Phylogenomics Resolves Ev

Current Biology 23, 2058–2062, October 21, 2013 ª2013 Elsevier Ltd All rights reserved http://dx.doi.org/10.1016/j.cub.2013.08.050

Reportolutionary

Relationships among Ants, Bees, and Wasps

Brian R. Johnson,1 Marek L. Borowiec,1 Joanna C. Chiu,1

Ernest K. Lee,2 Joel Atallah,1 and Philip S. Ward1,*1Department of Entomology and Nematology, University ofCalifornia, Davis, One Shields Avenue, Davis, CA 95616, USA2Sackler Institute for Comparative Genomics, AmericanMuseum of Natural History, Central Park West at 79th Street,New York, NY 10024, USA

Summary

Eusocial behavior has arisen in few animal groups, most

notably in the aculeate Hymenoptera, a clade comprisingants, bees, and stinging wasps [1–4]. Phylogeny is crucial

to understanding the evolution of the salient features ofthese insects, including eusociality [5]. Yet the phylogenetic

relationships among the major lineages of aculeate Hyme-noptera remain contentious [6–12].We address this problem

here by generating and analyzing genomic data for a repre-sentative series of taxa. We obtain a single well-resolved

and strongly supported tree, robust to multiple methods ofphylogenetic inference. Apoidea (spheciform wasps and

bees) and ants are sister groups, a novel finding that contra-dicts earlier views that ants are closer to ectoparasitoid

wasps. Vespid wasps (paper wasps, yellow jackets, and rel-atives) are sister to all other aculeates except chrysidoids.

Thus, all eusocial species of Hymenoptera are containedwithin twomajor groups, characterized by transport of larval

provisions and nest construction, likely prerequisites for the

evolution of eusociality. These two lineages are interpolatedamong three other clades of wasps whose species are pre-

dominantly ectoparasitoids on concealed hosts, the inferredancestral condition for aculeates [2]. This phylogeny pro-

vides a new framework for exploring the evolution of nest-ing, feeding, and social behavior within the stinging

Hymenoptera.

Results and Discussion

Aculeate (stinging) Hymenoptera are behaviorally diverse, en-compassing both solitary and eusocial species and exhibitinga variety of life history strategies including parasitoidism,predation, omnivory, and pollenivory [2, 13]. Multiple lines ofevidence provide strong support for the monophyly of theAculeata [9, 10], but relationships among the major lineageswithin this group have been a matter of continued uncertainty[7–9, 11, 12]. The position of ants, the most species-rich andecologically dominant of all eusocial insects, has been partic-ularly problematic [7–9, 11, 12] (Figure 1).

Advances in next-generation sequencing have unleashedthe potential of genomic data to clarify many previously intrac-table parts of the Tree of Life [14–16]. Here we addressed theproblem of aculeate Hymenoptera phylogeny by generatingtranscriptome data for ten representative species in nine fam-ilies and genomic data for one key taxon (Apterogyna) forwhich RNA was unavailable (Table 1; see also Table S1

*Correspondence: psward@ucdavis.edu

available online). We then combined these data with the pub-lished genome sequences of three bee species and three antspecies, a transcriptome from one additional bee species,and genomic data from Nasonia vitripennis, a nonaculeatehymenopteran used as an outgroup. Orthology identificationand matrix assembly was accomplished with the OrthologIDpipeline [17]. This yielded multiple partitioned amino acidmatrices, with different levels of gene representation acrossthe 18 ingroup taxa, ranging from a 5,214-gene matrix(3,001,657 amino acid sites) to a stringently filtered matrix of308 genes (175,404 sites) and only 14.98% missing data (seeExperimental Procedures and Supplemental ExperimentalProcedures).Table 1 shows the summary statistics for all the transcrip-

tomes and for the genomic assembly. For the transcriptomes,we identified a range of protein sequence numbers, with thescoliid wasp Crioscolia alcione having the largest transcrip-tome size and the sweat bee Lasioglossum albipes the small-est. Whether variation in the transcriptome sizes representsactual variation in the number of genes present in these spe-cies or whether it represents variation in the quality of theassemblies is uncertain. The genomic library of ApterogynaAZ01 (family Bradynobaenidae) had only 1,717 genes, with amedian amino acid length of 450. This reflects the relativelylow coverage of sequencing (183) and the apparent degrada-tion of this older sample. Nevertheless, the Apterogynasequence data were sufficient to reliably place this taxonwithin the hymenopteran tree.Phylogenetic analyses produced a fully resolved tree of the

aculeate Hymenoptera with robust support at all nodes (Fig-ure 2). The same tree topology and relative branch lengthswere obtained under a variety of analytical procedures,including partitioned maximum likelihood (ML) analyses andBayesian analyses of concatenated data sets, as well asspecies tree estimates (Figures 2 and S1). All nodes in thetopology have ML bootstrap support of 100% and Bayesianposterior probabilities of 1.0. Under species tree analyses,most nodes are also strongly supported, although supportvalues drop for some of the deeper nodes in the tree (Figures2 and S1). Most procedures employed the 308-gene matrix,but we also ran ML analyses of three other matrices of varyingsize and completeness (525, 3,018, and 5,214 genes, respec-tively), with the same results (Figures S1A–S1C).As expected [6] we found that the cuckoo wasp (Argochry-

sis) is sister to all other aculeates, and that ants, bees, apoids,and vespid wasps are all monophyletic. We recovered thevespid wasps (represented by a nonsocial pollen wasp, Pseu-domasaris, and a eusocial paper wasp, Mischocyttarus) assister to all aculeates except the cuckoo wasp, a result thatis in agreement with some other recent molecular studies[8, 11] although in strong conflict withmorphology-based trees[7] in which vespids are nestedwell within the aculeate phylog-eny, as sister to scoliid wasps (Figure 1A).Of particular interest is the finding that ants are sister to

Apoidea, a novel result that emphasizes a greater affinity ofants to the predatory wasps that characterize the earliestbranching lineages of Apoidea than to scoliids, bradynobae-nids, tiphiids, and other ectoparasitoid wasps with whichthey have been associated previously [1, 7, 9, 12]. This result

Figure 1. Previous Hypotheses of Phylogenetic Relationships among Ants,

Bees, and Stinging Wasps

Morphological analysis of Brothers [7] (A) and molecular phylogenetic

studies of Pilgrim et al. [8] (B) and Heraty et al. [9] (C). Major groups have

differently colored branches. Vespid wasps * = Vespidae + Rhopalosomati-

dae; scoliid wasps * = Scoliidae + Bradynobaenidae.

