Population genetics of incipient speciation in two species of jumping spiders (Salticidae: Habronattus) on the sky islands of southeast Arizona

Item Type text; Dissertation-Reproduction (electronic)

Authors Masta, Susan Elaine

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 02/10/2021 15:02:36

Link to Item http://hdl.handle.net/10150/284016 INFORMATION TO USERS

This manuscript has been reproduced from the microfilm master. UMI films the text directly firom the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, vs^e 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 deletioiL

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

Photographs included in the original manuscript have been reproduced xerographically in this copy. Kgher quality 6" x 9** black and white photographic prints are available for any photographs or illustrations appearing m this copy for an additional charge. Contact UMI directly to order. UMI A Bdl & Howell bifijnnation Conqany 300 North Zed> Road. Ann Aibor MI 48I06-I346 USA 313/761-4700 800/521-0600

POPULATION GENETICS OF INCIPIENT SPECIATION IN TWO SPECIES OF JUMPING SPIDERS (SALTIODAE: HABRONATTUS) ON THE SKY ISLANDS OF SOUTHEAST ARIZONA

by Stisan Elaine Masta

A Dissertation Submitted to the Facidty of the

DEPARTMENT OF ECOLOGY AND EVOLUTIONARY BIOLOGY

In Partial Fulfillment of the Requirements For the Degree of

DOCTOR OF PHILOSOPHY

In the Graduate College

THE UNIVERSITY OF ARIZONA

1999 DMT Ntanber: 9927470

UMI Microform 9927470 Copyriglit 1999, by UMI Company. All rights reserved.

This microform edition is protected against miauthorized copying mider Title 17, United States Code.

UMI 300 North Zeeb Road Ann Arbor, MI 48103 2

THE UNIVERSITY OF ARIZONA « GRADUATE COLLEGE

As members of the Final Examination Committee, we certify that we have Susan Elaine Masta read the dissertation prepared by entitled Populat-ion Ctenet-.ics of Tncipient Sppriattinn in 7W»

Species nf .1imipin9 Spiders H;:ihmnat-i-ng^

on tJie .qky Islands of Snirt-hf^asi- Ari^nna

and recommend that it be accepted as fulfilling the dissertation [3octor of PhilosoE*iy requirement for the Degree of

f Date Wayne Maddison ^ ^ , , ^ Michael Nachman Date C-C\., ~~2 I "2^/ ^ ^ 8 Nancy Moran Date

Lucinc^ McDade Date T[^ l/V nilthf J. Bruce W&lsh Date

Final approval and acceptance of this dissertation is contingent upon the candidate's submission of the final copy of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

H/1^/11 Disserta^on Director Date 3

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at The University of Arizona and is deposited in the University Library to be made available to borrowers imder the rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made- Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In eiU other instances, however, permission must be obtained from the author. 4

ACKNOWLEDGEMENTS

I am grateful to the many people who generously shared their time, thoughts, and expertise with me. I am especially indebted to my committee. My advisor, Wayne Maddison, first introduced me to the wonders of jimiping spiders and the amazing diversification of Habronattus pugillis in southeast Arizona. His willingness to share his knowledge and insight into phylogenetic biology helped clarify my thoughts, and greatly improved this dissertation. Luanda McDade was always available for lively discussions on phylogenetics and the Arizona sky islands, and for general advice on making it through graduate school. Her enthusiasm for science continues to inspire me. Nancy Moran has been an incredible mentor - not only has she provided scientific insight and opportunities for me, she generously allowed me to carry out the molecular aspects of this dissertation in her laboratory. Michael Nachn:\an edso generously allowed me the full use of his laboratory, and provided many hours of stimtdating discussions on population genetics. Bruce Walsh provided genetic and statistical expertise, and a sense of humor. I am also indebted to the members of the Maddison laboratory. Greta Binford, Gita Bodner, Peggy Gerba, Eileen Hebets and Marshal Hedin were a wealth of arachnological toowledge and assistance. Marshal also provided the initial "spider primers" that allowed my first foray into spider DNA sequencing. I am particularly grateful to Gabe Bien-Willner and James Borgmeyer who provided conscientious care in helping maintain jimiping spiders in the laboratory. The participants in two discussion groups - the Population Genetics and Molecular Evolution and the Phylogenetics Discussion Group - have provided countless hours of thought-provoking discussion over the years. Joana Silva and Patrick O'Grady were invaluable not only in these discussions, but also in helping implement many analyses. I am especially thankful for the many graduate students and postdoctoral fellows at the University of Arizona who have brightened my life in graduate school. They have been some of the brightest, most fun-loving, and generous people I have known. Their friendship has truly made graduate school a memorable experience. I am also grateful for generous financial support from a number of sources. A Doctoral Dissertation Improvement Grant from the National Science Foundation provided research support, and the Flinn Foundation Genetics Program provided fellowship and research support. I was also generously supported by the NSF-funded Research Training Grant in the Analysis of Biological Diversification, by the Department of Ecology and Evolutionary Biology, and by the Graduate College at the University of Arizona. And finally, it is difficult to express my gratitude to my husband. Jay Withgott. He provided me with tireless assistance - in collecting spiders, discussing science, reading dissertation chapters, and taking care of day-to-day tasks. His friendship, love, and encouragement provided the impetus that made completing this dissertation possible. 5

DEDICATION

This dissertation is dedicated to the people working to preserve the sky

island ecosystems of Arizona. 6

TABLE OF CONTENTS

ABSTRACT 7

1 INTRODUCTION 9

2 PRESENT STUDY 14

A HISTORICAL POPULATION GENETICS OF MORPHOLOGICALLY DIVERGENT POPULATIONS OF THE SPIDERS HABRONATTUS OREGONENSIS AND H. PUGILUS (SALTIODAE) IN SOUTHEAST ARIZONA 17

B PHYLOGEOGRAPHY OF HABRONATTUS PUGEUS (ARANEAE: SALHCIDAE) ON THE SKY ISLANDS OF SOUTHEAST ARIZONA.... 81

C UNUSUAL EVOLUTION OF MITOCHONDRIAL RNA GENE SEQUENCES IN SPIDERS (ARANEAE: SALTIODAE: HABRONATTUS) 123

REFERENCES 156 7

ABSTRACT

The population genetic forces that promote spedation, although well understood theoretically, are poorly known in nature. This dissertation focuses on the population genetics of allopatric speciation, using a system of jumping spiders (Araneae: Saltiddae) whose populations are subdivided among the disjimct patches of moimtain woodland habitat called "sky islands" in southeastern Arizona. I studied two spedes of saltidds that apparently share similar histories of range fragmentation but differ greatiy in their amoimt of rntraspedfic phenot5q5ic divergence. Using sequence data from neutrally evolving mitochondrial genes, I investigated the population genetic factors influencing divergence. Analyses of gene trees for Habronattus oregonensis and H. pugillis revealed that neither gene flow, effective population size, mutation rate, nor differences in divergence time can explain the interspecific difference in phenotypic divergence. Instead, selection - in these animals, presimiably sexual selection - must have acted differentially on traits encoded by nuclear lod to produce the discrepancy. A phylogeographic study of populations of H. pugillis may help clarify the influence of post-Pleistocene vegetational change on organisms dependent upon montane woodlands. Gene trees suggest limited migration between mountain ranges, but offer stronger evidence for incomplete lineage sorting. The trees provide no clear indication of the chronological sequence of woodland fragmentation, but suggest an old geographic division between northern and southern populations. Dates estimated for population 8

divergence range from 26,000 to 291,000 years ago, but rely on molecular clock estimates from non-arachnid arthropods. Divergence estimates based on vegetation change data would require that the mutation rate be considerably faster in these spiders than in non- arachnid arthropods. Whereas there is no fossil-based molecular clock calibration for arachnids to judge whether this is likely, analyses of mitochondrial sequences from three Habronattus species do reveal other highly unusual features. For example, secondary structures that were inferred from DNA sequences of tRNA genes lack the "PFC arm, and therefore are predicted not to form the standard tRNA cloverleaf. In addition, the 3' half of the gene encoding ribosomal 16S RNA appears to fold to a normal arthropod-like secondary structure, but the 5' half is extremely divergent and tnmcated with respect to other arthropods. 9

CHAPTERl

INTRODUCTION

Although the goal of many evolutionary biologists is to understand the factors that influence the process of spedation, knowledge of the factors that have promoted spedation in natural systems has remained elusive. We understand even less which factors have acted to increase the rate of spedation in some spedes-rich groups of organisms as compared to spedes- depauperate groups.

Phenotypic divergence and spedation have dassically been viewed as resulting from adaptive genetic change (Darwin 1859; Futuyma 1998), and many biologists still assume a pre-eminent role for natural and sexual selection in speciation. There is a good theoretical framework for expecting that sexual selection may be important in rapid diversification (Darwin 1871;

Fisher 1958; Lande 1981; West-Eberhard 1983). However, whether sexual selection has actually promoted spedes diversity in extant organisms has not been thoroughly examined. The studies that have been done have used two approaches. One approach has been to determine whether sexual selection is ciirrently acting in recently diverged populations (Kaneshiro and Boake 1987;

Kaneshiro and Kurihara 1981). The second approach has been to examine macroevolutionary phylogenetic trends using sister-group comparisons. In 10 this second approach, sexual dimorphism (and therefore presumably sexual selection) has been correlated with higher levels of taxonomic diversity in both passerine birds (Barraclough, Harvey, and Nee 1995) and fish (Mesnick

1996). However, there are problems with these approaches. The first approach assumes that evidence that sexual selection is ctirrently acting can be used as evidence that it was important in the spedation process. The second approach assumes that a specific trait is an indicator of sexual selection, so that a correlation between that trait and taxonomic diversity provides evidence that sexual selection promoted spedation. This dissertation proceeds from the stance that these assumptions may not always be appropriate. First, evidence that sexually selected traits exist in populations does not necessarily imply that sexual selection was important in diversification. Second, it is necessary to examine other population genetic factors before concluding that only selection has driven spedes divergence.

When allopatric populations become separated from one another, four population genetic factors may act either to promote or to hinder population divergence: mutation rate, migration, genetic drift, and selection. A myriad of biological (including social) and environmental factors (e.g., breeding system, vagility, habitat fragmentation) may influence which of these population genetic forces actually promotes divergence. Individually, each of these population genetic factors has been studied in diverging populations

(reviews in Avise 1994), but few studies have examined all factors. 11

The ability to study all these population genetic factors simultaneously

has become feasible only since the advent of rapid DNA sequencing

techniques, and since the development of coalescent theory. Coalescent

theory (reviewed in Hudson 1990) combines the discipline of systematics,

which has traditionally been used for studying macroevolutionary patterns,

with the field of population genetics, which examines microevolutionary

patterns and processes. When gene sequences are used to reconstruct a

phylogenetic tree showing how gene copies within a species are related, it

may be possible to infer the population genetic forces that have acted on that

gene. If the gene is evolving neutrally, the processes that have affected its

history may reflect the factors that affected all genes within the species, such

as effective population size and migration. The fact that neutral genes may

reflect a species' history has allowed researchers to track historical migration

and phylogeographic patterns in a wide variety of organisms, and to measure

historical effective population sizes (reviewed in Avise 1994; Avise et al.

1988).

In Appendix A, I present a genealogical framework for testing

competing population genetic hypotheses to explain apparent differences in

rates of phenot)q?ic divergence. I compare mitochondrial genealogies from

multiple populations of two closely related species of jumping spider

(Araneae: Saltiddae) that are hj^othesized to share similar histories of

isolation of subpopulations by climate-induced range fragmentation. Despite 12 these apparently similar histories of subdivision, Habronattus oregonensis

(Peckham and Peckham), shows little morphological or behavioral divergence among subpopulations, while H. pugillis Griswold is substantially more divergent among subpopulations. Because most differences among populations occur only in male traits, many of which are displayed in courtship, it appears that sexual selection may be driving this phenotypic divergence (Maddison and McMahon pers. comm. 1998). However, the entire genus Habronattus is characterized by extreme sexual dimorphism and elaborate male courtship displays, both indicative of sexual selection

(Peckham and Peckham 1909; Griswold 1987). If sexual selection is important in facilitating phenotypic divergence and subsequent speciation, it seems surprising to find evidence of it facilitating divergence in only one of these two Habronattus species. Other population genetic factors may be important in this apparent discrepancy in the rates of phenotypic divergence.

In Appendix B, I examine the mitochondrial phylogeographic patterns of 13 phenotypically divergent populations of H. pugillis. I use the phylogeographic patterns to try to elucidate the pattern of historical landscape change that may have caused initial fragmentation of populations, and to determine whether inter-moimtain range migration has occurred. To determine whether phenotypic divergence has been rapid, I attempt to date the isolation of these populations, using an arthropod mitochondrial molecular clock. Estimates of the times of divergence are three to 50 times greater than that expected based on vegetation change data. This indicates either that the vegetation dating based on packrat midden records (Van

Devender 1977; Van Devender and Spaulding 1979) does not reflect vicariance events relevant to these salticids, or that the mutation rate may be much higher in the mitochondria of these arachnids than in other arthropods. It is possible that both explanations may be valid.

Appendix C investigates molecular evolutionary patterns in the mitochondria of these arachnids. It compares the gene arrangement and inferred gene secondary structures in these salticids to those known from non-arachnid arthropods. The mitochondrial gene arrangement appears similar to that of insects and crustaceans, and differs from the gene arrangement previously described in other chelicerates. The inferred secondary structures of the RNA encoding genes involved in the formation and function of ribosomes are highly unusual. The tRNAs are tnmcated, and do not appear to fold into a canonical cloverleaf shape. The large subimit of ribosomal RNA is also considerably truncated with respect to other arthropod ribosomal RNAs. 14

CHAPTER 2 PRESENT STUDY

The methods, results, and conclusions of this study are presented in the papers appended to this dissertation. The following is a summary of the most important findings in these papers. Phenotypic divergence and spedation have classically been viewed as resulting from genetic changes in populations due to natural and sexual selection (Darwin 1859, Futuyma 1998). Sexual selection theory provides a compelling scenario that predicts rapid rates of spedation in organisms in which sexual selection acts (Fisher 1958; Lande 1981). However, little empirical evidence exists in organisms studied thus far to address whether this has actually occurred, or to adequately rule out other population genetic factors that may have accelerated spedation. The present study is one of the first to examine systematically all other population genetic factors that may have promoted incipient spedation in a system in which it appears that sexual selecting may be acting. In Appendix A, I compare multiple populations of two sexually dimorphic species of jxmiping spider {Habronattus oregonensis and H. pugillis) that appear to be differentiating phenotypically at very different rates. Comparisons reveeiled that neither gene flow, effective population size, mutation rate, nor differences in divergence times could explain the interspecific difference in amoimt of phenotypic divergence. Therefore, selection acting differentially on the nuclear lod responsible for phenotype must have produced the discrepancy. It is prestuned that sexual selection has 15 been the driving force in diversification, because only male spiders differ phenotypically among motmtain ranges, particularly in traits involved in courtship. In Appendix B, I examine the mitochondrial phylogeographic patterns of 13 morphologically divergent populations of H. pugillis, to try to elucidate patterns of historical landscape change that may have facilitated initial divergence of populations. I find evidence that smaller mountain ranges contain populations that no longer share migrants with neighboring ranges, which is usually considered the first stage in allopatric speciation. These same individuals form mitochondrial clades nested within clades from the neighboring larger ranges, a pattern concordant with a peripheral isolates mode of speciation (Mayr 1954). Although the trees provide no clear indication of the chronological sequence of woodland fragmentation, the time since isolation was evaluated for some mountain populations. The dates appear to be far older than would be predicted based upon data from fragmentation of woodland habitat. To determine if unusual patterns of evolution of mitochondrial sequences may exist in these salticids, I examined patterns of molecular evolution of Habronattus mitochondrial DNA in Appendix C. A highly unusual secondary structure was found for the two tRNA genes sequenced, both of which appear to lack a PPC arm. This type of modified tRNA structure has been found in the mitochondria of only one other organism - nematodes. A second novel finding in these arachnids is that the 5' end of the 16S gene has been truncated with respect to other arthropods. This change was also found in nematodes, suggesting that changes in tRNA structure and 5' 16S structure may be evolutionarily linked. Results from Appendix C suggest that further study of arachnid mitochondrial DNA may provide valuable new insights into the processes of molecular evolution. 17

APPENDIX A

HISTORICAL POPULATION GENETICS OF MORPHOLOGICALLY

DIVERGENT POPULATIONS OF THE SPIDERS HABRONATTUS

OREGONENSIS AND H. PUGILUS (SALTICIDAE) IN SOUTHEAST ARIZONA 18

HISTORICAL POPULATION GENETICS OF MORPHOLOGICALLY DIVERGENT POPULATIONS OF THE SPIDERS HABRONATTUS OREGONENSIS AND H. PUGILUS (SALTICIDAE) IN SOUTHEAST ARIZONA

Susan E. Masta Department of Ecology and Evolutionary Biology University of Arizona Tucson, Arizona 85721 [email protected]

[LRH: SUSAN E. MASTA] [RRH: POPULAHON GENETICS OF SPECIATION] 19

ABSTRACT

Studies of the spedation process have often examined the role of selection in divergence of allopatric populations, but inadequately examined the other population genetic factors that may drive differentiation. I investigated the roles of genetic drift, mutation, and migration in intraspecific divergence for each of two spedes of jumping spiders (Araneae: Saltiddae) that apparently share similar histories of range fragmentation but differ greatly in amoimt of intraspecific phenotjrpic divergence. I sequenced mitochondrial genes coding for NADH dehydrogenase subimit 1 (NDl) and the large subimit of ribosomal RNA (16S) from 56 individuals of each spedes from five mountain ranges. A statistical test of neutrality performed on the NDl gene yielded results consistent with neutral evolution. Mitochondrial genealogies were constructed and compared to determine the relative roles that gene flow, effective population size, and mutation rate have played in shaping morphological and behavioral divergence. Comparisons revealed that neither gene flow, effective population size, mutation rate, nor differences in divergence times could explain the interspecific difference in amoimt of phenotj^ic divergence. Therefore, selection must have acted differentially on nuclear lod to produce the discrepancy. It is presumed that sexual selection has been the driving force in diversification because male spiders differ among mountain ranges in traits shown to females during courtship. Divergence times estimated using an arthropod mitochondrial molecular dock are not concordant with estimated vicariance history, thus casting doubt on both the suitability of this molecular dock for arachnids, and whether relevant vicariance events occurred in the early Holocene. 20

Key Words speciation, sexual selection, phenotypic divergence, population genetics, intraspecific genealogies, Salticidae, test of neutrality

Phenotypic divergence and speciation have classically been viewed as resulting from adaptive genetic change (Darwin 1859), and many biologists still assume a pre-eminent role for natural and sexual selection in speciation. Speciation has often been studied by examination of the macroevolutionary patterr\s that have resulted from microevolutionary processes. Such a macroevolutionary approach has often led to the discussion of only one microevolutionary process that may have caused the pattern, usually sexual or natural selection. This is despite the fact that within populations, migration, mutation rate, and population size can all play important roles in partitioning genetic variation. There is little basis from empirical studies for discounting all of these processes in favor of selection. Numerous population genetic studies have examined the role of single factors in differentiation (reviews in Avise 1994); however, few studies have systematically examined all population genetic factors that may have contributed to morphological divergence. Developments in the field of coalescent theory, combined with the ability to sequence a large nvunber of individuals rapidly, may allow us to close the rift between macro- and microevolutionary approaches. Coalescent theory combines the discipline of systematics, which has traditionally been used for studying macroevolutionary patterns, with the field of population genetics, used for examining microevolutionary processes (reviewed in Hudson 1990). 21

When gene sequences are used to reconstruct a phylogenetic tree showing how gene copies within a species are related, it may be possible to make inferences about the population gaietic forces that have acted on that gene. If the gene is evolving neutrally, the processes that have affected its history may reflect the factors that affected all genes within the species, such as effective population size, and migration. The fact that neutral genes may reflect a species' history has allowed researchers to track historical migration and phylogeographic patterns in a wide variety of organisms (reviewed in Avise 1994), and to measure historical effective population sizes (e.g. Avise et al. 1988, Kliman and Hey 1993, Hey and Kliman 1993). In this paper I use a genealogical framework to evaluate competing population genetic hypotheses to explain apparent differences in rates of phenotypic divergence. I compare mitochondrial genealogies from multiple populations of two closely related species of jumping spider (Araneae: Salticidae) that are hypothesized to share similar histories of isolation of subpopulations by climate-induced range fragmentation. Despite these apparently similar histories of subdivision, one species, Habronattus oregonensis (Peckham and Peckham), shows little morphological and behavioral divergence among subpopulations; another, H. pugillis Griswold, is substantially more divergent among subpopulations. Because most differences among populations are only in male traits, it appears that sexual selection may be implicated in driving this phenotypic divergence (Maddison and McMahon pers. comm.). The genus Habronattus is characterized by extreme sexual dimorphism and elaborate male courtship displays, both indicative of sexual selection (Peckham and Peckham 1909; Griswold 1987). If 22 sexual selection is important in facilitating morphological divergence and subsequent spedation, it is surprising that it appears to be facilitating divergence in only one of these closely related species. This raises the possibility that other population genetic factors, such as differences in migration rate, may explain this apparent difference in the rates of phenotjrpic divergence. I investigate which population genetic processes may account for phenotypic divergence among H. pugillis and H. oregonensis subpopulations, and which processes may accoimt for the interspecific difference in divergence rate. Five questions are examined: (1) Is morphological similarity in H. oregonensis due to a greater amoimt of inter-population migration than in H. pugillis? (2) Did each H. pugillis population undergo population bottlenecks relative to H. oregonensis that may have contributed to their greater amount of phenotypic divergence? (3) Is the generation time shorter in H. pugillis, so that in a given amount of time, they may have accrued more mutations responsible for phenotypic divergence than in H. oregonensis? (4) If no evidence is found to support the above hj^otheses, can we infer that a greater amoimt of selection has acted on the lod controlling the phenotypic characters in H. pugillis? (5) Is there actually a difference in the rates of phenotypic divergence, or is the divergence in H. pugillis much older than that of H. oregonensis? 23

STUDY SYSTEM

The "sky islands" of southeast Arizona are forested mountain ranges separated by a "sea" of desert. These isolated patches of forest consist of the northern most extent of the Madrean evergreen woodlands that extend south through the Sierra Madre Occidental of Mexico (Brown and Lowe 1980); these woodlands are replaced at the higher elevations by north-temperate mixed conifer forest (Pase and Brown 1982). The fragmented forest landscape present in southeast Arizona today, however, contrasts sharply with the plant commimity that was present as recently as 8,000 to 11,000 years ago. During the early Holocene, the climate in southeast Arizona was cooler and more mesic than it is today. Douglas fir {Pseudotsuga menziesii) and pine {Pinus) woodlands are ciuxently restricted to elevations above 2,100 meters, but evidence from packrat middens indicates that Douglas fir formerly grew at elevations as low as 1,555 meters (Van Devender and Spaulding 1979). Oak- juniper (Quercus-Juniperus) woodlands are currently restricted to elevations above approximately 1,600 meters in southeast Arizona, but jtmiper woodlands were formerly found at elevations as low as 260 meters (Van Devender 1977). Thus, mountain ranges were presumably connected by contiguous stretches of woodland. The subsequent warming and drying climate forced these plant communities upslope, and this climate-induced fragmentation of woodland habitats isolated populations of organisms dependent on the more mesic woodleind environments. Habronattus pugillis and H. oregonensis are species inhabiting oak and conifer woodlands. 24 respectively, and are completely absent from the Sonoran Desert (Maddison and Masta pers. observation). Male individuals of H. oregonensis are distinguished from other Habronattus by swollen, nearly hairless tibia on their first legs (Peckham and Peckham 1909). The species' distribution extends from southem British Columbia to southem Arizona (Griswold 1987); it is typically foimd in moimtainous habitat. On the sky islands of southeast Arizona, H. oregonensis is found only in montane conifer forests at elevations above 2,100 m. (pers. obs.). Males in southeast Arizona differ slightly among moimtain ranges in coloration of the first two legs, palps, and abdomen (see Fig. 1). Within ranges, males are identical in coloration, whereas females do not differ morphologically within or among moimtain ranges (pers. obs.). Comparison of videotaped courtship displays of multiple individuals from each mountain range showed that all forms display the same courtship behavior (impublished data). Individuals of H. pugillis inhabit the disjimct Madrean oak woodlands in southeast Arizona, with a range extending south along the western slope of the Sierra Madre to Nayarit, Mexico (Griswold 1987). Maddison and McMahon (submitted 1998) foimd considerable variation among sky island populations of H. pugillis males, both in morphology (see Fig. 2) and in courtship behavior. Some of the morphological differences among ranges include presence and extent of an eyestripe, presence and extent of fringe and striping on the first leg, clypeias and palp coloration, and width of the carapace. Courtship behavior of males appears imiform within a range, yet varies considerably among ranges. For example, courting males from the 25

Santa Rita Motmtains slowly circle their shiny, nearly hairless palps, whereas males from the Galiuro Motmtains hold their black palps motionless, and males from the Santa Catalina Motmtains vibrate their heavily fringed palps. These differ^ces, and a stiite of other courtship and morphological differences, make it easy to determine from which range a male was collected. In contrast to the marked differences among males, the females appear similar within and among moimtain ranges.

