Phylogeny of the Xeromelissinae (Hymenoptera: Colletidae) Based Upon Morphology and Molecules*

Home , Colletidae, Hylaeus affinis

Apidologie 39 (2008) 75–85 Available online at: c INRA/DIB-AGIB/ EDP Sciences, 2008 www.apidologie.org DOI: 10.1051/apido:2007063 Original article

Phylogeny of the Xeromelissinae (Hymenoptera: Colletidae) Based upon Morphology and Molecules*

Eduardo A.B. Almeida1,3,LaurencePacker2,BryanN.Danforth1

1 Comstock Hall, Department of Entomology, Cornell University, Ithaca, NY, 14853 USA 2 Department of Biology, York University, Toronto, Ontario, M3J 1P3, Canada 3 Current address: Departamento de Zoologia, Universidade Federal do Paraná, Cx. Postal 19020, CEP: 81531-980, Curitiba, P.R. Brazil

Received 18 June 2007 – Revised 15 October 2007 – Accepted 30 October 2007

Abstract – We present the results of a combined analysis of 248 morphological characters and sequences from 3 genes for 29 species of Xeromelissinae and 7 outgroup taxa including representatives of the colletid subfamilies Colletinae, Euryglossinae, Hylaeinae, Paracolletinae, and Scrapterinae. The paracolletine genus Trichocolletes was used to root the tree. The results agree with most of those obtained in an earlier, entirely morphological analysis. Noteworthy are (1) the paraphyly of Chilimelissa in relation to Xeromelissa,and(2) the lack of sister group relationship between Hylaeinae and Xeromelissinae. Other than minor rearrange- ments resulting from swapping adjacent nodes, the only major diﬀerence is the placement of one species of Chilicola, C. aenigma, which no longer groups within C. (Chilioediscelis), but instead appears to be closer to Xenochilicola. The inﬂuence upon phylogenetic results caused by highly morphologically autapomorphic taxa is discussed.

bee / Colletidae / phylogeny / Neotropical / Xeromelissinae

1. INTRODUCTION There are four currently recognized genera, the phylogenetic relationships among which, ac- Xeromelissinae is a subfamily of Colletidae cording to Packer’s (2008) analysis of a large of moderate size (approximately 200 species) number of morphological characters, are sum- all of which are restricted to the New World. marized in Figure 1. As the name implies, these bees are generally Only Chilicola has been considered large found in xeric habitats, mostly in temperate ar- and/or diverse enough to warrant division into eas of southern South America. Xeromelissine subgenera, of which there are now fifteen. bees are typically small to minute, and gener- Chilimelissa, as commonly recognized (e.g. ally slender. Michener, 2000) is comparatively speciose The taxonomy of Xeromelissinae has had and diverse, although no formal subdivision a complex history, summarized by Packer into subgenera has yet been made as phy- (2008). No tribal divisions are currently ac- logenetic analysis of it remains incomplete. cepted within the subfamily: the previous Xenochilicola comprises three species, and tribal level classification having been dis- Xeromelissa is monotypic and only includes X. pensed with by Michener and Rozen (1999). wilmattae Cockerell. Packer (2008) found Xe- Corresponding author: Eduardo A.B. Almeida, romelissa to render Chilimelissa paraphyletic, [email protected] and as a consequence proposed a revised * Manuscript editor: Bryan Danforth generic classification for Xeromelissinae. Online material is available at: The purpose of the present paper is to test http://www.apidologie.org Packer’s classification of the Xeromelissinae

Article published by EDP Sciences and available at http://www.apidologie.org or http://dx.doi.org/10.1051/apido:2007063 76 E.A.B. Almeida et al.