Genomic-Based Phylogeny of Ants, Bees, and Wasps2059

receives significant statistical support (p < 0.01) under the Shi-modaira-Hasegawa test [23] against five alternate phyloge-netic placements, including ants as sister to scolioid wasps,and ants as sister to scolioid wasps plus apoids (Table S3).Thus, morphologically generalized apoid wasps such as Am-pulicidae (cockroach wasps) and Sphecidae (digger wasps,mud dauber wasps, and relatives) may provide more insightinto the early evolution of stem-group ants than the ectopara-sitoid wasps that have previously served as models for theorigin of ants.

Our results also provide newperspective on the lower Creta-ceous fossil Cariridris bipetiolata, originally claimed to be theoldest fossil ant [24] but later reinterpreted to be a spheciformwasp, probably belonging to Ampulicidae [25]. Although thisappeared to represent a major reassignment of the fossil,our discovery that ants and apoids are sister taxa helps toexplain difficulty in the placement of Cariridris [26] and sug-gests that it is best treated as a lineage close to the root ofthe ant-apoid tree, perhaps not assignable with certainty toeither branch.

The sister group of [ants + Apoidea] is [Scoliidae + Bradyno-baenidae sensu stricto]. This more inclusive clade was alsorecovered in some other molecular studies [8, 11], but with a

different branching order, such that the scolioid lineage wasidentified as sister to Apoidea. Our new results motivate asearch for features in common between ants and apoids—not shared with scolioid wasps—that might predispose thisgroup toward the evolution of sociality. The most obviousbehavioral commonality is the collection and transport ofresources (arthropod prey or pollen) to a constructed nest, atrait also sharedwith vespidwasps, the other group containingeusocial species. Scoliid wasps, in contrast, are ectoparasi-toids on concealed hosts [2, 27]. (The life history of Bradyno-baenidae remains to be elucidated.) It has long been arguedthat nest construction and provisioning are key prerequisitesfor the evolution of eusociality [1, 2]. The finding that antsand apoids are sister taxa suggests that this favorable combi-nation of traits arose only once in their common ancestor,rather than separately from ectoparasitoid predecessors inthe ant and apoid lineages, emphasizing that the precondi-tions for eusociality are rare and contingent.Our phylogeny reveals a well-supported clade of tiphioid

and pompiloid wasps, to the exclusion of the scolioids. Thiscomports with an earlier molecular study [8], except that wefind that the tiphiid wasp (Brachycistis) and the chyphotinewasp (Chyphotes) are sister taxa, to the exclusion of pompi-loids (spider wasps and velvet ants), a result inconsistentwith previous findings. Further taxon sampling is neededwithin this clade to test the monophyly and placement oftiphiid wasps (Tiphiidae) and velvet ants (Mutillidae), but ourresults confirm an earlier inference [8, 11] that the family Bra-dynobaenidae is not monophyletic, with true bradynobaenids(represented in our data set by Apterogyna) being sister toScoliidae, whereas the subfamily Chyphotinae (representedhere by Chyphotes) is part of the tiphioid complex. Bradyno-baenid-like wasps share a number of morphological features[7], some unique, and these must be interpreted as examplesof convergence between two distantly related clades, perhapsgenerated in part by the independent loss of wings in femalesof both groups. It should be noted that scoliid females arewinged, as are some members of the tiphioid-pompiloidclade, so winged females are the ancestral condition for bothclades.The phylogenetic results presented here support the

following scenario of behavioral evolution in aculeateHymenoptera (Figure 3). The ancestral aculeate wasp wasan ectoparasitoid, attacking and paralyzing concealed hostsand leaving its offspring in or near the host cavity [2]. Intwo major lineages (ants + Apoidea, and Vespidae), thisbehavior became modified as wasps adopted a more activepredatory lifestyle, with increased importance of prey trans-port, nest construction, and parental care. More specializedfeeding habits (pollenivory) were acquired later. Eusocialbehavior evolved multiple times within both of these lineages[4, 28, 29]. The three remaining clades of aculeates (chrysi-doids, scolioids, and the tiphioid-pompiloid clade) havelargely retained ectoparasitoid habits, except for pompilids[30], and no examples of eusociality are known in thesegroups.This is the first comprehensive phylogenomic analysis of

aculeate Hymenoptera. It demonstrates the utility and feasi-bility of employing transcriptome data to resolve outstandingproblems in insect phylogeny. The new tree provides a robustframework for investigating the evolution of nesting, feeding,and social behavior within the stinging Hymenoptera, andfor exploring genomic signatures of changes in thesecharacteristics.

Table 1. Species of Hymenoptera Sampled, and Summary Statistics for the Transcriptome Assemblies and the Genome Assembly

Species Common Name Assembly Size (MB)

Number of Sequences

(>300 aa)

Mean Protein

Length (aa)

Apterogyna ZA01 (Bradynobaenidae) bradynobaenid wasp 284.2 1,717 450

Chyphotes mellipes (Bradynobaenidae) bradynobaenid wasp 137.8 26,513 698

Argochrysis armilla (Chrysididae) cuckoo wasp 91.6 18,777 690

Stigmatomma oregonense (Formicidae) dracula ant 74.8 13,867 721

Sphaeropthalma orestes (Mutillidae) velvet ant (wasp) 84.5 15,820 707

Pepsis grossa (Pompilidae) spider wasp 150.3 27,549 703

Crioscolia alcione (Scoliidae) scoliid wasp 219.2 45,155 689

Sceliphron caementarium (Sphecidae) mud dauber wasp 161.3 33,499 659

Brachycistis timberlakei (Tiphiidae) tiphiid wasp 84.3 15,036 627

Mischocyttarus flavitarsis (Vespidae) paper wasp 76.8 15,907 676

Pseudomasaris vespoides (Vespidae) pollen wasp 103.7 21,543 678

Lasioglossum albipes (Halictidae) sweat bee 46.7 9,417 585

The assembly for Apterogyna is a partial genome assembly; all others are transcriptome assemblies. In addition to the assemblies derived from new

sequence data generated in this study, we also assembled a transcriptome for the sweat bee Lasioglossum albipes based on raw paired-end short reads

downloaded from the NCBI Sequence Read Archive (SRR578269). MB, megabases; aa, amino acids. See also Tables S1 and S2.