METHODS

Collection protocol Individuals of H. oregonensis and H. pugillis were collected from each of five mountain ranges in southeastern Arizona (see Fig. 3). The species co- occtir on four of these ranges (the Santa Catalina, Santa Rita, Galiuro, and Huachuca Mountains). For each species, a fifth mountain range was chosen in which only one species occurs - the Baboquivari Mountains for H. pugillis, and the Chiricahua Mountains for H. oregonensis. Within each motmtain range, individuals were collected from two to five locations, spanning each species' known distribution on that range. In addition, eight individuals from three locales in the Sierra Madre, Sonora, Mexico, were sampled for H. pugillis, and six individuals of H. oregonensis were collected from two sites in northern California and one in northern Arizona. The collecting locales outside of Arizona are in large areas of continuous habitat for each species; therefore it was expected that these populations have not experienced 26 population bottlenecks, and may possibly be the source of the initial colonizers of the sky islands.

DNA Extraction and Sequencing Voucher specimens of individuals from each collecting locality were placed in 80% EtOH, and the remaining spiders were stored at -80° C. Total genomic DNA was isolated from ten individuals from each mountain range for each species, using a modified SDS exfraction protocol. Together with the individuals collected from areas of continuous habitat, DNA was extracted from a total of 56 individuals of each species. The mitochondrial genes encoding the 5' end of NADH dehydrogenase subunit 1 (NDl) and the 3' end of the large subimit ribosomal RNA (16S) gene, were amplified by the polymerase chain reaction (PCR), using primers designed specifically to amplify these two species. The 3' end of the mitochondrial gene encoding cytochrome oxidase subunit 1 (COI) was also amplified and sequenced for a subset of the H. oregonensis and H. pugillis individuals. Descriptions of DNA isolation, primers, PCR amplification and sequencing are given elsewhere (see Appendix C). As only one sfrand of DNA was sequenced, the accuracy of the sequences was verified in two ways. All sequences were manually aligned in SeqApp 1.9 (Gilbert 1994), and all sites differing from the other sequences were reexamined to see if the base had been scored correctly. Secondly, 18 of the individuals were sequenced both manually and with the automated sequencer. There was complete agreement among all sequences. 27

Phylogenetic and Population Genetic Analyses Phylogenetic analyses were conducted using the beta version of PAUP* 4.0d64 (Swofford 1998). Several tree-building strategies were used. The number of sequences precluded the use of exact methods (branch-and-boimd or exhaustive searches); therefore unweighted heuristic parsimony searches were conducted on both the H. oregonensis and H. pugillis data sets, using the default options (TBR, ACCTRAN, etc.), and saving a maximum of 50,000 trees. Because each heuristic parsimony search will find only one island of trees (Maddison 1991), searches for multiple tree islands were conducted by performing replicate parsimony searches imtil 50 searches had foimd trees of the same length or shorter than the single heuristic parsimony search. Starting trees were built with random addition sequences, and tree-bisection- recormection (TBR) branch-swapping. Only optimal trees were saved from each replicate, from which a strict consensus tree was constructed. One htmdred bootstrap replicates were performed on the ND1-16S data sets, saving 5,000 trees per replicate. The bootstrap method assumes that the characters are independent and identically distributed (Felsenstein 1985). The assumption of identical distribution of characters appears to be violated for these sequences sampled from within populations, because the shape parameter for the gamma distribution is low (see Results), indicating significant rate heterogeneity between sites. To make the data conform to this assimiption, only parsimony-informative characters were included in the bootstrap analyses for the ND1-16S data sets. One himdred bootstrap replicates were performed on both the COI data set alone, and the combined 28

COI- ND1-16S data set, using all characters, and saving 1,000 trees per replicate. Neighbor-joining analyses (Saitou and Nei 1987) with both uncorrected p distances and the HKY85 (Hasegawa et al. 1985) model of evolution were performed. Regions of sequences with missing data were not used in the pairwise comparisons. Maximum likelihood analyses (Felsenstein 1981) were conducted with subsets of the taxa. The goed of this seeirch strategy was to reconstruct bassil relationships which may be obscured in parsimony analyses due to the high A+T bias present in the data set, and the potentially different patterns and rates of evolution present in the NDl and 16S genes. Two or three individuals were selected to represent each mountain range. Individuals were chosen that had the greatest pairwise distance within the moimtain range; in addition, each clade that resulted from the parsimony analyses was represented by at least one individual. Searches used the transition- transversion ratio, base frequency, and the gamma shape parameter (Yang 1993) estimated by the program. This search method has been found by Yang (1994) and Yang et al. (1994) to provide a good estimate of branch lengths. To test the monophyly of H. pugillis and H. oregonensis four divergent sequences from the other species' data set were included in each phylogenetic analysis. All trees were rooted with Habronattus geronimoi, which is outside the clade that contains H. pugillis and H. oregonensis (Hedin and Maddison, pers. comm. 1998). The McDonald-Kreitman test of neutrality (McDonald and Kreitmein 1991) compares the ratio of silent to replacement sites of a protein-coding 29

gene, within and between species. Neutral theory predicts that these ratios should be the same if the gene has been evolving neutrally. Unlike other statistical tests of selection, the McDonald-Kreitman test does not assume lack of population structure, an assumption clearly violated in these populations of spiders collected from isolated moimtain ranges. This test was performed on the NDl and the COl genes. The first 395 bp of the 5' end of NDl, and the last 717 bp of the 3' end of COl, were compeired separately for both species of Habronattus. Replacement and silent site mutations were scored as either polymorphic within a species, or fixed between species. Tajima's D statistic (Tajima 1989) compares the difference between two different estimators of the parameter Nfi, where N is the effective population size and \i is the neutral mutation rate. One estimator is based on the number of segregating sites and the other is based on the average number of pairwise nucleotide differences in a sample. Under equilibrium conditions with respect to mutation and drift, and when the gene is evolving neutrally, these two estimators should be the same. To determine if there was population subdivision within a single moimtain range, Tajima's D and the confidence limits of D were calculated for the ND1-16S sequences for populations from each moimtain range using the program DnaSP 2.5 (Rozas and Rozas 1997). Braverman at al. (1995) have investigated the power of Tajima's D to detect genetic hitchhiking, and found the expected values of D to be large and negative under hitchhiking models. Because mitochondria are non-recombining, it is expected that selective sweeps in the mitochondria may therefore be detected with Tajima's D. 30

Fu and Li (1993) proposed two tests, D* and F*, that utilize different estimators of theta, or the product of the effective population size and the mutation rate, to look for departures from neutral evolution. D* is based on the difference between the niunber of singletons in a sample and the total number of mutations. F* is based on the difference between the number of singletons and the average number of nucleotide differences between pairs of sequences. Simonsen et al. (1995) compared the power of Tajima's D statistic and Fu and Li's D* and P tests (Fu and Li 1993). They concluded that Tajima's test is generally more powerful against detecting departures from equilibriimi due to selective sweeps, population bottlenecks and population subdivision. However, they also foimd that the critical values used by Fu and Li (1993) may falsely reject the hypothesis of neutral evolution (Simonsen et al. 1995), and proposed new critical values. The program DnaSP 3.0 (Rozas and Rozas 1999) performs Fu and Li's D* and F* tests, but uses the Simonsen et al. (1995) critical values. Both D* and F* were calculated using DnaSP 3.0. One method of estimating long-term effective population sizes is to calculate the average pairwise nucleotide diversities, k (Nei 1987, eq. 10.6).

Under equilibriimi conditions and when mutations are neutral, k estimates the parameter 2Ne|J., the product of the effective population size for mitochondria and the mutation rate per nucleotide site per generation. The average pairwise genetic distance of populations of individuals in each moimtain range was calculated for 654 bp of the ND1-16S region, excluding all sites with missing data, using the program DnaSP 2.5. Genealogies may be used for estimating the amoimt of gene flow between geographically disjunct populations. The minimum number of migration 31 events, s, required to explain non-monophyly of individuals from a given geographic locale can be reconstructed on a gene tree, as suggested by Slatkin and Maddison (1989). Greater values of s are expected with greater amounts of gene flow. The geographic range from which each individual was collected was coded as a discrete character, and locality was then traced onto both the H. oregonensis and H. pugillis neighbor-joining trees, using MacClade 3.07 (Maddison and Maddison 1997). Because it is necessciry to use a fully resolved tree to calculate s, this precluded using coi\sensus trees. Therefore, to examine variation in s due to variation in tree reconstruction, s was subsequently calculated over 100 neighbor-joining bootstrap replicates of the ND1-16S gene trees for each species. MacClade was then used to graph the value of s versus the nvimber of trees with that value. Two different estimates of the amoimt of gene flow, as measured by Nm, were calculated. Slatkin and Maddison (1989) showed that the distribution of s is a function of Nm. Values of s were converted to estimates of Nm as described (Slatkin and Maddison 1989), discarding populations with unequal sample sizes (the Chiricahua population for H. oregonensis and the Baboquivari population for H. pugillis). Non-genealogical based estimates of migration (Nm) and population subdivision (Fsf) were also calculated from the mitochondrial sequences. Fst was estimated by estimating a measure of the effective population size in the total population and the average population size on each of the mountain ranges. Nm was then calculated following the standeird formula given by Wright (1951). These estimates also excluded the Chiricahua and Baboquivari populations in order to make the two different estimates of Nm comparable. 32

To determine if the rates of mitochondrial sequence evolution differed between H. oregonensis and H. pugillis, a relative rate test (Sarich and Wilson 1973) was performed using the ND1-16S genes. The average number of nucleotide substitutions per site, Dxy, between H. geronimoi and all individuals of both H. pugillis and H. oregonensis, and between all members of H. pugillis and H. oregonensis, was calculated in IDnaSP, using the method of Nei (1987 equation 10.20), with a Jukes-Cantor correction (Jukes and Cantor 1969). The value of Dxy on the branches leading to both H. oregonensis and H. pugillis was then calculated.

RESULTS AND DISCUSSION

Sequences and Tests of Neutrality A total of about 812 bp of ND1-16S sequence was obtained for each individual. The H. pugillis sequences contained a 4- or 8-bp insertion between the NDl and 16S genes, and both species had single bp indels within 16S. Indels were not coded as characters for use in phylogeny reconstruction, nor was information from these regions used in pairwise sequence comparisons. The sequences of the 56 individuals of H. oregonensis contained 55 unique haplotypes, whereas 54 haplotypes of H. pugillis were present among the 56 individuals sequenced (see Appendix). The pairwise sequence divergence was higher in H. pugillis (2.2%) than in H. oregonensis (1.3%), when all individuals were compared (table 2). 33

If the gene used for tree reconstruction is under selection, both the topology and branch lengths in the resultant tree may be affected (reviewed in Hudson 1990). Additionally, methods to estimate effective population sizes, such as 7t, rely on neutral evolution of the gene being sampled (Nei 1987). For these reasons, it is useful to test for selection on the gene before attempting to infer historical genealogical relationships. The results of the McDonald- Kreitman test performed on the NDl and COl genes are consistent with neutral evolution (see Table 1) for both H. pugillis and H. oregonensis. An assiunption of methods used to calculate effective population size is that the population sampled is not subdivided into non-randomly mating subpopulations. Values of Tajima's D and Fu and Li's D8 and F* indicate that populations within each moimtain range do not deviate from equilibrium conditions, except for H. oregonensis from the Chiricahua Moimtains (see Table 2). A subdivided population tends to produce positive values for all three of these test statistics (Simonsen et. al. 1995). The negative value of Tajima's D for this population could be caused by either an accumulation of deleterious mutations in the ND1-16S molecule, directional selection, or a population bottleneck (Tajima 1989; Simonsen et al. 1995). Because H. pugillis does not inhabit the Chiricahua Mountains, the deviation from equilibrium for this population of H. oregonensis does not affect the comparisons of effective population sizes calculated for each species that will be discussed later. 34

Phylogenetic Analyses Searches for multiple islands of trees yielded only trees of the same length or longer than the single parsimony searches for the H. oregonensis and H. pugillis ND1-16S data sets. Strict consensus trees constructed from all replicates with the same shortest tree lengths yielded the Seime topologies as the consensus trees from the single searches that saved 50,000 most parsimonious trees, and therefore are not shown. Whereas this method of comparing strict consensus trees from many replicates to trees inferred from a single search does not directly yield information on other islands of trees of the same length, it gives confidence that relationships present in each island are also represented by the strict consensus tree. This consensus tree therefore makes the most conservative statement about relationships. The strict consensus tree of 50,000 most parsimonious trees for H. oregonensis (see Fig. 4), includes five basal clades that consist only of individuals from southeast Arizona (labeled A-E). Three of these clades represent monophyletic groups containing all and only individuals from the mountain range from which they were collected (B, C, and D). Monophyly of individuals from a given geographic locale is violated by only two individuals: Orl7C from the Chiricahuas which is placed with the Santa Catalina individuals, and OrSSSC from the Santa Catalinas, whose placement is imresolved. Relationships among the five southeast Arizona clades are not resolved by the strict consensus free, and collapse to a single polytomy. The strict consensus tree for H. pugillis (see Fig. 5) supports three basal clades consisting only of individuals from southeast Arizona (A,B and C). In contrast to the H. oregonensis free, only one range (the Baboquivari 35

Mountains) contains individuals that form a monophyletic clade (clade A). The B and C clades include individuals from more than one geographic locale. The basal relationships of the reconstructed clades in the H. pugtllis tree are more fully resolved than those in the H. oregonensis tree, although the nodes of these H. pugillis clades lack strong bootstrap support (i.e. < 50%). The greater resolution of basal nodes in the H. pugillis genealogy may be due to more parsimony-informative characters (69, versus 41 in H. oregonensis). To test this hj^othesis, COI sequences for a subset of individuals from each clade of H. oregonensis were analyzed. For this subset of individuals there are 32 parsimony-informative sites for the ND1-16S genes, and 39 for the COI data set, yielding a combined total of 71 informative sites. Strict consensus trees from both the COI data set alone and the combined COI and ND1-16S data set continue to show poor resolution of the basal nodes (see Fig. 6), with less than 50% bootstrap support for the branch leading to the divergence of the H. oregonensis Huachuca and Santa Rita clades. There is good support for the Galiuro, Santa Catalina, and Chiricahua clades being more closely related to each other than they are to the Huachuca or Santa Rita clades. Although the combined ND1-16S-COI analysis provides more resolution for some basal relationships of the H. oregonensis sky islcind clades, it collapses the basal relations of the H. pugillis clades to a polytomy. The lack of clear resolution of basal relationships in both species may therefore be due both to a paucity of parsimony-informative characters and to character conflict. However, all analyses are congruent with regard to relationships within terminal clades, thus increasing confidence in the results. 36

To discern if relationships among the basal clades could be uncovered with a more realistic model of evolution than unweighted parsimony, maximum likelihood analyses were performed. Maximum likelihood searches were stopped after about 25,000 branch rearrangements. Nine trees of equal likelihood were found for H. oregonensis, and 21 trees for 4 H. pugillis. All of the trees foimd for H. oregonensis (see Fig. 7) indicate that first the Huachuca and then the Santa Rita clades diverged before the other three mountain ranges, in agreement with the combined parsimony analysis. The nine trees differ only in the relationships among the Santa Catalina, Galiuro and Chiricahua individuals. For H. pugillis, a trichotomy is formed by three sky island clades (see Fig. 8) in each of the 21 trees. These are the Seime three southeast Arizona clades that were found in the parsimony analysis. The twenty one trees differ only in relationships of individuals within the Santa Rita-Huachuca-Santa Catalina clade. The fact that there is a basal polytomy in each H. pugillis maximum likelihood tree indicates there are no characters to distinguish the order of branching of these clades. This indicates that lack of resolution in each tree is caused by few character changes rather than by conflict among characters. It is therefore impossible with this data to infer the putative order of the vicariance events that isolated the H. pugillis populations. Because adding more parsimony-informative characters fails to resolve fully the basal relationships, and because likelihood estimates for branch lengths are short, the relationships among some clades may indeed be a genuine pol)^omy; that is, one that will remain imresolved regardless of how much additional data is acquired. This pattern of a basal polytomy is 37 consistent with a rapid and simultaneous divergence of the moimtain clades, as would be expected under a scenario of climate-induced range fragmentation. Climate change that caused contraction of woodland habitats should have created numerous habitat islands in southeast Arizona nearly simultaneously. Another way to increase confidence that genealogical relationships have been correctly reconstructed is to find that multiple tree-finding algorithms converge on the same tree (Kim 1993). Neighbor-joining trees of both the H. oregonensis and H. pugillis ND1-16S data constructed using uncorrected p- values (tree not shown), and with the HKY model of evolution, 5deld topologies similar to those of the parsimony trees (see Figs. 9a, 9b). The neighbor-joining trees support the same terminal clades as the parsimony consensus trees and, like the parsimony trees, have low bootstrap support for resolving basal relationships among the terminal clades.

Hypothesis 1: Migration One hypothesis to accoimt for the greater morphological and behavioral divergence among populations of H. pugillis than among populations of H. oregonensis is that H. pugillis populations have experienced historically less migration and gene flow among mountain ranges. Populations that become genetically isolated when migration and gene flow from neighboring populations ceases will begin to diverge over time. In the parsimony strict consensus trees, the lack of reciprocal monophyly of each geographic population indicates that migration may have occurred among mountain ranges for both species of Habronattus. (Alternatively, incomplete lineage 38

sorting may have generated this pattern; this possibility will be discussed later.) To quantify the relative magnitude of historical migration events within the two species' genealogies, each individual spider's geographic collection locality was mapped onto the neighbor-joining tree, and the value of s was calculated. Figures 9a and 9b indicate that it is necessary to infer more migration for H. pugillis than for H. oregonensis in order to explain the incongruence of geography and genealogy. This is the reverse of the result expected under hypothesis 1, that lack of gene flow is enabling a more rapid rate of phenotypic divergence of H. pugillis. Thus, the data do not support the hypothesis, and migration carmot explain the observed interspecific difference in divergence rates. If all five mountain ranges and the continuous habitat range to the south or north contained reciprocally monophyletic populations, then the value of s would be five, indicating five migration events are necessary to explain the current geographic distribution of haplotj^es from initisil pemmixia. Six to ten migration events (see Fig. 9a graph) are necessary to explain the colonization of H. oregonensis onto the sky island ranges, depending on tree reconstruction. For the H. oregonensis tree in Figure 9a, the value of s is seven and monophyly of geographic locality is violated only by spiders OrlZC collected in the Chiricahua Moimtains, and OrSSSC collected in the Santa Catalina Moimtains. Trees with all values of s always show individual Orl7C placed within the Santa Catalina clade, whereas the placement of OrSSSC varies. This result is surprising: the Chiricahua Mountains are the most isolated range in this study and I had predicted that migration would not occur between it and any other range. 39

Multiple individuals in the H. pugillis genealogy cause geographic localities to be non-monophyletic. Between 11 and 19 historical migration events are needed to explain this pattern (Figure 9b graph). Some of the inferred migration would be over short geographic distances (e.g., between the Huachuca and Santa Rita Moimtains). However, other reconstructed migration events would be over very great distances (e.g., it is about 480 kilometers between the Galiuro Mountains and the locality in the Sierra Madre where individual PglOSM was collected). It seems imlikely that an individual could have migrated such an aerial distance. Even though migration between ranges may be imlikely for either species, estimates of Nm were calculated for both species. The estimates of Nm based on calculating s on the gene trees in Figures 9a and 9b are between 0.5-1.0 for H. oregonensis, and 7-10 for H. pugillis. The estimates of Nm calculated from Fst values are .077 for H. oregonensis and 0.45 for H. pugillis. Although the absolute value of the numbers differ depending on how Nm is calculated, they both are in concordance in that the estimates are much higher for H. pugillis. Patterns suggesting migration among populatioiis, however, may instead be caused by incomplete lineage sorting, and it is not possible from a genealogy alone to distinguish between these two potential causes of non- monophyly of populations. Incomplete lineage sorting may occur if the time since isolation of subpopulations from one another is short, and insufficient time has passed for the genes to stochastically sort into clades corresponding to geographic locale. Probabilities based on computer simulations (Neigel and Avise 1986) and theoretical predictions (Tajima 1983) suggest there is a high 40 probability of incomplete lineage sorting of mitochondrial DNA when the time since isolation of populations is less than 2N to 4N generations. Based on these estimates, if the spider populations have been separated for 10,000 years, and if they have one generation per year, historical effective population sizes of less than 2,500 to 5,000 females per moimtain range would be required for there to be a high probability of reciprocal monophyly of mountain clades. Although no census has been conducted of these spiders, observations made over five years of collecting suggest much larger population sizes on the moimtain ranges in this study (pers. obs.)- Regardless of whether migration or incomplete lineage sorting explains non-monophyly of geographic clades, the above evidence argues against a greater degree of gene flow or incomplete lineage sorting in H. oregonensis. Thus, some other population genetic factor must be responsible for the greater morphological and behavioral divergence of H. pugillis.