Figure 1. Summary of phylogenetic relationships among xeromelissinae genera according to the analysis of the morphological data set by Packer (2008: Figs. 1 and 2).

by comparing results among morphological, we use the paracolletine genus Trichocolletes to molecular and total evidence analyses. root the entire tree, including Colletes. This decision is based upon results of a higher level phylogeny of the entire Colletidae (Almeida, 2007). 2. METHODS In instances where only one sex is known for a species that was used in the molecular data set, that 2.1. Taxon sampling sex was coded for the morphological characters and a closely related species was used for the other sex. Different taxa were available in a form suitable This was necessary twice: females of Chilicola tri- for DNA sequencing than were used in the pre- carinatoides Packer and males of Cc. liliana Packer vious morphological analysis. Consequently, there are unknown. Females of Cc. tricarinata Packer and are fewer representative subgenera of Chilicola,but males of Cc. olmue Toro and Moldenke, respec- more species of Chilimelissa, than in the earlier tively, were used in their place. Similarly, there are a study and the remaining two genera are represented few cases where the species that was sequenced was by only a single species each. Chilicola is abbrevi- very closely related to one that was studied mor- ated as “Cc.”, Chilimelissa as “Cm.”, and Colletes phologically, these taxa were combined in the to- as “Co.”. There are also differences in the outgroups tal evidence analysis. Thus Cc. unicarinata Packer used for molecular and morphological analyses. In and Cc. chubutense Packer; Cc. andina Toro and some cases (such as Hylaeus affinis (Smith)), the Moldenke and Cc. araucana Toro and Moldenke; same outgroups available for DNA were scored Cc. mantagua Toro and Moldenke and Cc. vicugna for the morphological characters. In others, only Toro and Moldenke; Cc. brzoskai Michener and Cc. congeners could be used, with the result that Col- (Oroediscelis) sp. were used for morphological and letes cunicularius L., Scrapter nitida Friese, Euh- molecular data, respectively. esma halictoides (Rayment), and Callohesma cal- In total, 36 taxa were included in the analyses, liopsiformis (Cockerell) were used for morphology 7 representing outgroups. The complete list of taxa whereas Co. bicolor Smith, S. niger Lepeletier and and their provenances are provided in Table S1 (on- Serville, E. crabronica (Cockerell), E. platyrhina line material). (Cockerell), and Ca. calliopsella (Cockerell) were sequenced. Packer (2008) used Colletes to root the entire tree (which included representatives of three 2.2. Morphological characters other subfamilies related to the Xeromelissinae – Hylaeinae, Euryglossinae, and Scrapterinae1). Here We simply scored all exemplars for the same 1 The choice of usage of Scrapterinae Melo and characters as were used by Packer (2008) and do not Gonçalves over Scraptrinae Ascher and Engel is reiterate the characters or their states here. In cases discussed in Appendix 1 (online material). where additional character states were required for Phylogeny of Xeromelissinae 77

