<I>Centris</I> and <I>Epicharis</I>

ORIGINAL ARTICLE

doi:10.1111/evo.12689

Gain and loss of specialization in two oil-bee lineages, Centris and Epicharis (Apidae)

Aline C. Martins,1,2 GabrielA.R.Melo,2 and Susanne S. Renner1,3 1Department of Biology, University of Munich, 80638, Munich, Germany 2Department of Zoology, Federal University of Parana,´ PB 19020, Curitiba, PR 81531-980, Brazil 3E-mail: [email protected]

Received October 17, 2014 Accepted May 21, 2015

It is plausible that specialized ecological interactions constrain geographic ranges. We address this question in neotropical bees, Centris and Epicharis, that collect oils from flowers of Calceolariaceae, Iridaceae, Krameriaceae, Malpighiaceae, Plantaginaceae, or Solanaceae, with different species exploiting between one and five of these families, which either have epithelial oil glands or hair fields. We plotted the level of oil-host specialization on a clock-dated phylogeny for 22 of the 35 species of Epicharis and 72 of the 230 species of Centris (genera that are not sister genera) and calculated geographic ranges (km2) for 23 bee species based on collection data from museum specimens. Of the oil-offering plants, the Malpighiaceae date to the Upper Cretaceous, whereas the other five families are progressively younger. The stem and crown groups of the two bee genera date to the Cretaceous, Eocene, and Oligocene. Shifts between oil hosts from different families are common in Centris, but absent in Epicharis, and the direction is from flowers with epithelial oil glands to flowers with oil hairs, canalized by bees’ oil-collecting apparatuses, suitable for piercing epithelia or mopping oil from hair fields. With the current data, a link between host specialization and geographic range size could not be detected.

KEY WORDS: Bees, geographic ranges, molecular clock dating, oil flowers, phylogenetics, plant–insect interactions.

Individual phytophagous insect species rarely use the full range and fatty oil. The oil is produced by epithelial glands or oil hairs of their food plants, and the relationship between the range sizes found in the flowers of 2000 species in well over 100 genera of insects and their food plants is not a simple linear one (Stewart in 11 families on all continents except Antarctica (Vogel 1974, et al. 2015). Range expansion may be facilitated by temporary 1988; Rasmussen and Olesen 2000; Machado 2004; Renner and escape from natural enemies, by the exploitation of novel food Schaefer 2010). In the New World, the oil bee/oil plant “system” plants that permit a broader set of habitats to be used, or by cli- extends from the southern United States to Patagonia and involves mate change, making changes in distributions difficult to attribute seven plant families and four groups of bees, Centris (Fig. 1A– to particular factors. A testable expectation in this context is that C), Epicharis, Tapinotaspidini, and Tetrapedia (Rasmussen and foraging specialization should limit range expansions and should Olesen 2000; Sigrist and Sazima 2004; Martins et al. 2013). be associated with narrower ranges than occupied by related less Considering that there are only 450 oil-bee species world- specialized species. Bees are phytophagous insects because they wide in few genera and subfamilies and yet 2000 species of plants feed their larvae with the gamete-containing pollen grains of flow- in numerous genera and 11 families with oil-offering flowers, ering plants. Plants have evolved numerous strategies to reduce colonization of new floral oil hosts by bees over evolutionary the proportion of their pollen that ends up in larval guts, includ- time is the expected scenario, and this is supported by molec- ing offering additional or alternative rewards (Simpson and Neff ular clock dated bee and plant phylogenies indicating very few 1981). One such alternative reward is floral oil, which is collected ancient oil-flower lineages and numerous younger ones (Ren- by the females of some 450 species of bees in 18 genera in Ap- ner and Schaefer 2010). Selection by bees on preadapted flowers inae and Melittinae that provision their cells with a mix of pollen that could be “recruited” as new oil sources would have involved

C 2015 The Author(s). Evolution C 2015 The Society for the Study of Evolution. 1835 Evolution 69-7: 1835–1844 ALINE C. MARTINS ET AL.

