desert’s flora in general.

The goals of this study were to determine if the family’s phylogeny corresponds to its distribution, especially with regard to the disjunction around the Atacama Desert. If the phylogeny and distributions do correspond, the next step is to determine how the climatic

98 and geological histories of the region have affected the evolution and colonization history of the family. To this end, the physical history of the region and estimated divergence times of lineages in Malesherbiaceae were examined.

Phylogeny and distribution of Malesherbiaceae

The study of the biogeography of a taxon necessitates examination of the fit of the distributional pattern of the group to its phylogenetic history. If more closely related members of the taxon are also found in geographically related areas, the phylogeny provides correlational evidence that the physical history of a taxon’s range has influenced its evolution. Therefore, the first step in understanding the biogeography of Malesherbiaceae is to determine if the phylogeny and distributions of the species create a pattern interpretable in light of the history of the arid regions of Chile, western Argentina, and Peru.

The primary break in the distribution of Malesherbiaceae occurs around the driest part of the Atacama Desert in northernmost Chile (Figure 5.1). This disjunct distribution is strongly reflected in the phylogeny; all of the species found in Peru, as well as Af. auristipulata, M. tenuifolia, and Af. tocopillana, three rare species native to northem Chile, form a monophyletic, terminal clade (Figures 5.2 & 5.3), henceforth referred to as subg.

Malesherbia. The remaining species form a paraphyletic grade basal to the subg.

Malesherbia and distributed from central to northem Chile (the “Chilean species”). Such a distribution suggests that Malesherbiaceae originated in Chile and that the family was introduced to Peru once. The fact that the introduction to Peru occurred only once, although the geographic distance is not great, indicates that northem Chile acted as a barrier to the spread of Malesherbiaceae.

This area appears to have impacted the history of the flora in general. Rundel et al.

(1991) noted a large turnover in the composition of the coastal flora just south of the

Peruvian-Chilean border. They attributed this tumoverto the absence of coastal hills or

99 mountains capable of catalyzing fog formation. Coincident with the political border as well is the change in precipitation regime from the Peruvian tropical summer rains to the Chilean mediterranean winter rains (Borgel 1973, Arroyo et al. 1988). The widespread nature of the disruption of the flora indicates that the biogeographic barrier between Peru and Chile is ancient.

The distributions of the lineages comprising subg. Malesherbia and the Chilean group also reveal a concordance between the phylogeny and geography. The Chilean group is divided into four clades (Figures 5.2). The most basal lineage in the family is restricted to mediterranean central Chile and the adjacent semi-arid area to the north (Figures 5.2 & 5.3).

The two constituent species of subg. Cyanpetala are geographically separated, but some populations have morphologically mixed individuals, which suggests they recently diverged from each other.

The single-species subg. Parvistella has a very large distribution stretching from

Santiago, Chile to northem Chile (Figure 5.3). This subgenus has some disjunct populations to the south in Province Neuquén, Argentina (37°30’). One population was discovered intermingled with a southern population of M. tenuifolia of subg. Malesherbia.

Subgenus Parvistella's range includes the coast, Andean foothills, and inland valleys for much of its distribution, but in the far north it is largely coastal. Although M. humilis is the only constituent species, it is composed of five intergrading varieties with especially high morphological diversity in the Elqui Valley of north-central Chile (30°S).

A third clade, subg. Albitomenta, is composed of two disparate species (Figure 5.2).

One species is found in semi-arid Andean foothills, whereas the other is native further to the north and to the east, in very arid mountains up to 3770 m in Chile and Argentina (Figure

5.3). These divergent ranges are reflected in the very different morphologies of the two species as well. Unsurprisingly, there is very low support for the monophyly of this

100 subgenus. The distributions of this subgenus and those of subg. Parvistella and subg.

Cyanpetala overlap to some degree, although not extensively. Each of these Chilean lineages characterizes different ranges with regard to latitude and elevation.

The sister clade to the Peruvian species, subg. Xeromontana (Figure 5.2), occupies montane and very dry desert habitats (Figure 5.3). In contrast to subg. Parvistella, for much of its north-south range, it is primarily found in the Andes straddling the border of Chile with Argentina. Only in the dry north on the edge of the Atacama Desert does its distribution expand westward to include lower elevations. This clade is essentially isolated from the other subgenera.

Peruvian subg. Malesherbia is divided into two clades, one with strong and the other with very weak support (Figure 5.2). The strongly supported clade is located in the inter-

Andean valleys of central, coastal Peru and the central Peruvian Andes. The poorly supported clade, composed of four well defined lineages, has a general distribution in the broad valleys and dry ravines of southern Peru and northem Chile (Figure 5.3).

It is clear, then, that the phylogeny of Malesherbiaceae reflects the distributions of the species, in particular the disjunction around the Atacama Desert of northem Chile. This suggests strongly that the physical history of the area to which the family is endemic has influenced its evolution. Examination of this history may therefore yield insight into factors affecting the evolution of Malesherbiaceae.

Physical history of arid western South America and the Andes

To put current understanding of the physical history of arid Chile, westem Argentina, and Peru into perspective, it is necessary first to examine current topological and climatic conditions in the area and the processes that maintain climatic stability and aridity, the

101 dominant influence on the region’s flora (diCastri & Hajek 1976, Messerii et al. 1997). The development of these forces and thus the paleociimate of the region will then be examined in more detail.

The cold Humboldt Current running along the west coast of South America plays a major role in creating the coastal desert (Bdrgel 1973, Caviedes 1973, Lydolph 1973). At central and southern Chile, westerlies are diverted north along the coast by the Andes. These winds, in combination with the Coriolis Effect (the deflection of moving objects away ftom the equator), produce an upwelling of the current’s cold water. The upwelling water cools the air above it as it moves onshore, resulting in a stratified atmosphere in which layers of warmer air rest on layers of cooler air. This condition, known as an inversion, is extremely stable. Although this strong inversion inhibits precipitation and incursions of moister air, cloud and fog formation is common.

The Humboldt Current is also responsible for the remarkably homogeneous temperatures along the coast. Mean daily and yearly temperatures scarcely differ from Lima,

Peru (12“S) to La Serena, Chile (29°54’S), and daily and yearly fluctuations in temperature are minimal (Paskoff 1973). In addition, these temperatures are low in comparison to other areas at similar latitudes (Paskoff 1973, Linacre & Geerts 1997).

While the Humboldt Current creates an inversion near the coast, the eastern Pacific anticyclone, a high-pressure cell, determines the patterns of winds reaching the coast. It lies over the southem PaciEc Ocean at approximately 32°S in austral summer and 23°S in austral winter. The northem position of the high pressure cell in the winter allows the moist polar Aont to advance toward the north of Chile during that time (Paskoff 1973, Arroyo et al. 1988). The summer position of the anticyclone prevents rainfall over Chile; therefore,

90% of the rainfall from 27°S to 3 1°S falls during the winter, which is a typical of a mediterranean climate (Simpson 1979).

102 In extreme northeastern Chile and in Peru, precipitation falls during the summer, when the southem position of the anticyclone cannot block moist easterlies. This typically tropical pattern of rainfall extends south as far as 24°-26°S (Borgel 1973, Arroyo et al. 1988,

Messerii et al. 1997); the southem extensions of this precipitation regime are usually limited to high elevations (Borgel 1973).

The Atacama Desert of northem Chile lies on the peripheries of the influences of both the tropical and mediterranean rains. During summer, when the anticyclone lies in the south, the northem edge of the anticyclone generally prevents tropical rains from falling in northem Chile. In winter, the more northerly position of the anticyclone also prevents the polar front from creeping into this region. Hyperaridity is thereby maintained (Caviedes

1973).

The final influence on the region’s climate is the Andes Mountains, which intercept moist Amazonian winds, causing the air to lose its moisture on the eastem slopes (Arroyo et al. 1988). As the air descends the westem slopes, a subsidence inversion forms (the

“rainshadow effect”). The Andean subsidence inversion desiccates the west coast from

30°S north almost to the equator, undoubtedly amplifying the effects of the Humboldt

Current and eastem Pacific anticyclone.

Because the flora is shaped largely by the aridity of the region, the study of the area’s biogeography is tied to the development of the climate, which is intimately related to the formation of the Humboldt Current and the eastem Pacific anticyclone and the rise of the

Andes. The most critical factors appear to be the formation of the Humboldt and the anticyclone; Lydolph (1973) found that the absolute aridity characteristic of the Atacama

Desert today cannot exist without both of these influences.

The Humboldt Current is derived from the West Wind Drift, an eastward-flowing circumantarctic current. These currents are cmcial factors in the global climate, as they decouple Antarctica’s climate from the warmer north, leading to a steep thermal gradient

103 from the equator poleward (Kennett 1980, Brain 1984). The position of the West Wind

Drift today causes it to be largely diverted northward after hitting the South American coast.

Zinsmeister (1978) postulated that the present northerly position of this current was set when the Antarctic Shackleton Sea, which formerly diverted the West Wind Drift’s waters further south, closed due to the formation of the west Antarctic ice sheet.

Evidence of the timing of the formation of this polar ice cap lies in marine fossils, which are sensitive to temperature changes. Microfossil evidence from deep sea drilling cores suggests that the permanent east Antarctic ice cap formed by the middle Miocene (14-11 mya), with an increase in ice volume in the late Miocene to early Pliocene (5 mya) when the west Antarctic ice cap formed (a period known as the Terminal Miocene Event) (Kennett

1980, Brain 1984, Flohn 1984, Mercer 1984). Southem Peruvian fossils (deMuizon & deVries 1985) and central Chilean molluscan fossils (Zinsmeister 1978) show a turnover in the late Miocene to very early Pliocene, also supporting a late Miocene drop in ocean temperatures. Studies of Peruvian diatomaceous deposits reveal an increased upwelling of cold, nutrient-rich waters between the middle and late Miocene (Marty et al. 1987, Alpers &

Brimhall 1988). The northward migration of diatoms in the late Miocene suggests that the

West Wind Drift was forced north, which likely resulted in a similar displacement of westerly prevailing winds (Kennett 1980).

Not only is the Antarctic ice cap responsible for the formation of today’s Humboldt

Current, but it also resulted in modem global climatic conditions (Simpson 1983). The north to south thermal gradient is the basis for modem wind and pressure systems, including westerlies, trade winds, and the anticyclones (Flohn 1984, Markgraf et al. 1992).