Current Biology Vol 23 No 202060

Experimental Procedures

Taxon Sampling

Eleven species from key families across the aculeate Hymenoptera were

chosen for the generation of new phylogenomic data (Tables 1 and S1).

These represent all the major lineages of stinging Hymenoptera that have

been considered in previous hypotheses of phylogenetic relationships

[6–12]. Ten species were collected in the field, while one rare species,

Apterogyna ZA01, was available only as preserved specimens in ethanol.

We included this second representative of Bradynobaenidae because of

instability in the phylogenetic position of bradynobaenid wasps in previous

studies [7, 8, 11, 12]. We supplemented our data on these eleven species

with transcriptome data on one bee species (Lasioglossum albipes) from

the NCBI Sequence Read Archive and published genome assemblies of

three other bee species (Megachile rotundata, Bombus terrestris, and

Apis mellifera) and three ant species (Harpegnathos saltator, Linepithema

humile, and Pogonomyrmex barbatus).

Sequencing and Assembly

For the fresh collected samples, cDNA libraries were prepared, while for the

ethanol-preserved sample, a DNA library was prepared (further details in

Supplemental Information, including Table S2). Samples were pooled and

sequenced on an Illumina HiSeq 2000 (100 bp paired-end). Transcriptomes

were assembled using the Trinity software package [31], while ABySS was

used for the genome assembly [32]. After translation of contigs into amino

acid sequences, orthology was evaluated using a prerelease version 2.0

of the OrthologID pipeline [17]. OrthologID takes complete gene sets from

all taxa and assigns them into gene clusters. It then generates a parsimony

tree for each gene cluster and extracts one or more sets of orthologous

genes. Orthologous sets of genes were then assembled into multiple parti-

tioned matrices with different levels of taxon representation per gene,

including (1) a 5,214-genematrix with 3,001,657 amino acid sites and at least

9 ingroup taxa represented per gene partition, (2) a 3,018-gene matrix with

1,653,740 sites and at least 16 ingroup taxa represented per partition, and

(3) a matrix that has every gene partition represented across all ingroup

taxa with the additional requirement that they be single-copy in five publicly

available ingroup genomes (525 partitions and 298,968 sites). For Bayesian

inference and species tree estimation, a fourth, smaller 308-gene matrix

with 175,404 sites was used. Of the 3,332,676 cells (19 taxa3 175,404 sites)

in this 308-genematrix, 73.42%are coded as amino acids, 11.60%are gaps,

and 14.98% are missing.

Phylogenetic Analysis

For all four matrices, we performed partitioned (by gene) ML analyses with

G-distributed rate heterogeneity over sites using RAxML v7.4.2 [18, 19]. The

best protein substitutionmodel for each gene partition was selected individ-

ually using the ‘‘ProteinModelSelection.pl’’ script [19] over 36 different

models. For Bayesian inference, we used PhyloBayes MPI v1.3b [20, 21],

Figure 2. Maximum-Likelihood Tree of Aculeate

Hymenoptera Derived from a 308-Gene Matrix

This tree was estimated with RAxML [18, 19]

based on a 308-gene matrix with 175,404 amino

acid sites. Support values are given in the

following order: (1) posterior probabilities from a

separate Bayesian analysis with PhyloBayes

[20, 21], (2) RAxML bootstrap percentages based

on 1,000 replicates, and (3) bootstrap percent-

ages from a separate species tree analysis with

STAR [22]. Unlabeled nodes have maximum

support values (1/100/100). Scale bar indicates

number of substitutions per site. From the top

of the tree downward, species are as follows:

N. vitripennis (Pteromalidae), A. armilla (Chrysidi-

dae), P. vespoides (Vespidae), M. flavitarsus

(Vespidae), B. timberlakei (Tiphiidae),

C. mellipes (Bradynobaenidae), P. grossa (Pom-

pilidae), S. orestes (Mutillidae), Apterogyna

ZA01 (Bradynobaenidae), C. alcione (Scoliidae),

P. barbatus (Formicidae), L. humile (Formicidae),

S. oregonense (Formicidae), H. saltator (Formici-

dae), S. caementarium (Sphecidae), L. albipes

(Halictidae), M. rotundata (Megachilidae),

A. mellifera (Apidae), and B. terrestris (Apidae).

Major lineages are color coded using the same

scheme as in Figure 1. See also Figure S1.

Figure 3. Evolution of the Aculeate Hymenoptera

Blue-green branches represent parasitoidism;

orange branches represent nest construction

and predation (with pollenivory and omnivory as

derivative states thereof). Asterisks designate lin-

eages containing eusocial species. Ants are

entirely eusocial, but this is not true of all species

of Vespidae and Apoidea. Biological information

is from various sources, summarized in Gauld

and Bolton [2] and Huber [13]. Names of super-

families aremodified fromPilgrim et al. [8]. Place-

ment of Rhopalosomatidae is based on Pilgrim

et al. [8] and Debevec et al. [11]. Images courtesy

of Alexander Wild and Kurt Schaefer.

Genomic-Based Phylogeny of Ants, Bees, and Wasps2061

with CAT-GTR as the amino acid replacement model, in an unpartitioned

analysis of the 308-gene matrix. A species tree was estimated on the basis

of average ranks of gene coalescence events, as calculated in STAR [22].

The input for this analysis was 100 bootstrap replicate trees of each of

308 genes, built under maximum likelihood in RAxML. We also inferred a

species tree with PhyloNet [33], which uses the parsimony-based criterion

of minimizing deep coalescences [34]. We used 308 input trees with boot-

strap support values generated in RAxML.