Hypothesis 2: Effective Population Size Southeast Arizona, at the northern edge of the Sonoran Desert, experiences periodic droughts. Higher elevations in the mountains are somewhat buffered from droughts that affect the lower slopes, because moisture tends to condense when it reaches the peaks, causing more precipitation at higher elevations (Fritts et al. 1965). It is plausible that H. pugillis, residing in the lower-elevation oak woodlands, may go through periodic population bottlenecks caused by droughts. The ancestral population of H. pugillis that may have existed across southern Arizona when woodlands were widespread could have corisisted of 41

many different male morphological types. In the more recent isolated populations, repeated bottlenecks or single, severe population crashes could have caused fixation of one of these morphological tjrpes on each mountain range. Alternatively, males in the ancestral population could have been uniform, and bottlenecks may have caused an accelerated rate of morphological divergence and incipient speciation via a genetic revolution (Mayr 1954), or restructuring of co-adapted gene complexes (Carson 1975, Templeton 1980, Carson and Templeton 1984). To determine if population bottlenecks may be associated with the imique

morphologies of H. pugillis within each mountain range, I estimated k, a measure of historical effective population size for each species on each mountain range. Comparison of the nucleotide diversities within moimtain ranges between the two species shows consistently larger values for H. pugillis than H. oregonensis (see Table 2). Assuming the mutation rate is equivalent for the two species, the greater amoimt of nucleotide diversity present in H. pugillis indicates that it has maintained historically larger effective population sizes than has H. oregonensis on each of the four moimtain ranges where both species co-occur. (The largest nucleotide diversity for H. oregonensis is found in the Chiricahua Mountains; however this population violates the assumption of equilibrium as measured by

Tajima's D, so this calculation of k may not accurately reflect historical population size.) The it values for the four moimtain ranges on which the spiders co-occur give opposite values to that predicted by the h5^othesis that population bottlenecks have spurred phenotypic divergence in H. pugillis. 42

It could be argued that alterrmtive scenarios exist, e.g., that H. pugillis populations diverged via bottlenecks long ago, and that a limited amoimt of very recent migration has resulted in introgression of neutral mitochondrial genes but not of g^es for sexually selected traits. However, given the very large effective population size foimd in the unfragmented population in the Sierra Madre (see Table 2), it seems most parsimonious to conclude that large effective population sizes on the Arizona ranges are remnants of an old widespread polymorphic population (see Appendix B). In addition, imder a scenario of old bottlenecks and recent migration, one might expect to find strong geographic clustering in the tree; however, the clustering is limited and seems largely not due to migration (see Appendix B).

Hypothesis 3: Mutation Rate A third hypothesis to explain why H. pugillis is more phenotypically divergent among mountain ranges than H. oregonensis is that it has a shorter generation time. For instance, if populations of each species became subdivided at the same time, but H. pugillis had a generation time half as long as H. oregonensis, then it could have accrued twice as many substitutions per imit time. This would presimiably lead to greater morphological and behavioral divergence. The average nucleotide diversity is an estimate of 2NeM., with |i measured

in units of mutations per generation. A shorter generation time would have

the effect of inflating the value of k observed for H. pugillis if we did not have prior knowledge that generation times of the two species differed and therefore did not scale [L accordingly. Comparisons of neutral genetic 43

divergence within each species allow me to estimate k across individuals from all collecting locales. Table 2 indicates a higher average value of 7t for H. pugillis, but since this is influenced by both effective population size and mutation rate, it is difficult to tell from these data alone whether a difference in mutation rate or a difference in effective population size has caused the discrepancy. Two approaches were taken to try to distinguish effects of effective population size and mutation rate. First, individuals of both species were collected from different motintain ranges, mated in the lab, and their offspring raised to adulthood. Under identical laboratory conditions, there was no difference in their generation time from mating to egg to adult (approximately one year). However, in the mountains in southeast Arizona, H. pugillis lives in warm lower-elevation woodlands that remain free of snow nearly all year. In contrast, H. oregonensis lives at higher elevations that are snow-covered or frozen for several months of the year. Thus, although the laboratory data suggest there are not genetically based differences in maturation time, there may be greater ecological constraints on the maturation time of H. oregonensis. Unfortimately, there is at present no information on whether H. oregonensis is able to produce one generation per year in the moimtains of southeast Arizona. A second approach to determine whether more generations of H. pugillis have existed since the time of the vicariance events that isolated the mountain populations is to compare the amount of genetic divergence of each species from a third, outgroup, species. To determine this, a relative rate test was performed on the ND1-16S sequences. As illustrated in Figure 10, if 44

there is any trend toward a speedup in substitution rate, it is in the lineage leading to H. oregonensis. Therefore, I conclude that H. pugillis does not show a faster mitochondrial mutation rate due to a shorter generation time- It therefore seems imlikely that a faster mutation rate exists at nuclear loci to explain the species' greater amotmt of phenotypic divergence.

Hypothesis 4: Selection Population genetic factors that can influence the rate of divergence among populations are effective population size, mutation rate, and the amoimt of selection acting on each population. Migration interacts with these factors by limiting or preventing genetic differences from becoming fixed in a population. Given the lack of evidence for substantial differences in migration, effective population size, or mutation rate between H. oregonensis and H. pugillis to explain the difference in phenotypic divergence, it is reasonable to infer that selection has acted on nuclear loci to a greater extent in H. pugillis than in H. oregonensis. Migration does not appear to be more common in H. oregonensis, and generation time does not appear to be shorter in H. pugillis. There also appears to be no evidence for population bottlenecks or foimder-effect speciation driving divergence in H. pugillis. On the contrary, H. pugillis appears to have had historically larger effective population sizes on each mountain range than H. oregonensis. These larger population sizes may have allowed selection - either natural or sexual selection - to overcome the effects of genetic drift and sweep adaptive morphological and behavioral changes to fixation within moimtain ranges. 45

This presents a quandary, however, in that while directional selection is most effective in large populations, if it is very effective it will cause a decrease in the effective population size. Since this study examined only the effective population size of the maternally inherited mitochondria, I cannot determine if there has been a reduction in the effective population size of nuclear loci in H. pugillis relative to H. oregonensis, as would be expected if sexual selection due to female choice caused fewer H. pugillis males to contribute to the gene pool. Very little is known about the mating system of either of these spiders, and it is unknown if polygjmy exists in both species, as may be expected in organisms with apparently sexually selected traits. One possible explanation for the greater divergence of H. pugillis may be that its variable male traits are imder less constraint than are those of H. oregonensis. The male ornament that distinguishes H. oregonensis from all other Habronattus is the swollen tibia on its first leg. There is a high mortality rate associated with molting in spiders (Foelix 1982), and ornaments that enlarge the size of an appendage may be costly, especially during molting. In contrast, the variable ornaments of H. pugillis males do not alter the size of the appendages, and therefore may be more evolutionarily plastic. While the data support the notion that sexual selection has acted to a greater degree in H. pugillis, hence causing more substitutions in nuclear loci controlling phenotypic traits, altemative explanations do exist. It is possible that the actual mutation rates may vary among different regions of the genome, and that patterns of such variation may differ between species. If this is the case, then conclusions drawn from patterns evident at unlinked. 46 neutrally evolving mitochondrial genes may not reflect patterns in nuclear lod. Similarly, genetic architecture of the nuclear gene regions may differ between species. However, these alternatives seem imlikely, given the close phylogenetic relationship between H. pugillis and H. oregonensis.

Hypothesis 5: Time of Divergence of Mountain Populations It appears that H. pugillis populations are diverging phenotypically at a more rapid rate than H. oregonensis populations on the same mountain ranges. However, an alternative explanation is that the divergence of H. pugillis populations is actually much older than that of H. oregonensis. This explanation seems unlikely for several reasons. First, historical vegetation patterns in southeast Arizona suggest that habitats of both species were affected simultaneously by climate-induced vicariance events, with H. oregonensis' higher-elevation habitat subdividing slightly before that of H. pugillis. Second, if H. oregonensis secondarily invaded the sky islands after habitat fragmentation by ballooning on aerial webs (e.g. Bristowe 1939; Homer 1975; Vugts and Van Wingerden 1976), then we should see evidence of this migration in the gene tree. Instead, the genealogy of H. oregonensis does not show patterns of more recent migration (see h5q)othesis 1), but rather shows distinct sky island clades as compared to the genealogy of H. pugillis. This indicates that there has been minimal migration among most ranges and that enough time has passed since the ranges were colonized that the populations now show fixed differences. It seems imlikely that H. oregonensis would have initially colonized the mountains by migration after habitat fragmentation, then subsequently ceased migrating. 47

Another piece of evidence that would support the notion that the H. oregonensis populations are at least as old as H. pugillis populations would be to find similar divergence times between sky island clades. However, a requirement for dating the vicariance event(s) that isolated the sky island populations of these spiders is a reliable estimation of phylogenetic relationships among clades. Because lineage sorting is incomplete or migration has occurred in either species - especially H. pugillis - it is not possible to date all the nodes separating the moimtain popxilations. If we do not explicitly consider tree topology when calculating divergence times, we may date divergence times as being much older than the actual vicariance event, if we use clades in which lineage sorting has not gone to completion. Fortunately, because three of the moimtain ranges are reciprocally monophyletic for H. oregonensis, at least those nodes can be dated. To date the vicariance events, I used the maximum and minimum genetic divergences between individuals from different mountain ranges, divided by twice the substitution rate. Because the mitochondrial mutation rate in spiders is imknown, I used the estimate from insect and crustacean arthropods of 1.15% per million years (Brower 1994). The estimated times of divergence are based on maximum likelihood estimates from Figures 7 and 8. Molecular clock estimates indicate that the earliest divergence of H. oregonensis in the southeast Arizona ranges was 956,000 to 1,087,000 years ago (Table 3), when the Huachuca populations last shared a common ancestor with the northern populations of H. oregonensis. The divergences calculated for the split of the Santa Rita and the Huachuca populations are surprisingly 48 large - over half a million years ago, as is the split between the Santa Rita and Galiuro populations. Because none of the H. pugillis sky island clades are reciprocally monophyletic, the dating method used with H. oregonensis clades above cannot be applied to H. pugillis. In fact, in lieu of monophyletic taxa, technically there is no vicariance event to date. Clearly, then, the nature of the data make it impossible to make a direct interspecific comparison of divergence dates. However, two approaches can provide an imperfect and approximate comparison. The first involves dating coalescence times of H. pugillis populations on individual moimtain ranges using the larger data set used in Appendix B. Estimates from these analyses (see Appendix B) range from 26,000 to 291,000 years, for the Mule, Black, and Sierrita Mountains. These clades are not directly comparable to the clades measured for H. oregonensis (see Table 3) because they are different mountain ranges, but they do suggest that divergence dates for H. pugillis, if not yoimger than those for H. oregonensis, are at least similar. The second method for estimating divergence of H. pugillis, using solely the data in this paper, is to determine the most recent time a common ancestor was shared between a member of the hypothesized ancestral population, and an Arizona individual whose co- ancestry is not due to recent inter-moimtain migration. This would presumably provide a date when H. pugillis populations in southeast Arizona were panmictic, and predate vicariance events that isolated populations on different moimtain ranges. The individuals matching this description, Pg29GL from the northerrimost Galixiro Moimtains and PglOSM from contiguous habitat in the Sierra Madre, show an estimated divergence time of 391,000 years ago. This coalescence time may be older than the actual time of H. pugillis population fragmentation on the Arizona sky islands, but in the absence of a clear vicariance event giving rise to monophyletic groups, it is not reasonable to attempt to infer minimum estimates for the ages of population fragmentation. Again, this second estimate indicates that H.pugillis divergence is either younger than that of H. oregonensis, or at least similar. It must be admitted that, given the inability to directly compare interspecific estimates, it is possible that H. pugillis divergence is older. However, it seems clear that even if this were the case, divergence times for the two species are quite similar - too similar, in fact, to explain the interspecific difference in phenotypic divergence that is the subject of this paper. While a large variance is expected in estimates based on only a few individuals, these niunbers calcuiated for the vicariance events in H. oregonensis are about 50 times larger than the expected time of population splitting based on patterns of historical vegetation change. It is possible that the vicariance events relevant to jximping spider divergence are not reflected in the packrat midden data. The packrat midden dating of isolation of these woodlands was done only on a single range in this study, the Santa Catalina Mountains (Van Devender 1977), and therefore habitat fragmentation of the other ranges may have occurred longer ago. An additional possibility is that arachnids have higher mitochondrial substitution rates than other arthropods, inflating the date of the inferred vicariance events. Clearly, it is necessary to obtain a reliable method of estimating the rate of substitution in arachnids before firm dates can be put on these vicariance events. However, 50 the evidence from vicariance dating, packrat middens, and the gene tree topologies indicate that divergences among H. oregonensis populations are at least as old as those among H. pugillis populations. Thus, H. pugillis' greater phenotypic divergence is apparently not due to a greater overall period of population subdivision.

CONCLUSIONS AND GENERAL IMPUCATIONS

This study illustrates cin approach that may be teiken in some cases to separate the effects of population genetic factors influencing phenotypic divergence and incipient speciation. By comparing populations that share similar histories of isolation, it may be possible to distinguish what factors have played a role in shaping genetic divergence. Both H. pugillis and H. oregonensis co-occur on four mountain ranges, jdelding essentially four replicates in which to study the divergence process. A fifth motmtain range represents the extremes of each species' distribution, and therefore gives a replicate of the effects of isolation by distance. Because these species' populations are still in the process of diverging, the factors that have acted since population isolation may be the factors responsible for promoting speciation. Often in studies of speciation, the time since species have formed has been so long that traces of population genetic factors that promoted speciation have been erased. In such situations, we may only see evidence of selection that is currently acting in these systems, and incorrectly infer that this same selective force drove speciation. However, other population genetic factors may easily explain interspecific discrepancies in rates of phenotypic evolution. In the current study, without first specifically 51

addressing these other, nonselective, factors, it would be difficult to claim that a greater degree of selection has driven phenotypic divergence in H. pugillis as compared to H. oregonensis. Previous genealogical studies with numerous species have foimd that effective population sizes calculated fi:om mitochondrial genetic diversity are often lower than would be expected from censuses of current population size (reviewed in Avise et al. 1988, Avise 1994). However, as pointed out by Harrison (1991) and Avise et al. (1988), the apparent reduction in effective population size of mitochondria may actually be caused by directional selection acting on the mitochondria, and not due to a population bottleneck that could have effected nuclear loci as well. The finding that the mitochondrial protein coding genes used in this study are evolving neutrally allows us to rule out that directional selection has caused the apparent decrease in the mitochondrial effective population size of H. oregonensis as compared to H. pugillis. It is reassuring to find evidence to support the notion that - at least in some taxa - mitochondria may indeed be a good molecular marker for recovering population history. In this study, it appears that morphological and behavioral divergence are uncoupled from mitochondrial genetic divergence and thus do not reflect the true time of population subdivision. This de-coupling of phenotypic and neutral molecular evolution is consistent with theoretical predictions (Zuckerkandl and Pauling 1962, Kimura 1983) and empirical observations in a wide variety of taxa (reviewed in Avise 1994, Wilson et al. 1977; but see Omland 1997). It is predicted that selection acting on phenotypic characters may accelerate or slow their evolution as compared to that of neutrally 52

evolving lod. A surprising result of this study is that the de-coupling is much stronger in one of the two species, even though they are closely related and occupy similar ecological niches. Adaptation to different environments does not appear to explain phenotypic differences among ranges, as habitats on all ranges appear similar in terms of vegetation and substrate. Instead, the pronounced morphological and behavioral differences appear to be sexually selected characters, since these differences are present only in males. Another surprising conclusion is that H. pugillis still appears to be segregating old mitochondrial genetic variation, while maintaining uniform morphologies and behaviors within motmtain ranges. This is most striking for populations from the Santa Catalina and Galiuro Mountains. Morphologically these are the most different of any two populations in this study, and a briefly trained observer could easily identify the origin of a spider from either range. To a proponent of the biological species concepts (Mayr 1940), these H. pugillis populations would each be considered separate species, based on their morphological uniformity within each range and on the extrinsic geographic barriers promoting reproductive isolation. However, their mitochondrial DNA sequences do not jdeld monophyletic clades, in violation of the phylogenetic species concept (Cracraft 1983). In contrast, H. oregonensis shows little retention of ancestral mitochondrial poljnnorphisms among motmtain ranges, but displays little or no difference in courtship behavior and morphology among ranges. Although the Galiuro and Santa Catalina populations of H. oregonensis form reciprocally monophyletic clades on the gene tree, concordant with the phylogenetic species concept, they are essentially indistinguishable from each 53 other phenotypically. The contrast between incipient speciation patterns in the two closely related species, H. pugillis and H. oregonensis, highlights the difficulties in attempting to define exactly what a species is.

LITERATURE CITED

Avise, J. C. 1994. Molecular markers, natural history and evolution. Chapman «Sc Hall, New York.

Avise, J. C., R. M. Ball, and J. Arnold. 1988. Current versus historical population sizes in vertebrate species with high gene flow: a comparison based on mitochondrieil DNA lineages and inbreeding theory for neutral mutations. Molecular Biology and Evolution 5:331- rtAA

Braverman, J. M., R. R. Hudson, N. L. Kaplan, C. H. Langley, and W. Stephan. 1995. The hitchhiking effect on the site frequency spectrum of DNA polymorphisms. Genetics 140:783-796.

Bristowe, W. S. 1939. The Comity of Spiders. Ray Society, London.

Brower, A. V. Z. 1994. Rapid morphological radiation and convergence among races of the butterfly Heliconius erato inferred from patterns of mitochondrial DNA evolution. Proceedings of the National Academy of Science 91:6491-6495.

Brown, D. E., and C. H. Lowe. 1980. Biotic commuiuties of the southwest. Tempe, Forest Service, U.S. Department of Agriculture.

Carson, H. L. 1975. The genetics of speciation at the diploid level. The American Naturalist 109:83-92.

Carson, H. L., and A. R. Templeton. 1984. Genetic revolutions in relation to speciation phenomena: the foimding of new populations. Aimual Review of Ecology and Systematics 15:97-131. 54

Clary, D. O., and D. R. Wolstenholme. 1985. The mitochondrial DNA moleciile of Drosophila yakuba: nucleotide sequence, gene organization, and genetic code. Journal of Molecular Evolution 22:252- 271.

Cracraft, J. 1983. Species concepts and spedation analysis. Pages 159-187 in Current Ornithology (R. F. Johnston, eds.). Plenirai Press, New York.

Darwin, C. 1859. The origin of species by means of natural selection. John Murray, London.

Felsenstein, J. 1981. Evolutionary trees from DNA sequences: a maximum likelihood approach. Journal of Molecular Evolution 17:368-376.

Felsenstein, J. 1985. Confidence limits on phylogenies: An approach using the bootstrap. Evolution 39:783-791.

Foelix, R. F. 1982. Biology of spiders. Harvard University Press, London.

Fritts, H. C., D. G. Smith, J. W. Cardis, and C. A. Budelsky. 1965. Tree-ring characteristics along a vegetation gradient in northern Arizona. Ecology 46:393-401.

Fu, Y.-X., and W.-H. Li. 1993. Statistical tests of neutrality of mutations. Genetics 133:693-709.

Gilbert, D. 1994. SeqApp.

Griswold, C. E. 1987. A revision of the jumping spider genus Habronattus F.O P.-Cambridge (Araneae; Salticidae), with phenetic and cladistic analyses. University of California, Entomology 107:1-344.

Harrison, R. G. 1991. Molecular changes at speciation. Annual Review of Ecology and Systematics 22:281-308.

Hasagawa, M., H. Kishino, and T. Yano. 1985. Dating of the human-ape splitting by a molecular clock of mitochondrial DNA. Journal of Molecular Evolution 21:160-174.

Hedin, M. C. 1997. Speciational history in a diverse clade of habitat- specialized spiders (Araneae: Nestiddae: Nesticus): inferences from geographic-based sampling. Evolution 51:1929-1945. 55

Hey, J., and R. M. Kliman. 1993. Population genetics and phylogenies of DNA sequence variation within the Drosophila melanogaster species complex. Molectilar Biology and Evolution 10:804-822.

Homer, N. V. 1975. Annual aerial dispersal of jumping spiders in Oklahoma (Araneae, Salticidae). The Journal of Araclmology 2:101-105.

Hudson, R. R. 1990. Gene genealogies and the coalescent process. Pages 1-44 in Oxford Surveys in Evolutionary Biology (D. Futuyma and J. Antovics, eds.). Oxford University Press, Oxford.

Jukes, T. H., and C. R. Cantor. 1969. Evolution of protein molecules. Pages 21- 132 in Mairunalian Protein Metabolism (H. N. Mimro, eds.). Academic Press, New York.

BGm, J. 1993. Improving the accuracy of phylogenetic estimation by combining different methods. Systematic Biology 42:331-340.

Kimura, M. 1983. The Neutral Theory of Molecular Evolution. Cambridge University Press, Cambridge, England.

Kliman, R. M., and J. Hey. 1993. DNA sequence variation at the period loctis within and among species of the Drosophila melanogaster complex. Genetics 133:375-387.

Maddison, D. R. 1991. The discovery and importance of multiple islands of most-parsimonious trees. Systematic Zoology 40:315-328.

Maddison, W., and M. McMahon. submitted 1999. Differentiation of the jumping spider Habronattus cf. pugillis Griswold among the montane "sky islands" of Arizona. Systematic Biology.

Maddison, W. P., and D. R. Maddison. 1997. MacClade: Analysis of phylogeny and character evolution. Stmderland, Massachusetts, Sinauer.

Mayr, E. 1940. Speciation phenomena in birds. The American Naturalist 124:249-278.

Mayr, E. 1954. Change of genetic environment and evolution. Pages 157-180 in Evolution as a process (J. Huxley, A. C. Hardy and E. B. Ford, eds.). Macmillan, New York. 56

McDonald, J. H., and M. Kreitman. 1991. Adaptive protein evolution at the Adh locus in Drosophila. Nature 351:652-654.

Nei, M. 1987. Molecular Evolutionary Genetics. Columbia University Press, New York.

Neigel, J. E., and J. C. Avise. 1986. Phylogenetic relationships of mitochondrial DNA under various demographic models of speciation. Pages 515-534 in Evolutionary Processes and Theory. Academic Press.

Omland, K. E. 1997. Correlated rates of molecular and morphological evolution. Evolution 51:1381-1393.

Pase, C. P., and D. E. Brown. 1982. Rocky Mountain (Petran) and Madrean montane conifer forests. Pages 43 in Desert Plants: Biotic commimities of the American southwest- United States and Mexico (D. E. Brown, eds.). University of Arizona.

Peckham, G. W., and E. G. Peckham. 1909. Revision of the Attidae of North America. Transactions of the Wisconsin Academy of Sciences, Arts and Letters XVI, part 1:355-646.

Rozas, J., and R. Rozas. 1997. DnaSP. Barcelona, Spain.

Rozas, J., and R. Rozas. 1999. DnaSP 3.0. Barcelona, Spain.