Therefore, the eastem Pacific anticyclone that affects climatic patterns of westem South

America so strongly has been in place since at least the mid-Miocene, when the polar ice cap formed. Flohn (1984) noted that cooler conditions in Antarctica led to strengthening of

104 trade winds and upwelling and to a more northerly position of the anticyclone. Increased upwelling and depressed ocean surface temperatures accompanying these effects result in decreased evaporation and, therefore, increased aridity for adjacent land masses.

The rise of the Andes is also important to the development of aridity along the west coast of South America because these mountains stabilize the position of the eastem Pacifîc anticyclone (Alpers & Brimhall 1988), direct westerlies northward, and create a massive subsidence inversion. Several ranges comprise the Andean chain. Only the Cordillera

Oriental, the Cordillera Occidental, the altiplano (a high elevation plateau), and the Principal

Cordillera are of interest for our discussion.

The Cordillera Occidental stretches from 5°S to 14°S, where it abuts the altiplano

(Figure 5.4). This range initially uplifted in the Cretaceous, with further orogeny in the

Pliocene. The final uplift likely occurred between the first and second Pleistocene glacial periods (Simpson 1975). The Cordillera Oriental lies east of the Cordillera Occidental, reaching from 5°S to about 25°S. This range borders the altiplano's eastem edge and arose in three pulses during the Cretaceous, the Miocene-Pliocene, and the Pleistocene (Simpson

1975). The Principal Cordillera runs along the westem margin of the altiplano from southem Pern to about 40°S in central Chile. During the Miocene and the Pliocene there were major uplifts. In the central Andes, early Pliocene uplifts were particularly strong, raising the mountains some 1000-3000 m to their present height (Peterson 1958; Simpson

1975,1979; Paskoff 1977). The altiplano is a high elevation plateau sandwiched by the higher Cordillera Oriental and Principal Cordillera in southem Peru, Bolivia, northem to central Chile, and Argentina. It began to uplift in the Miocene, but major uplifts occurred after the mid-Pliocene. The altiplano, the Principal Cordillera, and the Cordillera Oriental create an extremely wide high elevation zone between approximately 18°S and 27°S. These four portions of the Andes thus experienced signifîcant uplifting during the Pliocene

(Peterson 1958), when they likely began to affect greatly the climate of the region. By 2-5

105 mya, the Andes had reached elevations over 2000 m; Alpers & Brimhall (1988) noted that elevations of 2000-3CXX) m are necessary to produce a significant rainshadow effect in similar mountain ranges.

The geological histories of South America and Antarctica suggest a paleoclimatic chronology for today’s arid westem South American region. Since the Antarctic ice cap appears to be responsible for formation of the modem Humboldt Current and ultimately for modem climatic conditions, including the formation of the eastem Pacific anticyclone, the cold Pacific waters and the presence of the anticyclone by the end of the Miocene or early

Pliocene strongly suggest that coastal Pern and northem Chile were arid by that point. Data conceming weather-induced mineral enrichment from further inland In northem Chile concur with this estimate. These data suggest that the area was semi-arid from the Eocene through the early Miocene, with intervals of high precipitation, especially in the early

Miocene. They found that continuously arid conditions began in the mid-Miocene (Alpers

& Brimhall 1988). Mortimer and Saric’s (1975) data also agree, noting that aggradation of the longitudinal valley ceased about 9 mya.

A warm spell appears to have opened the Pliocene (Mercer 1984, Paskoff 1973), and a warmer ocean may have led to more mesic conditions along the westem side of South

America. Fossils of large vertebrates like Equus have been found in the Atacama Desert, indicating that there the climate at some time during the Pliocene was more mesic than today

(Arroyo et al. 1988). Paskoff (1973) concluded that the climate in the zone between 30° and

33°S (an area known as the Norte Chico) was tropical during the Pliocene. The mid-

Pliocene saw a retum to cooler oceanic temperatures (Mercer 1984), and, by the end of the

Pliocene, the climate of the Norte Chico was semiarid, judging from erosion pattems

(Paskoff 1973). Major uplifts of the Andes in the Pliocene would have intensified arid

106 conditions via a strong subsidence inversion, especially west of the wide high elevation zone between 18°S and 27°S. In northem Chile, water from the Andes cut the deep transverse valleys and outlets to the sea (Mortimer & Saric 1975).

The climate of the Pleistocene was characterized by fluctuating conditions, which greatly influenced the Andean biota (Simpson Vuilleumier 1971; Simpson 1975,1979,1983;

Simpson & Todzia 1990; Marquet 1994). In Peru, Simpson ( 1975) deduced from the distributions of the modem flora, fossil evidence, and reconstructions of past oceanic temperatures that the westem slopes of the Peruvian Andes experienced alternating humid and arid conditions. Glacial periods, she noted, were humid at higher elevations and arid on the lower slopes and coast. Aridity predominated at higher elevations during interglacial periods (like modem times); during these times, sea fogs ameliorated aridity along the coast, although rainfall does not appear to have contributed much to overall precipitation. These precipitation changes seem to have affected the flora more directly than depressed temperatures and lowered vegetation zones (Simpson 1975).

Fluctuations between humid and arid conditions also characterized Pleistocene Chile

(Arroyo et al. 1988, Paskoff 1973). They appear to be symptoms of oscillations between glacial phases and melting phases in Antarctica. Glacial periods led to northward shifts of the polar front by 5°-7°, and these shifts were likely concurrent with weakening of the anticyclone. Together, these forces resulted in latitudinal compression of arid regions

(Paskoff 1977, Heusser 1984). Not all glacial periods seem to have resulted in similar climates, however—during at least one glacial maximum ( 18(X)0 BP), the anticyclone moved over central South America, resulting in very arid conditions for that region (Lauer &

Frankenberg 1984). In the changing climate of the Pleistocene, aridity seems to have remained constant only from 27° - 30°S, an area perhaps reached by neither the northem summer rains nor the southern polar front.

107 Paskoff (1977) has given the most comprehensive review of events of the Chilean

(Quaternary, which likely influenced the evolution of Malesherbiaceae greatly. Because the country covers such a large latitudinal range, he divided northem and central Chile into four areas (Figure 5.4). I will follow his divisions in this discussion. They are the hyperarid north, known as the Norte Grande i\%°-2TS); the marginal desert (27°-30°S); semiarid

Chile (30°-33°S, the Norte Chico); and central Chile (33°-39°S).

The Norte Grande experienced little glaciation, and glaciers appear to have descended only to about 40(X) m (Simpson 1979, Messerii et al. 1993). Pleistocene fluctuations in climate appear to have affected the Altiplano significantly. Glacial periods seem to correspond to pluvial periods, when lakes formed on the Altiplano (Paskoff 1977, Seltzer

1990). Dry lake beds (salares) dated to the Pleistocene also dot the longitudinal valley below the Andes. Moreno etal. ( 1994) argue that the water source for these paleolakes was groundwater and streams from the mountains, rather than increased precipitation at low elevations. Messerii et al. (1993), Kessler ( 1984), and Grosjean ( 1994) agree that increased precipitation on the Altiplano does not necessarily translate into increased precipitation on the westem slopes of the Andes and at lower elevations. Messerii et al. (1993), in fact, maintain that increased precipitation in the Altiplano corresponded with increased aridity below 3500 m as more rain or snow fell at upper elevations. Mortimer and Saric (1975), moreover, found little evidence of Quatemary topographical changes, suggesting that the

Norte Grande was very arid throughout that period. Paskoff ( 1977), however, holds that there were periods of increased precipitation during the Pleistocene in the Norte Grande.

Latorre (pers. comm.) agrees with Paskoff, citing his studies of Pleistocene middens near

Salarde Atacama (23°S), which show a much richer annual flora than is found today. He argues that the presence of a diverse annual flora indicates higher rainfall.

Evidence from the Pleistocene for the marginal desert is scanty, but it suggests that the climate remained dry. The Norte Chico to the south of the marginal desert, however.

108 experienced many climatic changes during the Pleistcxzene. Temperate rain forest remnants near the coast, the most notable of which is Fray Jorge (30°S), suggest that the Valdivian forest migrated north S°-6° during the Pleistocene and retreated during the Holocene

(Simpson Vuilleumier 1971; Paskoff 1973, 1977; Simpson 1979; Pérez & Villagrân 1994).

Paleosoil evidence suggests the Pleistocene was marked by alternating dry and humid conditions, with humid periods resembling the rainy provinces to the south (Paskoff 1973,

1977). Repeated erosion and renewed buildup of alluvial fans suggest that torrential flooding characterized times of greater precipitation (Paskoff 1977), whereas at least one dry period was more arid than the current one, a finding based on paleodune data (Paskoff

1973). Morraines as low as 3000 m indicate that glaciers advanced several times during the

Pleistocene (Briiggen 1950, Paskoff 1977). Paskoff ( 1977) noted that the glacial climate in the Andean foothills from 1000-3000 m in this area must have been more humid.

Central Chile (33°-39°S) lies on the periphery of the distribution of Malesherbiaceae.

Here glaciation evidence clearly shows there were three Pleistocene glacial advances, with glaciers dropping to 28(X) m during the last period (Paskoff 1977). Fossils from Lago

Taguatagua (34°30’S), which today has a temperate climate with a winter rain regime and eight-month dry season, indicate that the climate at the end of the last glacial period was warmer with fluctuating precipitation. Molluscan fossils suggest a climate more like Chile’s modem southern lake region. The Pleistocene-Holocene boundary shows an abrupt climate transformation to more arid modem conditions (Nunez et al. 1994). Heusser (1984) determined from pollen data that this area was wetter during the Late Pleistocene as well.

Potential divergence times: perspectives from the moiecuiar ciock

Key to understanding how past climates have affected evolution of Malesherbiaceae is the correlation of paleoevents with cladogenic events in the inferred phylogeny.

Traditionally, biogeographers have carefully examined the physical history of the area of

109 interest, then applied their findings to the taxa in question. For example. Raven and Axelrod

(1974), in their groundbreaking study on continental drift and its effects on the fauna and flora, postulated that the separation of South America from Africa logically explains distributions and relationships of fîsh, anurans, caecilians, turtles, ratite birds, Annonales,

Tumeraceae, Gentianales, and Santanales, among others. This traditional method has given great insight into biogeography, but it is difficult to use when the area studied is within a continent and barriers are not so obvious as oceans or the region’s history is complex enough that it is not clear how specific paleoevents may have affected plant evolution. In such cases, an absolute or relative date is essential. Fossils provide excellent evidence of past distributions and morphologies and can often be given at least a relative date. However, not all groups, including Malesherbiaceae, have a fossil record. For such situations, workers have turned to molecular data for some evidence of times of divergence (Kim et al. 1996,

Lee et al. 1996, Francisco-Ortega 1997, Li et al. 1997, Baum et al. 1998, Zink et al. 1999).