To evaluate alternate phylogenetic hypotheses against our best-scoring

ML tree, we employed the Shimodaira-Hasegawa test [23]. Five constraints

were considered (Table S3), and separate constrained partitioned analyses

were conducted using RAxML on the same 308-gene matrix used to

generate our ML tree (Figure 2). The five best trees satisfying the respective

constraints were then subjected to the Shimodaira-Hasegawa test in

RAxML (‘‘-f h’’ option) to determine whether they were significantly worse

than our best unconstrained ML tree.

Accession Numbers

Illumina reads have been deposited in the NCBI Sequence Read Archive

with the accession number SRP020476. The matrices, partition files, and

gene trees have been deposited in Dryad (http://doi.org/10.5061/dryad.

jt440).

Supplemental Information

Supplemental Information includes one figure, three tables, and Supple-

mental Experimental Procedures and can be found with this article online

at http://dx.doi.org/10.1016/j.cub.2013.08.050.

Acknowledgments

This work was funded by the University of California, Davis.We thank James

Pitts for provision of the Apterogyna (Bradynobaenidae) specimens and

three anonymous reviewers for helpful comments that improved the

manuscript.

Received: June 26, 2013

Revised: August 1, 2013

Accepted: August 21, 2013

Published: October 3, 2013

References

1. Wilson, E.O. (1971). The Insect Societies (Cambridge: Harvard

University Press).

2. Gauld, I.D., and Bolton, B., eds. (1988). The Hymenoptera (London:

Oxford University Press).

3. Holldobler, B., and Wilson, E.O. (1990). The Ants (Cambridge: Harvard

University Press).

4. Bradley, T.J., Briscoe, A.D., Brady, S.G., Contreras, H.L., Danforth, B.N.,

Dudley, R., Grimaldi, D., Harrison, J.F., Kaiser, J.A., Merlin, C., et al.

(2009). Episodes in insect evolution. Integr. Comp. Biol. 49, 590–606.

5. Hughes, W.O.H., Oldroyd, B.P., Beekman, M., and Ratnieks, F.L.W.

(2008). Ancestral monogamy shows kin selection is key to the evolution

of eusociality. Science 320, 1213–1216.

6. Ronquist, F. (1999). Phylogeny of the Hymenoptera (Insecta): the state

of the art. Zool. Scr. 28, 3–12.

7. Brothers, D.J. (1999). Phylogeny and evolution of wasps, ants and bees

(Hymenoptera, Chrysidoidea, Vespoidea and Apoidea). Zool. Scr. 28,

233–249.

8. Pilgrim, E.M., von Dohlen, C.D., and Pitts, J.P. (2008). Molecular phylo-

genetics of Vespoidea indicate paraphyly of the superfamily and novel

relationships of its component families and subfamilies. Zool. Scr. 37,

539–560.

9. Heraty, J., Ronquist, F., Carpenter, J.M., Hawks, D., Schulmeister, S.,

Dowling, A.P., Murray, D., Munro, J., Wheeler, W.C., Schiff, N., and

Sharkey, M. (2011). Evolution of the hymenopteran megaradiation.

Mol. Phylogenet. Evol. 60, 73–88.

10. Sharkey, M.J., Carpenter, J.M., Vilhelmsen, L., Heraty, J., Liljeblad, J.,

Dowling, A.P.G., Schulmeister, S., Murray, D., Deans, A.R., Ronquist,

F., et al. (2012). Phylogenetic relationships among superfamilies of

Hymenoptera. Cladistics 28, 80–112.

11. Debevec, A.H., Cardinal, S., and Danforth, B.N. (2012). Identifying the

sister group to the bees: a molecular phylogeny of Aculeata with an

emphasis on the superfamily Apoidea. Zool. Scr. 41, 527–535.

12. Wilson, J.S., von Dohlen, C.D., Forister, M.L., and Pitts, J.P. (2013).

Family-level divergences in the stinging wasps (Hymenoptera:

Aculeata), with correlations to angiosperm diversification. Evol. Biol.

40, 101–107.

13. Huber, J.T. (2009). Biodiversity of Hymenoptera. In Insect Biodiversity:

Science and Society, R. Footit and P. Adler, eds. (Oxford: Wiley-

Blackwell), pp. 303–323.

14. Hedin, M., Starrett, J., Akhter, S., Schonhofer, A.L., and Shultz, J.W.

(2012). Phylogenomic resolution of paleozoic divergences in harvest-

men (Arachnida, Opiliones) via analysis of next-generation transcrip-

tome data. PLoS ONE 7, e42888.

15. Smith, S.A., Wilson, N.G., Goetz, F.E., Feehery, C., Andrade, S.C.S.,

Rouse, G.W., Giribet, G., and Dunn, C.W. (2011). Resolving the evolu-

tionary relationships of molluscs with phylogenomic tools. Nature 480,

364–367.

Current Biology Vol 23 No 202062

16. Oakley, T.H., Wolfe, J.M., Lindgren, A.R., and Zaharoff, A.K. (2013).

Phylotranscriptomics to bring the understudied into the fold: monophy-

letic ostracoda, fossil placement, and pancrustacean phylogeny. Mol.

Biol. Evol. 30, 215–233.

17. Chiu, J.C., Lee, E.K., Egan, M.G., Sarkar, I.N., Coruzzi, G.M., and

DeSalle, R. (2006). OrthologID: automation of genome-scale

ortholog identification within a parsimony framework. Bioinformatics

22, 699–707.

18. Stamatakis, A. (2006). RAxML-VI-HPC: maximum likelihood-based

phylogenetic analyses with thousands of taxa and mixed models.

Bioinformatics 22, 2688–2690.

19. Stamatakis, A. (2012). RAxML GitHub repository. https://github.com/

stamatak/standard-RAxML/.

20. Lartillot, N., Lepage, T., and Blanquart, S. (2009). PhyloBayes 3: a

Bayesian software package for phylogenetic reconstruction andmolec-

ular dating. Bioinformatics 25, 2286–2288.