the different suite of species used, these are de- low melting point agarose gels (FMC, Rockland, scribed in Appendix 2 (online material). Similarly, Maine) overnight at 4 ◦C. DNA was recovered from we only present the data matrix for those taxa not gel slices using the Promega Wizard PCR Preps included in the previous analysis (Appendix 3: on- DNA Purification kit. Gel purification was unneces- line material). sary for PCR products that produced a single prod- uct: both fragments of 28S rRNA and the upstream 1100 bp fragment of EF-1α. Automatic DNA se- 2.3. Choice of molecular data quencing was performed using the Applied Biosys- tems Automated 3730 DNA Analyzer employing Molecular data were collected from three gene Big Dye Terminator chemistry and AmpliTaq-FS loci that have been providing robust results for in- DNA Polymerase at Cornell University Life Sci- sect phylogenetic studies (Danforth, 1999; Danforth ences Core Laboratories Center. et al., 2004). Elongation factor-1 alpha, F2 copy (EF-1α) and the large subunit 28S rRNA locus (28S rRNA), regions D1–D5 were chosen to resolve 2.5. Data analysis deeper relationships among outgroup and ingroup taxa, and within Xeromelissinae as well. These 2.5.1. Alignment genes have been used to successfully recover Ter- tiary to Cretaceous age divergences in bee phylo- Alignments were generated using similarity cal- genies (e.g., Danforth et al., 1999, 2004, 2006a, culated at the nucleotide level (“-n”) with DI- b). EF-1α has been the most widely used nuclear ALIGN 2.2 (Morgenstern, 1999) and corrected protein-coding gene for insect phylogenetics (see manually for obvious alignment errors using Mac- Danforth et al., 2004, pp. 310–311 for comments on Clade v. 4.08 OSX (Maddison and Maddison, 2005) this gene) The third gene sampled was cytochrome and Winclada 1.00.08 (Nixon, 2002). For EF-1α, oxidase 1 (COI), a mitochondrial protein-coding the honey bee (Apis mellifera) sequence was used gene known for its utility in species-level phyloge- to establish reading frames and intron/exon bound- netic studies of insects (e.g. Danforth, 1999). aries. In cases where multiple sequences were avail- Primer information for each gene is presented in able for the same species, the sequences were Table S2 (online material). merged after being resolved in the same clade in preliminary analyses and resulting partial polymor- phisms were kept as such. 2.4. DNA extraction, PCR, and sequencing 2.5.2. Phylogenetic analyses using Genomic DNA was extracted using phenol- parsimony chloroform protocols (Doyle and Doyle, 1990, adapted by Danforth, 1999) but without use of liq- Raw data files were edited with Winclada uid nitrogen and RNase. Tissue was taken from the (Nixon, 2002) and this program was used to ex- thoracic musculature and/or legs depending on the port to TNT version 1.1 (Goloboff et al., 2004). Se- rarity and size of available specimens. The phenol- quence indels were treated as missing data. Ratchet, chloroform-isoamyl alcohol stage was performed in sectorial, drift and tree fusing with “collapse trees Phase-Lock Gelr 2.0 mL Eppendorf tubes to facil- after search” and “find minimal length” set to 10 itate the separation of phenol from the remainder found the same most parsimonious trees. Symmet- and thus increase the final DNA yield. PCR am- rical resampling (Goloboff et al., 2003) was per- plifications of the genes listed above were done for formed on unweighted results with 10 000 iterations 35 cycles under the following conditions: an initial and a probability of character weight change (up or denaturation at 94 ◦C for 60 s, followed by 35 cy- down) of 33%. Symmetric resampling allows esti- cles under the following conditions: an initial denat- mation of group support without being biased by uration at 94 ◦Cfor90s,followedby35cyclesof differential character (or character state) weights, denaturation at 94 ◦C, annealing at 48–58 ◦C, and which affect results obtained with jackknifing and extension at 72 ◦C – specific conditions for each bootstrapping (see Goloboff et al., 2003). Support locus amplified are listed in Table S2. Prior to se- is indicated on the cladograms using GC (Group quencing, most PCR products were gel-purified in supported/Contradicted). For a particular node, this 78 E.A.B. Almeida et al.

Table I. Overview of the datasets. Number of Informative Total Information/ characters characters information number of characters EF-1a exons 1098 267 1416 1.29 EF-1a introns 586 240 1345 2.30 28S rRNA 1565 135 655 0.38 COI 663 257 1490 2.24 molecular combined 3912 899 4906 1.24 morphology 248 425 2143 8.61 COMBINED 4160 1324 7049 1.68