Saitou, N., and M. Nei. 1987. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecxilar Biology and Evolution 4:406-425.

Sarich, V. M., and A. C. Wilson. 1973. Generation time and genomic evolution in primates. Science 179:1144-1147.

Simon, C., F. Frati, A. Beckenbach, B. Crespi, H. Liu, and P. Flook. 1994. Evolution, weighting, and phylogenetic utility of mitochondrial gene sequences and a compilation of conserved polymerase chain-reaction primers. Aimals of the Entomological Society of America 87:651-701.

Simonsen, K. L., G. A. Chiarchill, and C. F. Aquadro. 1995. Properties of statistical tests of neutrality for DNA polymorphism data. Genetics 141:413-429. 57

Slatkin, M., and W. P. Maddison. 1989. A dadistic measure of gene flow inferred from the phylogenies of alleles. Genetics 123:603-613.

Swofford, D. L. 1998. PAUP*: Phylogenetic analysis using parsimony. Sunderland, Massachusetts, Sinauer Associates.

Tajima, F. 1983. Evolutionary relationship of DNA sequences in finite populations. Genetics 105:437-460.

Tajima, F. 1989. Statistical method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics 123:585-595.

Templeton, A. R. 1980. The theory of speciation via the foimder principle. Genetics 94:1011-1038.

Van Devender, T. R. 1977. Holocene woodlands in the southwestern deserts. Science 198:189-192.

Van Devender, T. R., and W. G. Spaulding. 1979. Development of vegetation and climate in the southwestern United States. Science 204:701-710.

Vugts, H. F., and W. K. R. E. Van Wingerden. 1976. Meteorological aspects of aeronautic behavior of spiders. Oikos 27:433-444.

Wilson, A. C., S. S. Carlson, and T. J. White. 1977. Biochemical evolution. Aimual Review of Biochemistry 46:573-639.

Wright, S. 1951. The genetical structiire of populations. Annals of Eugenics 15:323-354.

Yang, Z. 1993. Maximimi likelihood estimation of phylogeny from DNA sequences when substitution rates differ over sites. Molecular Biology and Evolution 10:1396-1401.

Yang, Z. 1994. Maximum likelihood phylogenetic estimation from DNA sequences with variable rates over sites: approximate methods. Journal of Molecular Evolution 39:306-414.

Yang, Z., N. Goldman, and A. Friday. 1994. Comparison of models for nucleotide substitution used in maximum-likelihood phylogenetic estimation. Molecular Biology and Evolution 11:316-324. 58

Zuckerkandl, E., and L. Pauling. 1965. Evolutionary divergence and convergence in proteins. Pages 97-166 in Evolving Genes and Proteins (V. Bryson and H. J. Vogel, eds.). Academic Press, New York. Appendix H. oregonensis NDJ-16S haplotypes

1 10 20 30 40 50 60 70 RO 00 100) ( ...... 1

HborlSC --ATA'rGAGAAGaGAAAAAGAAATAAGGAC;GTAAATGAGGAA'rAAGAAAAA(3ATAi;'rAGrm"ITGAAAAAAGATAAA(JTCi(iTC0GAAAt:A(!AA'r'raT'n' Hbor2SC -G Hbor3SC -- r. Ilhorfl.'if.' -(! c; JIborl2t;C -G Hborl3SC -0 G G HborlGSC -G G Hborl7SC -G. Hbor32SC -G.... Hbor33SC -A HborlGL -G..W..G 0.. , ...G.... Hbor2GL Hbor3GL GG.... HborSGL, GG HborQGL -G .G HborlOGL _ _ HborllGL -G.....G llborl7(iI, lllidt 1 fl( !I. Hbor2!i(:L, HborlR » _ Hbor2R -G..,, G ...... O...,, .A. . Hbor3R G Hbor4R -G G .. .G Hbor6R GG Hbor7R 0. . ,A. . .. .G HborlOR -G Hborl3R G HborlSR « _ Hborl6R -G HborlH --.C.. G G...... G .. .Q Hbor2H -A.C.. G .,.GW Hbor3H GA.C,. G G. . . , , .(J Hbor6H+Hbor7H --.C.. G G. , , ...G..,. ...CG.G. , . ,G Hborl4H --.C.. G G. . . . , ,G . . . .G...... CO.n. .. .G HborlSH --.C,. Hborl6ll -A.C.. llborl7H -A.C.. 'Ji ac tt s s s s 3 s s s a ae cr o* trXXX tr tr trX tr trX trX xxxsxxxxxxxxxxxxx 0tr 0"0 ocr ocr crO ocr otr oo* ocr ocr o o O O 0 0 o o o oooooooooooooooootrcTD'crcrtrcro'O'crD-crirtrcrcrtr >-»•1 H* I-* »-• so>1 •1tn Hu« lO•1 M CJ UJ M ^l-» M M *1w •1to l-» OD O w o a o o O o u» to -J 0\ CJ to COCD CO CO CO SSSCoOOvoco\o«Nji^OOOOnao or or Or rO r r r r r* oCO oCO oCO oCO oCO oCO o o o o 5o»3xnoooooo sc

1i I I I lOiI I I It 1lOOOOOi i I Q i n I 1i> i o O O O n o ^

:::::::::: ::;:::5 n ^ O 7) S o 1C o b ^ OO-?: >9 to O a o O nnMMHMnnni o iiO a a C3 O o n o 0 O • O I I I ^ a >o C3 a o n a a o •••*•••••••••*•*•• o >9 UnniHiMHMMI o a a o o Q o a o

cn o 0C50i00000i* '* * 4> • • OiOOOOOi* OOO* t*I - •11-3 W IIII* - II-.,.. ||.... i.I. i>iQ 1 I .I ..... I I .... I . I

a o o a o

o a a o a an

> > o

a Q a

> > > >>>>>> oO

09 1 110 120 130 140 150 ) I 1

Hbor25GL. ....T. ....G...... G... HborlR A...... T. Hbor2R A ....T. Hbor3R A...... T, ....W.AG.... Hbor4R A...... T. Hbor6R A...... T. Hbor7R A.. . . HborlOR A.. . . HborlBR A HborlSR A Kboriea A.. . . HborlH ....T. Hbor2H A...... T. , A .A.. . Hbor3H ....T. Hbor6H+Hbor7H ....T. .A...... C. .K.. . Hborl4H T. • A.. . G. , HborlSH Hborl6H ....T. Hborl7H HborlBH Hbor2C Hbor3C HborSC ....T.. . .C ..A. . Hbor6C ....T.. . .C Hbor7C Hborl4C ....T.. . .C ..A. . .T. , , ....CM... Hborl7C ,.T,., W. -- Hborl9C ....T.. . .C ..A. . Kbor2BC ....T. .. .0 ..A. . Hbor29C HborlGC Tft w.. . Hbor2GC ....T. A.. .A. . ..T.G.... KborlSK HborlMR Hbor2MR ....T. .. A...... K..M HborSMR ...- All H. oregonensis names abbreviated in the text by deleting 'Hb'. Locality abbreviation: MR=Mad River CA SH=Mt. Shasta CA; GC=N. Grand Cyn AZ; GL= Galiuro Mts; SC=flanta Catalina Mts; H=Huachuca Mts; SR=nanta Rita Mts; C=Chiricahud Mtii. (lUPAC symliols used for uncertainty) *0IQ 'OIQ *0UJ •0IQ tJIB •Qto •8IQ UJ to 00 LP UJ to MCAU3MCOCOQCPC9C5DC0OL71 tU U> §f?so f? CO o> in CO:O:SSPO?oooooooo

I I I I I I I I I I I I I I I I • 8 I J I I I o o o o n > > i > b o a 8 o o a o •9 I •a h > > b 0 • a a > > > is5 § >>5 Si> S•3 S 5"3 S o n O o a 1 o o o a a n o o o o o a a b n n n > o o

o . , . . G G G G

> > > > > > > > i P > > > > > > > > > > > > > > > b b I § S § § S ^ S § § § •a "•9 ^ •3 ">0 •3 •3 "•a ^ '>9

> > > > > > > > > > > > > >

o o n 1 > > > > > > > > > > > > > > > o o o n s

b a 8 o 5 > > > > > > > 5 > > >>>>>> > > > > > > b o s c o a a o o o o o o Q O o a "8 o o o o • b oO

Z9 ( 10 20 30 40 50 60 70 80 90 101 ( ......

Hbpgl5SC GAT. Hbpg20SC Hbpg24SC .. .G ....A. .AT. , A A , .A A. , G . ,A, , Hbpg25SC G .. .G ....A. .AT. A A ..A A. . n ..A. . Hbpg26SC Hbpg28SC G .. .G ....A. A A ..A A. , n . .A. . HbpgllGL G .. .G ...GA. • AT. A A , .A A. , n A Hbpg29GL GCGG.....C..A..G.. G...A. ..T. . A .. .G. G.C. A ..C.... .A .GGW.A.. . 'I>C HbpgBOGL ..T. . , .A.. A A ..A.G.... .AA. . HbpgSlGL _ ^ _ .. .R ,.T. ,,,R., . .R.. A A ..A.G.... .A. . , Hbpg43GL G .. .G .. .GA, , .AT. A A ..A A. , n . .A , Hbpg50GL .. .G .. .GA, , .AT. A A ..A A. , n ..A. , Hbpg52GL G .. .G ...GA,. .AT. A A . .A A. , n . .A. . Hbpg55GL GC .. .G ...GA.. .AT. A A ..A A. n ..A. , Hbpg56GL .. .G ...GA,. .AT. A A ..A A. n . ,A. , H!I:^66GL .. .G .. .GA, • AT. A A ..A A. , n ..A. , H.geronimoi A.G..G..AAGGTGA.T. .G GATG .A,,A r. .A • A. TAAGT..GAT...G. .GGTC.GATAG,,.(•?, , AT . . A. .

[ 110 120 130 140 150 160 nO 180 1 ( ...... 1

'HbpglBQ+Hbpg2BQ' ATMTTTTTATTGTTAAAATTGATAAAATTCTTTTTCAACTTATTGAACCTAACGATATTTTATTATAGTATA'l'TTTTAASACMC- - Hbpg3BQ T.-- Hbpg4BQ W.-- HbpgSBQ Hbpg7BQ C GA Y HbpgSBQ C C GA Hbpg9BQ C GA W HbpglSR C T AA C GT HtpgSSR C T AA Hbi»g7SR C T S AA C... C T AA C... HfcpgSSR C T AA C -- Hb^llSR C T AA Hbpg25SR C T AA G- 'Hbpg36SR+37SR' C T AA C G- Hbpg46SR C T AA C G- HbpglH C T AA T.G- Hbpg2H C T AA C. . . .G- 110 120 130 140 150 160 170 180 ] ( J

Hbpg3H .C. .AA. -.0- HbpgVH .C. .AA. HbpgBH .C. .AA. . . .G- Hbpg21H .C. .AA. . ..G- Hbpg25H .C. .AA. • T--- Hbpg26H .C. .AA. . . .G- Hbpg27H .C. .AA. . ..GT Hbpg34H .C. HbpglSM .C. .A. .C. .G..A. .AT Hbpg2SM • C. .A. .C. • G..A. .AT Hbpg3SM .C. .A. .C. .G..A. .AT HbpgSSM .C. .A. .C. .G..A. .A. .AT HbpgSSM GC.G T. .C. .G. • G. • A. .AT HbpglOSM G A • C. .G. .A..A. .AT G..G .C. .A. .AT Hbpgl2SM .C. .A. .TA. .A.. m^isc .G. .C. .T. .TA. .TC Hbpg2SC • G. .C. .T. TA C.. .T.AC Hbpg3SC .T. • C. .G.TA..C C HbpglOSC • A. .C. AA HbpglSSC .A. .C. .AA. C GT Hbpg20SC .A. .C. .AA. C GT Hl7pg24SC .G. .C. .T. .TA. Hbpg25SC .G. .C. .T. .TA. .AC Hbpg26SC .C. .AA. .GT Hbpg28SC .G. .C. ,T, .TA. .AC HbpgllGL .G. .c. .W.. .T. .TA. Hbpg29GL. .c. .V4W. .T. .A.. Kbpg30GL .G. .c. .GT. .A.. .CG AT Ht^3lGL .G. .W.MWR TC.C..G.A.C.AA G T C .A. . Hbpg4 3GL .G. C.. , . ..T.., .T. .TA. HbpgSOGL .G. C.. , ...T.., .T.WT TA. Ht3pg52GL .G. C.. , ...T.., .T .C.TA HbpgSSGL .G. C.. , .,.T.., .T ,,.TA ..T.A- Hbpg56GL .G. C.. , ...T.., .T..- .. .TA A- Hbpg66GL .G. C.. , ...T.., .T ...TA .W. A- H.geronimoi G.GTAA.AAG.GT., .AA.CX3., . .T. G. • T.AG, , .A,.T., ...AAGAT.,A....C...TTTA All H. pugillis names abbreviated in the text by deleting 'Hb' Locality abbreviation: SM=Sierra Madre, Mexico; GL= Galiuro Mts; SC=Santa Catalina Mts; H=Huachuca Mts; SR=Santa Rita Mts; BQ= Baboquivari Mts, (lUPAC symbols used for uncertainty) 65

Table 1 McDonald-Kreitman test Number of replacement and silent nucleotide differences within and between species

NDl Within Between H. piigillis H. oregonensis Replacement 6 2 2 Silent 70 36 7 p value 0.16 0.141

COl Within Between H. piigillis H. oregonensis Replacement 5 1 1 Silent 52 43 12 p value 0.415 0.358

Fisher s exact tests were used to test the null hypothesis that the ratio of replacement to silent differences is the same within a species as between species. The null hypothesis was not rejected in ciny comparison. Table 2

DNA sequence variation in ND1-16S in H. oregonensis Location n K % Tajima's D F*

Sta Catalina 10 0.331 -1.412 -1.813 -1.927 Galiuro 10 0.085 0.775 0.175 0.358 Santa Rita 10 0.092 -1.562 -1.916 -2.077 Huachuca 10 0.133 -0.657 -0.686 0.796 Chiricahua 10 0.399 -1.768' -2.093" -2.267" N.AZ+N.CA" 5 1.229 all localities'" 55 1.297

DNA sequence variation in ND1-16S in H. pugillis

Location n 7t % Tajima's D D*

Sta Catalina 10 1.913 0.054 -0.724 -0.662 Galiuro 10 1.373 -1.518 -1.354 -1.524 Santa Rita 10 0.231 0.264 -0.024 0.166 Huachuca 10 0.527 -0.849 -1.144 -1.328 Baboquivari 8 0.202 0.585 0.876 1.06

Sierra Madre 8 2.683 all localities 56 2.208 ® significant p<0.01 '' significant p<0.05 HborlSH was excluded from the analysis because only NDl was sequenced. 67

Table 3

Divergence Times of H. oregonensis

CLADES COMPARED MIN AND MAX DIVERGENCE Huachuca and northern AZ and CA 956,000 to 1,087,000 years ago Huachuca and Santa Rita 522,000 to 696,000 years ago Santa Rita and Galiuro 391,000 to 652,000 years ago

Coalescence Times of H. oregonensis

CLADE COALESCENCE TIME Huachuca 173,900 years ago Santa Rita 173,900 years ago Galiuro 173,900 to 217,400 years ago

Estimates are based on the maximum likelihood values calculated over the tree described (see Fig. 7). 68

Figure Legends

1. Photos of H. oregonensis males from each mountain range. The major differences are in coloration of the hairs under the first tw'o legs, being rust colored in the Santa Catalina and Galiuro males, and gray in the Santa Rita, Huachuca and Chiricahua males. 2. Photos of H. piigillis males from each mountain range. Note the differences in presence and extent of eyestripe, length and coloration of hairs on the first leg, and amount and coloration of scales on the palps. 3. Map outlining the distribution of oak woodlands (after Brown and Lowe 1980) on the mountains from which H. oregonensis and H. piigillis were collected. The light gray dots correspond to collecting localities for H. oregonensis and the dark gray dots for H. piigillis. The names of the mountain ranges have been abbreviated on the map, and in all subsequent taxa names as follows: SC = Santa Catalina; GL = Galiuro; C = Chiricahua; H = Huachuca; SR = Santa Rita; BQ = Baboquivari. 4. Parsimony strict corisensus tree of 50,000 trees for the ND1-16S gene of H. oregonensis. Bootstrap values > 50% are above the branches, tree length = 218. The clades A-E are described in the text. Horizontal bars along the right margin delineate collecting locality. Note that not all individuals from a locality are contained within the bar. 5. Parsimony strict consensus tree of 50,000 trees for the ND1-16S gene of H. piigillis. Bootstrap values > 50% are above the branches, tree length = 302. The clades A-C are described in the text. Horizontal bars along 69

the right margin delineate collecting locality. Note that not all individuals from a locality are contained within the bar. Parsimony strict consensus tree of 216 trees for the COl and ND1-16S genes from a subset of H. oregonensis and H. piigillis individuals. . Bootstrap values > 50% are above the branches, tree length = 386. One of nine maximum likelihood trees found for the ND1-16S gene of H. oregonensis. The -In likelihood score = 2076; estimated transition/transversion = 2.284; estimated gamma shape parameter = 0.1725. The base composition is A = 0.378; C = 0.084; G = 0.147; T = 0.392. One of 21 maximum likelihood trees foimd for the ND1-16S gene of H. pugillis. The -In likelihood score = 2297; estimated transition/transversion = 3.658; estimated gamma shape parameter = 0.1106. The base composition is A = 0.367; C = 0.106; G = 0.133; T = 0.395. Neighbor-joining tree for the ND1-16S gene of H. oregonensis. Geographic locality is traced onto the tree as an unordered character. The number of steps required, s, for the reconstruction over 100 neighbor-joining bootstrap trees is graphed in the lower left comer. Neighbor-joining tree for the ND1-16S gene of H. pugillis. Geographic locality is traced onto the tree as an unordered character. The number of steps required, s, for the reconstruction over 100 neighbor-joining bootstrap trees is graphed in the lower left comer. Relative rate test for the ND1-16S genes of H. geronimoi, H. oregonensis, and H. pugillis. Values of a, b and c were calculated from the number of nucleotide substitutions between these three species. Figure 1 Habronattus oregonensis forms

Santa Catalina Galiuro

Chiricahua Santa Rita Huachuca Figure 2 Habronattus pugillis forms

Santa Catalina Galiuro

Baboquivari Santa Rita Huachuca Figure 3

^ North

SC 10 miles

Tucson I

Sonora, Mexico Arizona 73

H. oregonensis Figure 4 ND1-16S Parsimony Strict Consensus 64 75 •PglBQ 98 leng& = 218 •Pg56GL •OrISC •Oi2SC values >50% •OrSSC a above branches •OtSSC c 58 -Oi2SC •t;a ~® -Orl3SC c -t: -Orl65C CO U -OrlTSC 67 • Or32SC -OrlTC -OiSSSC -OrlGL •OtSCL 72 -OrSGL -Ot9GL 2 -OrllGL s -Or3GL "5 -OrlOGL U 60 -Orl7GL -OiglSGL • Oi25GL • OrlSR • Oi2SR -OrSSR -Oi4SR (S 72 •O16SR c S -OrTSR cne; -OrlOSR -Orl3SR 100 -OrlSSR -OrieSR -OrlH -Or2H a -Or3H u -Or6H+7H X3 65 .Orl4H esu -OrlSH 3 • Orl6H E -OrlTH -OrlSH -Or2C -Or3C ce -OtSC 3 -OtfiC 65 -OrTC g -Orl4C W -Orl9C u -Oi28C -Ot29C -OrlGC -Or2GC 65 -OrlSH N< - Oi5MR

cd \oON

o ui o a>

s js J jp js js»^ J? J? J?/S iff 09.(ff iff (!?(!?09 9 (ff il?(l?(ff (!?o?o?(!?"B.'S o?9J

Huachuca Santa Santa Galiuro Sierra Baboquivari Rita Catalina Madre 75

Figures H. oregonensis and H. pugillis C5rlH COl + ND1-16S 100 C>t2H Parsimony Strict Consensus of 216 trees Oi3H OrlTH Huachuca length = 386 Or2SC bootstrap values > %50 OiSSC above branches 85 Orl6SC Santa Catalina 64 OtSZSC C)rgl7C 68 OrlGL

88 OiSGL Oi9GL

84 Galiuro 96 OrllGL 77 OrSOL Or33SC (S 3 OrgTC JZ 100 S Or29C *c IS 100 OrlSR u 61 OrSSR 100 Santa OrlOSR Rita Ot4SR < 96 OrlSH u •OrlMR 2 •PglSM 88 100 •PglBQ •PgTBQ 59 •PglSC 100 100 -P^4SC • Pg56GL 97 -PgTSR 100 -PgSSR Santa Rita -PgllSR - H-gerociimot 76

Maximum Likelihood tree H. oregonensis Figure? (one of 9 trees)

-ki likelihood = 2076.1^ 0.013 PglBQ branch lengths estimated with 0.006 Ts/Tv = Z284 0.025 gamma shape parameter =0-1725 0.001 •PglSM freq. A=0378 C=0.084 G=0.147T=0392 0.036 0.010

0.007 Pg56GL

0.003 — OrSSC

0.003 0.001 r C)i32SC 0.001 ^Orl7C 0.001 0.001 0.001 rOaCL

0.001 l.OOOe-08 Or3GL

Orl7GL 0.003 0.004 0.001 rOr3C 0.009

„'T-.Or29C 0.001 0.003 ^C>r33SC 0.003

•3R 0.012 0.003 -OrTR 0.101 0.001

0.001 9^C)r2H 0.006 l-OreH 0.045 0.001 0.003 ^OlMR

0.003 — OrlGC

H.geionisioi 77

HgureS

H. pugUlis 0.005 Pg3BQ 0.007 Maxiinuni Likelihood tree 0.004 (one of 21 trees) PgSBQ -In likelihood = 2297.431 5:2^PglSM branch lengths estimated with 0.020 Ts/Tv = 3.658 gamma shape parameter =0.1106 ^^Pgl2SM freq. A=0367 C=0.106 G=0.133 T=0395

0.028 PgSSC 0.003

I Pg9SR 0.008 0J04 Pg34H

0.001 0.008 p-Pg46SR 0.001 0.003 Pg?7H

0.003 0.004 0.005 PgSH 0.019 0.008 PglSSC l.OOOe-08 PglSC 0.009 0.003 Pg52GL

°|2^PglOSM 0.018 0.007 •Pg29GL

0.006 •OrlGL 0.001 0.007 •OrlH

0.040 0.014 OrlMR 0.001

o.o5r°'® 0.070 •H.geianimoi Number of trees ooiocnooiooio eg (3 s>VO iin PgiBQ ID PgSSR ID PglSM ID Pg56GL — cSisc Or3SC Orl2SC Oi2SC QrlTSC OrSSC Orl3SC QrieSC Or32SC OrlTC OrSBSC OrlOGL Orl7GL OrlSGL Or25GL OrlGL OrllGL Or9GL OrZGL Or3GL Or2C OrSC Or7C Orl4C Orl9C Oi28C Oi29C Or6C OrSC OrlSR OrSSR Orl6SR OrlOSR Qr2SR O16SR Or7SR Or4SR OrlSSR OrlSSR OrlH OiBH Orl6H Orl7H OrlSH Orl4H OrlSH Or6H+7H OOH OrlGC Qz2MR OrlSH OiSMR OrlMR Oi2GC H. geronimcn Figure 9b

locale unordered 12 > I S. Catallna i> ""I Galluro •• Huachuca HI Baboquivarl •• S. Rita HI Sierra Madre ^3 equivocal

s 80

Figure 10 Relative Rate Test

H. geronimoi

H. pugUlis

H. oregonensis

Average niunber of nudeotide substitutions per site between populations

H. geronimoi vs H. pugillis 0.09366 (SD = 0.01242) H. geronimoi vs H. oregonensis 0.11620 (SD = 0.01554) H. pugillis vs H. oregonensis 0.08582 (SD = 0.00244)

a = 0.03164 b = 0.05418 c = 0.06202 81

APPENDIX B

PHYLOGECX^RAPHY OF HABRONATTUS PUGILUS (ARANEAE:

SALTICIDAE) ON THE SKY ISLANDS OF SOUTHEAST ARIZONA 82

Phylogeography of Habronattus pugillis (Araneae: Saltiddae) on the sky islands of southeast Arizona

Susan Masta Department of Ecology & Evolutionary Biology University of Arizona Tucson, Arizona 85721 [email protected]

RH: Phylogeography of jumping spiders in Arizona

Key Words: Habronattus pugillis, phylogeography, vicariance, sky islands, migration, lineage sorting 83

Abstract

Studies of historical landscape change may provide a framework for studying vicariance events in populations of organisms, while phylogeographic studies of organisms may in turn aid interpretation of the history of landscape change. Such is the case in the "sky island" moimtain ranges of southeast Arizona, where climate-induced contraction of woodlands fragmented populations of woodland-dwelling species. A phylogeographic study of montane populations of the jumping spider Habronattus pugillis may help clarify the history of vegetational change. Using mitochondrial sequence data, I investigate phylogeographic patterns in 13 isolated populations of this incipiently speciating spider, thus providing the first such population genetic analysis in the Arizona sky island system. Gene trees show mitochondrial sequences from individuals from neighboring moimtain ranges sometimes form a clade, suggesting recent migration among populations, but geographic information offers stronger evidence that incomplete lineage sorting accoimts for most cases of non- monophyletic groupings. Smaller ranges tend to have populations of spiders whose mitochondrial DNA is completely sorted into clades corresponding to the moimtain range from which they were collected. The gene trees provide no clear indication of the chronological sequence of woodland fragmentation, but suggest an old north-south geographic divide. Dates estimated for divergence of spider populations vary from 26,000 to 291,000 years ago, but are dependent on molecular clock estimates from insects, which may be inadequate for arachnids. However, the dates calculated from the 84 mitochondrial sequence divergences are much older than the presumed 10,000-year-old vicariance event that fragmented the woodland habitat between mountain ranges.