21. Lartillot, N., Rodrigue, N., Stubbs, D., and Richer, J. (2013). PhyloBayes

MPI: phylogenetic reconstruction with infinite mixtures of profiles in a

parallel environment. Syst. Biol. 62, 611–615.

22. Liu, L., Yu, L.L., Pearl, D.K., and Edwards, S.V. (2009). Estimating

species phylogenies using coalescence times among sequences.

Syst. Biol. 58, 468–477.

23. Shimodaira, H., and Hasegawa, M. (1999). Multiple comparisons of log-

likelihoods with applications to phylogenetic inference. Mol. Biol. Evol.

16, 1114–1116.

24. Brandao, C.R.F., Martins-Neto, R.G., and Vulcano, M.A. (1990). The

earliest known fossil ant (first southern hemisphere Mesozoic record)

(Hymenoptera: Formicidae: Myrmeciinae). Psyche (Camb. Mass.) 96,

195–208.

25. Dlussky, G.M., and Rasnitsyn, A.P. (2003). Ants (Hymenoptera:

Formicidae) of Formation Green River and some other Middle Eocene

deposits of North America. Russ. Entomol. J. 11, 411–436.

26. Ohl, M. (2004). The first fossil representative of the wasp genus

Dolichurus, with a review of fossil Ampulicidae (Hymenoptera:

Apoidea). J. Kans. Entomol. Soc. 77, 332–342.

27. Clausen, C.P. (1940). Entomophagous Insects (New York: McGraw-Hill).

28. Hines, H.M., Hunt, J.H., O’Connor, T.K., Gillespie, J.J., and Cameron,

S.A. (2007). Multigene phylogeny reveals eusociality evolved twice in

vespid wasps. Proc. Natl. Acad. Sci. USA 104, 3295–3299.

29. Cardinal, S., and Danforth, B.N. (2011). The antiquity and evolutionary

history of social behavior in bees. PLoS ONE 6, e21086.

30. Evans, H.E., and Shimizu, A. (1996). The evolution of nest building and

communal nesting in Ageniellini (Insecta: Hymenoptera: Pompilidae).

J. Nat. Hist. 30, 1633–1648.

31. Grabherr, M.G., Haas, B.J., Yassour, M., Levin, J.Z., Thompson, D.A.,

Amit, I., Adiconis, X., Fan, L., Raychowdhury, R., Zeng, Q., et al.

(2011). Full-length transcriptome assembly from RNA-Seq data without

a reference genome. Nat. Biotechnol. 29, 644–652.

32. Simpson, J.T., Wong, K., Jackman, S.D., Schein, J.E., Jones, S.J., and

Birol, I. (2009). ABySS: a parallel assembler for short read sequence

data. Genome Res. 19, 1117–1123.

33. Than, C., Ruths, D., and Nakhleh, L. (2008). PhyloNet: a software pack-

age for analyzing and reconstructing reticulate evolutionary relation-

ships. BMC Bioinformatics 9, 322.

34. Maddison, W.P. (1997). Gene trees in species trees. Syst. Biol. 46,

523–536.

Current Biology, Volume 23

Supplemental Information

Phylogenomics Resolves Evolutionary

Relationships among Ants, Bees, and Wasps

Brian R. Johnson, Marek L. Borowiec, Joanna C. Chiu, Ernest K. Lee, Joel Atallah,

and Philip S. Ward

Supplemental Inventory

Figure S1. Aculeate Hymenoptera phylogeny inferred under different phylogenetic methods,

related to Figure 2

Table S1. Species of aculeate Hymenoptera sampled: collection data and composition of pooled

material from which RNA or DNA was extracted

Table S2. Number of raw and quality-controlled reads for each library

Table S3. Evaluation of alternative tree topologies using the SH-test as implemented in RAxML

Supplemental Experimental Procedures

Supplemental References

Figure S1.

Figure S1. Aculeate Hymenoptera Phylogeny Inferred under Different Phylogenetic

Methods, Related to Figure 2

(A-C) Maximum likelihood (ML) trees derived from 5,214-gene matrix, 3,018-gene matrix, and

525-gene matrix, respectively, estimated with RAxML. All nodes have maximum (100%)

bootstrap support. Scale bar indicates substitutions per site.

(D) Majority rule consensus tree from Bayesian analysis of 308-gene matrix, estimated with

PhyloBayes. All nodes have 1.00 posterior probability. Scale bar indicates substitutions per site.

(E) Species tree resulting from STAR analysis. Only bootstrap percentages less than 100% are

indicated. “Branch lengths” are average ranks of coalescences, not estimated substitutions per

site.

(F) Species tree (cladogram) resulting from PhyloNet analysis. Only bootstrap percentages less

than 100% are indicated.

Table S1. Species of Aculeate Hymenoptera Sampled: Collection Data and Composition of

Pooled Material from which RNA Was Extracted (DNA for Apterogyna)

Species

Voucher

Specimen Code

Date of

Collection GPS Coordinates Composition

Apterogyna ZA01 CASENT0106304 Oct 2004 -33.43 22.25 2 males

Chyphotes mellipes CASENT0106316 3-Jul-2012 39.29020 -118.41550 2 males

Argochrysis armilla CASENT0106319 14-Jul-2012 39.42435 -120.23679 3 females

Stigmatomma oregonense CASENT0106313 21-Jun-2012 39.99652 -120.99261 2 workers, 6 larvae

Sphaeropthalma orestes CASENT0106318 10-Jul-2012 39.77403 -120.07391 2 males

Pepsis grossa CASENT0106314 2-Jul-2012 39.28784 -119.27530 1 female, 1 male

Crioscolia alcione CASENT0106317 8-Jul-2012 39.84368 -119.43436 1 male

Sceliphron caementarium CASENT0106321 26-Jul-2012 38.54030 -121.75628 1 female, 1 male

Brachycistis timberlakei CASENT0106315 2-Jul-2012 39.28899 -119.27316 3 males

Mischocyttarus flavitarsis CASENT0106320 26-Jul-2012 38.54030 -121.75628 1 female, 1 male

Pseudomasaris vespoides CASENT0106312 19-Jun-2012 39.31914 -120.65916 1 female, 1 male

Each voucher specimen code is linked to more detailed collection data on AntWeb

(www.antweb.org). Voucher specimens have been deposited in the Bohart Museum of

Entomology (UCDC).