calculates the difference between the frequency of 3. RESULTS the group and the most frequently found contradic- tory arrangement. GC values can vary from –100 3.1. Data set to +100, representing maximum contradiction (the alternative grouping is favored in all resampled ma- GenBank accession numbers for all se- trices) to maximum support (the original grouping quences used in this study are presented in Ta- found in all resampled matrices) (Goloboff et al., ble S3 (Supplementary Materials). 2003). The complete combined matrix contained 4160 characters: 3912 molecular and 248 morphological. The information content (sensu 2.5.3. Bayesian phylogenetics Farris) was higher for the morphological com- ponent as was the mean relative information of its constituent characters, i.e. informa- The best-fit model of evolution for each of the tion/number of characters (Tab. I). Informa- four partitions (introns and exons of EF-1α, 28S tion of a character is a quantity defined as its rRNA, and COI,) was statistically tested. Decision maximum number of steps minus its minimum theory (DT), Akaike information content (AIC), number of steps (this is also the denominator and Bayesian information content (BIC) were em- for the retention index [Farris, 1989]). ployed to seek for a balance between model com- plexity and its suitability for each data partition. Two computer programs were used to shed light on 3.2. Phylogenetic results the most appropriate model(s) for the data: (1) DT- ModSel (Minin et al., 2003 – DT model selection) The morphological data gave nine most par- and (2) MrAIC.pl 2.2 (Nylander, 2004 – AIC and simonious trees of length 1215, ci = 46 and BIC). ri = 69 (strict consensus shown in Fig. 2). All Bayesian searches were conducted with the of the unresolved nodes are within the clade serial version of MrBayes 3.1.2 (Altekar et al., formed by Chilimelissa and Xeromelissa.This 2004; Huelsenbeck and Ronquist, 2005) through is not so surprising considering that the main the Computational Biology Service Unit at the Cor- objective of the original morphological data nell Theory Center. Searches were run for 3 × 106 matrix was to assess generic level relationships generations on two sets of 10 chains each. The initial 2000 trees were discarded after examining the within the subfamily and subgeneric relation- variation in log likelihood scores over time. Con- ships within Chilicola – a more extensive suite vergence was also assessed using the potential scale of characters for Chilimelissa would likely re- factor for the parameters. A partitioned model for sult in greater resolution. The new result is en- the four loci (exons and introns of EF-1α treated tirely congruent with the previous analysis. separately) of the concatenated dataset was applied The molecular data yields eight most par- using the following unlinked models: (1) EF-1α,ex- simonious trees with length of 3703 steps, ons: HKY+I+G; (2) EF-1α, introns: GTR+I+G; (3) ci = 49 and ri = 62 (strict consensus shown 28S rRNA: SYM+I+G; (4) COI: GTR+G+SSI. in Fig. 3). While the relationships among Phylogeny of Xeromelissinae 79

Figure 2. Strict consensus cladogram of the nine most parsimonious trees based on the morphological matrix analyzed with equal weights; tree length = 1215, ci = 46, ri = 69. Species of Xeromelissinae are marked in bold; names of colletid subfamilies sampled as outgroups for the analysis are provided after the species names. the genera of Xeromelissinae remain congru- either: Xenochilicola mamigna and Chilicola ent with those of the morphological analy- aenigma form a clade, sister to the remaining sis (Fig. 2), the phylogenetic pattern among Chilicola species (Fig. 4). outgroups and subgenera of Chilicola are dif- The total evidence analysis resulted in three ferent. In particular, the molecular data place equally most parsimonious trees with length the euryglossines with the scrapterines, which of 4963 steps, ci = 48, ri = 63 (strict con- is congruent with previous molecular analy- sensus shown in Fig. 5). This analysis re- ses (Almeida, 2007). Chilicola does not ap- verses most of the unusual features of the pear as monophyletic here, with Xenochili- results from molecular data to less unex- cola mamigna, Chilicola aenigma,andthe pected patterns. The subfamily level result is clade comprised by the remaining Chilicola (Paracolletinae, Colletinae (Scrapterinae (Hy- species forming a trichotomy. Many of the laeinae (Euryglossinae + Xeromelissinae)))). subgenera of Chilicola are also in unexpected Relationships among the genera are identical relationships. Within the clade consisting of to that found in the morphological cladistic Chilimelissa and Xeromelissa,however,the analysis and the Bayesian molecular analysis, pattern is congruent with that of the previ- except for the placement of Cc. aenigma. ously published result (Packer, 2008), except that Cm. rozeni and Cm. australis are now in a clade (along with three species not previously 4. DISCUSSION included) separate from other Chilimelissa. The Bayesian analysis of the molecular data Our results are largely in good agreement does not support the monophyly of Chilicola with the purely morphological analysis of 80 E.A.B. Almeida et al.

Figure 3. Strict consensus cladogram of the eight most parsimonious trees based on the molecular matrix analyzed with equal weights; tree length = 3703, ci = 49, ri = 62. Numerals represent GC support values calculated from 10000 replications using TNT (P = 33). Species of Xeromelissinae are marked in bold; names of colletid subfamilies sampled as outgroups for the analysis are provided after the species names.