Introduction

Populations of organisms that are isolated on islands offer ideal opportimities to study the processes of morphological and genetic differentiation that result from natural selection or genetic drift. When populations become isolated from one another, they will begin to diverge as new mutations and old polymorphisms become fixed on each island. Migration between populations may be sufficient enough to oppose differentiation of populations by preventing genctypic fixation. Alternatively, gene flow may help to spread new adaptive gene combinations. Determining how much gene flow actually occurs during the process of differentiation and incipient speciation in different organisms has been the focus of many studies (reviewed in Avise 1994). However, in many of these studies, separating the effects of gene flow from those of historical association is problematic. When populations have separated recently, parental genes may still be shared by both populations. Historical association (incomplete lineage sorting) and gene flow both may leave similar genetic patterns. One potential way to separate these two factors is to use natural history and geographic information to infer whether populations may be exchanging migrants. Because islands contain geographically discrete populations, methods for distinguishing between gene flow and incomplete lineage 85 sorting based upon geographic locale and isolation by distance may be particularly useful in island systems. The "sky islands" of southeast Arizona are forested mountain ranges separated by a sea of desert and grassland. They contain the northernmost extent of the Madrean evergreen woodlands, which extend south into northern Mexico (Brown and Lowe 1980; Brown 1982). The disjimct patches of these woodlands on Arizona's sky islands are believed to have had a greater degree of connectivity during the late Pleistocene to early Holocene, as recently as 10,000 years ago, when the climate was cooler and more mesic. This presumption is based on remains from packrat middens (Betancourt, Van Devender, and Martin 1990) that indicate that woodland habitat may have spread contiguously across lower elevations at that time, effectively linking moimtain ranges. Today, the isolated mountain habitats support populations of plants and animals that would have been affected by climate- induced range fragmentation as the woodlands moved upslope with warming and drying climate. Like other island ecosystems, southeast Arizona's mountain ranges provide ideal opportimities to study the processes of population differentiation caused by genetic drift and selection, and to examine the role of migration during this process. Because the sky island habitats have subdivided relatively recently, populations of organisms may not yet be completely differentiated into distinct species, thereby providing opportimities to study the process of incipient spedation. Surprisingly, few studies have utilized this natural laboratory to study population differentiation. Earlier studies have focused on the relationship between island biogeography (MacArthur and Wilson 1963,1967) and species richness. 86 with particular emphasis on mammals (e.g. Lomolino, Brown and Davis 1989). More recent work has included descriptions of morphological divergence among moimtain popvdaticns of the plcint Castilleja austromontana (Slentz, Boyd and McDade, pers. comm.). Little is known about genetic divergence among these ranges, except in two populations of the plant Lilium parryi (Linhart and Premoli 1994). None of these studies attempted to date the isolation of populations on different ranges. Consequently, very little is known about the order and timing of fragmentation of the montane woodlands from continuous woodland habitat. If woodland fragmentation is recent, it may be difficult to distinguish between ongoing migration among ranges and incomplete lineage sorting when the genetic connections of populations are examined. Dating of woodland fragmentation in southeast Arizona is based on packrat midden data from a few locations at the northwestern edge of the Madrean sky islands (Van Devender 1979). Surprisingly, evidence of oaks {Quercus spp.) which are typical of the Madrean flora, have not been foimd in these particular middens. This may be due either to insufficient sampling or to a recent invasion of Madrean flora from the south. If the latter is the case, then we may expect to see evidence of a recent northward range expansion of organisms associated with Madrean woodlands, with little genetic subdivision among individuals from different moimtain ranges. If Madrean woodlands were present in southeast Arizona during the Pleistocene, but simply not foxmd in the two middens sampled, it may be that Madrean woodlands were patchy and isolated from each other much longer ago. If this 87 is the case then we may expect to find that gene sequences from populations from given mountain ranges form monophyletic groups, if lineage sorting has gone to completion and there is no gene flow. Therefore, a phylogeographic study (sensu Avise et al. 1987) of organisms living in these fragmented woodlands may shed light on the historical associations of woodlands in southeast Arizona. This might in turn be used in turn to make phylogeographic predictions for other organisms dependent on these island habitats. Habronattus pugillis Griswold (Araneae: Salticidae) inhabits the Madrean woodland habitats, residing in oak woodlands at elevations above 1600 meters. It is completely absent from the lower-elevation desert and grasslands. This ground-dwelling jimiping spider has been collected from Nayarit (Griswold 1987), through the western Sierra Madre Occidental of Sonora and Chihuahua, Mexico, to its northernmost limit on the sky islands of southeast Arizona (Bodner, Masta, and Maddison, pers. comm.). The striking divergence in morphology emd courtship display among males living on different moimtain ranges in Arizona has been attributed to sexual selection (Maddison and McMahon, submitted). If we assiune that isolation of these populations occurred when woodlands retreated to higher elevations about 10,000 years ago, or via recent expansion of Madrean woodlands into Arizona, then this species of jiunping spider appears to be phenotypically differentiating at a rapid rate on each moimtain range. Alternatively, isolation of these populations may have occurred much earlier them the postulated end-Pleistocene vicariance, yielding a less remarkable rate of differentiation. Whether this divergence is actually rapid and recent can be 88

inferred from studies of genetic differentiation. If DNA sequences from populations on each moimtain range form genetically discrete clades, molecular clock-based estimates may be made of times since population isolation. In this paper I use mitochondrial DNA sequence data from populations on thirteen southeast Arizona moimtain ranges and three localities in the Sierra Madre Occidental of Mexico to construct a genealogy of H. pugillis. Geographic information is then used, along with the sequence data, to determine if migration or incomplete lineage sorting best explains the data. The genealogy is then used to make inferences about the order and timing of fragmentation of H. pugillis populations on the sky islands of southeast Arizona.

Material and Methods

Individuals of H. pugillis were collected from 13 moimtain ranges in southeast Arizona (see Fig. 1), and three locations of continuous oak habitat in the Sierra Madre of Sonora, Mexico, for a total of 86 H. pugillis individuals. Two other species of Habronattus were used to provide polarity in the genealogy - four individuals of H. oregonensis and one individual of H. geronimoi, both of which also inhabit the sky islands. A list of individuals with their collecting localities are given in the Appendix. Total genomic DNA was isolated from 90 individuals. Mitochondrial DNA was isolated from eggs of one individual H. pugillis from the Sierrita Mountains (HbpglSI). The region of the mitochondrion spanning the 5' end 89 of NADH dehydrogenase subunit 1 (NDl) through the 3' half of the large subunit ribosomal RNA gene (16S) was PCR-amplified and sequenced using primers designed specifically for Habronattus. Protocols for DNA isolation, PCR amplification, and sequencing have been described previously (Appendix C). Sequences were aligned manually in SeqApp 1.9 (Gilbert 1994); identical sequence haplotypes were merged, using the search and merge function of MacClade 3.07 (Maddison and Maddison 1997). The average nucleotide diversity values (ic) within a moimtain range were calculated with the program DnaSP (Rozas and Rozas 1997). Phylogenetic analyses were conducted using PAUP* 4.0d64 (Swofford 1998), using several approaches. A heuristic parsimony analysis was performed with indels coded as missing data, and using all default options (ACCTRAN, TBR, simple addition sequence). Ehie to the large number of most parsimonious trees, the search was limited to saving 5,000 trees. Because a single heuristic search does not guarantee that the shortest tree will be foimd, multiple searches were conducted to search for islands of shorter trees as suggested by Maddison (1991). One hundred replicate searches were conducted, limiting each search to saving only 50 most parsimonious trees. Neighbor-joining analyses (Saitou and Nei 1987) were also performed with the HKY85 (Hasegawa, Kishino, and Yano 1985) model of evolution. One hundred bootstrap replicates (Felsenstein 1985) were conducted to assess confidence in the trees, using both parsimony and neighbor-joining criteria. Recent work has shown that H. geronimoi lies outside the clade that contains H. pugillis and H. 90

oregonensis (Hedin and Maddison, pers. comm.); therefore it was used to root all trees. Because H. pugillis is closely associated with Madrean oak woodlands, the geographic area that these woodlands occupy on each moxmtain range may reflect the potential population sizes of these spiders on each range. The current distribution of Madrean Evergreen vegetation was measured from a scanned image of Brown and Lowe's (1980) vegetation map, and the area occupied by this habitat was calculated with NIH Image 1.61. Distances between ranges were measured with a ruler to determine closest distances between oak woodlands on different ranges. To examine possible historical woodland cormections between ranges, I drew maps depicting the hypothetical distribution of woodlands, at elevations 500,1000, and 1500 feet below their ciurent 5000-foot elevation, using U.S. Geological Survey topographic maps. To estimate the dates that populations on each range last shared a conunon ancestor, divergences within ranges were dated using a molecular clock rate of 2.3% per million years, as calibrated for mitochondria of arthropods (Brower 1994). Pairwise divergences were calculated by PAUP*, with the HKY85 model of evolution. Sites with missing data were excluded from the comparisons. To estimate the greatest amount of divergence, the two most divergent individuals from a population were compared. The time since individuals last shared a common ancestor was estimated as divergence divided by 2.3% per million years. 91

Results

A total of about 815 bp of ND1-16S sequence was obtained for 87 individuals. For another four individuals, 440 bp of NDl was sequenced (HbpglE, 4E, IT, and 2T). Sequences were manually aligned, and contained a single 4-8 bp indel, located between the NDl and 16S genes (Appendix C). Eighty-one unique haplotypes of H. pugillis were found among the 86 individuals sequenced, with a total of 160 variable sites. Three haplot3q)es were found in more than one individual. Two of these haplotypes were in individuals from the same mountain range, whereas the third was found in individuals from neighboring moimtain remges (the Santa Rita and Atascosa Mountains). The exact niunber of most parsimonious trees could not be determined, due to limitations of computer time and memory. The single heuristic search that was limited to saving 5,000 most parsimonious trees yielded a strict corisensus tree of length 368 (see Fig, 2). Results from the 100 replicate island searches that were performed yielded only trees of length 368. The strict consertsus tree constructed from these replicate searches supports the same clades as the single heuristic search consensus tree. Only three individuals differ in their placement on the two consensus trees, but their placement is not well supported on either tree, with bootstrap values less than 50%; therefore the tree constructed from the replicate searches is not shown. Neighbor-joining analysis with the HKY85 model of evolution yielded a tree (fig. 3) that is generally corisistent with the parsimony consensus trees. However, certain relationships of terminal clades differ. 92 depending on the search method. In particular, the individuals from Mexico (SMX) cluster with the Baboquivari Moimtain (BQ) individuals in the neighbor-joining tree, but not in the parsimony trees. The phylogenetic trees show five basal clades with greater than 50% bootstrap support that consist only of individuals from southeast Arizona. These clades are distinguished with crossbars in Figures 2 and 3. The remaining basal clades all contain individuals from Mexico. The map in Figure 4 is coded so that the five basal clades formed by individuals from southeast Arizona that were foimd in the parsimony and neighbor-joining analyses share the same shading on the map. The mitochondrial sequences from the Sierrita individuals form a clade, as do those from two Galiuro individuals. All other basal clades consist of individuals from two or more mountain ranges. Only a few of the more terminal clades consist entirely of individuals from a single mountain range. The Cerro Colorado, Black, and Mule Mountain sequences each form clades containing only individuals from those ranges, although not all of these clades have strong bootstrap support. The sequences from the Empires form a clade in the neighbor-joining tree, although there is less than 50% bootstrap support for it. In contrast, all other clades contain individuals from at least two ranges, or do not include all individuals from a range. 93

Discussion

Geographic structure

The phylogenetic trees and the geographic collecting locales indicate that there are dades assodated with the northern, southern and western mountain ranges (Fig. 2-4). This geographic structuring is most clear for the southern ranges, which form a large dade consisting of individuals from the Huachuca, Santa Rita, Empire, Mule, and Atascosa Moimtains. Many of the individuals collected from the northern ranges (the Santa Catalina, Galiuro, cind Black Mountains) form a separate dade, with weak support (54%) for a sister group relationship with the Cerro Colorado individuals to the southwest. Another basal dade is composed of individuals from the Tumacacori and Baboquivari Mountains in the southwest region. This geographic concordance in the gene tree may be indicative of historical or on­ going gene flow among neighboring ranges. Little is known about the biology of H. pugillis to allow a priori predictions about patterns of gene flow. Some spiders disperse aerially on silken threads, in a process known as ballooning (Bristowe 1939). While this process has been well documented for other groups of spiders (e.g. Freeman 1946; Greenstone et al. 1987), little is known about whether saltidds in general, and H. pugillis specifically, can accomplish this feat. If H. pugillis can and does balloon between moimtain ranges, then there should either be complete panmixia of populations, which there is not, or an isolation by distance effect. Despite the geographic struct\ire noted above, there does not appear to be an isolation by distance effect. For ir\stance, individuals collected from the 94

two most distant localities (480 km. apart), the Galiuro Mountains and the Sierra Madre, are most closely related, yet distantly related to individuals from moimtain ranges between those two sites. This lack of complete geographical concordance may be due historical associations; before the woodland habitat became fragmented, spiders from southeast Arizona to the Sierra Madre may have belonged to one large panmictic population. The close genetic associations of the Galiuro individuals with the Sierra Madre individuals may reflect the previous geographic and genetic connections during a cooler and wetter geological period.

Migration versus Incomplete Lineage Sorting

Because migration and lineage sorting leave similar genetic pattems, utilizing information from both current geographic distribution and phylogenetic relatedness can be useful for separating lineage sorting from migration. Methods employing phylogenetic and geographic distance information to distinguish between lineage sorting and isolation by distance migration have been developed by Slatkin and Maddison (1990), Slatkin (1993), and Templeton, Routman and Phillips (1995). These methods require relatively large sample sizes from each collecting locality. Given the realities of collecting small, cryptic, seasonally active spiders from multiple ranges of very disparate sizes, I was unable to collect enough individuals in all ranges to justify using these methods. However, if migration of H. pugillis is limited or absent, there should be no pattern of isolation by distance, and moimtain ranges that are geographically isolated by great distances should be no more or less likely to harbor populations of spiders whose DNA sequences form a 95 single clade. Table 1 gives the aerial distances between each moimtain range and its nearest neighbors, and whether or not individuals from each range form a monophyletic clade in both the parsimony and neighbor-joining trees. I then asked whether ranges with spider sequences that formed monophyletic groups are more distant from other ranges. Average distance to the closest range for the monophyletic ranges was 19.5 km, and average distance to the closest ranges for the non-monophyletic ranges was 19.1 km (excluding the Tumacacori and Atascosa ranges, see footnote in table). Therefore, the average distances of monophyletic and non-monophyletic ranges are not significantly different (p>0.5) by the Wilcoxon-Maim-Whitney test Siegel and Castellan 1988, p. 128). Thus, ranges with sequences forming monophyletic clades are in as close a geographic proximity to neighboring ranges as those with individuals that do not form a clade. Therefore, there does not appear to be isolation by distance gene flow via aerial ballooning in these populations of H. pugillis. When two or more populations become separated and gene flow ceases, there will be stochastic extinction of neutral parental haplotypes in the daughter populations. The time necessary for all but one parental type to go extinct - that is, for lineage sorting to go to completion - is predicted to be 4N generations, where N is the effective population size (Neigel and Avise 1986; Tajima 1983). After that time, the daughter populations are expected to achieve reciprocal monophyly. Since incomplete lineage sorting may result from large ancestral population size, examination of population sizes of H. pugillis on each range may help determine if incomplete lineage sorting explains non-monophyly of moimtain populations. Since no censuses have 96 been taken of these small, cryptic spiders, I used the size of their potential habitat (oak woodlands) as a rough indicator of their population size on each mountain range. The area of oak habitat on each mountain range is given in Table 2, along with whether each range has populations that form geographically monophyletic clades. To test whether the size of the oak habitat is correlated with monophyly of the population, I performed a Wilcoxon-Maim-Whitney test. The test shows that ranges with monophyletic populations have significantly smaller areas of oak habitat (p = 0.0098). Because ranges have unequal sample sizes of spiders collected, I again performed a Wilcoxon-Marm-Whitney test to determine whether sample size was correlated with monophyly of sequences from a range. The relationship was not significant (p = 0.1301). In conclusion, it appears that geographic area of oak habitat is a fairly good predictor of whether or not populations from a range form a clade. Mountain ranges with oak habitats of less than 20 square miles generally contain populations that form monophyletic groups, whereas none of the individuals from larger ranges form exclusive clades. This provides evidence that incomplete lineage sorting may explain non-monophyly of populations from large moimtain ranges. We are now left with geographic evidence that carmot rule out migration between some neighboring ranges in the south, but makes aerial ballooning between small ranges appear unlikely. If genetic cormections between individuals from large ranges was due to aerial ballooning, I would have to postulate different migratory behaviors of spiders on different moimtain ranges. Instead, incomplete lineage sorting is likely 97 the reason genetic connections are shared among individuals from distant ranges. A closer examination of the geography and topography of the mountain ranges may further elucidate whether historically recent migration via continuous habitat between ranges is likely. The elevation of land lying between sky islands varies, with some ranges separated by very low desert and others by higher grassland. For this reason, it seems likely that some ranges were connected by corridors of continuous woodland more recently than others. Figure 5 illustrates that some mountain ranges would have been coimected by habitat for H. pugillis if the oak zone was only 500 feet lower, and that most ranges would have shared potential woodland cormections if oaks were previously present at elevations of 3500 feet. There are four north- to-south 3500 foot elevation corridors through southeast Arizona: a short corridor extending through the Baboquivari Moimtains furthest west, one through the Atascosa and Sierrita Mountains, a long corridor through the Santa Rita and Santa Catalina Moimtains in the center, and one through the Mule and Galiuro Moimtairis in the east. All of these corridors are coimected in the south by elevations of at least 3500 feet in southern Arizona and northern Mexico (data not shown). From this historical perspective, we may predict that four clades would be found in the genealogy, corresponding to these four corridors. Indeed, there is evidence for genetic cormection between individuals in the central corridor encompassing the Huachuca, Santa Rita, and Santa Catalina Mts. If the oak woodland was only 500 feet lower than it is at present, the Huachuca and Santa Rita ranges would be completely connected by continuous habitat for H. pugillis. Presently, individuals from 98 these two ranges are genetically indistingiiishable by their mitochondrial sequences, indicating that a woodland corridor probably existed relatively recently. However, the presumed historic connections carmot explain many of the other relationships in the genealogy. For example, there is a low elevation (below 3500 feet) riparian corridor that dissects the Galiuro and Santa Catalina Moimtains, yet these two ranges share similar haplotypes. Because the data in Tables 1 and 2 indicate that aerial migration is not likely between ranges, the genetic relatedness of individuals from these two ranges is probably due to old associations and not recent migration.