Table S2. Number of Raw and Quality-Controlled Reads for Each Library

Species Raw Reads Filtered Reads 1 Filtered Reads 2

Apterogyna ZA01 27,243,036 25,293,951 25,293,951

Chyphotes mellipes 14,529,464 14,362,151 13,744,283

Argochrysis armilla 17,264,088 17,074,979 16,346,892

Stigmatomma oregonense 13,648,445 13,457,982 12,773,283

Sphaeropthalma orestes 14,493,142 14,307,149 13,714,095

Pepsis grossa 18,098,742 17,860,027 17,023,108

Crioscolia alcione 14,585,832 14,388,055 13,707,012

Sceliphron caementarium 18,861,014 18,594,250 17,615,417

Brachycistis timberlakei 13,463,088 13,295,450 12,714,614

Mischocyttarus flavitarsis 16,910,898 16,742,131 16,148,909

Pseudomasaris vespoides 16,675,940 16,502,109 15,858,504

Lasioglossum albipes 19,412,393 19,388,645 19,388,645

“Filtered reads” refers to the number of reads remaining after quality control using fast-toolkit

and cutadapt. Reads 1 and 2 refer to forward and reverse reads, respectively.

Table S3. Evaluation of Alternative Tree Topologies Using the SH Test as Implemented

in RAxML

Constraint Tree Likelihood1 D(LH) Significantly Worse (p < 0.01)

(ants, all aculeates except Argochrysis) -2515334.7 -753.3 ± 97.1 Yes

(ants, scolioids + apoids) -2514722.3 -140.9 ± 47.3 Yes

(ants, scolioids) -2514758.4 -176.9 ± 46.9 Yes

(ants, tiphioid-pompiloid wasps) -2515486.6 -905.2 ± 83.1 Yes

(ants, vespids) -2515424.9 -843.4 ± 97.2 Yes

D(LH) is the difference in log likelihood units between the best constrained tree and the best

unconstrained tree. 1 Likelihood Score of best ML tree (Figure 2) = -2514581.4.

Supplemental Experimental Procedures

Sampling of Taxa

Samples of ten species of aculeate Hymenoptera were collected at field sites in California and

Nevada, United States, in June and July 2012 (Table S1). Taxa were targeted to cover all major

families of Aculeata. After collection, live insects were immobilized by freezing, cut into several

pieces, briefly dipped in 95% ethanol to reduce the hydrophobic quality of the cuticle, and then

immersed in RNAlater. Specimens were kept at 4°C for 24 hours to encourage perfusion of

RNAlater, and then stored at -20°C until extraction. The sample of Apterogyna ZA01

(Bradynobaenidae) represents a special case, since fresh collection of this rare African insect was

not possible. We therefore extracted DNA from two male specimens collected in 2004 and

preserved in 95% ethanol.

Library Construction, Sequencing, and Quality Control

For all samples, one to several pooled specimens were used for RNA or DNA extraction. Table

S1 shows the number and sex of each individual from each species used in the study. For

samples for which transcriptomes were to be generated, total RNA was extracted using Trizol

according to the manufacturer’s instructions. RNA quality and integrity were checked with the

Nanodrop 1000 and Bioanalyzer 2100. Total RNA was used for library prep using the Illumina

TruSeq v2 kit according to the manufacturer’s instructions. For the Apterogyna ZA01

(Bradynobaenidae) DNA library, DNA was extracted using the ChargeSwitch® gDNA Mini

Tissue Kit according to the manufacturer’s instructions. The additional RNAase step was

conducted to remove RNA. DNA concentration was determined using Qubit, and an Illumina

genomic library was then made using the Illumina TruSeq v2 kit according the manufacturer’s

instructions. Both cDNA and DNA libraries were checked for integrity with the Bioanalyzer

2100 before being pooled into one lane on the Illumina HiSeq 2000. Each library had a unique

Illumina barcode. Because genome assembly requires more reads than transcriptome assembly,

we biased the pooling such that we obtained more reads from the genomic library. 100 bp paired-

end sequencing was conducted. Low quality bases and adapter contamination were removed

from reads using the fastx toolkit and the cutadapt software package [S1].

Table S2 shows the number of raw and processed reads for each library. In addition to the

new genomic and transcriptomic data sets generated, raw Illumina paired-end short reads from

whole body extractions were downloaded from the NCBI SRA archive (SRR578269) for the

sweat bee, Lasioglossum albipes (family Halictidae). This raw data was quality controlled using

the same procedure as the newly generated short reads and was treated identically with respect to

transcriptome assembly. Protein sequences of the leafcutter bee Megachile rotundata (primarily

from the Megachile genome study) were downloaded from NCBI.

Transcriptome Assembly

De novo transcriptomes were generated with the Trinity short read assembler [S2]. For each

assembly, the “transcripts_to_best_scoring_ORFs.pl” script, included with the Trinity software,

was used to identify the longest open reading frame in each contig and translate it into a protein

sequence. Further identification of unigenes was not necessary as the OrthologID pipeline (see

below) automatically estimates the single best ortholog for each gene and discards other

sequences.

Genome Assembly: Apterogyna

The rationale for this portion of the study was that a complete genome of Apterogyna AZ01 was

unnecessary for the phylogenetic problem at hand. Rather we required a sufficiently

comprehensive and phylogenetically informative assembly of protein coding genes. Hence, while

short reads alone are typically insufficient for a complete genome assembly, we hypothesized

that they would be sufficient to assemble enough protein coding sequence to allow for a rigorous

phylogenomic placement of this key taxon.

The genome assembly was performed with the ABYSS software package v1.3.5 [S3].