Figure 1. Oil-producing flowers of the Neotropics and Centris females collecting oil. (A) Centris aenea on oil-offering Byrsonima (Malpighiaceae). (B) Centris tarsata on oil-offering Krameria (Krameriaceae). (C) Centris trigonoides on oil-offering Angelonia inte- gerrima (Plantaginaceae). (D) Nierembergia species (Solanaceae). (E) Cypella herbertii (Iridaceae). (F) Calceolaria species (Calceolariaceae). Photos: (A)–(E) Antonio Aguiar, (F) Oscar Perez.´ oil-host broadening and polyphagous stages, possibly correlated Epicharis with 35 species occurring from 23° in the north to 34° with geographic range expansion, similar to what has been sug- in the south (Fig. 2A), both in the Apidae (Martins et al. 2014a). gested for phytophagous butterflies (Janz et al. 2006; Janz and Females in both genera feed on nectar from many kinds of flow- Nylin 2008; Nylin et al. 2013). ers, but feed their brood with pollen mixed with oil from a limited Here we address the question of host breadth evolution, in- number of flower types (Vinson et al. 1996, 1997; Aguiar and creasing specialization, and its possible relation to range size in Gaglianone 2003). The pollen and oil come from the flowers of two genera of Neotropical oil bees, Centris with 230 species oc- Malpighiaceae (Figs. 1A), Calceolariaceae (Fig. 1F), Iridaceae curring from 39° in the north to 47° in the south (Fig. 2B) and (Fig. 1E), Krameriaceae (Figs. 1B), Orchidaceae, Plantaginaceae

1836 EVOLUTION JULY 2015 GAIN AND LOSS OF SPECIALIZATION IN TWO OIL-BEE LINEAGES, CENTRIS AND EPICHARIS (APIDAE)

known, the family assignments of their oil flowers are known from field observations (Table S1). For a species to be able to exploit a particular oil flower, it has to have the right type of oil- collecting apparatus, suitable either for piercing epithelia under which floral oil has accumulated or for mopping up oil from fields of minute hairs that secrete it. Epithelial glands exist in Malpighi- aceae, Krameriaceae, and Orchidaceae; oil-producing hair fields in Calceolariaceae, Iridaceae, Plantaginaceae, and Solanaceae. Bees that collect oil from epithelial glands have rigid setae on the tarsi of their forelegs and middle legs to break the epithelium. Bees that collect oil from oil-producing hair fields instead have spoon-like setae on the foreleg tarsi, quite different from the setae required for piercing epithelia (Vogel 1974; Neff and Simpson 1981). By combining data on the actual oil host of different species with clock-dated phylogenies for Centris and Epicharis as well as their main oil-host clades, we here reconstruct the evolutionary direction of oil-host use and the evolution or loss of specialization (i.e., oil foraging on one, two, three, or more plant families). For a subset of species, we infer whether the number of oil hosts relates to geographic range size.

Material and Methods TAXON SAMPLING AND MOLECULAR CLOCK DATING We sampled 64 species of Centris with pointed or spoon-shaped setae on four legs, the condition found in the majority of species in this genus, four of the 10 species with setae on only the forelegs (C. hyptidis, C. thelyopsis, C. cineraria, C. anomala), and four species that have no oil-collecting apparatuses. All Epicharis have oil combs on both pairs of front legs, and we sampled 22 species of this genus. We also included 20 corbiculate bee species. The DNA data matrix is from Martins et al. (2014a), and includes 4300 aligned nucleotides from four ribosomal and protein-coding nuclear markers; namely, 28S, LW-Rhodopsin, elongation factor 1α-F2-copy, and RNA polymerase II. Divergence times were es- timated using the Bayesian approach implemented in BEAST 1.7 (Drummond et al. 2012), with a Yule tree prior as appropriate for species-level work, the GTR + G substitution model, which best fit the data as found with jModelTest v. 2 under the AIC (Darriba et al. 2012), and the uncorrelated lognormal (UCLN) re- Figure 2. Geographic ranges of Epicharis, Centris, and subclades laxed clock model. We used two calibration fossils. One was Apis of Centris discussed in the text. (A) Epicharis;(B)Centris; (C) range lithohermaea from the Middle Miocene Chojabaruˆ formation in of the Wagenknechtia clade (labeled in Fig. 4); (D) range of the Japan (it dates to the Langhian, a geological stage with an age hyptidis clade (Fig. 4). of 13.8–16 Mya). This is the oldest fossil assigned to the crown group of Apis, specifically the dorsata group (Engel 2006). We (Fig. 1C), and/or Solanaceae (Fig. 1D). For the purpose of this used this fossil to constrain the crown age of A. dorsata and A. study, specialists are species taking oil from only one family, cerana using a log-normal prior with a median of 1.26, SD of 0.5, whereas generalists are capable of exploiting the oil glands of and an offset of 15 Mya, which let ages fall between 13.8 and 16 more than one plant family. Although the nectar, pollen, and oil Mya. The other fossil was Kelneriapis eocenica, a stingless bee sources used by individual species Centris or Epicharis are not from Eocene Baltic Amber (Lutetian, with an age of 41.2–47.8