Order of Fragmentation

If there were a sequence of vicariance events that isolated the populations of H. pugillis on the Arizona moimtain ranges, we may expect to see this sequence reflected in the genealogy, with populations isolated longest ago basal to more recently isolated populations. Such a pattern has been foimd in genealogies of multiple organisms from the Hawaiian Islands, with a stepwise progression from basal to terminal taxa reflecting the stepwise colonization history of the islands (e.g. Fleischer, Mcintosh and Tarr 1998). The parsimony analysis of H. pugillis sequences yields a consensus tree that does not resolve the relationships of the major moimtain clades, but instead forms a basal polytomy. The neighbor-joining free contains short branches leading to the Seime clades that peirsimony was tmable to resolve, indicating a paucity of informative molecular characters (there are 160 variable sites among the 81 H. pugillis haplotypes). This indicates that these mitochondrial sequences are unable to resolve completely the order in which the moimtain 99

populations were fragmented. Alternatively, the basal polytomy may reflect a rapid divergence. A hard polytomy, or star phylogeny, may be expected if the woodlands fragmented into their current arrangement within a very short period of time. Simultaneous fragmentation of some of the ranges, in particular the Huachuca, Santa Rita, Empire and Whetstones, may have occurred. However, the paucity of molecular characters in this data set does not allow discrimination between a nearly simultaneous fragmentation, or a slower, ordered series of vicariance events. Some relationships among ranges can be resolved- All individuals from the Black Moimtains are nested within the clade consisting of Santa Catalina and Galiuro individuals. Likewise, all individuals from the Mule and Empire Mountains are nested in the larger clade composed mostly of Huachuca and Santa Rita individuals. This pattern of a small, peripheral population derived from a larger population is consistent with a peripheral isolates mode of differentiation (Mayr 1954). In contrast, populations sampled from the Santa Catalina and Galiuro Mountains show deep divergences among individuals, incompatible with a population bottleneck expected under a peripheral isolates mode of differentiation. Individuals from both these ranges each have three different ancestral haplotypes, located in different places on the gene tree, as well as a high nucleotide diversity (Table 2). These moimtain ranges likely harbored a large amoimt of old genetic diversity that existed before the woodlands contracted, because of their large size. Populations of H. pugillis that currently reside in the continuous oak habitat of the Sierra Madre Occidental may reflect what the population 100 structure of H. pugillis was like in southeast Arizona before climate warming. Individuals collected from three locations in this range have an immense genetic diversity (Table 2). Two individuals collected less than 10 meters apart (8SMX and lOSMX) show over 5% mitochondrial sequence divergence. This much genetic diversity over such a small spatial scale is consistent with a very large effective population size. The large genetic divergences seen among individuals in the Galiuro and Santa Catalina Moimtains may reflect a previously very large effective population size of H. pugillis in Arizona. Colonization from south to north has been inferred for many organisms based upon Pleistocene gladation patterns that suggest northern areas were iminhabitable during glacial maxima (reviewed in Rand 1948; Hewitt 1996). If the Madrean woodlands recently invaded southeast Arizona from Mexico (e.g. within the past 10,000 years), we may expect to see a similar pattern of coloruzation for H. pugillis. If this is the case, then individuals collected from continuous habitat from the Sierra Madre of Mexico may be expected to have a basal position in the genealogy. This is not the case; their position is unresolved in the parsimony consensus tree, and nested within Arizona clades in the neighbor-joining tree. Additionally, if the southeast Arizona populations are the result of a rapid northward range expansion, we may expect the leading edge of colonizers to be somewhat genetically depauperate, compared to individuals whose population has maintained a large effective population size (Hewitt 1996). However, the large effective population size and lack of a derived phylogenetic relationship in the northenunost Santa Catalina and Galiuro populations argues agair\st the 101 hypothesis that the northern mountain ranges were most recently colonized by spider popidations from the south. Lack of evidence for recent H. pugillis expansion into southeast Arizona does not rule out that the Madrean oak woodlands expanded here within the past 10,000 years. An alternative explanation may be that the current affinity of H. pugillis for Madrean oak habitat is relatively recent in evolutionary history, and that the species formerly inhabited other woodland habitats in southern Arizona.

Phylogeography

While the gene trees do not resolve the relationships of all individuals, a number of conclusions can be drawn about the phylogeography of H. pugillis in southeast Arizona. The two large mountain ranges near the northern end of H. pugillis' range are phylogenetically closely related, and support large populations of these spiders. The gene tree suggests an unexpected phylogeographic conclusion: perhaps an early climate-induced range fragmentation separated woodland ranges primarily along a latitude of 32 degrees, running between the Whetstone and Santa Catalina Mountains and the Mule and Galiuro Mounteiins. A large population of H. pugillis may have existed in the northern woodlands for an extended period during glacial cycles, with possible secondary contact with more southerly H. pugillis populations when the climate became suitable. This phylogeographic break is not mirrored by the present day distribution of Madrean woodlands (Brown and Lowe 1980). If H. pugillis shows such a pattern, we may predict that other woodland-dependent organisms will show this same phylogeographic pattern 102 of northern populations genetically distinct from southern populations. There is some evidence of such a divide from the closely related H. oregonensis. Sequences from individuals from ranges in the north (Galiuro and Santa Catalina Moimtains) cluster together, while those from the ranges in the south (Santa Rita and Huachuca) are most alike (Appendix A). However, as in the H. pugillis phylogeny, there is not strong bootstrap support for these nodes. It remains to be seen if this phylogeographic divide is present in other organisms.

Age of clades

Even though we may not be able to determine the precise order of fragmentation of these populations from the gene free, we can still make inferences about how long ago some of the sky island populations stopped sharing mitochondrial genes with other populations. Both the Sierrita and Black Moimtains have populations that form well supported clades. By determining the maximum amount of divergence within these populations, we can estimate the minimum amoimt of time since individuals within each range last shared a common ancestor. This will put a lower limit on the amount of time individuals within a range have been genetically isolated from other motmtain ranges. The greatest pairwise divergence of mitochondrial sequences from the Sierrita Moimtains is 0.67%. If there is 2.3% divergence/million years (Brower 1994), then the mitochondria last shared a common ancestor about 291,000 years ago. The same calculation for the Black Mountains yields 171,000 years, and for the Mule Mountains, 116,000 years. The most recent common ancestor between the Sierra Madre 103

and southern Arizona individuals is 440,000 years, from comparison of the divergence between Hbpg29GL and HbpglOSMX. Estimating the age of fragmentation of the montane woodlands based upon these mitochondrial sequences and a molecular clock calibrated for other arthropods is problematic for several reasons. First, the molecular clock was calibrated only for insects and a decapod. There is no clock based on arachnid divergences, and it is possible that their mitochondrial DNA evolves at a very different rate than insect DNA. Additionally, all the arthropod divergence dates used to calibrate the molecular clock were obtained from vicariance events, some of which may not be any more reliable than the dating of woodland fragmentation in the southwest deserts. Furthermore, there is much rate variation across mitochondrial genes in general (reviewed in Avise 1994), and the mitochondrial genes used in this study in particular (data not shown). Since these mitochondrial regions were specifically selected because they have a high substitution rate, and therefore coiild be used to distinguish closely related individuals, the calculated rates of divergence for these spiders may be higher than rates calculated from genes used for interspecific studies. An example that illustrates the problems with intraspecific vicariance dating using regions of the 16S gene is given in Fleischer, Mcintosh and Tarr (1998). Using dates established for the well- studied Hawaiian Islands, they determined that the sequence divergence in Laupala crickets (Shaw 1996) ranged from 2.4 to 10.2% per million years. If this higher divergence calibration is used for these spider sequences, it pushes the divergence times far forward, so that clades may be as yoimg as 26,000 years (Mule Mts.) to 66,000 years (Sierrita Mts.). Clearly, molecular clock 104 dating for these arachnids is not yet reliable enough to be used to accurately assess the age of fragmentation of the woodlands in southeast Arizona. However, it does seem clear that the divergences between populations of spiders on different mountain ranges must be older than 10,000 years. Perhaps the woodlands that may have connected some of the mountain ranges at that time did not provide suitable habitat for H. pugillis. This study has assessed the phylogeographic cormections of organisms inhabiting multiple sky islands in southeast Arizona. It depicts a complex history of fragmentation and divergence, which may not have been predicted merely by examining the geography and vegetation patterns of the area. It will be interesting to see if future phylogeographic work on other organisms in the same region uncovers similar patterns of geographic splits between north and south, with northern woodland populations maintaining historically large effective population sizes.

Literature Cited

Avise, J. C. 1994. Molecular markers, natural history and evolution. Chapmein & Hall, New York.

Avise, J. C., J. Arnold, R. M. Ball, E. Bermingham, T. Lamb, J. E. Neigel, C. A. Reeb, and N. C. Saunders. 1987. Intraspecific phylogeography: the mitochondrial DNA bridge between population genetics and systematics. Pages 489-522 in Armual Review of Ecology and Systematics eds.). Annual Reviews Inc., 105

Betancourt, J. L., T. R. V. Devender, and P. S. Martin. 1990. Packrat Middens: the last 40,000 years of biotic change. The University of Arizona Press, Tucson.

Bristowe, W. S. 1939. The Comity of Spiders. Ray Society, London.

Brower, A. V. Z. 1994. Rapid morphological radiation and convergence among races of the butterfly Heliconius erato inferred from patterns of mitochondrial DNA evolution. Proceedings of the National Academy of Science 91:6491-6495.

Brown, D. E. 1982. Desert Plants: Biotic communities of the American southwest- United States and Mexico. University of Arizona,

Brown, D. E., and C. H. Lowe. 1980. Biotic commimities of the southwest. Tempe, AZ., Forest Service, U.S. Department of Agriculture.

Felsenstein, J. 1985. Coixfidence limits on phylogenies: An approach using the bootstrap. Evolution 39:783-791.

Fleischer, R. C., C. E. Mcintosh, and C. L. Tarr. 1998. Evolution on a volcanic conveyor belt: using phylogeographic reconstructions and K-Ar-based ages of the Hawaiian Islands to estimate molecular evolutionary rate. Molecular Ecology 7:533-545. 106

Freeman, J. A. 1946. The distribution of spiders and mites up to 300 ft. in the air. Journal of Animal Ecology 15:69-74.

Gilbert, D. 1994. SeqApp.

Greenstone, M. H., C. E. Morgan, A.-L. Hultsch, R. A. Farrow, and J. E. Dowse. 1987. Ballooning spiders in Missouri, USA, and New South Wales, Australia: family and mass distributions. The Journal of Arachnology 15:163-170.

Griswold, C. E. 1987. A revision of the jumping spider genus Habronattus F.O P.-Cambridge (Araneae; Salticidae), with phenetic and cladistic analyses. University of California, Entomology 107:1-344.

Hasegawa, M., H. Kishino, and T. Yano. 1985. Dating of the human-ape splitting by a molecular clock of mitochondrial DNA. Journal of Molecular Evolution 21:160-174.

Hewitt, G. M. 1996. Some genetic consequences of ice ages, and their role in divergence and speciation. Biological Journal of the Linnean Society 58:247-276. 107

Linhart, Y. B., and A. C. Premoli. 1994. Genetic variation in central and disjunct populations of Lilium parryi. Canadian Journal of Botany 72:79-85.

LomoHno, M. V., J. H. Brown, and R. Davis. 1989. Island biogeography of montane forest mammals in the American southwest. Ecology 70:180- 194.

MacArthur, R. H., and E. O. Wilson. 1963. An equilibrium theory of insular zoogeography. Evolution 17:373-387.

MacArthur, R. H., and E. O. Wilson. 1967. The theory of island biogeography. Princeton University Press, Princeton, N.J.

Maddison, W. P., and D. R. Maddison. 1997. MacClade: Analysis of phylogeny and character evolution. Simderland, Massachusetts, Sinauer.

Maddison, D. R. 1991. The discovery and importance of multiple islands of most-parsimonious trees. Systematic Zoology 40:315-328.

Mayr, E. 1954. Change of genetic environment and evolution. Pages 157-180 in Evolution as a process (f. Huxley, A. C. Hardy and E. B. Ford, eds.). Macmillan, New York. 108

Neigel, J. E., and J. C. Avise. 1986. Phylogenetic relationships of mitochondrial DNA under various demographic models of spedation. Pages 515-534 in Evolutionary Processes and Theory eds.). Academic Press,

Rand, A. L. 1948. Gladation, an isolating factor in spedation. Evolution 2:314-321.

Rozas, J., and R. Rozas. 1997. DnaSP. Barcelona, Spain,

Saitou, N., and M. Nei. 1987. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecular Biology and Evolution 4:406-425.

Siegel, S., and J. Castellan N. John. 1988. Nonparametric statistics for the behavioral sdences. R. R. Dormelley & Sons Company, Mexico.

Shaw, K. L. 1996. Sequential radiations and patterns of spedation in the Hawaiian cricket genus Laupala inferred from DNA sequences. Evolution 50:237-255.

Slatkin, M. 1993. Isolation by distance in equilibrium and non-equilibrium populations. Evolution 47:264-279. 109

Slatkin, M., and W. P. Maddison. 1990. Detecting isolation by distance using phylogenies of genes. Genetics 126:249-260.

Swofford, D. L. 1998. PAUP*: Phylogenetic analysis using parsimony. Sunderland, Massachusetts, Sinauer Associates.

Tajima, F. 1983. Evolutionary relationship of DNA sequences in finite populations. Genetics 105:437-460.

Templeton, A. R., E. Routman, and C. A. Phillips. 1995. Separating population structure from population history: a cladistic analysis of the geographical distribution of mitochondrial DNA haplotj^es in the Tiger Salamander, Ambystoma tigrinum. Genetics 140:767-782.

Van Devender, T. R., and W. G. Spaulding. 1979. Development of vegetation and climate in the southwestern United States. Science 204:701-710. 110

Appendix CODE MOUISTTAIN RAIVK3E ! LOCATION

Hbp<]1B Black Mountains. AZ Jewell Well Hbpg2B Black Mountains. AZ Jewell Well Hbpa3B 1 Black Mountains, AZ Jewell Well HbpgISC Santa Catalina Mts. AZ Bear Cyn. Hbpg2SC Santa Catalina Mts, AZ Bear Cyn. HbpgSSC Santa Catalina Mts. AZ Bear Cyn, HbpglOSC Santa Cataiina Mts, AZ Rnger Rock Trail HbpqlSSC Santa Catalina Mts. AZ Rnqer Rock Trail Hbpg20SC Santa Catalina Mts. AZ Rnger Rock Trail Hbpa24SC Santa Catalina Mts. AZ Rnger Rock Trail i-ibpg25SC 'S£tnta Cataiina Mts, AZ Molino Basin. Sta Catalinas Hbpa26SC Santa Catalina Mts. AZ iMolino Basin. Sta Catailinas Hbpg28SC Santa Catalina Mts, AZ Molino Basin. Sta CatEilinas HbpqlQ BaixKiuivari Mts, AZ Elkhom ranch. East cental Hbpq2Q Baboquivari Mts. AZ Elkhom ranch. East cental H^gSQ i Baboquivari Mts, AZ Elkhom ranch. East cental Hbpg4Q i Baboquivari Mts. AZ Elkhom ranch. East cental HbpqSQ 1 Baboquivari Mts, AZ Elkhom ranch. East cental Hbpq7QK ' Baboquivari Mts, AZ Kitt Peak HbpqSQK | Baboquivari Mts, AZ Kitt Peak Hbpq9QK j Bsiboquivari Mts. AZ Kitt Peak Hbpq1SR 1 Santa Rita Mts, AZ Gardner Cyn HbpqSSR Santa Rita Mts. AZ Gardner Cyn Hbpq7SR Santa Rita Mts. AZ Madera Canyon HbpqSSR 1 Santa Rita Mts, AZ Madera Canyon Hbpq9SR | Santa Rita Mts, AZ Madera Canyon HbpqIISR Santa Rita Mts, AZ Rorida Cyn Hbpq25SR Santa Rita Mts. AZ Rorida Cyn Hbpq36SR Santa Rita Mts. AZ .5 mi abve Josephine Sddle Hbpq37SR Santa Rita Mts. AZ .5 mi abve Josephine Sddle Hbpq46SR Santa Rita Mts. AZ .5 mi below Josephine Peak HbpqIH Huachuca Mts. AZ Reef Mine Cmqrd Hbpq2H iHuachuca Mts, AZ Reef Mine Cnfiqrd HbpqSH 1 Huachuca Mts, AZ ; Reef Mine Cmqrd Hbpq7H Huachuca Mts, AZ Montezuma Cyn E Monmnt HbpqSH Huachuca Mts. AZ Garden Cyn Hbpq21H Huachuca Mts, AZ Garden Cyn Hbpq25H Huachuca Mts. AZ | Montezuma Rd near Copper ( Hbpq26H Huachuca Mts. AZ Sunnyside Cyn Hbpq27H Huachuca Mts. AZ ISunnvside Cvn Appendix Hbpg34H Huachuca Mts, AZ : 1 mi up Oversite Cyn HbpglE Empire Mts, AZ Michod's ranch Hbpg4E Empire Mts. AZ Michod's ranch Hbpg6E Empire Mts, AZ Michod's ranch HbpQ7E Empire Mts, AZ Michod's ranch HbpgiT Tumacocori Mts, AZ east central canyon HbpQ2T Tumacocori Mts. AZ east central canyon Hbpg3T Tumacocori Mts. AZ east central canyon Hbpg4T Tumacocori Mts. AZ east central canyon HbpglCC Cerro Colorado Mts, AZ! north slope Hbpg2CC Cerro Colorado Mts, AZ north slope HbpglW Whetstone Mts, AZ French Joe Cyn Hbpg2W Whetstone Mts, AZ French Joe Cyn HbpglM Mule Mts, AZ 1 mi. NW of Bisbee Hbpg7M Mule Mts. AZ 1 mi. NW of Bisbee Hbpg9M Mule Mts. AZ 1 mi. NW of Bisbee HbpglOM Mule Mts, AZ 1 mi. NW of Bisbee HbpgllM Mule Mts. AZ 1 mi. NW of Bisbee HbpgIA Atascosa Mts, AZ Sycamore Cyn Hbpg2A Atascosa Mts, AZ Sycamore Cyn HbpgSA Atascosa Mts, AZ Sycamore Cyn HbpglSMX Sierra Madres,MX 22.5mi W Yecora Hbpg2SMX Sierra Madres.MX 22.5mi W Yecora HbpgSSMX ' Sierra Madres.MX 22.5mi W Yecora HbpgSSMX Sierra Madres.MX 22.5mi W Yecora HbpgSSMX Sierra Madres.MX 19.3mi W Yecora HbpglOSMX! Sierra Madres.MX 19.Smi W Yecora HbpgllSMX Sierra Madres.MX 25.8mi E Maycoba Hbpg12SMX Sierra Madres.MX near Yecora HbpglSI Sierrita Mts, AZ east cyn, 1.5 mi up road HbpgSSI Sierrita Mts, AZ east cyn. 1.5 mi up road Hbpg14SI Sierrita Mts, AZ east cyn, 1.5 mi up road HbpgSOSI Sierrita Mts, AZ east cyn, 1.5 mi up road Hbpg34SI Sierrita Mts. AZ N of Placer Peak Hbpg46SI Sierrita Mts. AZ N of Placer Peak Hbpg49Sl Sierrita Mts. AZ N of Placer Peak Hbpgl 1GL i Galiuro Mts. AZ High Crk Cyn. (6660 ft) Hbpg29GL 1 Galiuro Mts. AZ High Crk Cyn, (6660 ft) HbpgSOGL Galiuro Mts. AZ High Cri< Cyn. (6660 ft) HbpgSIGL Galiuro Mts. AZ High Crk Cyn. (6660 ft) Hbpg43GL Galiuro Mts, AZ High Crk Cyn. (6480 ft) HbpgSOGL Galiuro Mts. AZ Crest Trail. (6860 ft) Appendix Hbpg52GL Galiuro Mts. AZ Crest Trail. (6860 m Hbpg55GL ! Gaiiuro Mts. AZ i Crest Trail, (6940 ft) Hbpg56GL GeUiuro Mts. AZ i Crest Trail, (6960 ft) Hbpg66GL Galiuro Mts. AZ Crest Trail. (7020 ft) Table 1

Distance to nearest Madrean woodlands

Mountain Range Mono­ Closest range/km 2^^ closest range/km 3"* closest range/km phyletic Cerro Colorado yes Sierrita/15 Atascosa/18 Tumacacori/20 Black yes Santa Catalina/23 Galiuro/48 Sierrita/100 Empire no Santa Rita/10 Whetstone/15 Huachuca/38 Sierrita yes Cerro Colorado/15 Baboquivari/28 Santa Rita/32 Whetstone no Empire/15 Santa Rita/20 Huachuca/20 Mule yes Huachuca/25 Whetstone/50 Santa Rita/70 Baboquivari no Cerro Colorado/25 Sierrita/28 Atascosa/28 Atascosa* no Tumacacori/0-10 Cerro Colorado/18 Santa Rita/25 Tumacacori* no Atascosa/0-10 Cerro Colorado/20 Santa Rita/22 Huachuca no Whetstone/20 Mule/25 Santa Rita/30 Santa Rita no Empire/10 Whetstone/20 Tumacacori/22 Santa Catalina no Black/23 Galiuro/31 Santa Rita/47 Galiuro no Santa Catalina/31 Black/48 Whetstone/62

* These mountain ranges are indicated as having continuous Madrean woodlands connecting them on the Brown and Lowe (1980) map. They do not, but exact measures of the distance between woodlands are not known. Because different male morphological forms inhabit the different ranges, there is probably a distinct boundary to their respective ranges. However, this area has not been sampled thoroughly enough to know their exact distribution. Therefore, these two ranges were excluded from the analysis of whether ranges with monophyletic groups of sequences are more distant from other ranges. Table 2

Geographic area covered by Madrean oak woodlands on each mountain range

Mountain Range Square miles of oak n n % Monophyletic woodlands Cerro Colorado 1" 2 1.015 yes Black 1.5" 3 0.261 yes Empire 2.0 4 0.126 no Sierrita 5.8 7 0.396 yes Whetstone 11.3 2 3.689 no Mule 18.8 5 0.106 yes Baboquivari 22.1 8 0.303 no Santa Catalina 26.7 10 1.997 no Santa Rita 31.4 10 0.231 no Atascosa 20.2*^ 3 3.076 no Tumacacori 20.2*^ 4 no Galiuro + Winchester 41./ 10 + 0 2.239 no Huachuca + Patagonia 93.46'' 10 + 0 0.59 no

" The amount of oak habitat on these ranges estimated by personal observation, since the presence of oak woodlands is not recorded on the Brown and Lowe (1980) map. '' These mountain ranges are indicated as having continuous oak woodlands connecting them on the Brown and Lowe map, so the entire woodland was measured. ' Different morphological types exist in the Atascosa and Tumacacori Mts. Where the range of one morphological type ends and the other begins is not known. Therefore the size of the range of each was estimated to be half the size of the entire oak woodlands. ^ These ranges are indicated as having continuous oak woodlands between them on the vegetation map. In fact, they do not, but the map was measured for consistency of method. 116

Figure Legends

1. The outline of the areas inhabited by Madrean oak vegetation (after Brown and Lowe 1980) on southeast Arizona mountains. Collecting locales of H. pugillis are depicted with drcles. The mountain range names are abbreviated as follows: BL=Black; SC=Santa Catalina; GL=Galiuro; BQ=Baboquivari; SI=Sierrita; CC=Cerro Colorado; T=Tvimacacori; A=Atascosa; SR=Santa Rita; E=Empire; WET=Whetstone; H=Huachuca; M=Mule Mountains

2. Parsimony strict consensus of 5000 trees. Bootstrap values greater than 50% are above the branches. The colored slashes are the clades referred to in the text.

3. Neighbor-joining tree with HKY model of evolution. Bootstrap values greater than 50% are above the branches. Branch lengths are relative to the number of changes.

4. Map of Madrean oak vegetation, overlaid with filled circles corresponding to the parsimony tree clades. A single circle on a moimtain range indicates that all individuals belong to a single clade. Multiple circles indicate individuals collected from that range belong to multiple clades. 117

5. Partial topographic map of southeast Arizona. Light grey areas are 3500-4000 feet; medixun grey areas are 4000-4500 feet; dark grey areas are 5000 feet and above. All unshaded areas are below 3500 feet elevation. North BL

10 miles

Tucson I if"'

SI

SR CC • 0

Sonora, Mexico Arizona 119

2BQ Figure 2 63 Parsimony Strict Consensus 7n of 5000 trees ^ 91 length = 368 CI = 0.6087 35. RI = 0.8519

-26H 88 m. •lA

jaz. •9M 33- :p

91 l+37ai+5A

3A. JLQ3. 63

inn -fia. rc(_ .31 54

_Z£.