Packer (2008). The relationships among the datasets tested (morphological, molecular, outgroups in our combined analysis is identi- and combined), regardless of the kind of cal to that found in the earlier study, with one phylogenetic analysis performed, strongly exception: the generic relationships among support the paraphyly of Chilimelissa in the ingroup are the same and the subgeneric relation to Xeromelissa. This change is be- (Chilicola) and species (Xeromelissa) phylo- ing made formally by Packer (2008). genetic patterns are mostly congruent. (iii) The resurrection of the subgenus Oedis- The following points of some classificatory celisca from synonymy with Oediscelis, import can be stated more firmly than previ- from which it is distantly located on the ously on the basis of the combined data: cladogram. It also groups with Anoedis- celis as in the earlier study and the resur- (i) Our combined data argue against En- rection is made formally in Packer (2008). gel’s (2005) suggestion that Euryglossi- (iv) The erection of the new subgenus for nae and Hylaeinae are sister groups. A Chilicola liliana Packer, Cc. olmue Toro larger study of colletid subfamily relation- and Moldenke, and their relatives (see ships (Almeida, 2007), as well as a recent Packer, 2008). study of family-level relationships in bees (v) The resurrection of Heteroediscelis as a (Danforth et al., 2006b), strongly suggests subgenus distinct from Oediscelis (Packer, that Xeromelissinae is sister to Hylaeinae. 2008). (ii) The synonymization of Chilimelissa with Xeromelissa (Packer, 2008), the latter The differences between the two sets of re- rendering the former paraphyletic. All sults are largely minor, involving swapping Phylogeny of Xeromelissinae 81

Figure 4. Bayesian majority-rule phylogram based on partitioned model with parameters estimated separately for (1) exons of EF1α, (2) introns of EF1α, (3) 28S rRNA, and (4) COI. Branch support is given by posterior probabilities. derived from 58 000 trees; the first 2000 trees were discarded. (harmonic mean: –lnL = 22713.90). Species of Xeromelissinae are marked in bold; names of colletid subfamilies sampled as outgroups for the analysis are provided after the species names. subgenera of Chilicola on adjacent nodes. to Chilicola rather than to [Geodiscelis + Thus, the following pairs exchange places in (Chilimelissa + Xeromelissa)] (Fig. 6). This the combined analysis in comparison to the possibility was mentioned by Packer (2008) previously published result: Oediscelis and who showed that the pattern suggested here Heteroediscelis;[Oediscelisca + Anoediscelis] was, with the original data, actually only and [Cc.chubutense + Pseudiscelis]; Oroedis- one step longer than the most parsimonious celis and Cc. liliana (Fig. 62). Two aspects result obtained by morphology alone. The of the current result suggest more important closer relationship of Xenochilicola to Chili- differences from Packer’s (2008) study. The cola had already been suggested by the clas- first is that Xenochilicola becomes sister taxon sification in which Xeromelissinae was subdi- vided into Chilicolini and Xeromelissini (e.g. 2 Cc. chubutense is used here to represent the taxon Michener, 1995), and by the analysis by Toro represented by Cc. unicarinata by Packer (2008); and Moldenke (1979). and, similarly, Cc. liliana corresponds to the taxon represented by Cc. olmue by Packer (2008), in each The other major difference is more surpris- case we use a very closely related species that is ing. Packer (2008) found that Cc. aenigma available for molecular analysis. The morphological was nested within the subgenus Chilioedis- data were recoded as required. celis; the combined data place it as sister 82 E.A.B. Almeida et al.