Estimated divergence times here will be used simply as supplementary information for a more traditional biogeographic study (in this chapter), rather than as the primary source of information for biogeographic inferences.

A robust phylogeny of Malesherbiaceae, using Piriqueta and Tumera (Tumeraceae) as outgroups, was constructed using parsimony analysis (Gengler, chapter 4). The strict consensus of the four most parsimonious trees (Figure 5.2) formed the basis for testing the molecular clock hypothesis by relative rates tests.

Kimura 2-parameter distances were calculated utilizing the program MEGA l.O (Kumer et al. 1993) for use in two different relative rates tests. The relative rates test of Tajima

(1993) compares two individual sequences with a reference sequence. A modified version

(Xiang et al. 1998) of Li’s (1997) test compares the rates of evolution of two lineages (i.e., clades) with a third reference lineage using the average sequence divergences of the lineages. Sister lineages or sequences to the pair of lineages or sequences being compared

no always served as the reference sequences. In the Tajima test, therefore, as many separate tests were conducted as references were available in the sister lineage (e.g. a pair of sequences with a sister lineage represented by three sequences was tested three times using each sister sequence as a reference). Where more than one accession’s sequences were identical, only one accession was included in the Li test (Figure 5.5).

The modified Li test rejected the molecular clock for subg. Parvistella and its sister clade (containing subg. Albitomenta, subg. Xeromontana, and subg. Malesherbia) using subg. Cyanpetala as the reference. This test also rejected the clock for subg. Xeromontana and subg. Malesherbia using subg. Albitomenta as a reference. All other lineage comparisons, including the comparison of the basal subg. Cyanpetala with the rest of the family using the outgroups as a reference, failed to reject the clock hypothesis.

Since the Tajima test compares the rates of two individual sequences, it may be able to detect the specific source of rate variation found in the Li test. Because examination of large numbers of sequences may lead to spurious rejections of the molecular clock (Sanderson

1998), test stringency was increased; only those comparisons significant at the 0.05 level regardless of reference sequence were considered to be rejections of the molecular clock hypothesis. Using this criterion with subg. Cyanpetala as the reference sequences, all comparisons of subg. Parvistella to the {M. angustisecta - M. tenuifolia) clade were significant. In addition, all comparisons were significant between subg. Parvistella and M. tocopillana and two of the three accessions of M. ardens. The comparisons of M. humilis

(Stuessy 9883) with the other accession of M. ardens (KMG 185) and with M. turbinea and M. tubulosa (KMG 354) were not significant, but comparisons with all other members of subg. Parvistella using all reference sequences were significant.

I l l Using reference sequences from the (A/, fasciculata - M. lactea) clade, the (M. angustisecta - M. tenuifolia) clade again was a source of rate heterogeneity when compared with M. lirana, M. rugosa, M. campanulata, and M. lanceolata. No other comparison was significant for all four accessions in the reference group.

The modified Li method used therefore revealed rate heterogeneity between subg.

Xeromontana and subg. Malesherbia and between subg. Parvistella and its sister clade (A/. fasciculata - M. tenuifolia). In both comparisons, the (M. angustisecta - M. tenuifolia) clade was implicated as a major source of rate heterogeneity by the more detailed Tajima test

Significantly, the molecular clock was not rejected in comparisons between members of subg. Parvistella and Chilean subg. Xeromontana. The (Af. tubulosa - M. splendens) clade also did not appear to contribute to the rate heterogeneity detected by the Li test using either subg. Cyanpetala or Albitomenta as references. The rate of ITS evolution seems, therefore, to be significantly different only in the (Af. haemantha -Af. tenuifolia) clade, and especially so in the (Af. angustisecta - Af. tenuifolia) clade.

It is unclear what may have caused the difference in mutation rate that seems to have arisen in these taxa. Generation time differences have been implicated in rate changes in animals (Wu & Li 1985, Li & Tanimura 1987), although the effect of such differences in plants is unknown (Gaut et al. 1997). Bousquet et al. (1992) also suggested that the rate of genetic change is positively related to the rate of spéciation. However, the habits of Af. angustisecta, Af. arequipensis, and Af. tenuifolia fail to suggest that they possess signifîcantly faster or slower generation times, since they are small shrubs like other species of Malesherbia. This differently evolving clade also is not particularly speciose. The cause of the rate change therefore remains obscure.

Although some Peruvian taxa seem to be evolving at different rates than some Chilean taxa when using particular reference taxa, the molecular clock was not rejected for other comparisons, so divergence times may be calculated for those lineages. These calculations

112 require either an internal calibration, usually using fossil or biogeographic evidence, or the mutation rate of another taxon with a similar life history. Because no fossil evidence exists for Malesherbiaceae and to avoid the circular logic of using paleoevents to calculate divergence times to examine those very events, divergence times were estimated using mutation rates from an external taxon. Although the effect of generation time in plants Is unknown (Gaut et al. 1997), the choice of a short-lived perennial as an external taxon would eliminate that variable. Moreover, small population size has been suggested to affect substitution rates, at least in the rbcL gene, by increasing the likelihood of bottlenecks

(Bousquet et al. 1992), so an external taxon with relatively small populations or a history of bottlenecks would be appropriate. Two genera in Asteraceae, Robinsonia and Dendroseris, are short-lived perennials endemic to the Juan Fernandez Islands of Chile. Dendroseris exists in small populations, and Robinsonia appears to have undergone bottlenecks in its history (Crawford et al. 1992). The mutation rate of the ITS regions (or Robinsonia was calculated to be 1.57% per million years (Sang et al. 1995). Dendroseris has a lower calculated rate of 0.78% per million years (Sang et al. 1994). Both rates were calibrated using the geological age of the oldest island. Crawford et al. (1998) suggest that

Dendroseris may have arrived on the islands after Robinsonia, so the Robinsonia estimate may be more realistic.

Using the mutation rate for Robinsonia and the averages of the sequence divergence

values for the lineages of Malesherbia places the first split in the family (between subg.

Cyanpetala and the rest of the family) at 4.0 mya (Figure 5.2). The split between subg.

Albitomenta and its sister clade is estimated to be 3.6 my old. The divergence between the two major clades of subg. Malesherbia may have occurred 2.1 mya, whereas the two clades of subg. Xeromontana split an estimated 1.9 mya. These dates would indicate that evolution of Malesherbiaceae has been influenced primarily by Plio-Pleistocene events.

113 If these calculations are reasonably close to the actual evolution of the family,

Malesherbiaceae likely diverged bom Tumeraceae in the early Pliocene or the late Miocene.

Unfortunately, without a complete understanding of the relationships among the various genera in Tumeraceae and Malesherbiaceae and divergence values derived from a more slowly evolving gene, it is not possible to estimate with confîdence the age of the family using the molecular clock hypothesis.

Proposed reconstruction of the evolutionary history of Malesherbiaceae

Because the phylogeny and distribution of species of Malesherbiaceae appear to be correlated, I will assume that the physical history of coastal western South America and the adjacent Andes have influenced the evolutionary history of the family. Reconstmction of the biogeography of the family requires the determination of the sequence of phylogenetic events, the duration of time over which these events occurred, and the timing of these events.

The phylogeny of Malesherbiaceae establishes a sequence of events in the evolutionary history of the family; the duration of time and timing of events will be determined in conjunction with the physical history of arid westem South America. I will attempt to place the evolution of the family within a specific time frame by establishing the origin of the family and of the modem species. From these “anchor points,” I will then correlate phylogenetic events leading to the formation of modem species with paleohistorical events.

All species of Malesherbiaceae today are xerophytic, so it is most parsimonious to assume the common ancestor of modem species was also a xerophyte. The development of aridity is therefore the critical factor informing the estimate of the age of the family. A conservative estimate of the age of the family will therefore place the divergence from

Tumeraceae with the formation of the desert. It appears that coastal westem South America has been more or less arid since the Eocene, but the area became consistently arid in the middle to late Miocene. Malesherbiaceae could be as old as 40 my, but that is almost

114 certainly an overestimate of the family’s age, since the early Miocene was a rather wet period in the area’s history. The best conservative estimate of the age of the family is the late Miocene and perhaps early Pliocene, when the region was quickly drying as Antarctic ice built up. The molecular clock evidence does not conflict with this assessment; if the first cladogenic event in the family occurred 4 mya (mid-Pliocene), the family must have broken from Tumeraceae well before that time—likely the Miocene.

The second anchor point for the time frame of the evolution of Malesherbiaceae is the development of modem species. The low confidence indices, lack of resolution among some closely related species, the low sequence divergence values, and the lack of evidence supporting monophyly of different accessions of a single species (Figure 5.2, Figure 5.6) suggest that most modem species in four of the five subgenera are the results of a recent rapid radiation. Subgenus Cyanpetala consists of two morphologically and molecularly very similar species, but the relationship between the two is unclear. Subgenus Parvistella, although containing only M. humilis, is diversified into five varieties covering a large distribution. It is uncertain, unfortunately, whether this variation is a reflection of genetic differences or great phenotypic plasticity. Five barely diverged species comprise the more species-rich of the two clades of subg. Xeromontana. Lack of sequence divergences among the four species within the (Af. scarlatiflora Af.- weberbaueri) lineage in subg.

Malesherbia demonstrates the close relationships among those terminal taxa. Finally, within each of the lineages having more than one species that comprise the other subclade in subg.

Malesherbia, morphological and molecular diversities are low. It should be noted that, although closely related species are morphologically similar, taxon limits are generally clear.

Sequence divergence values estimated using the Kimura 2-parameter model reveal a family-wide pattern of low divergence among sister species, with two exceptions (Figure

5.6). Average divergence values calculated by comparing one accession of a species against an accession of each other species in its immediate clade are very low within all the terminal

115 clades mentioned previously (0.0018 -0.(X)36). The subclade in subg. Xeromontana containing M. obtusa, Af. lirana, Af. rugosa, Af. campanulata, and Af. lanceolata has a high divergence value for a terminal clade (0.0144). Malesherbia obtusa, however, appears to be diverged from the remaining species in the clade, since sequence divergences between it and the other species are all relatively high. Exclusion of that species showed the others to be little diverged, suggesting that Af. obtusa has a different mutation rate or that the other species are more closely related to each other than to Af. obtusa.

These phylogenetic and sequence divergence data suggest that a recent widespread event or events promoted spéciation of Malesherbiaceae in Chile and Peru. Subgenus

Xeromontana, a native of the Norte Grande, marginal desert, Norte Chico, and central Chile and adjacent Argentina, is a particularly useful clade for placing times on cladistic events.