B 96 :Z3. I-f46SI-f49SI {f^ .£2 p 100 [ Eugieioniinoi 120

Figures Neighbor joining tree HKY85 model of evolution

IWET 105MX 29GL ISMX

IISMX .8SMX

J.21HLleat-^STSR-I-SA ISR 3SR 2WEr IE 4E 7E 6E IISR 25SR L-46SR — 27H •tj 15SClOSC 20SC r26SC I I , I ilOM Ll9M - I-IIM IH 96I- TSR25H 8SR t• 7H >4,— ICC 34H .2CC IB 100 J—• 3B 1- 232r -2SC 24SC 25SC 28SC28S Irn^ 11 43GL 50GL 55GL 56GL 66GL b•52GL ISC 34SI-f46SI^^SIISI -SSI 14SI 30SI .3SC I jjrivOrlGL 1Q0 L— •OrlH •OrlMR • H.geronimoi North Figure 4 B

10 miles

Tucson J ij"" icf

M # 122 123

APPENDIX C

UNUSUAL EVOLUTION OF MITOCHONDRIAL RNA GENE SEQUENCES IN

SPIDERS (ARANEAE: SALTIODAE: HABRONATTUS) 124

Unusual evolution of mitochondrial RNA gene sequences in spiders (Araneae: Salticidae: Habronattus)

Susan Masta Department of Ecology & Evolutionary Biology University of Arizona Tucson, Arizona 85721 telephone (520) 621-1889 fax (520) 621-9190 [email protected]

Key Words: Habronattus, mitochondrial genes, large subunit rRNA structure, tRNA structure, gene arrangement, arthropod systematics

Running Head: Molecular evolution of arachnid mtDNA 125

Abstract

Analyses of mitochondrial DNA sequences from three species of Habronattus jumping spiders (Chelicerata: Araneae: Salticidae) reveal unusual tRNA secondary structures and gene arrangements, providing new information on both tRNA evolution and arthropod systematics. Sequences from the protein-coding genes NADH dehydrogenase subimit 1 (NDl), cytochrome oxidase subunit I (COI), and subimit n (COU) were obtained, along with tRNA^^', tRNA'-®"^^"^^ and large subunit ribosomal RNA (16S) sequences; these revealed several imusual features. First, inferred secondary structures of transfer and tRNA^®' lack the PFC arm and the variable arm, and therefore do not form a standard cloverleaf structure. In place of these arms is a 5-6 nucleotide TV-replacement loop, previously described from nematode mitochondrial tRNAs (Wolstenholme et al. 1987). Second, the proposed secondary structure of the 3' end of 16S is found to be similar to that reported for insects, but the 5' end is extremely divergent and truncated about 300 nucleotides with respect to Drosophila yakuba. Third, initiation codons consist of ATY (ATT and ATC) and TTG for NDl and COII, respectively. The use of TTG as a start codon in mitochondria is the first such report in arachnids. Finally, tRNA gene arrangement in Habronattus differs from that reported for the chelicerate Limulus polyphemus, thus contradicting previous claims that tRNA gene arrangements could be safely used to distinguish the major groups of arthropods. 126

Introduction

Knowledge of the patterns and processes of evolution of animal mitochondrial genomes has been imfolding at a rapid rate in the past two decades (reviewed in Moritz, Dowling and Brown 1987; Avise 1991; Wolstenholme 1992). In addition to information provided by complete sequences of genomes, much of our understanding of mitochondrial evolution has resulted from the use of mitochondrial sequences to reconstruct the phylogenetic relationships of groups of organisms. A large database of sequences now exists for many invertebrate and vertebrate species, with over 44,000 mitochondrial sequences published in GenBank. In spite of this large database, there are still substantial gaps in our knowledge of the composition, organization, and patterns of evolution of mitochondrial genes for many taxonomic groups. Among invertebrates, there has been an emphasis on reconstructing arthropod relationships with mitochondrial DNA sequences, particularly within the taxonomically diverse Insecta. Much less work has focused on the other groups within Arthropoda - chelicerates, crustaceans, and myriapods - and subsequently is less known about patterns of molecular evolution in their mitochondrial genomes. Animal mitochondrial genomes contain a number of unusual features that distinguish them from the nuclear genomes of eukaryotes, and from other related prokaryotes. Only in mitochondrial genomes have tRNAs been discovered that do not have the canonical cloverleaf secondary structure (reviewed in Dirheimer et al. 1995). The genetic codes of metazoan mitochondria are highly modified, and utilize unusual initiation codons 127

(reviewed in Wolstenholme 1992). While there are typically 22 tRNA, 13 protein-coding, and two ribosomal genes in metazoan mitochondria, the arrangement of these genes differs greatly between taxa (reviewed in Wolstenholme 1992). Because so few mitochondrial-based molecular studies have been reported for arachnids, it is probable that new evolutionary patterns may still be undiscovered. In this paper I report on the molecular evolutionary patterns observed in six mitochondrial genes of Habronattus jumping spiders (Chelicerata: Araneae): NADH dehydrogenase subimit 1 (NDl), cytochrome oxidase subxmit I (COI) and subunit H (COII), tRNA^^, tRNA'^"^^ , and the gene encoding the large subimit of ribosomal RNA (16S). The secondary structures of both tRNA genes and ribosomal 16S are then deduced from DNA sequence data. Lastly, I discuss potential tRNA gene rearrangements in the mitochondria of these arachnids relative to other arthropods.

Materials and Methods DNA extraction Nineteen individuals of Habronattus oregonensis were collected from five moimtain ranges in southeast Arizona, and one location in northern California; thirteen H. pugillis individuals were collected from seven ranges in southeast Arizona and one location in the Sierra Madre Occidental of Mexico. One individual of H. geronimoi was collected from southeast Arizona, for a total of 33 individuals. Total genomic DNA was isolated using a modified SDS extraction protocol. Four legs from each spider were placed in 200 pi lysis buffer (100 128 mM Tris pH 9.1, 50 mM EDTA, 0.5% SDS, 50 mM NaCl, 200 jig proteinase K/ml) in a microfuge tube and ground using a disposable plastic pestle. The DNA was incubated at 45°C for three to five hours, extracted with phenol, phenol-chloroform and chloroform, ethanol precipitated, resuspended in water, and stored at -20°C. Mitochondria enriched DNA was isolated from one H. pugillis and two H. oregonensis individuals, using a modified mitochondrial DNA preparation. Either ten eggs or the abdomen from a single spider was ground in cold STE (250 mM sucrose; 30 mM Tris pH 8.0; 50 mM EDTA pH 8.0) in 100 (j1 glass homogenizers. Cell debris and nuclei were removed by pelleting at 3500 rpm, 4°C for 8 minutes. The supernatant was then transferred to a clean microfuge tube, and spun at 12,000 rpm, 4°C, for 20 minutes to pellet the mitochondria and remaining nuclei. The pellet was resuspended in TE, pH 8.0, and membranes were lysed by addition of 1/lOth of a volimie of 10% SDS, followed by the addition of 1/lOth of a volimie of 5M potassium acetate. The mixture was allowed to sit overnight at O'C to precipitate the nuclear DNA and then centrifuged at 12,000 rpm, 4°C, for 30 minutes. The supernatant containing mitochondrial DNA was then treated with proteinase K, phenol- chloroform extracted, and ethanol precipitated. The resulting mitochondrial DNA pellet was then resuspended in water, and stored at -20°C.

PCR and Sequencing The mitochondrial genes encoding NDl and 16S ribosomal RNA, were amplified by the polymerase chain reaction (PCR), using primers designed specifically to amplify H. oregonensis and H. pugillis. The primer HbNDl (5- 129

TGAGCTACTCTTCGAATAGC-3') corresponds to positions 12,257 to 12,276 on the J strand in Drosophila yakuba mitochondrial NDl (Clary and Wolstenholme 1985). The primer Hbl6S (5'-TTACGGAAGTGCACATATCG- 3") corresponds to positions 14,226 to 14,245 on the N strand of D. yakuba, located in the small subunit ribosomal RNA (12S) coding region. Amplification by PGR was performed in 50 (il reaction volumes, using final concentrations of 0.52 [iM of each primer, 3 mM MgClz, 0.2mM of each dNTP, IX PGR buffer, 1 imit of Taq polymerase and l|jJ DNA. The amplification conditions were: denaturation at 95 °C for 30 seconds, annealing at 58 °C for 1 minute, with a 1 minute extension at 72°C for 35 cycles. The 3' end of GOI and the 5' end of COn were amplified using the primers Gl-J-2309 (5'-nTATGCTATAGTTGGAATTGG-3') designed by M. Hedin (pers. comm.) based upon sequences from multiple species of Habronattus, and the universal primer C2-N-3389 (Simon et al. 1994). The names of the primers correspond to their positions in the mitochondrial sequence of D. yakuba (Clary and Wolstenholme 1985). PCR solution conditions identical to those for ND1-16S amplification were used, except that the primers were at a final concentration of 0.4|iM. The cycling conditions for these reactior\s were: 95°C for 30 seconds, 50°C for 1 minute, and 72°C extension for 1 minute, for 35 cycles. The PCR products were purified by either excising the PCR band from a 1% agarose gel, followed by GeneClean purification, or by Qiagen^" QIAquick PCR spin-colimm purification to remove primers and tmincorporated dNTPs. 130

The PCR products were sequenced manually with the GibcoBRL dsDNA Cycle Sequencing system end-labeling kit, or on an ABI automated sequencer at the University of Arizona sequencing facility. Both of the PCR amplification primers were used to sequence NDl through 16S. The 5' end of con was sequenced using the PCR primer C2-N-3389, but the 3' end of the adjacent COI gene was sequenced using the internal primer Cl-N-2971 (5'- ATTGAGTAAAAGTATGATC-3')- With the exception of primer C2-N-3389, the universal arthropod primers as described in Simon et al. (1994) do not work well or at all in these spiders, so it was necessary to design all primers specifically for Habronattus. Only one strand of DNA was sequenced, therefore two measures were taken to assess the reliability of the sequence data. All sequences were manually aligned in SeqApp 1.9 (Gilbert 1994), and all sites differing from the other sequences were reexamined to see if the base had been correctly scored. Because the individuals are closely related (many intra-population samples), this type of polymorphism data allows assessment of sequencing accuracy. Secondly, although most sequences were obtained with an ABI automated sequencer, 18 of the sequences were obtained both manually and with the automated sequencer for the 5' end of NDl and the 3' end of 16S. There was complete agreement among all sequences. Sequences were aligned meinually for ciU nucleotide sequences from Habronattus and all protein sequences from other arthropods. Habronattus protein coding sequences were translated with MacClade 3.1 (Maddison and Maddison 1997) using the Drosaphila mitochondrial genetic code, and aligned with the protein sequences of NDl, COI and COn for three other arthropods. 131

Drosophila yakuba, Limulus polyphemus, and Ixodes hexagonus, whose sequences were obtained from GenBank. The nucleotide sequence of D. yakuba large subimit rRNA was initially aligned with Habronattus using the programs Blast 2.0 (Altschul et al. 1997) in NCBI and PileUp in GCG, respectively. These alignments failed to align sequence motifs that were predicted to be conserved between the two species, therefore it was necessary to fold the ribosomal sequences into their presumed secondary structures to look for conserved motifs. All folding of the DNA sequences into their secondary structures was done manually, using the proposed 16S rRNA structures for D. yakuba (Gutell and Fox 1988), and the chelicerates Haemaphylis cretica and Cupiennius salei (Black and Piesman 1994; Huber et al. 1993) for comparisons. Final alignment of Habronattus and D. yakuba 16S sequences was done manually, using a beta test version of MacClade 3.5 (provided by D. Maddison), based on conserved stem and loop regions present in the folded secondary structure. Folding of DNA sequences coding for tRNAs was performed manually, using the tRNA structures for L. polyphemus (Staton, Daehler and Brown 1997) for comparisons. Codon usage for the three protein coding genes was calculated using the program MEGA (Kumar, Tamura and Nei 1993). Calculation of A+T content and sequence divergence was performed using the beta version of PAUP* 4.0d64 (Swofford 1998). 132

Results

Sequence alignment Sequences ranging from 800 to 1600 nt (entire length) were obtained for the PCR fragment spanning the region of the mitochondria containing the 5' end of NDl, 16S, tRNA^®' and part of 12S, from a total of 33 individuals. There was 40 nt of sequence overlap for eight H. oregonensis individuals when sequences from opposite strands were aligned, yielding eight full-length sequences of 16S. Approximately 700 nt of sequence was obtained for three H. oregonensis individuals for the 3' end of COI. Three himdred nt of the 5' end of COn were sequenced for two individuals of H. oregonensis. The Appendix presents a stunmary of the 60 sequences obtained. There are several reasons to be confident that these sequences are mitochondrial, rather than mitochondrial-like sequences that have been transposed to the nucleus and have become pseudogenes. First, PCR amplifications yielded a single band. Second, the sequences did not have many ambiguous bases, as would be expected if the primers amplified multiple different copies from the two genomes. Third, the sequences appeared to encode functional products: the protein-coding sequences all translated, without intervening stop codons, and the RNA-coding sequences folded into plausible secondary structures. There was no sequence variation in either of the tRNA regions corresponding to the anticodon arms, which may be expected to be highly conserved, among the 33 individuals that were sequenced. If the sequences were pseudogenes, it is not expected that these features would be conserved. Lastly, the sequences obtained from the 133

mitochondrial DNA extractions were identical to the sequences from the total genomic DNA extractions. Alignment of the ND1-16S gene sequences in Habronattus resulted in a single region containing an indel of either four or eight nucleotides, a simple ATTA repeat. The presence of this four-nucleotide repeat is foimd only in individuals of H. pugillis collected from two geographic locales. It appears, for reasons discussed below, that this is an insertion located between coding regions. Other single- or few-nucleotide indels are present in 16S, but not in NDl.

There were no indels present within Habronattus in the aligned COI and con sequences. When the amino add sequences of the 5' end of COn from D. yakuba are aligned with Habronattus, the corresponding first 97 codons have a amino acid sequence similarity of 33J°/o. Amino add sequences of the 3' end of COI, when aligned with D. yakuba, show 35% sequence similarity.

Secondary structure The inferred secondary structure of the 3' half of the 16S molecule has many structural features in common with D. yakuba mitochondrial large subimit rRNA (Gutell and Fox 1988). These two taxa share highly conserved stem and loop region sequence motifs (fig. 1), and show 70% sequence similarity for the first 675 nt of the 3' end of the gene. These same regions are conserved in other chelicerates for which the 3' end of 16S has been sequenced (Black and Piesman 1994, Huber et al. 1993). Compensatory mutations are present in stem regions of Habronattus where the sequence differs from D. yakuba; this provides evidence that these sequences encode functional ribosomal RNA. The putative rBlNA transcription termination 134

sequence was identified in the tRNA adjacent to the 3" end of 16S. The stop signal for mitochondrial 16S transcription is believed to be the conserved heptamer TGGCAGA (Valverde, Marco, and Garesse 1994). This sequence is located at positions 8 to 14 of the adjacent gene, where it is also located in insects (Valverde, Marco, and Garesse 1994). While the 3" end of 16S shares similarities with D. yakuba, the first 675 nt of the 5' half of the 16S region shares only about 40% sequence similarity. The D. yakuba 16S sequence is substantially longer, by over 300 nucleotides, than the aligned Habronattus sequence. Additionally, there do not appear to be conserved sequence motifs with D. yakuba, and I was imable to fold the 5' sequence into a secondary structure resembling D. yakuba. It is thus unclear if the 5' half of Habronattus 16S rRNA has a secondary structure that is very similar to D. yakuba. Two transfer RNA sequences were identified and folded into their putative secondary structures. The structures of and tRNA^^' are depicted in Figure 2. Both of these tRNAs appear to lack the variable arm and the TM'C arm found in the canonical tRNA. Instead, these arms are replaced by a TV-replacement loop, as previously described in nematodes (Wolstenholme at al. 1987; Okimoto and Wolstenholme 1990). However, there is almost complete conservation of the anticodon arm and of the DHU arm in between Habronattus and L. polyphemus; these differ by only four nucleotides. The secondary structure of tRNA^®' also shows conservation of the anticodon stem and loop and DHU stem with L. polyphemus, but to a lesser degree, differing at ten nucleotide positions, but with complementary base changes in the anticodon stem. Like tRNA^®"^^^, 135 tRNA^®' has much less sequence conservation witii L. polyphemus in the regions corresponding to the "PPC, variable and amino add acceptor arms. The amino acid acceptor arm that is illustrated for overlaps with the first four nucleotides of the NDl gene. While it may be possible to fold the DNA sequence into a structure resembling the classical tRNA cloverleaf, it would require that there be substantially more overlap with the NDl gene (fig. 3). Immediately downstream of the sequence encoding the amino add acceptor arm in tRNA^'' is a region with high intra- and interspecific sequence variation (fig. 4). To fold this sequence into a structure resembling a doverleaf would require that the amino add acceptor arm be highly divergent both within populations and among Habronattus spedes. It would be surprising to find this much variation within a gene that is required for normal translation of all mitochondrial genes. Furthermore, to maintain a cloverleaf structure would require either a large number of unusual base pairings, or substantial editing of the transcripts.

Genetic code When NDl in Habronattus is translated using the Drosophila mitochondrial genetic code (de Bruijn 1983), it appears that the protein is tnmcated by three amino adds with respect to D. yakuba. My aligrmient of the chelicerate L. polyphemus NDl (Boore et al. 1995) indicates the same tnmcation relative to D. yakuba (fig. 5). In Habronattus, three lines of evidence indicate that the start position for NDl corresponds to ATT or ATC, and not the more typical ATG or ATA codons. First, there is no codon for ATG or ATA present in any of the Habronattus individuals until 44 codons 136

downstream from the ATT or ATC codons, making it extremely unlikely that those codons are used as initiation codons. Secondly, if the NDl protein in H. oregonensis is the same length as the D. yakuba protein, the first codon is a stop codon (TAA). Lastly, codons for identical amino adds in L. polyphemus begin two codons past the proposed start codon of Habronattus. ATG or ATA do not appear to be initiation codons in the COn gene either. The Habronattus codon in the position corresponding to the first codon of D. yakuba, L. polyphemus, and L hexagonus, is TTG (fig. 5), which normally codes for leucine in other arthropods. A start codon, or an ATN codon, is located 16 codons away. If this ATN codon was the actual standard start codon, the Habronattus COn protein would be considerably tnmcated with respect to these other arthropods. Additionally, an amino acid motif that is conserved with L polyphemus (FQD), begins eight codons before the first ATN codon. The adjacent COI gene terminates with TTA, one of the two stop codons that have been recognized in invertebrate mitochondria. When the Drosophila mitochondrial genetic code is used to translate the three protein coding gene sequences, all possible codons are present (data not shown). As is commonly found in the mitochondria of other invertebrates, codons ending in guanine are infrequently used in these arachnids, and there is variation in codon use among genes.

A+T bias There is a high A+T bias in all regions of the Habronattus mitochondria that were sequenced. The A+T content for the four genes is: 77% for NDl, 78% for 16S, 71% for COI, and 73% for COn. A high A+T bias is typical of 137

most arthropod mitochondrial genomes, and these values are similar to the 78.6% A+T bias reported for D. yakuba (Clary and WoIsterJiolme 1985), the 75% reported for mites (Navajas et al. 1996), and the 83% for 12S rRNA in other spiders (Croom, Gillespie, and Palumbi 1991). The transition to traiisversion ratio is low, ranging from about 1:2 to 1:4 across genes. Such low ratios may be expected when comparing closely related individuals, and with high base composition bias.

Gene Arrangement There has been an apparent gene rearrangement in the location of in the mitochondria of the Habronattus lineage of arachnids as compared to L. polyphemus (Staton, Daehler and Brown 1997) and four myriapods (Boore, Lavrov and Brown 1998). Only a single tRNA, that coding for is located between the NDl and ribosomal 16S genes in Habronattus (fig. 6). In contrast, the chelicerate L. polyphemus and myriapods have two tRNA genes, tRNA^®"^^'^' and tRNA^'®"^^'^, between these genes. Other arthropods have located between the genes coding for COI and con (Boore et al. 1995; Boore, Lavrov and Brown 1998). There is no between COI and COH in Habronattus, and the location of this tRNA in these spiders remains imknown. The arrangement of the remainder of the genes sequenced in Habronattus appears to be conserved with respect to L. polyphemus and other arthropods. Figure 6 shows that the ribosomal 12S and 16S genes are separated by tRNA^®'. It appears that there is 4 nt of overlap of the 3' end of tRNA^^^^ with the adjacent 5' end of the NDl gene, if the iiutiation of NDl begirt with the first 138

aligned isoleuc±ie in Figure 3. If actually has a doverleaf secondary structure, it would require 20 nucleotides of overlap with the NDl gene. The 5' end of begins immediately after a 4-8 nt insertion in H. pugillis, and therefore the tRNA does not overlap with 16S. These 4-8 nts are an insertion, as they are not present in either H. oregonensis or H. geronimoi, which is outside the clade containing H. pugillis and H. oregonensis. This region is probably a non-coding spacer region since the insertion lies outside of the tRNA.

Discussion

Secondary structure The putative secondary structures of both tRNA^®' and tRNA^"^'^^ have unusual features. Both appear to lack the T^C arm, and both have some nucleotide mis-pairings in their amino-acyl stem, especially at the base. This mis-pairing is also seen in nematodes, whose tRNAs often have T-T pairings at the base of the amino-acyl stem (Wolstenholme et al. 1987; Okimoto et al. 1992). However, there is variation among individuals from the same species and population in the presence and number of mis-pairings of nucleotides in the amino-acyl stem. This type of variation among individuals may be expected to be foimd in regions whose nucleotide sequence is free to vary as long as an appropriate secondary structure is maintained. Such mis-pairings may represent intermediate stages of base substitutions, after a mutation has occurred in one location in a stem region, but before there has been a complementary mutation in the opposing region of the stem. 139

Since only DNA sequence data has been obtained to support the unusual secondary structure of the tRNA genes, there may be some doubt as to whether they form viable tRNA structures that lack PFC arms. However, if these spider mitochondrial tRNAs have "PPC and variable arms typical of other organisms, then it would be necessary for these 3' regions and the amino-acyl acceptor stem to have many nucleotide mis-pairings in the cloverleaf structure. Transfer RNA is characterized by modified bases, with modification occurring after incorporation into the ribonucleotide chain (Soil and RajBhandary 1995). It may be that base modification occurs in both of these tRNAs post-transcriptionally, so that the typical tRNA cloverleaf form is maintained. However, the kind of modification necessary to convert sequences that vary at the intra-population level to uniform cloverleaf formations has not been previously described. Alternatively, there may be post-transcriptional RNA editing to remove or change the bases so as to conform to a typical cloverleaf structure. Editing of transfer RNAs has been documented in mitochondria from other orgarusms (Lonergan and Gray 1993; Yokobori and Paabo 1995). However, becaxise of both the overlap with an adjacent gene (tRNA^"^'"^^ with NDl), and the highly divergent sequences that begin one nucleotide after the proposed end of tRNA^®', it seems unlikely that editing can explain why these imusual sequences are fotmd in Habronattus mitochondria. Truncated large subvmit rRNA genes have also been found in the mitochondria of nematodes that have tRNAs which lack the T*FC arm (Wolstenholme et al. 1987; Okimoto and Wolstenholme 1990). These authors suggested that correct functioning of tRNAs with a TV-replacement loop may 140

require changes in ribosomal RNA structure. My findings of tRNAs with a TV-replacement loop, accompanied by a tnmcated 16S gene, are consistent with that idea. Some of the mitochondrial tl^As of pulmonate gastropods also appear to lack the "PFC arm (Yamazaki et al. 1997), but their large subunit rRNA appears not to be tnmcated. An alternative explanation for why it appears that approximately 300 nucleotides of the 5' end of 16S are missing is that the gene is fragmented, and the remainder of the gene lies elsewhere within the mitochondria. This type of mitochondrial genome rearrangement has been foimd in algae (e.g. Nedelcu 1997), and may be more extensive than previously recognized. However, fragmented mitochondrial genes have never been reported from metazoans, and genomes from taxonomically diverse metazoans have been completely sequenced.