Figure 5. Strict consensus cladogram of the three most parsimonious trees inferred from a combined data set composed of 248 morphological characters and molecular data from three gene loci: EF-1α, 28S rRNA, and COI.; tree length = 4963, ci = 48, ri = 63. Numerals represent GC support values calculated from 10000 replications using TNT (P = 33). Species of Xeromelissinae are marked in bold; names of colletid subfamilies sampled as outgroups for the analysis are provided after the species names. to all remaining Chilicola (Fig. 2) and the subgenera, particularly those of the male hind molecular data alone suggest that it is sis- leg, which are considerably modiﬁed in related ter to Xenochilicola mamigna (Fig. 4). In the taxa but are not sexually dimorphic (other than morphological analysis (Fig. 2), the grouping for the scopal hairs) in Cc. aenigma. Assum- of this species within Cc. (Chilioediscelis)was ing the close relationship of this species to supported most strongly by (1) the robust and Xenochilicola to be correct, it is possible to curved hind tibial spurs; (2) the reduced in- see some similarities between some of the ner tooth on the hind tarsal claws; and (3) unique states for Cc. aenigma and those found the absence of corbiculate structure of the fe- to be synapomorphic for the two species of male sternal scopa. These all appeared as ro- Xenochilicola included in the morphological bust synapomorphic evidence for monophyly study. Resolution of this discrepancy will re- of the subgenus Cc. (Chiloediscelis) includ- quire reanalysis of the morphological traits ing Cc. aenigma in the original morphologi- of these bees, gathering additional molecular cal analysis. However, Cc. aenigma is a highly data and incorporation into the data matrix of autapomorphic species with numerous unique a recently discovered Patagonian bee species states, particularly of the male genitalia, that that superﬁcially appears somewhat interme- could not be homologized with those of any diate between Cc. aenigma and Xenochilicola other exemplar included in the study. Further- (Genaro and Packer, unpublished data). more, it lacked some of the synapomorphies These results suggest an interesting con- that united Cc. (Chilioediscelis) with related trast to the long branch problem that is well Phylogeny of Xeromelissinae 83

Figure 6. Summaries of phylogenetic hypotheses relationships within Xeromelissinae inferred from a morphological dataset by Packer (2008: Fig. 2): cladogram on the left; and the result of the combined matrix formed by molecular and morphological data of the present study (abridged from Fig. 5): cladogram on the right. Dashed lines represent incongruence between the two trees. Species of Chilimelissa, Geodiscelis, Xenochilicola,andXeromelissa and subgenera of Chilicola missing from one of the original studies were removed from the summary cladograms, while preserving the relationships among the remaining taxa. Chilicola liliana and Cc. chubutense are included in the present study to represent two newly described subgenera (see Packer, 2008 and comments in the text).

known to dog molecular phylogenetic anal- of Canada to L. Packer. A National Geographic yses (reviewed by Bergsten, 2005). Here we Grant in Aid of Research (NGS grant No. 6946- have a morphologically highly autapomorphic 01) provided travel funds for the collection of species that came out nested deeply within the some important specimens for this study. E.A.B. phylogeny in, what we now believe to be, the Almeida is grateful to Fundação de Coordenação wrong position, as a result of sharing a few de Aperfeiçoamento de Pessoal de Nível Inferior convergences. This cautions against accepting (CAPES: Brazilian Ministry of Education) for a the phylogenetic position of morphologically Ph.D. scholarship. The following people are ac- outlying taxa uncritically. An alternative inter- knowledged for arranging loans or gifts of speci- pretation is that the molecular data is respon- mens from their collections or loans from collections under their jurisdiction, J.S. Ascher, J.M.F. sible for the attraction of Cc. aenigma to be Camargo, B.N. Danforth, G. Else, Z. Falin, T. sister of Xenochilicola (Figs. 3, 4) or to the Griswold, R. Hoebeke, J. Huber, C.D. Michener, J. base of the Chilicola-clade (Fig. 5). The rel- Neﬀ, F.D. Parker, A. Roig-Alsina, J.G. Rozen, L. atively long branch of Xenochilicola mamigna Ruz, F.A. Silveira, F. Vivallo, K. Walker and the late (Fig. 4) reﬂects many autapomorphies, which H. Toro. is the typical cause of the long-branch attraction phenomenon. Une phylogénie des Xeromelissinae (Hymenop- tera : Colletidae) basée sur les caractères mor- phologiques et moléculaires. ACKNOWLEDGEMENTS Colletidae / abeille / phylogénie / Xeromelissi- This research was funded by Systematic Biol- nae / région néotropicale ogy grants to B.N. Danforth (DEB-0211701 and DEB-0412176) and a discovery grant from the Nat- Zusammenfassung – Eine Phylogenie der ural Sciences and Engineering Research Council Xeromelissinae (Hymenoptera: Colletidae), 84 E.A.B. Almeida et al.