Because Af. campanulata and Af. lanceolata, two montane species endemic to the Andes of the Norte Chico, grow in areas previously occupied by Pleistocene glaciers (Paskoff 1977), they almost certainly evolved after the retreat of these glaciers. The very close relationship between these species and Af. lirana and Af. rugosa as well as the very low sequence divergence values among the four indicate that they are all approximately the same age, suggesting they evolved at the end of the Pleistocene or beginning of the Holocene. If we further assume that similar divergence values indicate more or less similar times of divergence, we can extrapolate that modem species in subg. Cyanpetala from the Norte

Chico and central Chile and subg. Malesherbia of Peru also evolved in the early Holocene or late Pleistocene.

The increasing aridity following the glacial periods of the Pleistocene may have triggered the split in subg. Cyanpetala. There is ample evidence that the foothills of central

Chile and the Norte Chico, the current range of this taxon, were much more humid during glacial times. No species of Malesherbia today tolerates such moist conditions, so the range of this basal lineage during much of the Pliocene and Pleistocene almost certainly was

116 greatly reduced or pushed much to the north, where the climate may have been somewhat drier. Such unfavorable conditions may also have driven some species in this subgenus to extinction, which may explain why this old clade has only two species today. The advent of

modem dry conditions in central Chile and the Norte Chico have only recently allowed the expansion of Malesherbiaceae in those areas. This scenario may be testable. If the ancestor

of the two modem species did find refuge in the north, it is likely that the more southem of

the two. M. Unearifolia, will show signs of recent divergence. There are in fact populations

of plants having traits of both M. Unearifolia and M. paniculata, suggesting that perhaps

they are not completely distinct morphologically. Population genetics studies may shed

more light on this situation.

The stimuli for spéciation in Peru appear to be more complex. Only M. arequipensis of

the modem species of subg. Malesherbia is not a narrow endemic; it is unlikely that any of

the other species were once more widespread. The well supported (A/, scarlatiflora - M.

weberbaueri) clade contains four easily distinguished species with almost identical ITS

sequences. All four are natives of isolated inter-andean valleys in central Pern and near the

city of Lima. Long-distance dispersal (perhaps via the capsules, which have recurved valves)

among these isolated valleys after they developed arid climates following Pleistocene climate

fluctuations likely triggered spéciation in this group.

In the other Pemvian clade, there are three examples of disjunct, very closely related

species that appear to have diverged at the end of the Pleistocene. Malesherbia ardens is an

endemic to the Moquegua Valley ( I7°S) of southem Peru, where it is relatively common. Its

sister species, M. tocopillana, is found only near Tocopilla, Chile (22.05°S), where it

appears only rarely. Malesherbia tenuifolia, an endemic to a few dry washes in the Atacama

Desert, is sister to two southern Peruvian species. Malesherbia auristipulata grows only in

dried, cracked mud in the Azapa valley of northem Chile, and its sister, M. turbinea, is a

very rare endemic to the dry shore of a lake in southem Pern. With the exception of M.

117 arequipensis, all of these species are narrow endemics, and what appear to be appropriate habitats exist between populations of the sister species. It is therefore very difficult to imagine that the ancestors of these species had broader distributions disrupted by the increased aridity at the end of the Pleistocene. Long distance dispersal from Peru to Chile may be the best explanation for this distribution pattern.

Using these data placing the divergence of sister species in the late Pleistocene or early

Holocene and the divergence of the family itself from Tumeraceae in the late Miocene to early Pliocene establishes a time frame within which the timing of the development of the subgenera can be studied. The relationships among the subgenera are completely resolved, but supports for these relationships are low (Figure 5.2), suggesting that the subgenera developed rapidly.

If Malesherbiaceae itself evolved after the mid-Miocene desiccation of westem South

America, perhaps around the Terminal Miocene Event (5 mya), this proposed early rapid radiation must have occurred in the Pliocene or very early Pleistocene. Again, because

Malesherbiaceae are xerophytic plants, we will assume that climate changes affecting aridity levels impacted the evolution of the subgenera. Data indicate that after the Terminal Miocene event there was a Pliocene mesic period, which probably did not provide early

Malesherbiaceae with ideal conditions for rapid expansion. The large number of mutations and morphological differences separating Malesherbiaceae from Tumeraceae may be attributable to extinctions and lack of diversifîcation of the group, both which may have been promoted by the more humid conditions of the early Pliocene. This mesic period was followed by increased aridity (Mercer 1984, Paskoff 1973). The entire Andean Cordillera in the area of interest also uplifted during the Pliocene. Such broad-scale uplifts likely intensified aridity on the mountains’ westem flanks (Simpson 1983), which may have stimulated the formation of the major lineages of Malesherbiaceae.

118 It was at this time that the most important biogeographical feature of the family, the disjunction around the Atacama Desert, developed. It is clear from the phylogeny that the family evolved in Chile and later spread to Peru. The Atacama Desert is a biogeographical barrier: only one lineage is in Peru, there are no Malesherbiaceae in the Norte Grande other than those that have most likely arrived recently, and the areas north and south of the desert do not share many taxa. The area may act as a barrier to dispersal and range expansion on a number of fronts. First, the deep transverse gorges carved during the Pliocene likely impede dispersal via land animals. Second, the area is extraordinarily arid, which is certainly a modem barrier to dispersal and range expansion. Finally, and perhaps most importantly, the season in which precipitation falls changes in this area from austral winter to summer. It is likely difficult for species which are adapted to respond closely to precipitation regimes to shift to a climate with a different rainfall pattern. The shift to the arid tropical climate of Pern from the mediterranean one of Chile coincides with morphological changes in

Malesherbiaceae that are so great that early taxonomists placed the Chilean and Peruvian species in separate genera. These morphological features and the accumulation of many mutations supporting the monophyly of subg. Malesherbia indicate that the common ancestor of the species did not begin to split into the modem Peruvian lineages until long after the Pliocene arrival to the region.

In Chile, Pliocene events also influenced subg. Albitomenta, the only subgenus weakly supported by ITS and morphological data. Malesherbia fasciculata, a shrub characterized by its long intemodes and compact, head-like inflorescences, is native to the Andean foothills of the Norte Chico, whereas M. lactea, a cushion-like plant, grows in very arid montane regions over 3700 m from the Norte Grande to the Norte Chico and across the border into the Argentinean Andes. The sequence divergence value between these species is

119 an order of magnitude greater than those between sister species in other subgenera. These data suggest that they diverged very soon after splitting from the (subg. Xeromontana - subg. Malesherbia) clade in the Pliocene.

Some time after the proposed Pliocene radiation of the family, the Peruvian lineage radiated. There is very little support for splitting the Peruvian lineage into two subclades; there is also little resolution of the relationships among the clades of weakly supported group. These five lineages therefore emerged more or less simultaneously, spreading

Malesherbiaceae into the deep, dry valleys near Lima and in the central Andes (the M. tubulosa - M. splendens lineage), the Peruvian-Chilean border region (Af. auristipulata and

M. turbinea), the Andean foothills (Af. haemantha), and broad southem Peruvian and northem Chilean river valleys [the (Af. ardens -Af. tocopillana) and (Af. arequipensis Af. - tenuifolia) clades]. Late Pliocene to early Pleistocene uplifts of the Andes in Pern formed new habitats, both in valleys and mountains, which Malesherbiaceae appear to have colonized.

These late Pliocene to early Pleistocene events may also have affected subg.

Xeromontana. Malesherbia densiflora and Af. deserticola, although closely related, are morphologically more distinct from each other than sister species are in the subgenus’ other clade. In addition, their sequence divergence values resemble those of the distantly related species of subg. Albitomenta more closely than those of closely related sister species.

These data suggest that this species pair diverged much before late Pleistocene to early

Holocene events resulted in most modem species, probably the late Pliocene to early

Pleistocene.

The molecular data support this reconstruction of the development of the family after it split from Tumeraceae. The first split in the family was placed at 4.0 mya, and the split between subg. Albitomenta and (subg. Xeromontana subg. - Malesherbia) clade likely occurred 3.6 mya (Hgure 5.2). This places the family’s fîrst radiation in the mid-Pliocene.

120 The Ave Peruvian lineages are estimated to have evolved 2.1 mya (latest Pliocene to early

Pleistocene). In addition, the split between the two major clades in subg. Xeromontana was estimated to have happened 1.9 mya (early Pleistocene).

Conclusions

The distribution of Malesherbiaceae over much of arid westem South America and the

availability of geological and paleoclimatic data for the area offer an excellent opportunity to

study the biogeography of a relatively small, isolated, continental area. The concordance

between the distributions of the species and geography strongly indicates that the region’s

history has influenced the evolution of Malesherbiaceae since its estimated late Miocene

divergence from Tumeraceae. Sequence divergence data and cladogram topology informed

the reconstruction of the biogeographic history of the family, which found Malesherbiaceae

to be influenced heavily by both Pliocene and Pleistocene climatic changes.

The biological importance of the disjunction between the Pemvian, summer-rain and

Chilean, winter-rain Malesherbiaceae adds further support to other studies (Rundel et al.

1991, Arroyo et al. 1988) finding the Atacama Desert of northem Chile to be an important

biogeographic barrier in arid westem South America. Future phylogeny-based studies of

the biogeography of the entire desert are necessarily limited to the few taxa which are found

both north and south of the Atacama Desert. Because the floras north and south of the

Atacama Desert are well distinguished, however, biogeographic studies limited to either

region are possible. The isolation of the arid regions in Chile and Pern from each other and

from other dry regions in South America make the arid coastal and adjacent Andes ideal for

the study of continental biogeography.

121 Literature Cited

Alpers, C.N. & G.H. Brimhail. 1988. Middle Miocene climatic change in the Atacama Desert, northem Chile: evidence from supergene mineralization at La Escondida. Geological Society of America Bulletin 100: 1640-1656.

Arroyo, M.K.T., F A. Squeo, J.J. Armesto & C. Villagrân. 1988. Effects of aridity on plant diversity in the northem Chilean Andes: results of a natural experiment. Annals of the Missouri Botanical Garden 75: 55-78.

Ayala, F.J. 1997. Vagaries of the molecular clock. Proceedings of the National Academy of Sciences USA 94:7776-7783.

Baum, D.A., R.L. Small, & J.F. Wendel. 1998. Biogeography and floral evolution of baobabs (Adansonia, Bombacaceae) as inferred from multiple data sets. Systematic Biology 47: 181-207.