EVOLUTION JULY 2015 1837 ALINE C. MARTINS ET AL.

Mya), and we used this fossil to constrain the stem age of the topologies compared with the published phylogenetic assessments African Meliponini (in our sampling: Hypotrigona, Meliponula, based on the same data. Except for minor differences in unsup- Plebeina, Axestotrigona) to 44 million years, using a log-normal ported nodes, trees were similar to the previously published ones prior with a median of 1.67, SD of 0.5, and an offset of 45 Mya, (Figs. S1–S3). which let ages fall between 44 and 50 Mya. The Markov Chain Divergence times for the oil-offering plant clades were again Monte Carlo (MCMC) process was run for 50 million generations, estimated with BEAST 1.7, using a Yule tree prior, the GTR + sampling every 10,000th generation (using tree subsampling and G substitution model, and the UCLN clock model. We used this several independent runs). Convergence of chains was checked relaxed clock model because, for each matrix, the UCLN.stdev in Tracer (Rambaut et al. 2014) by making sure that all effec- values in Tracer were ࣙ0.5. We used published plastid and nuclear tive sample sizes (ESS) were >200. The EES values indicate the substitution rates for calibration of the plant phylogenies, specif- number of effectively independent draws from the posterior in the ically 0.0007 substitution/site/Ma for the plastid regions (Palmer sample, and this statistic can help to identify autocorrelation and 1991) and 0.00427 s/s/Ma for the ITS region, which is intermedi- poor mixing. TreeAnnotator (part of the BEAST package) was ate between the lowest (0.00038, the shrub Hamamelis)andthe then used to create maximum clade credibility trees, and trees highest (0.00834, the herb Soldanella) ITS substitution rate com- were visualized in Figtree (Rambaut 2014). piled in a review (Kay et al. 2006). The prior on each rate was For the Malpighiaceae, we used the phylogenetic tree and dat- a gamma distribution with an initial value of 1.0, and rates were ing results of Xi et al. (2012); that this family has long offered oil unlinked among the nuclear and plastid partitions. For the Krame- in its flowers is clear from an Eocene fossil flower, Eoglandulosa riaceae and Nierembergia matrices, the MCMC process was run warmanensis (Taylor and Crepet 1987), that already possesses the for 10 million generations. For Iridaceae, we carried out two runs calyx glands also seen in extant malpighs. For the Angelonieae of 20 million generations each. For the Calceolaria matrix, we (Plantaginaceae), we used the phylogenetic tree and dating re- carried out four runs, ranging from 20 to 80 million generations sults of Martins et al. (2014b, TreeBASE accession 16619; 44 each, totaling 210 million generations. Chains were sampled ev- taxa, 2085 aligned nucleotides of combined nuclear ITS and plas- ery 10,000th generation (after subsampling and combining sepa- tid markers). Oil production may have evolved four or five times rate runs for Iridaceae and Calceolaria in LogCombiner, part of in the Angelonieae, and in each case, Centris bees are