basierend auf morphologischen und molekula- Bergsten J. (2005) A review of long-branch attraction, ren Merkmalen. Die Xeromelissinae bilden eine Cladistics 21, 163–193. Subfamilie der Colletidae. Sie umfasst etwa 200 Danforth B.N. (1999) Phylogeny of the bee genus Art mittelgrosser Bienen, die in ihrer Verbreitung Lasioglossum (Hymenoptera: Halictidae) based alle auf die Neue Welt bechränkt sind. Wie der on mitochondrial cytochrome oxidase, Syst. Namen bereits besagt, handelt es sich hierbei um Entomol. 24, 377–393. Bienen, die im allgemeinen in Trockenhabitaten vorkommen, vor allem im südlichen Südamerika. Danforth B.N., Sauquet H., Packer L. (1999) Xeromelissinen sind typischerweise klein bis sehr Phylogeny of the bee genus Halictus klein und im allgemeinen von schlanker Gestalt. (Hymenoptera: Halictidae), based on parsi- In der vorliegenden Arbeit präsentieren wir die mony and likelihood analyses of nuclear EF-1α Ergebnisse einer kombinierten Analyse von 248 sequence data, Mol. Phylogenet. Evol. 13, morphologischen Merkmalen und den Sequenzen 605–618. von drei Genen. Die Analyse umfasst 29 Arten, Danforth B.N., Brady S.G., Sipes S.D., Pearson die alle Genera der Xeromelissinae repräsentieren, A. (2004) Single-copy nuclear genes recover sowie 7 Taxa mit Vertretern der Colletiden- Cretaceous-age divergences in bees, Syst. Biol. Subfamilien Colletinae, Euryglossinae, Hylaeinae, 55, 309–326. Paracolletinae und Scrapterinae als Aussengrup- Danforth B.N., Fang J., Sipes S. (2006a) Analysis of pen. Der molekulare Datensatz bestand aus den family-level relationships in bees (Hymenoptera: Sequenzen von zwei Kerngenen (Elongationsfaktor Apiformes) using 28S and two previously unex- 1 alpha (F2-Kopie) und 28S rRNA) und einem plored nuclear genes: CAD and RNA polymerase mitochondrialen Gen (Cytochromoxidase 1). Die II, Mol. Phylogenet. Evol. 39, 358–372. Wurzel des Stammbaums wurde mithilfe der Merk- male des Genus Trichocolletes (Paracolletinae) Danforth B.N., Sipes S., Fang J., Brady S.G. (2006b) definiert. Die Ergebnisse stimmen in den meisten The history of early bee diversification based on Punkten mit den Befunden einer früheren Analyse five genes plus morphology, Proc. Natl. Acad. Sci. überein. Bemerkenswert sind (1) die Paraphylie USA 103, 15118-15123 von Chilimelissa in Bezug zu Xeromelissa und Doyle J.J., Doyle J.L. (1990) A rapid total DNA prepa- (2) das Fehlen einer Schwestergruppenbeziehung ration procedure for fresh plant tissue, Focus 12, zwischen Hylaeinae und Xeromelissinae. Ausser 13–15. kleineren Veränderungen in der Stammbaumto- pologie, die aus der Verschiebung benachbarter Engel M.S. (2005) Family-group names for bees Knotenpunkte herrühten, lag der einzige grössere (Hymenoptera: Apoidea), Am. Mus. Novit. 3476, Unterschied in der Positionierung einer Art des 1–33. Genus Chilicola, C. aenigma. Diese gruppierte Farris J.S. (1989) The retention index and the rescaled nicht mehr innerhalb von C. (Chilioediscelis), son- consistency index, Cladistics 5, 417–419. dern erschien enger verwandt mit Xenochilicola. Folmer O., Black M., Hoeh W., Lutz R., Vrijenhoek Neben diesen Ergebnissen diskutieren wir den R. (1994) DNA primers for amplification of mi- Einfluss von morphologisch stark autapomorphen tochondrial cytochrome coxidase subunit I from Taxa mit wenigen gemeinsamen Merkmalen auf diverse metazoan invertebrates, Mol. Mar. Biol. die phylogenetischen Beziehungen mit anderweitig Biotechnol. 3, 294–299. weniger aussergewöhnlichen Arten. Goloboff P.A., Farris J.S., Källersjö M., Oxelman B., Biene / Colletidae / Phylogenie / neotropisch / Xe- Ramırez M.J., Szumik C.A. (2003) Improvements romelissinae to resampling measures of group support, Cladistics 19, 324–332. Goloboff P.A., Farris J.S., Nixon K.C. (2004) TNT: Tree analysis using new technology, com- REFERENCES puter program distributed by the authors at http://www.zmuc.dk/public/phylogeny/TNT/ (ac- Almeida E.A.B. (2007) Systematics and Biogeography cessed on 8 November 2007). of Colletidae (Hymenoptera : Apoidea), unpublished PhD Dissertation, Cornell University, Huelsenbeck J.P., Ronquist F. (2005) MRBAYES: Ithaca. Bayesian inference of phylogeny, Bioinformatics 17, 754–755. Altekar G., Dwarkadas S., Huelsenbeck J.P., Ronquist F. (2004) Parallel Metropolis-coupled Markov Maddison D.R., Maddison W.P. (2005) MacClade chain Monte Carlo for Bayesian phylogenetic in- 4.08. Sinauer Associates, Sunderland. ference, Bioinformatics 20, 407–415. Mardulyn P., Whitfield J.B. (1999) Phylogenetic sig- Belshaw R., Quicke D.L.J. (1997) A molecular nal in the COI, 16S, and 28S genes for infer- phylogeny of the Aphidiinae (Hymenoptera: ring relationships among genera of microgastrinae Braconidae), Mol. Phylogenet. Evol. 7, 281–293. (Hymenoptera: Braconidae): Evidence of a high Phylogeny of Xeromelissinae 85