Borgel, R. 1973. The coastal desert of Chile, 111-114. In Coastal Deserts: Their Natural and Human Environment, D.H.K. Amiren and A.W. Wilson, eds. University of Arizona Press, Tucson, AZ.

Bousquet, J., S.H. Strauss, A H. Doerksen & R.A. Price. 1992. Extensive variation in evolutionary rate of rbcL gene sequences among seed plants. Proceedings of the National Academy of Sciences USA 89: 7844-7848.

Brain, C.K. 1984. The Terminal Miocene Event: a critical environmental and evolutionary episode? Pp. 491-498. In Late Cainozoic Palaeoclimates of the Southem Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

Briiggen, J. 1950. Fundamentos de la Geologia de Chile. Instituto Geogràfico Militar, Santiago, Chile.

Caviedes, C. 1973. A climatic profile of the north Chilean desert at latitude 20° South. Pp. 115-121. In Coastal Deserts: Their Natural and Human Environment, D.H.K. Amiren and A.W. Wilson, eds. University of Arizona Press, Tucson, AZ.

Crawford, D.J, T. Sang, T.F. Stuessy, S.-C. Kim, & M. Silva O. 1998. Dendroseris (Asteraceae: Lactuceae) and Robinsonia (Asteraceae: Senecioneae) on the Juan Femandez Islands: similarities and differences in biology and phylogeny. Pp. 97-119. In Evolution and Spéciation of Island Plants, T.F. Stuessy and M. Ono, eds. Cambridge University Press, New York, NY.

------, T.F. Stuessy, D.W. Haines, M B. Cosner, M. Silva O., & P. Lopez. 1992. AUozj^e diversity within and divergence among four species of Robinsonia (Asteraceae: Senecioneae), a genus endemic to the Juan Femandez Islands, Chile. American Joumal of Botany 79: %2-966. deMuizon, C. & TJ. deVries. 1985. G eolo^ and paleontology of the late Cenozoic marine deposits in the Sacaco area (Peru). Geologische Rundschau 74:547-563. diCastri, F. & E.R. Hajek. 1976. Bioclimatologia de Chile. Universidad CatôUca, Santiago, Chile.

122 Doyle, J.J. & B.S. Gaut. 2000. Evolution of genes and taxa; a primer. Plant Molecular Biology 42: 1-23.

Fay, M. P., S.M. Swensen, & M.W. Chase. 1997. Taxonomic affinities of Medusagyne oppositifolia (Medusagynaceae). Kew Bulletin 52: 111-120.

Flohn, H. 1984. Climate evolution in the southem Hemisphere and the equatorial region during the Late Cenozoic. Pp. 5-20. In Late Cainozoic Palaeoclimates of the Southem Hemisphere, J.C. Vogel, ed. Aj \. Balkema, Rotterdam.

Francisco-Ortega, J., A. Santos-Guerra, A. Hines, & R.K. Jansen. 1997. Molecular evidence for a Mediterranean origin of the Macaronesian endemic genus Argyranthemum (Asteraceae). American Joumal of Botany 84: 1595-1613.

Gaut, B.S., L.G. Clark, J.F. Wendel, & S.V. Muse. 1997. Comparisons of the molecular evolutionary process at rbcL and ndh? in the grass family (Poaceae). Molecular Biology and Evolution 14:769-777.

Grosjean, M. 1994. Paleohydrology of the Laguna Lejia (north Chilean altiplano) and climatic implications for late-glacid times. Palaeogeography, Palaeoclimatology, Palaeoecology 109: 89-100.

Heusser, C.J. 1984. Late Quatemary climates of Chile. Pp. 59-83. In Late Cainozoic Palaeoclimates of the Southem Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

Kennett, J.P. 1980. Paleoceanographic and biogeographic evolution of the southem ocean during the Cenozoic, and Cenozoic microfossil datums. Palaeogeography, Palaeoclimatology, Palaeoecology 31: 123-152.

Kessler, A. 1984. The palaeohydrology of the Late Pleistocene Lake Tauca on the Bolivian Altiplano and recent climatic fluctuations. Pp. 115-122. In Late Cainozoic Palaeoclimates of the Southem Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

Kim, S.-C., D.J. Crawford, J. Francisco-Ortega, & A. Santos-Guerra. 1996. A common origin for woody Sonchus and five related genera in the Macaronesian islands: molecular evidence for extensive radiation. Proceedings of the National Academy of Sciences USA 93:7743-7748.

Kumer, S., K. Tamura, & M. Nei. 1993. MEGA: Molecular evolutionary genetics analysis, version 1.0. The Pennsylvania State University, University Park, PA.

Lauer, W. & P. Frankenberg. 1984. Late glaciation and the development of climate in southem South America. Pp. 103-114. In Late Cainozoic Palaeoclimates of the Southem Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

Lee, N.S., T. Sang, D J. Crawford, SJI. Yeau, & S.-C. Kim. 1996. Molecular divergence between disjunct taxa in eastem Asia and eastem North America. American Joumal of Botany 83: 1373-1378.

Linacre, E. & B. Geerts. 1997. Climates and Weather Explained. Routledge, New York, NY.

123 Li, J., A.L. Bogle, & A S. Klein. 1997. Interspecific relationships and genetic divergence of the disjunct genus L/^uiit/am/^arfHamamelidaceae) inferred from DNA sequences of plastid gene matK. Rhodora 99:229-240.

Li, W.-H. 1997. Molecular evolution, pp. 216-221. Sinauer Associates, Inc. Sunderland, MA.

& M. Tanimura. 1987. The molecular clock runs more slowly in man than in apes and monkeys. Nature 326:93-96.

Lydolph, P.E. 1973. On the causes of aridity along a selected group of coasts. Pp. 67-72. In Coastal Deserts: Their Natural and Human Environments, D.H.K. Amiran & A.W. Wilson, eds. University of Arizona Press, Tucson, AZ.

Markgraf, V., J.R. Dodson, A.P. Kershaw, M.S. McGlone, & N. Nicholls. 1992. Evolution of late Pleistocene and Holocene climates in the circum-Pacific land areas. Climate Dynamics 6: 193-211.

Marquet, P.A. 1994. Diversity of small mammals in the Pacific coastal desert of Peru and Chile and in the adjacent Andean area: biogeography and community structure. Australian Joumal of Zoology 42: 527-542.

M a^ , R.C., R.B. Dunbar, & M R. Allen. 1987. Tertiary diatomite distribution in the Pisco basin, Peru. Geological Society of America Abstracts 19:761.

Mercer, J.H. 1984. Late Cainozoic glacial variations in South America south of the equator. Pp. 45-58. In Late Cainozoic Palaeoclimates of the Southem Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

Messerli, B., M. Grosjean, G. Bonani, A. Biigi, M.A. Geyh, K. Graf. K. Ramseyer, H. Romero, U, Schottererer, H, Schreier, & M, Vuille. 1993. Climate change and natural resource dynamics of the Atacama altiplano during the last 18000 yn a preliminary sythesis. Mountain Research and Development 13: 117-127.

------, M. Grosjean, & M. Vuille. 1997. Water availability, protected areas, and natural resources in the Andean Desert altiplano. Mountain Research and Development 17: 229-238.

Moreno, P.I., C. Villagrdn, P.A. Marquet, & L. Marshall. 1994. Quatemary paleobiogeography of northem and central Chile. Revista Chilena de Historia Natural 67: 487-502.

Mortimer, C. & N. Saric. 1975. Cenozoic studies in northemmost Chile. Geologische Rundschau 64: 395-420.

Muse, S.V. 2000. Examining rates and pattems of nucleotide substitution in plants. Plant Molecular Biology 42:25-43.

Nuhez, L., J. Varela, R. Casamiquela, V. Schiappacasse, H. Niemeyer, & C. Villagrdn. 1994. Cuenca de Ta^atagua en Chile: el ambiente del Pleistoceno superior y ocupaciones humanos. Revista Chilena de Historia Natural 67:503-519.

124 Paskoff, R. 1973. The Plio-Quatemary climatic changes along the semlarid seaboard of Chile. F^. 147-151. In Coastal Deserts: Their Natural and Human Environments, D.H.K. Amiran and A.W. Wilson, eds. University of Arizona Press, Tucson, AZ.

------. 1977. Quatemary of Chile: the state of research. Quaternary Research 8:2-31.

Pérez, C. & C. Villagidn. 1994. bifluencia del clima en el cambio flonstico, vegetacional y edàfîco de los bosques de “olivillo” {Aextoxicon punctatum R. et Pav.) de la Cordillera de la Costa de Chile: implicancias biogeogrâficas. Revista Chilena de Historia Natural 67: 77-90.

Peterson, U. 1958. Structure and uplift of the Andes of Peru, Bolivia, and Chile and adjacent Argentina. Boletin de la Sociedad Geolôgica del Peru 33:57-129.

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

Rundel, P.W., M.O. Dillon, B. Palma, H.A. Mooney, S.L. Gulman, & J R. Ehleringer. 1991. The phytogeography and ecology of the coastal Atacama and Peruvian deserts. Aliso 13: 1-49.

Sanderson, M.J. 1998. Estimating rate and time in molecular phylogenies: beyond the molecular clock? Pp. 242-264. In Molecular Systematics of Plants II: DNA Sequencing, D.E. Soltis, P S. Soltis, & J.J. Doyle, eds. Kluwer Academic Publishers, Boston, MA.

Sang, T., D.J. Crawford, S.-C. Kim, & T.F. Stuessy. 1994. Radiation of the endemic genus Dendroseris (Asteraceae) on the Juan Femandez Islands: evidence from sequences of the ITS regions of nuclear ribosomal DNA. American Joumal of Botany 81: 1494-1501.

------, D J. Crawford, & T.F. Stuessy. 1995. ITS sequences and the phylogeny of the genus Robinsonia (Asteraceae). Systematic Botany 20: 55-64.

Seltzer, G.O. 1990. Recent glacial history and paleoclimate of the Pemvian-Bolivian Andes. Quatemary Science Review 9: 137-152.

Simpson, B 3 . 1975. Pleistocene changes in the flora of the high tropical Andes. Paleobiology 1: 273-294.

------. 1979. Quatemary biogeography of the high montane regions of South America. Pp. 157-188. In The South iWerican Herpetofauna: Its Origins, Evolution, and Dispersal, W.E. Duellman, ed. University of Kansas Museum of Natural History, Lawrence, KS.

------. 1983. An historical phytogeography of the high Andean flora. Revista Chilena de Historia Natural 56:109-122.

------& C.A. Todzia. 1990. Pattems and processes in the development of the high Andean flora. American Joumal of Botany 77: 1419-1432.