the main the BEAST package). Convergence was assessed as done for the pollinators (Martins et al. 2014b). For Calceolaria (Calceolari- bee phylogeny, and TreeAnnotator and FigTree were used as be- aceae), a genus of 200 species (Molau 1988), we used the matrix fore. Where possible, we cross-validated the obtained divergence of Cosacov et al. (2009), with 104 taxa and 1544 nucleotides times by comparing the ages of nodes in our chronograms with the of combined nuclear ITS and plastid matK.ForNierembergia ages obtained for the same nodes in other studies. For Calceolari- (Solanaceae), which has 21 species, we used the molecular matrix aceae, this was the family crown node, whose age was compared of Tate et al. (2009, 35 taxa, 1639 aligned nucleotides of ITS and to the age for the same node calculated by Nylinder et al. (2012: plastid rpl16), adding outgroup representatives from Petunia and rbcL 19 [6.5–40.2], matK 30.8 [14.8–51.3], atpB-rbcL 12.9 [5.1– Bouchetia.ForKrameria (Krameriaceae), a genus of 18 species, 24.8] Ma). For Nierembergia, this was the Solanaceae crown node, 10 in North and Central America and six from northern Colom- whose age was compared to that calculated for the same node by bia to east-central Brazil (Simpson 1989), we used the matrix of Dillon et al. (2009: 35 [21–48] Ma) and Bell et al. (2010: 38 Simpson et al. (2004), consisting of 800 nucleotides of ITS for [26–49] Ma). For the oil-producing Iridaceae, we compared our 17 Krameria plus the outgroups Guaiacum angustifolium, Kall- estimate for the crown group node of Iridaceae with that of Bell stroemia parviflora,andTribulus terrestris (Zygophyllaceae). For et al. (2010: 34 [20–49] Ma). For the Krameriaceae, we know of the Sisyrinchieae, Trimezieae, and Tigridieae (Iridaceae), which no previous published ingroup age estimate. are closely related (Goldblatt et al. 2008) and are the only Iri- daceae producing floral oil, we used a matrix of 100 taxa and 5905 ANCESTRAL STATE RECONSTRUCTION OF OIL-HOST aligned nucleotides of the plastid markers rps4, matK, rbcL, trnL- SPECIALIZATION F,andrps16 from Chauveau et al. (2012) and Lovo et al. (2012). Ancestral state reconstruction was carried out under maximum- Oil hairs are thought to have evolved twice in Sisyrinchium, once likelihood optimization in Mesquite version 2.75 (Madison and in Trimezieae, and seven times in Tigridieae, and species in three Madison 2009), with the input topology being the maximum clade of these clades are oil hosts for Centris (A. Aguiar, pers. comm.). credibility chronogram for the apine bees from Martins et al. For each plant matrix, maximum-likelihood tree searches and (2014a). Oil production is the ancestral condition in Malpighi- bootstrapping, with 1000 replicates, were performed in RAxML aceae (Davis and Anderson 2010; Davis et al. 2014), Krameri- (Stamatakis 2006) using the RAxML-GUI (Silvestro and Micha- aceae (Simpson et al. 2004), Calceolaria (Cosacov et al. 2009), lak 2012). The trees were again visualized in Figtree and their and Nierembergia (Tate et al. 2009). In Iridaceae, Orchidaceae,