The resulting assembly was 284.2 MB with an N50 of 1,065 bp and a maximum contig length of

9,143. The assembly is highly fragmented due to low coverage (for a short read dataset) and the

likely degradation of the sample, which was collected in ethyl alcohol eight years ago. The result

is essentially a “gene assembly” more than a complete genome. For this reason, we used the

“transcripts_to_best_scoring_ORFs.pl” script (from the Trinity assembly package) to identify the

longest ORF in each contig and translate it into a protein sequence. A more formal genome

bioinformatics process was not pursued, as our downstream bioinformatics analyses are designed

to identify orthologs from each species and discard chimeric genes.

Orthology Estimation and Data Matrix Assembly

Gene orthology was evaluated using a pre-release version 2.0 of the OrthologID pipeline. Similar

to the original version [S4], the latest version of OrthologID takes complete gene sets from all

ingroup and outgroup taxa as input and assigns them into gene clusters. OrthologID then

generates a parsimony tree for each gene cluster and extracts one or more sets of orthologous

genes according to the gene tree topology. In addition to improved execution pipeline on Sun

Grid Engine clusters, this version of OrthologID uses the MCL algorithm [S5, S6] for improved

clustering, and includes automated extraction of orthologs from gene trees into a partitioned

matrix in a single package. Using gene sets from our 19 species as input, OrthologID recovered

19,746 sets of orthologs from 11,119 gene clusters with at least one gene from 5 publicly

available genomes (Apis mellifera, Bombus terrestris, Harpegnathos saltator, Linepithema

humile, and Pogonomyrmex barbatus) in each cluster. For phylogenetic analysis, these

orthologous sets of genes or gene fragments were then assembled by OrthologID into multiple

partitioned matrices with different levels of taxon representation per gene, including (i) a 5,214-

gene matrix with 3,001,657 amino acid sites and at least 9 ingroup taxa represented per gene

partition, (ii) a 3,018-gene matrix with 1,653,740 sites and at least 16 ingroup taxa represented

per partition, and (iii) a matrix that has every gene partition represented across all ingroup taxa

with the additional requirement that they are single-copy in the 5 publicly available ingroup

genomes. This matrix has a total of 525 partitions and 298,968 sites. To further reduce this

matrix to allow for the most computationally intensive analysis, i.e. Bayesian inference, we

extracted a submatrix by examining the maximum likelihood tree estimate of each of the 525

partitions, calculating the total tree length and branch length variance for each partition and

choosing only the gene partitions that lay within the 68th percentile in both categories. We used

the 68th percentile because this reduced the matrix to a manageable size (~300 genes) for

Bayesian analysis. The resultant 308-gene matrix has 175,404 sites. Of the 3,332,676 cells (19

taxa × 175,404 sites) in this 308-gene matrix, 73.42% are coded as amino acids, 11.60% are

gaps, and 14.98% are missing. Equivalent percentages of missing data in the 5,214-gene, 3,018-

gene and 525-gene matrices are 36.77%, 21.53% and 15.16%, respectively.

Maximum Likelihood Analysis

We performed maximum likelihood (ML) analysis on the different matrices described above.

Partitioned analyses with -distributed rate heterogeneity over sites were performed using

RAxML v7.4.2 [S7, S8]. For each matrix, the best protein substitution model for each gene

partition was selected individually using the “ProteinModelSelection.pl” script [S8] over 36

different models. We used the MPI-AVX version of RAxML to perform rapid bootstrap on each

of the matrices, and the PTHREADS-AVX version to search for the best scoring trees. At least

250 bootstrap replicates were computed for each matrix. For the 308-gene matrix, 1,000

bootstrap replicates were performed. Identical topologies with bootstrap value of 100 at every

node were recovered for all 4 matrices with very small differences in relative branch lengths

(Figures 2, S1A-C), except for Apterogyna ZA01, which has a shorter relative branch length for

the smaller but more completely represented matrices.

Bayesian Inference

Bayesian inference was performed on the 308-gene matrix using PhyloBayes-MPI v1.3b [S9,

S10]. CAT-GTR is recommended in Lartillot et al. [S9] as the best overall model implemented

in PhyloBayes for large data sets and it was therefore chosen as the amino acid replacement

model for our analysis. We executed two independent chains of at least 7,000 cycles on 128 CPU

cores each. Convergence was determined by visually examining the trace plots of the

PhyloBayes summary statistics using the mcmcplots R library [S11]. Using the bpcomp program

of PhyloBayes on the two chains with a burn-in of 1,500 returned maxdiff and meandiff of 0,

indicating that no discrepancies were observed across all bipartitions after convergence. The

consensus of all trees pooled across both chains after burn-in was identical to the ML tree in

topology, with posterior probabilities of 1.0 across all nodes (Figure S1D).

Species Tree Inference

Although model-based methods of species tree estimation have been shown to outperform other

approaches, their computational costs are prohibitive for genomic-scale data sets [S12, S13]. For

that reason, we estimated the species tree with average ranks of gene coalescence events in

STAR [S14]. As input we used 100 bootstrap replicate trees of each of 308 genes from the small

matrix, built under maximum likelihood in RAxML. The analyses were carried out on the

STRAW web server [S15].

In addition we inferred a species tree in PhyloNet [S16], which uses the parsimony-based

criterion of minimizing deep coalescences [S17]. We used 308 input trees with bootstrap support

values generated in RAxML. PhyloNet allows for accounting of uncertainty in gene trees and

performs better when collapsing poorly supported splits [S16]. We thus set an arbitrary cutoff

value of 70, contracting all branches with lower bootstrap support and ran the analysis over

10,000 replicates. The resulting cladogram is presented in Figure S1F.

Supplemental References

S1. Martin, M. (2011) Cutadapt removes adapter sequences from high-throughput sequencing

reads. EMBnet journal 17, 10–12.

S2. Grabherr, M.G., Haas, B.J., Yassour, M., Levin, J.Z., Thompson, D.A., Amit, I.,

Adiconis, X., Fan, L., Raychowdhury, R., Zeng, Q., et al. (2011) Full-length

transcriptome assembly from RNA-Seq data without a reference genome. Nat. Biotech.

29, 644–652.