Simpson Vuilleumier, B. 1971. Pleistocene changes in the fauna and flora of South America. Science 173:771-780.

125 Tajima, F. 1993. Simple methods for testing the molecular evolutionary clock hypothesis. Genetics 135:599-607.

Wu, C.-I. & W.-H. Li. 1985. Evidence for higher rates of nucleotide substitution in rodents than in man. Proceedings of the National Academy of Sciences USA 82: 1741-1745.

Xiang, Q.-Y., D.J. Crawford, A.D. Wolfe, Y.-C. Tang, & C.W. DePamphilis. 1998. Origin and biogeography of Aesculus L. (Hippocastanaceae): a molecular phylogenetic perspective. Evolution 52:988-997.

Zinsmeister, W.J. 1978. Effect of formation of the west antarctic ice sheet on shallow-water marine faunas of Chile. Antarctic Joumal of the United States 13:25-26.

Zink, R.M., R.C. Blackwell-Rago, & F. Ronquist. 1999. The shifting roles of dispersal and vicariance in biogeography. Proceedings of the Royal Society of Biological Sciences London Series B 267:497-503.

126 L im a 1 2 Peru I

16

-a?

20

Pacific • - Ocean Chile/ 24

28

32

Santiago

36

Figure 5.1. Distribution of Malesherbiaceae. 127 Figure 5.2. Strict consensus cladogram of four most parsimonious trees derived from ITS sequence data. Bremer and Jackknife support values are reported above each branch; values below some of the branches are divergence times (mya) calculated using the molecular clock. Subgenera are labeled in bold-face, and their distributions are indicated in parentheses.

128 Piriqueta Tiimera M . Unearijbtia 5.100 Cyanpetala ______M. linamfolia 1.63 (mid-elevations of mediterranean Chile) M. Unearifolia M . paniculata M. humilis M . humilis 15.100 Parvistella 4.93 M. humilis 4.0 (coast & southem foothills of Chile) M . humilis M . humilis 5.100 M. fasciculata Albitomenta l.<50 M. fasciculata 3,81 (foothills & high elevations 5.99 M . lactea of Chile & w. Argentina) M. lactea 5.96 M . densiflora 1.63 M . deserticola M . deserticola 5,91 l.<50 M. obtusa 1.9 M . lirana v. lirana 3.6 M . lirana v . subglab. 5,99 1,<50 M . rugosa 1,<50 Xeromontana M. rugosa (Andes & very M . rugosa dry desert of 2,84 M . campanulata Chile & 1.63 M. lanceolata M . lanceolata 1,<50 l.<50 w. Argentina) M. lanceolata M . lanceolata M. tubulosa M. tubulosa M . scarlatiflora 5.97 M . scarlatiflora (inter-Andean M . scarlatiflora M . scarlatiflora valleys) M. scarlatiflora M . splendens 5.99 M . uteberbaueri 2.1 M . haemantha Malesherbia 2.77 M . auristipulata (Peru) M . auristipulata M . turbinea M . tocopillana 1.63 3.95 M . ardens (southem M. ardens Peru) M . ardens M . arequipensis M . arequipensis 5,100 M . arequipensis M . angustisecta 1,63 M . tenuifidia M . tenuifolia M . tenuifolia Figure 5.2 129 1 2 Peru

16

20 Pacific Ocean Chile 24

Subg. Cyanpetala

:: Suh^. Parvistella 28 Subg. Albitomenta

I Subg. Xeromontana

32 Subg. Malesherbia inter-andean valleys Subg. Malesherbia southem Peru &c northem Chile 36

Figure 5.3. Distributions of the subgenera of Malesherbiaceae. 130 Figure 5.4. The four high elevation regions in the range of Malesherbiaceae. The Norte

Grande, marginal desert, and Norte Chico are indicated (Paskoff 1977); central Chile

(33°-39°S) is not shown.

131 1 2

16

Pacific Ocean 20

Norte Grande 0 '

24 Cordillera Occidental

Principal Cordillera

Cordillera Oriental 28 Altiplano

Marginal desert

32 Norte Chico

36

Figure 5.4 132 Species Accession

Af. angustisecta KMG & Refulio R. 199 M. ardens KMG 184 KMG 185 KMG 187 M. arequipensis Dillon, Sagastegui, & Santisteban 4793 Af. auristipulata KMG 65 Af. campanulata Ricardi, Marticorena, & Matthei 1748 Af. densiflora KMG 115 Af. deserticola KMG 93 Dillon & Dillon 6013 Af. fasciculata KMG 151 Marticorena, Stuessy, & Baeza 9803 Af. haemantha Hutchinson 1278 (UC) Af. humilis var. parviflora KMG 190 Stuessy 9883 Stuessy 9892 Af. /acfca KMG 44 Stuessy 9876 Af. lanceolata Stuessy 12809 KMG 133 Af. Unearifolia KMG 36 Af. lirana van lirana Stuessy 9814 Af. lirana var. subglabrifolia Hjertling 6312(A) Af. obtusa KMG 119a Af. paniculata KMG 28 Af. rwgOM KMG 37 KMG 118 Af. scarlatiflora KMG & Roque 350 Af. splendens KMG 188 Af. tocopillana Dillon & Dillon 5719 Af. tubulosa KMG & Roque 354 KMG & Bedoya 362 Af. turbinea KMG & Refulio 198 Af. weberbaueri var. weberbaueri KMG 288

Figure 5.5. Accessions used to calculate average sequence divergences for the modified Li relative rates test All collections are deposited at OS except where noted.

133 Species Group Average divergence Range

M. paniculata 0.0036 M. humilis 0.0036 - M. tubulosa 0.0036 0 M. turbinea 0.0036 - M. tocopillana 0.0036 — M. angustisecta 0.0018 0 - 0.0036 M. obtusa ’ 0.0144 0.0108 - 0.0255 M. lirana ** 0.0024 0-0.0036 M. densiflora 0.0255 - M. lactea 0.0367 -

* * Divergences between M. obtusa and M. lirana var. lirana, M. rugosa, M. campanulata, and M. lanceolata ** Divergences between M. lirana var. lirana and M. rugosa, M. campanulata, and M. lanceolata.

Figure 5.6. Average sequence divergence values for terminal clades. The average divergence values between the listed species and a single accession of all the other species in its immediate clade were calculated. Differences in divergence values for the various accessions of a single species were minimal and are not considered here. For those terminal clades containing more than two taxa, the range of divergence values is given.

134 LIST OF REFERENCES

Alpers, C.N. & G.H. Brimhail. 1988. Middle Miocene climatic change in the Atacama Desert, northern Chile: evidence from supergene mineralization at La Escondida. Geological Society of America Bulletin 100: 1640-1656.

Arbo, M.M. 1995. Tumeraceae Parte I. Piriqueta. Flora Neotropica. Monograph 67. The New York Botanical Garden, Bronx, New York.

Arft, A M. & T.A. Ranker. 1998. Allopolyploid origin and population genetics of the rare orchid Spiranthes diluvialis. American Journal of Botany 85: 110-122.

Arroyo, M.K.T., F A. Squeo, JJ. Armesto & C. Villagrân. 1988. Effects of aridity on plant diversity in the northern Chilean Andes: results of a natural experiment. Annals of the Missouri Botanical Garden 75: 55-78.

Ayala, F.J. 1997. Vagaries of the molecular clock. Proceedings of the National Academy of Sciences USA 94: 7776-7783.

Baldwin, B.G. 1997. Adaptive radiation of the Hawaiian silversword alliance: congruence and conflict of phylogenetic evidence from molecular and non-molecular investigations. Pp. 103-128. In Molecular evolution and adaptive radiation, T.J. Givnish & K.J. Sytsma, eds. Cambridge University Press, Cambridge, UK.

Baum, D.A., R.L. Small, & J.F. Wendel. 1998. Biogeography and floral evolution of baobabs {Adansonia, Bombacaceae) as inferred from multiple data sets. Systematic Biology 47: 181-207.

Bernhard, A. 1999. Flower structure, development, and systematics in Passifloraceae and in A^arid (Flacourtiaceae). International Journal of Plant Science 160:135-150.

Bôrgel, R. 1973. The coastal desert of Chile. Pp. 111-114. In Coastal Deserts: Their Natural and Human Environment, D.H.K. Amiren and A.W. Wilson, eds. University of Arizona Press, Tucson, AZ.

135 Bousquet, J., S.H. Strauss, A.H. Doerksen & R.A. Price. 1992. Extensive variation in evolutionary rate of rbcL gene sequences among seed plants. Proceedings of the National Academy of Sciences USA 89: 7844-7848.

Brain, C.K. 1984. The Terminal Miocene Event: a critical environmental and evolutionary episode? Pp. 491-498. In Late Cainozoic Palaeoclimates of the Southern Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

Bretô, M.P., M.J. Asins, & E.A. Carbonell. 1993. Genetic variability in Lycopersicon species and their genetic relationships. Theoretical and Applied Genetics 86: 113-120.

Brüggen, J. 1950. Fundamentos de la Geologia de Chile. Institute GeogràHco Militar, Santiago. Chile.

Brummitt, R.K. & C.E. Powell. 1992. Authors of Plant Names. Royal Botanic Gardens, Kew.

Carlquist, S. 1974. Island biology. Columbia University Press, New York, NY.

Caviedes, C. 1973. A climatic profîle of the north Chilean desert at latitude 20° South. Pp. 115-121. In Coastal Deserts: Their Natural and Human Environment, D.H.K. Amiren and A.W. Wilson, eds. University of Arizona Press, Tucson, AZ.

Crawford, D.J., E. Landolt, D.H. Les, & E. Tepe. 1997. Allozyme variation and the taxonomy of Wolffiella. Aquatic Botany 58:43-54.

------T. Sang, T.F. Stuessy, S.-C. Kim, & M. Silva O. 1998. Dendroseris (Asteraceae: Lactuceae) and Robinsonia (Asteraceae: Senecioneae) on the Juan Fernandez Islands: similarities and differences in biology and phylogeny. Pp. 97-119. In Evolution and Spéciation of Island Plants, T.F. Stuessy and M. Ono, eds. Cambridge University Press, New York, NY.

------, A. Segâstegui A., T.F. Stuessy, & I. Sânchez V. 1993. Variaciôn aloenzimàtica en la rara especie endémica peruana Chuquiraga oblongifolia (Asteraceae). Amaldoa 1: 73-76.

------, T.F. Stuessy, D.W. Haines, M B. Cosner, M. SilvaO., & P. Lopez. 1992. Allozyme diversity within and divergence among four species of Robinsonia (Asteraceae: Senecioneae), a genus endemic to the Juan Fernandez Islands, Chile. American Journal of Botany 79:962-966.