diversification rate in this group of parasitoids, Nylander J.A.A. (2004) MrAIC.pl. Computer Mol. Phylogenet. Evol. 12, 282–294. program distributed by the author at: http://www.abc.se/∼nylander/mraic/MrAIC143.tgz Michener C.D. (1995) A classification of the bees of (accessed on 8 November 2007). the subfamily Xeromelissinae, J. Kans. Entomol. Soc. 68, 332–345. Packer L. (2008) Phylogeny and classification of Michener C.D. (2000) The Bees of the World, John the Xeromelissinae (Hymenoptera: Apoidea, Hopkins University Press, Baltimore. Colletidae) with special emphasis upon the genus Chilicola, Syst. Entomol., 33, 72–96. Michener C.D., Rozen J.G. Jr. (1999) A new [Published article online: 6 September, 2007: doi: ground-nesting genus of Xeromelissine bee from 10.1111/j.1365-3113.2007.00398.x]. Argentina and the tribal classification of the subfamily (Hymenoptera: Colletidae), Am. Mus. Novit. 3281, 1–10. Schulmeister S., Wheeler W.C., Carpenter J.M. (2004) Simultaneous analysis of the basal lineages of Minin V., Abdo Z., Joyce P., Sullivan J. (2003) Hymenoptera (Insecta) using sensitivity analysis, Performance-based selection of likelihood models Cladistics 18, 455–484. for phylogeny estimation, Syst. Biol. 52, 674–683. Toro H., Moldenke A. (1979) Revisión de los Morgenstern B. (1999) DIALIGN 2: improvement of Xeromelissinae Chilenos, An. Mus. Hist. Natur., the segment-to-segment approach to multiple se- Valparaiso 12, 95–182. quence alignment, Bioinformatics 15, 211–218. Nixon K.C. (2002) WINCLADA v.1.00.08, Ward P.S., Brady S.G. (2003) Phylogeny and bio- computer program distributed by the author geography of the ant subfamily Myrmeciinae at http://www.cladistics.com/wincDownload.htm (Hymenoptera: Formicidae), Invertebr. Syst. 17, (accessed on 8 November 2007). 361–386.