Genetic Code It is not surprising to find that ATY is the start codon for NDl. ATT was first found to be a start codon in the mitochondrial ND2 gene in himiar\s (Feamley and Walker 1987). From comparisons of other animal mitochondrial DNAs, it appears that all ATN codons may act as initiation codons in mitochondria (Bibb et al. 1981; Wolstenholme 1992). This has been suggested to be due to wobble in the third position of the codon (Barrell et al. 1980). It is more unusual to find TTG as a start codon in arachnids, as I did for con in these spiders. Among metazoans, TTG has only been foimd to be an initiation codon in nematodes (Okimoto, Macfarlane, and Wolstenholme 1990; Okimoto et al. 1992). However, all three of these proposed initiation 141 codons code for amino acids (leucine, isoleucine and methionine) that are similar — each hydrophobic with a neutral pH. Overlap of gene sequences appears to be common in metazoan mitochondria (Wolstenholme 1992). This usually occurs when gene products are encoded on opposite DNA strands, so that the transcripts do not overlap. However, it is more unusual for the genes to be encoded on the same strand as they are in these arthropods for and NDl. In some cases it appears there is overlap in tRNA genes (Staton, Daehler, and Brown 1997; Boore and Brown 1994). Relaxed constraints on tRNA structure, alternate processing, and post-transcriptional editing have been postulated as mechanisms that may allow truncated transcripts to exist.

Gene Rearrangements Mitochondrial gene order has been shown to be conserved in arthropods, with most rearrangements consisting of different placements of tRNA genes (Staton, Daehler, and Brown 1997). This conservation of gene order was used to infer phylogenetic relationships of arthropods by Boore et al. (1995; 1998). They conclude that the presence of only tRNA^"^'^'^ between NDl and 16S represents a shared derived loss of tRNA^®"^^^^ from that location among some groups of arthropods, and distinguishes crustaceans eind insects from chelicerates. My finding of no tRNA^®"^'"^^ between NDl and 16S in the chelicerate Habronattus refutes that view, showing that variation exists within chelicerates. The 16S-tRNA'^"^^'^-NDl gene arrangement presented here for Habronattus is identical with that known from crustaceans and insects (Boore, Lavrov, and Brown 1998). Recent reports by Black and Roehrdanz (1998) and Campbell and Baker (1998) indicate that mitochondrial 142 gene order amorig ticks (Arachnida) varies from that described for L. polyphemus. In light of this evidence, it appears that the arrangement of these mitochondrial tRNA genes is not sufficient for distinguishing chelicerates from other arthropods. The tRNAs may in fact be more evolutionarily mobile than assimied by Boore et al. (1995; 1998), and their future use in reconstructing relationships among arthropods should be considered more critically.

Acknowledgements

I thank R. GuteU, A. Nedelcu, R. Parker and D. Wolstenholme for helpful discussions about secondary structiu-e and gene arrangement. The manuscript benefited from comments by N. Moran, M. Nachman, and J. Silva. This work was supported by and NSF Doctoral Dissertation Improvement Grant to S. Masta and W. Maddison (DEB9520668), and a Flinn Foimdation Genetics Grant. 143

Literature Cited

Altschul, S. F., T. L. Madden, A. A. Schaffer, J. Zhang, Z. Zhang, W. Miller, and D. J. Lipman. 1997. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Research 25:3389-3402.

Avise, J. C. 1991. Ten unorthodox perspectives on evolution prompted by comparative population genetic ^dings on mitochondrial DNA. Armual Review of Genetics 25:45-69.

Barrell, B. J., S. Anderson, A. T. Bankier, et al. (15 co-authors). 1980. Different pattern of codon recognition by mammalian mitochondrial tRNAs. Proceedings of the National Academy of Science 77:3164-3166.

Black IV, W. C., and J. Piesman. 1994. Phylogeny of hard- and soft-tick taxa (Acari:Ixodida) based on mitochondrial 168 rDNA sequences. Proceedings of the National Academy of Science 91:10034-10038.

Black IV, W. C., and R. L. Roehrdanz. 1998. Mitochondrial gene order is not conserved in arthropods: prostriate and metastriate tick mitochondrial genomes. Molecular Biology and Evolution 15:1772-1785.

Bibb, M. J., R. A. Van Etten, C. T. Wright, M. W. Walberg, and D. A. Clayton. 1981. Sequence and gene organization of mouse mitochondrial DNA. CeU 26:167-180.

Boore, J. L., and W. M. Brown. 1994. Complete DNA sequence of the mitochondrial genome of the black chiton, Katharina tunicata. Genetics 138:423-443.

Boore, J. L., T. M. Collins, D. Stanton, L. L. Daehler, and W. M. Brown. 1995. Deducing the pattern of arthropod phylogeny from mitochondrial DNA rearrangements. Nature 376:163-167.

Boore, J. L., D. V. Lavrov, and W. M. Brown. 1998. Gene translocation links insects and crustaceans. Nature 392:667-668.

Campbell, N. J. H., and S. C. Barker. 1998. An imprecedented major rearrangement in an arthropod mitochondrial genome. Molecular Biology and Evolution 15:1786-1787. 144

Clary, D. O., and D. R. Wolstenholme. 1985. The mitochondrial DNA molecule of Drosophila yakuba: nucleotide sequence, gene organization, and genetic code. Journal of Molecular Evolution 22:252- 271.

Croom, H. B., R. G. Gillespie, and S. R. Palumbi. 1991. Mitochondrial DNA sequences coding for a portion of the RNA small ribosomal subunits of Tetragnatha madibulata and Tetragnatha hawaiensis (Araneae, Tetragnathidae). The Journal of Arachnology 19:210-214. deBruijn, M. H. L. 1983. Drosophila melanogaster mitochondrial DNA, a novel organization and genetic code. Nature 304:234-241.

Dirheimer, G., G. Keith, P. Dumas, and E. Westhof. 1995. Primary, secondary, and tertiary structures of tRNAs. Pages in tRNA: Structure, Biosynthesis, and Fimction (D. Soli and L. RajBhandary, eds.). ASM Press, Washington D. C.

FearrUey, I. M., and J. E. Walker. 1987. Initiation codons in mammalian mitochondria: differences in genetic code in the organelle. Biochemistry 26:8247-8251.

Gilbert, D. 1994. SeqApp. Program available from the author.

Gutell, R. R., and G. E. Fox. 1988. A compilation of large subunit RNA sequences presented in a structural format. Nucleic Acids Research 16:rl75-r313.

Huber, K. C., T. S. Haider, M. W.,. Miiller, B. A. Huber, R. J. Schweyen, and F. G. Barth. 1993. DNA sequence data indicates the polyphyly of the family Ctenidae (Araneae). The Journal of Arachnology 21:194-201.

Kumar, S., K. Tamura, and M. Nei. 1993. MEGA: Molecular Evolutionary Genetics Analysis. University Park, Pennsylvania, Pennsylvania State University.

Lonergan, K. M., and M. W. Gray. 1993. Editing of transfer RNAs in Acanthamoeba castellanii mitochondria. Science 259:812-816.

Maddison, W. P., and D. R. Maddison. 1997. MacClade: Analysis of phylogeny and character evolution. Sunderland, Massachusetts, Sinauer. 145

Moritz, C., T. E. Dowling, and W. M. Brown. 1987. Evolution of animal mitochondrial DNA: relevance for population biology and systematics. Annual Review of Ecology and Systematics 18:269-292.

Navajas, M., D. Foumier, J. Lagnel, J. Guiterrez, and P. Boursot. 1996. Mitochondrial COI sequences in mites: evidence for variations in base composition. Insect Molecular Biology 5:281-285.

Nedelcu, A. M. 1997. Fragmented and scrambled mitochondrial ribosomal RNA coding regions among green algae: a model for their origin and evolution. Molecular Biology and Evolution 14:506-517.

Okimoto, R., J. L. Macfarlane, and D. R. Wolstenholme. 1990. Evidence for the frequent use of TTG as the translation initiation codon of mitochondrial protein genes in the nematodes, Ascaris suum and Caenorhabditis elegans. Nucleic Acids Research 18:6113-6118.

Okimoto, R., J. L. Macfarlane, D. O. Clary, and D. R. Wolstenholme. 1992. The mitochondrial genomes of two nematodes, Caenorhabditis elegans and Ascaris suum. Genetics 130:471-498.

Okimoto, R., and D. R. Wolstenholme. 1990. A set of tRNAs that lack either the "PFC arm or the dihdrouridine arm: towards a minimal tRNA adaptor. The EMBO Journal 9:3405-3411.

Simon, C., F. Frati, A. Beckenbach, B. Crespi, H. Liu, and P. Flook. 1994. Evolution, weighting, and phylogenetic utility of mitochondrial gene sequences and a compilation of conserved polymerase chain-reaction primers. Aimals of the Entomological Society of America 87:651-701.

Soil, D., and U. L. RajBhandary. 1995. tRNA: structure, biosjmthesis, and function. ASM Press, Washington, D.C.

Staton, J. L., L. L. Daehler, and W. M. Brown. 1997. Mitochondrial gene arrangement of the horseshoe crab Limulus polyphemus L: conservation of major features among arthropod classes. Molecular Biology and Evolution 14:867-874.

Swofford, D. L. 1998. PAUP*: Phylogenetic analysis using parsimony. Sunderland, Massachusetts, Sinauer Associates. 146

Valverde, J., R. Marco, and R. Garesse. 1994. A conserved heptamer motif for ribosomal RNA tremscription termination in animal mitochondria. Proceedings of the National Academy of Science 91:5368- 5371.

Wolstenholme, D. R. 1992. Animal mitochondrial DNA: structure and evolution. Pages 173-216 in D. R. Wolstenholme and, K. W. Jeon, eds. Mitochondrial genomes. International Review of Cytology, Vol 141. Academic Press, New York.

Wolstenholme, D. R., J. L. Macfarlane, R. Okimoto, D. O. Clary, and J. A. Wahleithner. 1987. Bizarre tRNAs inferred from DNA sequences of mitochondrial genomes of nematode worms. Proceedings of the National Academy of Science 84:1324-1328.

Yamazaki, N., R. Ueshima, J. A. Terrett, and (15 co-authors). 1997. Evolution of pulmonate gastropod mitochondrial genomes: comparisons of gene organizations of Euhadra, Cepaea and Albinaria and implications of unusual tRNA secondary structures. Genetics 145:749-758.

Yokobori, S., and S. Paabo. 1995. Transfer RNA editing in land snail mitochondria. Proceedings of the National Academy of Science 92:10432-10435. Appendix

Sequence data

Approximate length of sequence Individual 3'COI 5* con 5'ND1-16S 3'12S-16S HbpglBQ 830 520 HbpgTBQ 830 520 HbpgTR 820 520 HbpgSR 820 520 HbpgllR 820 520 HbpglH 830 520 Hbpg2H 830 520 HbpglSMX 840 250 HbpgSSMX 840 250 HbpglSI 810 HbpglB 795 520 HbpglSC 830 520 Hbpg24SC 820 520 Hbpg56GL 675 840 640 H.geronimoi 970 425 HborlSC 850 480 HborSSC 700 300 740 HborlGL 810* 810 Hbor2GL 780* 820 148

Individual 3'COI ycon 5'ND1-16S 3'12S-16S HborSGL 730"^ 835 HborSGL 750=^ 840 Hbor9GL 780* 840 HborllGL 745 810* 810 HborlR 820 490 HborSR 730 490 Hbor4R 800* 820 HborTR 300 830 Hborl3R 850 HborlSR 825 HborlH 830 460 HborSH 830 260 Hbor2C 810 520 Hborl9C 830 480 HborlMR 800* 800 * = complete sequence Hbpg = Habronattus pugillis Hbor = Habronattus oregonensis BQ = Baboqiiivari Mts, AZ; R = Santa Rita Mts, AZ; H = Huachuca Mts, AZ; SI = Sierrita Mts, AZ; B = Black Mts, AZ; SC = Santa Catalina Mts, AZ; GL = Galiuro Mts, AZ; C = Chiricahua Mts, AZ; SMX = Sierra Madre, Mexico; MR = Mad River, CA 149

Figure Legends

Figiure 1. Proposed secondary structure of the 3' half of H. oregonensis 16S RNA, based upon the mitochondrial DNA sequence. Capital letters correspond to nucleotides conserved with respect to D. yakuba. Dashes represent Watson-Crick bonds, circles T-G bonds. The individual H. oregonensis 4R, for which there was the most overlap of sequences in two directions, was chosen to represent the folded structures in Habronattus.

Figure 2. Proposed secondary structures for and tRNA^®' based upon DNA sequence data from H. oregonensis 4R. Dashes represent Watson-Crick bonds.

Figure 3. Aligned Habronattus sequences corresponding to the tRNA^^"^^"'^^ gene.

Figure 4. Aligned Habronattus sequences corresponding to the tRNA^®' gene.

Figures. Proposed aligrunent of Habronattus NDl and COn amino acids with Drosophila yakuba, Limulus polyphemus, and Ixodes hexagonus.

Figiure 6. Mitochondrial gene arrangement in Habronattus. 150

u a u u a® u. u a u ? a a a u a g u a-u ®u-a u c U, Figure 1 u-a c ® a-u u c \ UtA. G-C \ U u a « aAAuC 'Ur* ' A A JJ666xj^C uUUag a ,uOOA, ^G-C ^ a u-a a. _ aOc a/ /c O a-u A AG a" G,/ O n -auuacug'®^ A / Ojj c A g u uu a 0a^ a a 9 u a G D u u c-g G G O u u a-u G-c A-U u a-u n-a A-U ua u-a c-g u-a u u-a CUU. u a. u-a a u u c a-u ^ u c A-ug a uc a-Ua^c u au , g-c«' u a' L U U c u—a u a u—a u a u—a u A-O a u—a a a-A a u a auua C-<3 u c u u a c A u u c A A C-G a a a auU U C-G uu G A a t A c VI auactaa^^^^^ GA A ucUUCU CU G AU o peptidyl ®u. - - G^U a o transferase WW*GCaUUA f Vfff c g GG center A Wau^^aa -U\,c C ^g\VS C o -u U<^ ^GA-ugO' U G GeU UeG u—a a ;^G Cq O-A A-O uuuuua C .A I I el M A GoU aagaauc a ^ A u" ^ A G ° AC tA A—u U-a ® a U-a g c couiacuaag1111 MI ir a ggatagattc a AnUU-A a a " U-A C-G A-U G-C c a ^u® Figure 2 tRNA Structures

tRNA leucine(CUN) tRNA valine

u u amino-acyl a amino-acyl u-a arm a-u arm a-u c a c NDl start a-u u-a a-u a c a-u a-u c c u u DHU arm a DHU arm u U A a A G U UCG a ACG u III a I I I AGC A u a TV-replacement loop A a TGc ^ TV-replacement loop g U-A U u-a U-A u-a A-U U-A G-C C-G A-U A-U a A c C U A U A UAG UAC

anticodon anticodon

tn 152

Figxxre 3

(CDN)

10 20 30 40 50 60 70] .]

aa etna DHD arm ancicodon arm TV aa arm I 1 I 1 I 1 I 1 BbpglBQ AMTAATTGGCflGAAAAATGCKTEAGAAIirrACSUVrCIAAirAAAflCnAIIIAZVrEAAATTTAirCSITMVrTAT HbpgTBQ Hbpg7SR BbpgSSR BbpgllSR HbpglH Hbpg2H HbpglSMX HbpgSSMX C G C C HfapglSI C HfcpglB C HbpglSC C Hbpg24SC C Hbpg56GL C H.geroniinoi ..0 TT T TA C C HborlSC ..C T TC C HborSSC ..C T TC C HborlGL ..C T TC C Hbor2GL ..C T TC C Hbor3GL ..C T TC C HborSGL ..C T TC C HborSGL ..C T TC C HborllCSL ..C T TC C HborlR ..C T TC C C Rbor3R ..C T TC C C Hbor4R ..C T TC C W.C HborTR ..C T TC C C Hborl3R ..C T TC C C HborlSR ..C T TC C C HborlH ..C T TC C C HborBH ..C T TC C C Hbor2C T.C T TC C HborlSC T.C T TC C HborlMR •GC T TC C C HHKUKKKKM gvoowwfojotoe So P

o o o o o o o O O O O O O O O O O O O O O O o

5 is %% %% poppppppppppppp gttgg???gg§§gg§? o o >- >• Figure 5 Translated NDl and COII

NDl amino acids ( 10 20 30 40 SO 60 70 80 90 100 110 120 } I ...... I D.yak MEPILSLIOSljLLIICVLVSVAFI/niliEIUOn/jyiQIRKOPIWVroLmiPQPFCDAIKliFTKEQTyPUiSN-YLSyyiSPIFSLFIiSLFVVAKMPFFVXLYSFNUXJLFFLCCTSLQVYTVHVAOWSSNSNyA Limulus HSPIVCyWlLICVLVOVAFiyTLLERSlLOyiQIR-- Ixodes HISyLSyiFLIHCVLVSVAFrrLLERKIUSyiHIRKaPNKVaPiaVFQPPSDMKLFSKEMNMHFyiN-MIFyilSSWLLILHHSUfPLFSWDyNQILIEYGHVFliLClSSIiSVYVlLFGCMSSNSKYA Hbor4R IKFIINYILXl.VSILISVAt'rriLBRKIIiQyiQIRKOPNKVQMMQIliQPFSOALKU'NKNLLSSQfrNFMISyiTPALSFlilSHLIIPIIl.FNr^SSLyDNKHNILLFFILSSISAySlLLICMSSNSKyS

COII amino acids [ 10 20 30 40 SO 60 70 80 90 100 110 ) ( ...... 1 D.yak MSlVWNLGLQDSASPLMEQLIFFHDHALLILVMITVLVayiWMLFFNNyVNRFUilGQLIQIIWriLPAI ILLFIALPSUtLLyLLOBINEPSVTLKSIOHQWyWSyEySOFDN Limulus MAIWSHLLFQDSASLUlBQHIFFHDHTMIII/rLI — Ixodes HVlVISNIHFSNSNSPIHEIQHIFFHDHSMLIIMMITIITLyHIINIMMNSPSSRFUtBCSQEIETIWrilPAITLIFIAFPSUU.LyMLOESFTPSITIKIIQHQWyWSyEl.SDFNI Hbor7SR LI>VWQSI,yFQOaVSSVMBKLIFI>HDm«IMIMIMFLVOyMLiaAYVGQSyiniOLFBOQEI.EMIV/rvi.PAVFISiyCFSPSSPIMFNSSVW Figure 6 Gene Arrangement

primer primer HbieS HbNDI

(0 c< ^ CO 16S ^ ° Z w Nni ^ 5' QC ~

395 nt 1021 nt

lOT- 10O

primer primer primer C1-J-2309 C1-N-2971 C2-N-3389

5' CXJii cot

717 nt 300 nt 156

INFERENCES

Avise, J. C. 1994. Molecular markers, natural history and evolution. Chapman & Hall, New York.

Avise, J. C., R. M. Ball, and J. Arnold. 1988. Current versus historical population sizes in vertebrate species with high gene flow: a comparison based on mitochondrial DNA lineages and inbreeding theory for neutral mutations. Molecular Biology and Evolution 5:331- ^TAA X X*

Barraclough, T. G., P. H. Harvey, and S. Nee. 1995. Sexual selection and taxonomic diversity in birds. Proceedings of the Royal Society of London 259:211-215.

Darwin, C. 1859. The origin of species by means of natural selection. John Murray, London.

Darwin, C. 1871. The Descent of Man, and Selection in Relation to Sex. Murray, London.

Fisher, R. A. 1958. The Genetical Theory of Natural Selection. Clarendon Press, Oxford.

Futuyma, D. J. 1998. Evolutionary Biology. Sinauer, Simderland, Massachusetts.

Griswold, C. E. 1987. A revision of the jimiping spider genus Habronattus F.O P.-Cambridge (Araneae; Saltiddae), with phenetic and cladistic analyses. University of California, Entomology 107:1-344.

Hudson, R. R. 1990. Gene genealogies and the coalescent process. Pages 1-44 in Oxford Surveys in Evolutionary Biology (D. Futujnna and J. Antovics, eds.). Oxford Uruversity Press, Oxford. 157

Kaneshiro, K. Y., and C. R. B. Boake. 1987. Sexual selection and spedation: Issues raised by Hawaiian Drosophila. Trends in Ecology and Evolution 2:207-213.

Kaneshiro, K. Y., and J. S. Kurihara. 1981. Sequential differentiation of sexual behavior in populatior\s of Drosophila silvestris. Pacific Science 35:177- 183.

Lande, R. 1981. Models of spedation by sexual selection on polygenic traits. Proceedings of the National Academy of Sdence 78:3721-3725.

Mayr, E. 1954. Change of genetic envirorunent and evolution. Pages 157-180 in Evolution as a process (J. Huxley, A. C. Hardy and E. B. Ford, eds.). Macmillan, New York.

Mesnick, S. L. 1996. Sexual selection and biological diversification: patterns and processes. Ph.D. dissertation. University of Arizona.

Peckham, G. W., and E. G. Peckham. 1909. Revision of the Attidae of North America. Transactions of the Wisconsin Academy of Sciences, Arts and Letters XVI, part 1:355-646.

Van Devender, T. R. 1977. Holocene woodlands in the southwestern deserts. Sdence 198:189-192.

Van Devender, T. R., and W. G. Spaulding. 1979. Development of vegetation and climate in the southwestern United States. Science 204:701-710.

West-Eberhard, M. J. 1983. Sexual selection, sodal competition, and speciation. Quarterly Review of Biology 58:155-183. IMAGE EVALUATION TEST TARGET (QA-3) /> w & <• e

150mm

IISA4GE. Inc 1653 East Main Street Rochester, NY 14609 USA Phone: 716/482-0300 Fax: 716/288-5989