Cronquist, A. 1981. An integrated system of classification of flowering plants. Columbia University Press, New York, NY.

136 Cullings, K.W. 1992. Design and testing of a plant-specifîc PCR primer for ecological and evolutionary studies. Molecular Ecology 1: 233-240

Davis, B.J. 1964. Dis-electrophoresis n. Methods and application to human serum proteins. Annals of the New York Academy of Science 21:404-427.

Davis, JJ. 1995. A phylogenetic structure for the Monocotyledons, as inferred from chloroplast DNA restriction site variation, and a comparison of measures of clade support. Systematic Botany 20: 503-527. deMuizon, C. & T.J. deVries. 1985. Geology and paleontology of the late Cenozoic marine deposits in the Sacaco area (Peru). Geologische Rundschau 74: 547-563. diCastri, F. & E.R. Hajek. 1976. Bioclimatologia de Chile. Universidad Catôlica, Santiago, Chile.

Doyle, J.J. & B.S. Gaut. 2000. Evolution of genes and taxa: a primer. Plant Molecular Biology 42: 1-23.

------& J.L. Doyle. 1987. A rapid DNA Isolation procedure for small quantities of fresh leaf material. Phytochemical Bulletin 19: 11-15.

Farris, J.S. 1988. Hennig86 version 1.5. MS-DOS program. Port Jefferson, New York.

------1996. Parsimony jackknifing outperforms neighbor-joining. Cladistics 12: 99-124.

Fay, M. F., S.M. Swensen, & M.W. Chase. 1997. Taxonomic affinities of Medusagyne oppositifolia (Medusagynaceae). Kew Bulletin 52: 111-120.

Felsenstein, J. 1978. Cases in which parsimony or compatibility methods will be positively misleading. Systematic Zoology 27:401-410.

Flohn, H. 1984. Climate evolution in the southern Hemisphere and the equatorial region during the Late Cenozoic. Pp. 5-20. In Late Cainozoic Palaeoclimates of the Southern Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

Francisco-Ortega, J., A. Santos-Guerra, A. Hines, & R.K. Jansen. 1997. Molecular evidence for a Mediterranean origin of the Macaronesian endemic genus Argyranthemum (Asteraceae). American Journal of Botany 84: 1595-1613.

------, R.K. Jansen, & A. Santos-Guerra. 1996. Chloroplast DNA evidence of colonization, adaptive radiation, and hybridization in the evolution of the Macaronesian flora. Proceedings of the National Academy of Science USA 93:4085-4090.

137 Gajardo, R. 1993. La vegetaciôn natural de Chile; claslficaciôn y distribuciôn geogrâfîca. Editorial Universitaria, Santiago, Chile.

Gaut, B.S., L.G. Clark, J.F. Wendel, & S.V. Muse. 1997. Comparisons of the molecular evolutionary process at rbcL and ndhF in the grass family (Poaceae). Molecular Biology and Evolution 14:769-777.

Gay, C. 1846. Historia fisica y politica de Chile. Tomo segundo. Paris.

Godt, M.J.W., B.R. Johnson, & J.L. Hamrick. 1996. Genetic diversity and population size in four rare southern Appalachian plant species. Conservation Biology 10: 796-805.

J. Walker, & J.L. Hamrick. 1995. Genetic diversity in a threatened wetland species, Helonias bullata (Liliaceae). Conservation Biology 9: 596-604.

------1997. Genetic diversity in the endangered lily Harperocallis flava and a close relative, Tofieldia racemosa. Conservation Biology 11: 361-366.

Goloboff, P.A. 1993. NONA, version 1.6. MS-DOS program. Published by the author, San Miguel de Tucumdn, Argentina.

------1993. Pee-Wee version 2.15. MS-DOS program. Published by the author, San Miguel de Tucumàn, Argentina.

Gottlieb, L.D. 1973. Genetic differentiation, sympatric spéciation, and the origin of a diploid species of Stephanomeria. American Journal of Botany 60: 545-553.

------1981. Electrophoretic evidence and plant populations. Progress in Phytochemistry 7: 1-46.

Grosjean, M. 1994. Paleohydrology of the Laguna Lejfa (north Chilean altiplano) and climatic implications for late-glacial times. Palaeogeography, Palaeoclimatology, Palaeoecology 109: 89-100.

Hamrick, J.L. & M.J.W. Godt. 1990. Allozyme diversity in plant species. Pp. 43-63. In Plant population genetics, breeding, and genetic resources, A.H.D. Brown , M.T. Clegg, A.L. Kahler, and B.S. Weir, eds. Sinauer Associates, Sunderland, Massachusetts.

& -----. 1997. Effects of life history traits on genetic diversity in plant species. Pp. 103-118. In Plant life histories: ecology, phylogeny, and evolution, J. Silvertown, M. Franco, & J.L. Harper, eds. Cambridge University Press, Cambridge, UK.

Hasegawa, M., H. Kishino, & T. Yano. 1985. Dating of the human-ape splitting by a molecular clock of mitochondrial DNA. Journal of Molecular Evolution 22: 160-174.

138 Hendy, M.D. & D. Penny. 1989. A framework for the quantitative study of evolutionary trees. Systematic Zoology 38:297-309.

Heusser, C.J. 1984. Late Quaternary climates of Chile. Pp. 59-83. In Late Cainozoic Palaeoclimates of the Southern Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

Hodges, S.A. 1997. Rapid radiation due to a key innovation in columbines (Ranunculaceae: Aquilegia) pp. 391-406. In Molecular evolution and adaptive radiation, T.J. Givnish & K.J. Sytsma, eds. Cambridge University Press, Cambridge, UK.

Huelsenbeck, J.P. 1997. Is the Felsenstein zone a fly trap? Systematic Biology 46:69-74.

------1998. Systematic bias in phylogenetic analysis: is the Strepsiptera problem solved? Systematic Biology 47:519-537.

Hunziker, A.T. & L A. Espinar. 1967. Nota aclaratoria sobre las Malesherbiaceae argentinas y una clave para su identificaciôn. Kurtziana 4: 83-86.

Johnston, I.M. 1929. Papers on the flora of northern Chile. Contributions from the Gray Herbarium of Harvard University 85:1-171.

Kennett, J.P. 1980. Paleoceanographic and biogeographic evolution of the southern ocean during the Cenozoic, and Cenozoic microfossil datums. Palaeogeography, Palaeoclimatology, Palaeoecology 31: 123-152.

Kessler, A. 1984. The palaeohydrology of the Late Pleistocene Lake Tauca on the Bolivian Altiplano and recent climatic fluctuations. Pp. 115-122. In Late Cainozoic Palaeoclimates of the Southern Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

Kim, S.-C, D J. Crawford, J. Francisco-Ortega, & A. Santos-Guerra. 1996. A common origin for woody Sonchus and fîve related genera in the Macaronesian islands: molecular evidence for extensive radiation. Proceedings of the National Academy of Sciences USA 93:7743-7748.

Kumer, S., K. Tamura, & M. Nei. 1993. MEGA: Molecular evolutionary genetics analysis, version 1.0. The Pennsylvania State University, University Park, PA.

Lauer, W. & P. Frankenberg. 1984. Late glaciation and the development of climate in southern South America. Pp. 103-114. In Late Cainozoic Palaeoclimates of the Southern Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

Leberg, P.L. 1992. Effects of population bottlenecks on genetic diversity as measured by allozyme electrophoresis. Evolution 46:477-494.

139 Lecointre, G., H. Philippe, H.L.V. Lê, & H. LeGuyader. 1993. Species sampling has a major impact on phylogenetic inference. Molecular Phylogenetics and Evolution 2; 205-224.

Lee, M.S., T. Sang, D.J. Crawford, S.H. Yeau, & S.-C. Kim. 1996. Molecular divergence between disjunct taxa in eastern Asia and eastern North America. American Journal of Botany 83: 1373-1378.

Lewis, P.O. & R. Whitkus. 1989. GeneStat (version 3.3) for microcomputers. American Society Plant Taxonomy Newsletter 2: 15-16.

Li, J., A.L. Bogle, & A S. Klein. 1997. Interspecific relationships and genetic divergence of the disjunct genus Liquidambar (Hamamelidaceae) inferred from DNA sequences of plastid gene morK. Rhodora 99:229-240.

Li, W.-H. & M. Tanimura. 1987. The molecular clock runs more slowly in man than in apes and monkeys. Nature 326: 93-96.

------. 1997. Molecular evolution, pp. 216-221. Sinauer Associates, Inc. Sunderland, MA.

Linacre, E. & B. Geerts. 1997. Climates and Weather Explained. Routledge, New York, NY.

Lydolph, P.E. 1973. On the causes of aridity along a selected group of coasts. Pp. 67-72. In Coastal Deserts: Their Natural and Human Environments, D.H.K. Amiran & A.W. Wilson, eds. University of Arizona Press, Tucson, AZ.

Markgraf, V., J R. Dodson, A.P. Kershaw, M.S. McGlone, & N. Nicholls. 1992. Evolution of late Pleistocene and Holocene climates in the circum-Pacific land areas. Climate Dynamics 6: 193-211.

Marquet, P.A. 1994. Diversity of small mammals in the Pacifrc coastal desert of Peru and Chile and in the adjacent Andean area: biogeography and community structure. Australian Journal of Zoology 42:527-542.

Marty, R.C., R.B. Dunbar, & M R. Allen. 1987. Tertiary diatomite distribution in the Pisco basin, Peru. Geological Society of America Abstracts 19:761.

Mercer, J.H. 1984. Late Cainozoic glacial variations in South America south of the equator. Pp. 45-58. In Late Cainozoic Palaeoclimates of the Southern Hemisphere, J.C. Vogel, ed. A.A. Balkema, Rotterdam.

140 Messerli, B., M. Grosjean, & M. Vuille. 1997. Water availability, protected areas, and natural resources in the Andean Desert altiplano. Mountain Research and Development 17:229-238.

------M. Grosjean, G. Bonani, A. Bilgi, M.A. Geyh, K. Graf, K. Ramseyer, H. Romero, U. Schottererer, H. Schreier, & M. Vuille. 1993. Climate change and natural resource dynamics of the Atacama altiplano during the last 18000 yr: a preliminary sythesis. Mountain Research and Development 13: 117-127.

Moreno, P.I., C. Villagrân, P.A. Marquet, & L. Marshall. 1994. Quaternary paleobiogeography of northern and central Chile. Revista Chiiena de Historia Natural 67: 487-502.