1838 EVOLUTION JULY 2015 GAIN AND LOSS OF SPECIALIZATION IN TWO OIL-BEE LINEAGES, CENTRIS AND EPICHARIS (APIDAE)

and Plantaginaceae, oil as a reward originated several times as clade visited by EpicharisorCentris is the Malpighiaceae, which explained above. Based on relevant literature and observations by originated in the Upper Cretaceous (Xi et al. 2012). This mutual- the first author (Table S1), we coded bees for the following oil- ism continues until the present, and there are no known cases of host states, which captured whether they were oligoleges (on one other oil hosts used by Epicharis, the stem lineage of which dates family) or polyleges (on >1 family): 0 = only Malpighiaceae; 1 to more or less the same time as Malpighiaceae (Fig. 4). Within = only Krameriaceae; 2 = only Angelonia;3= only Calceolaria; Centris, on the other hand, we inferred several host range expan- 4 = two, three, or four oil hosts (including Monttea and sporadic sions, in which single species of Centris now use malpigh oil in records on Orchidaceae, Iridaceae, and Nierembergia); 5 = un- addition to oils from up to four other plant clades (Fig. 3 and known oil host; 6 = non-oil-collecting. Because of the trait-state Table S1). At least four Centris have lost the four-legged lability found during the recent past, at least in Centris, we did not oil-collecting apparatus (discussed below). Without sharp oil- infer ancestral states further back in time than 40 million years. collecting setae on four legs, bees are not able to exploit malpigh flowers. PLANT DISTRIBUTION MAPS, BEE RANGE SIZES, With the present taxon sampling, we found several oil-host AND STATISTICAL COMPARISON OF RANGES shifts between plants in different families and even different The geographic ranges of some of the oil plant clades could be ob- orders. This involves even closely related bee species, such as tained from the following treatments: Angelonia (Barringer 1981), C. cineraria, which exploits Calceolaria and Sysirinchium,and Calceolaria (Molau 1988; Cosacov et al. 2009), Krameriaceae C. rhodophthalma, which exploits Malpighiaceae and Krameri- (Simpson 1989), Malpighiaceae (only Neotropical species, Davis aceae, whereas a third close relative has lost the oil-collecting et al. 2002), and Monttea (Sersic´ and Cocucci 1999). The ranges behavior (Fig. 3). The most recent common ancestor of Centris of the oil-offering species of Iridaceae, Nierembergia, and Orchi- probably obtained oil from the epithelial glands of Malpighiaceae daceae could not be mapped reliably. or Krameriaceae, the only two groups in existence 50 (40–59) Ranges for 23 species of Centris were obtained based on label My, perhaps followed by switches to Angelonia (Plantaginaceae), data of specimens deposited in the insect collection of the Depart- which has oil hair patches, at 25 (17–34) My (Fig. 3). Host broad- ment of Zoology, Universidade Federal do Parana (DZUP), as well ening to other oil hosts with oil hair fields, such as Calceolaria as taxonomic publications (Snelling 1974, 1984; Zanella 2002; (Calceolariaceae), Monttea (Plantaginaceae), and Nierembergia, Vivallo and Melo 2009; Vivallo 2013). Range sizes were calcu- all occurred during the Miocene (Fig. 3). Among the most recent lated using the measuring tool in QGIS 2.6.1 (www.qgis.org), Centris oil hosts are Cypella and Cipura in the Iridaceae, both following an approximate minimum concave polygon. Ranges with oil hairs (Table 1), and perhaps Orchidaceae (Table S1; not of Epicharis were not calculated because all species collect oil dated here). from the flowers of just one family, Malpighiaceae. To test for Losses of the oil-collecting behavior occurred around 15 Mya differences in range sizes between oil-host specialist and gener- in the common ancestor of Centris pallida and C. hoffmanseggiae alists, we categorized the 23 species into oligoleges (eight) and (clade Xerocentris), endemic to the deserts of the United States polyleges (15) and carried out a nonparametric ANOVA-based and Mexico (Fig. 3; Table S1), and some 5 to 2 Mya in C. mu- area rank (with species ranked by their distribution area). We also ralis and C. tamarugalis, both in xeric habitats of Argentina and assigned them to nine independent (nonnested) clades (Results) Chile. Another loss of the four-legged oil-collecting apparatus has and conducted another ANOVA on the area rank. This accounts occurred in C. anomala (Fig. 3: Paracentris II clade), a species for both clade membership and host specialization. endemic in Central America and Mexico that has well-developed pads of specialized mopping setae on fore and mid legs, as well on the distal sterna, adapted probably to mopping up oil from hair Results fields (Neff and Simpson 1981). OIL-HOST SPECIALIZATION AND GENERALIZATION IN CENTRIS AND EPICHARIS NO DETECTABLE CORRELATION BETWEEN OIL-HOST Figure 3 shows a chronogram for 22 of the 35 species of Epicharis SHIFT BREADTH AND GEOGRAPHIC RANGE SIZE and 72 of the 230 species of Centris, with the 20 representative The overall geographic ranges of the genera Epicharis and Centris corbiculate bee species reduced to a single line. The stem ages and of subgenus Wagenknechtia and the hyptidis clade of Centris inferred for the five most important oil-providing plant clades are are shown in Figure 2, and those of their main oil-host clades in shown at the bottom of the phylogenetic tree (the individual plant Figure 4. The 23 species (eight of them oligoleges, 15 polyleges) chronograms are shown in Figs. S4–S6), and the oil-host plant(s) for which we could calculate range sizes (in km2) are marked by visited by each species are shown next to the bee names (with the an asterisk in Figure 3, and the ranges as given in Table S2. A source information provided in Table S1). The oldest oil-offering nonparametric ANOVA that contrasted polyleges with oligoleges

EVOLUTION JULY 2015 1839 ALINE C. MARTINS ET AL.