S3. Simpson, J.T., Wong, K., Jackman, S.D., Schein, J.E., Jones, S.J., and Birol, I. (2009)

ABySS: A parallel assembler for short read sequence data. Genome Res. 19, 1117–1123.

S4. Chiu, J.C., Lee, E.K., Egan, M.G., Sarkar, I.N., Coruzzi, G.M., and DeSalle, R. (2006)

OrthologID: automation of genome-scale ortholog identification within a parsimony

framework Bioinformatics 22, 699–707.

S5. Enright, A.J., Van Dongen, S., and Ouzounis, C.A. (2002) An efficient algorithm for

large-scale detection of protein families. Nucleic Acids Res. 30, 1575–1584.

S6. Van Dongen, S. (2008) Graph clustering via a discrete uncoupling process. SIAM J.

Matrix Anal. Appl. 30, 121–141.

S7. Stamatakis, A. (2006) RAxML-VI-HPC: Maximum likelihood-based phylogenetic

analyses with thousands of taxa and mixed models. Bioinformatics 22, 2688–2690.

S8. Stamatakis, A. (2012) RAxML GitHub repository, https://github.com/stamatak/standard-

RAxML/.

S9. Lartillot, N., Lepage, T., and Blanquart, S. (2009) PhyloBayes 3: a Bayesian software

package for phylogenetic reconstruction and molecular dating. Bioinformatics 25, 2286–

2288.

S10. Lartillot, N., Rodrigue, N., Stubbs, D., and Richer, J. (2013) PhyloBayes MPI:

Phylogenetic reconstruction with infinite mixtures of profiles in a parallel environment.

Syst. Biol. 62, 611-615.

S11. Curtis, S. M. (2012) R package - mcmcplots v0.4.1

S12. Yang, J., and Warnow, T. (2011) Fast and accurate methods for phylogenomic analyses.

BMC Bioinformatics 12, S4. doi:10.1186/1471-2105-12-S9-S4.

S13. Whelan, N.V. (2011) Species tree inference in the age of genomics. Trends Evol. Biol. 3,

e5. doi:10.4081/eb.2011.e5.

S14. Liu, L., Yu, L.L., Pearl, D.K., and Edwards, S.V. (2009) Estimating species phylogenies

using coalescence times among sequences. Syst. Biol. 58, 468–477.

S15. Shaw, T. I., Ruan, Z., Glenn, T. C., and Liu, L. (2013). STRAW: Species TRee Analysis

Web server. Nucleic Acids Res. Advance Access, doi:10.1093/nar/gkt377

S16. Than, C., Ruths, D., and Nakhleh, L. (2008) PhyloNet: a software package for analyzing

and reconstructing reticulate evolutionary relationships. BMC Bioinformatics 9, 322.

doi:10.1186/1471-2105-9-322.

S17. Maddison, W.P. (1997) Gene trees in species trees. Syst. Biol. 46, 523–536.

Current Biology 23, 2565, December 16, 2013 ª2013 Elsevier Ltd All rights reserved

Phylogenomics Resolves Evolutionary

Erratum

Relationships among Ants, Bees, and Wasps

Brian R. Johnson, Marek L. Borowiec, Joanna C. Chiu, Ernest K. Lee, Joel Atallah, and Philip S. Ward*

(Current Biology 23, 2058–2062; October 21, 2013)In our recent paper on the phylogeny of aculeate Hymenoptera, we supplemented our primary data set with publicly availablegenome sequence data from four bee species, three ant species, and one wasp species. However, we did not give detailedcitations for these data, nor did we make a clear distinction between those species whose genomes had been formally pub-lished and those for which data had been made publicly available prior to publication. The published genomes were those ofthe wasp Nasonia vitripennis [1], the honeybee Apis mellifera [2], and the ants Harpegnathos saltator [3], Pogonomyrmexbarbatus [4], and Linepithema humile [5]. The prepublication genome data came from the bees Lasioglossum albipes (NCBISequence Read Archive SRR578269, as part of the Lasioglossum albipes WGS project, http://www.ncbi.nlm.nih.gov/bioproject/174755),Megachile rotundata (NCBI Protein database search, with data coming mostly from theMegachile genomesequencing project, http://www.ncbi.nlm.nih.gov/bioproject/66515), and Bombus terrestris (protein set from NCBI RefSeq andGenome Annotation projects, derived from genomic sequence generated by the Bumble Bee Genome Project, https://www.hgsc.bcm.edu/arthropods/bumble-bee-genome-project).

We apologize for any confusion created by the lack of explicit citations of these data sources.

References

1. Werren, J.H., Richards, S., Desjardins, C.A., Niehuis, O., Gadau, J., Colbourne, J.K., Werren, J.H., Richards, S., Desjardins, C.A., Niehuis, O., et al.;

Nasonia Genome Working Group (2010). Functional and evolutionary insights from the genomes of three parasitoid Nasonia species. Science 327,

343–348.

2. Honeybee Genome Sequencing Consortium (2006). Insights into social insects from the genome of the honeybee Apis mellifera. Nature 443, 931–949.

3. Bonasio, R., Zhang, G., Ye, C., Mutti, N.S., Fang, X., Qin, N., Donahue, G., Yang, P., Li, Q., Li, C., et al. (2010). Genomic comparison of the ants

Camponotus floridanus and Harpegnathos saltator. Science 329, 1068–1071.

4. Smith, C.R., Smith, C.D., Robertson, H.M., Helmkampf, M., Zimin, A., Yandell, M., Holt, C., Hu, H., Abouheif, E., Benton, R., et al. (2011). Draft genome of

the red harvester ant Pogonomyrmex barbatus. Proc. Natl. Acad. Sci. USA 108, 5667–5672.

5. Smith, C.D., Zimin, A., Holt, C., Abouheif, E., Benton, R., Cash, E., Croset, V., Currie, C.R., Elhaik, E., Elsik, C.G., et al. (2011). Draft genome of the

globally widespread and invasive Argentine ant (Linepithema humile). Proc. Natl. Acad. Sci. USA 108, 5673–5678.

*Correspondence: psward@ucdavis.edu

http://dx.doi.org/10.1016/j.cub.2013.11.056