Mort, M.E., P.S. Soltis, D.E. Soltis & M.L. Mabry. 2000. Comparison of three methods for estimating internal support on phylogenetic trees. Systematic Biology 49: 160-170.

Mortimer, C. & N. Saric. 1975. Cenozoic studies in northernmost Chile. Geologische Rundschau 64: 395-420.

Muse, S.V. 2000. Examining rates and patterns of nucleotide substitution in plants. Plant Molecular Biology 42: 25-43.

Nei, M. 1973. Analysis of gene diversity in subdivided populations. Proceedings of the National Academy of Sciences USA 70: 3321-3323.

------T. Maruyama, & R. Chakraborty. 1975. The bottleneck effect and genetic variability in populations. Evolution 29: 1-10.

Nunez, L., J. Varela, R. Casamiquela, V. Schiappacasse, H. Niemeyer, & C. Villagrân. 1994. Cuenca de Taguatagua en Chile: el ambiente del Pleistocene superior y ocupaciones humanos. Revista Chiiena de Historia Natural 67: 503-519.

Palmer, J.D. 1982. Isolation and structural analysis of chloroplast DNA. Nucleic Acids Research 10: 167-186.

Paskoff, R. 1973. The Plio-Quatemary climatic changes along the semiarid seaboard of Chile. Pp. 147-151. In Coastal Deserts: Their Natural and Human Environments, D.H.K. Amiran and A.W. Wilson, eds. University of Arizona Press, Tucson, AZ.

------. 1977. Quaternary of Chile: the state of research. Quaternary Research 8: 2-31.

Percy, D.M. & Q.C.B. Cronk. 1997. Conservation in relation to mating system in Nesohedyotis arborea (Rubiaceae), a rare endemic tree from St. Helena. Biological Conservation 80: 135-145.

141 Pérez, C. & C. Villagrân. 1994. Influencia del clima en el cambio florfstico, vegetacional y edàfico de los bosques de “olivillo” {Aextoxicon punctatum R. et Pav.) de la Cordillera de la Costa de Chile: implicancias biogeogrâficas. Revista Chiiena de Historia Natural 67: 77-90.

Peterson, U. 1958. Structure and uplift of the Andes of Peru, Bolivia, and Chile and adjacent Argentina. Boletm de la Sociedad Geologica del Peru 33: 57-129.

Poeppig, E. 1829. Schreiben des jezt in Chile reisenden Hm. Dr. Poppig. Notizen aus dem Gebiete der Natur- und Heilkunde 23: 290-293.

Purdy, B.G., R.J. Bayer, & S.E. MacDonald. 1994. Genetic variation, breeding system evolution, and conservation of the narrow sand dune endemic Stellaria arenicola and the widespread S. longipes (Caryophyllaceae). American Journal of Botany 81:904-911.

Qiu, Y.-L. & C.R. Parks. 1994. Disparity of allozyme variation levels in three Magnolia (Magnoliaceae) species from the southeastern United States. American Journal of Botany 81: 1300-1308.

Quiros, C.F., S B. Brush, D.S. Douches, K.S. Zimmerer, & G. Huestis. 1990. Biochemical and folk assessment of variability of Andean cultivated potatoes. Economic Botany 44: 254-266.

Rauh, W. 1985. The Peruvian-Chilean deserts. Pp. 239-267. In Hot deserts and arid shrublands, M. Evenari, I. Noy-Meir, & D.W. Goodall, eds. Elsevier, Amsterdam.

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

Ricardi, M. 1967. Revisiôn taxonômica de las Malesherbiaceas. Gayana Botânica No. 16: 1-139.

Richards, C. & P.L. Leberg. 1996. Temporal changes in allele frequencies and a population's history of severe bottlenecks. Conservation Biology 10: 823-839.

Rick, C M. & J.F. Fobes. 1975. Allozyme variation in the cultivated tomato and closely related species. Bulletin of the Torrey Botanical Club 102: 375-384.

------E. Kesicki, J.F. Fobes, & M. Holle. 1976. Genetic and biosystematic studies on two new sibling species of Lycopersicon from interandean Peru. Theoretical and Applied Genetics 47:55-68.

Rieseberg, L.H., S. Zona, L. Aberbom, & TD . Martin. 1989. Hybridization in the island endemic, Catalina mahogany. Conservation Biology 3: 1-7.

142 Rohlf, F J. 1998. NTSYSpc version 2.0. Exeter Software, Setauket, NY.

Rundel, P.W., M.O. Dillon, B. Palma, H.A. Mooney, S.L. Gulmon, & J R. Ehleringer. 1991. The phytogeography and ecology of the coastal Atacama and Peruvian deserts. Aliso 13: 1-49.

Sahley, C.T. 1996. Bat and hummingbird pollination of an autotetraploid columnar cactus, Weberbauerocereus weberbaueri (Cactaceae). American Journal of Botany S3: 1329-1336.

Sanderson, M.J. 1998. Estimating rate and time in molecular phylogenies: beyond the molecular clock? Pp. 242-264. In Molecular Systematics of Plants II: DNA Sequencing, D.E. Soltis, P.S. Soltis, & J.J. Doyle, eds. Kluwer Academic Publishers, Boston, MA.

Sang, T., D.J. Crawford, & T.F. Stuessy. 1995. ITS sequences and the phylogeny of the genus Robinsonia (Asteraceae). Systematic Botany 20: 55-64.

------, D.J. Crawford, S.-C. Kim, & T.F. Stuessy. 1994. Radiation of the endemic genus Dendroseris (Asteraceae) on the Juan Fernandez Islands: evidence from sequences of the ITS regions of nuclear ribosomal DNA. American Journal of Botany 81 : 1494-1501.

SAS. 1989. SAS/STAT user’s guide, version 6. Fourth edition, volumes 1 & 2. SAS Institute Inc., Cary, NC.

Schwaegerle, K.E. & B.A. Schaal. 1979. Genetic variation and founder effect in the pitcher plant Sarracema purpurea L. Evolution 33: 1210-1218.

Seltzer, G.O. 1990. Recent glacial history and paleoclimate of the Peruvian-Bolivian Andes. Quaternary Science Review 9: 137-152.

Siddall, M E. & M.F. Whiting. 1999. Long-branch abstractions. Cladistics 15:9-24.

Simpson, B.B. 1975. Pleistocene changes in the flora of the high tropical Andes. Paleobiology 1:273-294.

------& C.A. Todzia. 1990. Patterns and processes in the development of the high Andean flora. American Journal of Botany 77: 1419-1432.

------. 1983. An historical phytogeography of the high Andean flora. Revista Chiiena de Historia Natural 56: 109-122.

143 1979. Quaternary biogeography of the high montane regions of South America. Pp. 157-188. In The South American Herpetofauna: Its Origins, Evolution, and Dispersal, W.E. Duellman, ed. University of Kansas Museum of Natural History, Lawrence, KS.

Simpson Vuilleumier, B. 1971. Pleistocene changes in the fauna and flora of South America. Science 173:771-780.

Singh, S.P., R. Nodari, & P. Gepts. 1991. Genetic diversity in cultivated common bean: I. Allozymes. Crop Science 31: 19-23.

Sneath, P.H.A. & R.R. Sokal. 1973. Numerical Taxonomy. W.H. Freeman and Company, San Francisco, California.

Spencer, K.C. & D.S. Seigler. 1978. Cyanogenic glycosides of Malesherbia. Biochemical Systematics and Ecology 13:23-24.

Squeo, F.A., R. Osorio, & G. Arancio. 1994. Flor de Los Andes de Coquimbo: Cordillera de Doha Ana. Editorial Universitaria, Chile.

Swofford, D.L. 1999. PAUP*. Phylogenetic Analysis Using Parsimony (* and Other Methods). Version 4. Sinauer Associates, Sunderland, MA.

Tajima, F. 1993. Simple methods for testing the molecular evolutionary clock hypothesis. Genetics 135: 599-607.

Takhtajan, A.L. 1980. Outline of the classification of flowering plants (Magnoliophyta). The Botanical Review 46(3): 225-359.

Thompson, J.D., D.G. Higgins, & T.J. Gibson. 1994. Clustal W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position specific gap penalties and weight matrix choice. Nucleic Acids Research 22:4673-4680.

Tohme, J., D. Orlando G., S. Beebe, & M.C. Duque. 1996. AFLP analysis of gene pools of a wild bean core collection. Crop Science 36: 1375-1384.

Torres, R.A., M. Ganal, & V. Hemleben. 1990. GC balance in the internal transcribed spacers ITS 1 and ITS 2 of nuclear ribosomal RNA genes. Journal of Molecular Evolution 30: 170-181.

Trewartha, G.T. 1966. The Earth’s problem climates. The University of Wisconsin Press, Madison, Wisconsin.

Weeden, N.F. & J.F. Wendel. 1989. Genetics of plant isozymes. Pp. 46-72. In Isozymes in plant biology, D.E. Soltis & P.S. Soltis, eds. Dioscorides Press, Portland, Oregon.

144 Wendel, J.F. & N.F. Weeden. 1989. Visualization and interpretation of plant isozymes. Pp. 5-45. In Isozymes in plant biology, D.E. Soltis & P.S. Soltis, eds. Dioscorides Press, Portland, Oregon.

White, T J., T. Bruns, S. Lee, & J. Taylor. 1990. Amplification and direct sequencing of fungal ribosomal RNA genes for phylogenetics. Pp. 315-322. In PCR protocols; A guide to methods and amplifications. M. Innis, D. Gelfand, J. Sninsky, and T. White, eds. Academic Press, San Diego, CA.

Whitkus, R. & J.G. Packer. 1984. A contribution to the taxonomy of the Carex macloviana aggregate (Cyperaceae) in western Canada and Alaska. Canadian Journal of Botany 62: 1592-1607.

Wu, C.-I. & W.-H. Li. 1985. Evidence for higher rates of nucleotide substitution in rodents than in man. Proceedings of the National Academy of Sciences USA 82: 1741-1745.

Xiang, Q.-Y., D.J. Crawford, A.D. Wolfe, Y.-C. Tang, & C.W. DePamphilis. 1998. Origin and biogeography of Aesculus L. (Hippocastanaceae): a molecular phylogenetic perspective. Evolution 52: 988-997.

Zink, R.M., R.C. Blackwell-Rago, & F. Ronquist. 1999. The shifting roles of dispersal and vicariance in biogeography. Proceedings of the Royal Society of Biological Sciences London Series B 267:497-503.

Zinsmeister, W.J. 1978. Effect of formation of the west antarctic ice sheet on shallow- water marine faunas of Chile. Antarctic Journal of the United States 13: 25-26.

145