Figure 3. Divergence times for Centris and Epicharis, with the 20 corbiculate bee species reduced to one line (full tree in Martins et al. 2014a), together with the stem ages of their main oil-host clades. Gray and blue bars at nodes indicate 95% highest posterior density intervals (i.e., error ranges); 95% error ranges for the plant stem ages are also indicate by bars. Oil hosts are represented by the colored squares next to species names. Because of the state lability found at the tips, at least in Centris, we did not infer ancestral states further back than 40 million years. Bees for which we have geographic range data are marked with an asterisk. The numbers 1–9 indicate the nonnested clades used for phylogenetically informed comparisons.

1840 EVOLUTION JULY 2015 GAIN AND LOSS OF SPECIALIZATION IN TWO OIL-BEE LINEAGES, CENTRIS AND EPICHARIS (APIDAE)

Figure 4. Geographic ranges of the main oil hosts exploited by Centris and Epicharis. showed no significant difference (Kruskall–Wallis χ2 = 0.15, capable of exploiting the oil glands of more than one plant family. df = 1, P = 0.70). The nine nonnested clades to which the 23 Although this distinction disregards host shifts within plant fami- species were assigned for a phylogenetically informed analysis lies, which may occur during single foraging flights (A. Martins, are also shown in Figure 3. There were no significant associations pers. obs.), it is evolutionarily meaningful because host shifts between range size and specialization (F = 0.1; df = 1, 13; between oil-offering families require not only behavioral, but P = 0.76). also morphological, changes in bees’ oil-collecting structures. Exploiting oil hairs requires soft hair pads, whereas epithelial oil glands require stiff setae capable of piercing an epithelium (Vogel Discussion 1974, 1988; Neff and Simpson 1981). These different collect- SWITCHES BETWEEN OIL-HOST SPECIALIZATION ing apparatuses may entail an evolutionary canalization of the AND GENERALIZATION, BUT NO SIGNAL OF direction of host shifts because once a bee species has lost the OIL-HOST BREADTH BEING LINKED TO GEOGRAPHIC stiff setae required for exploiting Malpighiaceae and Krameri- RANGE SIZE aceae glands, a regain of these structures may be rare. Our recon- For the purpose of this study, oil-host specialists are bee species structions are handicapped by the incomplete species sampling taking oil from only one plant family, whereas oil generalists are in Centris (72 of 230) and lack of in-depth data on oil hosts for

EVOLUTION JULY 2015 1841 ALINE C. MARTINS ET AL.

Table 1. Inferred ages of clades with oil-offering flowers.

Mean Age (95% Highest Posterior Density Interval)

Clade Crown Group Stem Group Number of Oil Species Gland Type Malpighiaceae 60 (59–82)1 85 (73–100) 1000 Epithelial Krameriaceae 23 (19–28) 50 (40–59) 18 Epithelial Calceolaria (Calceolariaceae) 8 (6–10) 13 (9–18) 200 Trichomatic Nierembergia (Solanaceae) 8 (6–11) 20 (14–26) 21 Trichomatic Plantaginaceae I: Angelonia 17 (12–23)2 25 (17–34) 26 Trichomatic Plantaginaceae II: Monttea 3.5 (1–8)2 17 (8–28) 3 Trichomatic Iridaceae I: Sysirinchium 14 (11–17) 18 (14–22) 35 Trichomatic Iridaceae I: Cypella 4 (3–6) 5 (3–6) 30 Trichomatic Iridaceae I: Cipura 2 (1–4) 7 (4–10) 8 Trichomatic

Species numbers from the taxonomic references cited in Materials and Methods. 1Xi et al. (2012). 2Martins et al. (2014b). species that have not been sufficiently studied in the field. As far DATA ARCHIVING as known, Epicharis collect oil only from Malpighiaceae, fitting The TreeBASE accession number is 16619. with the consistent presence in this genus of pointed setae on four legs. We detected no significant difference in range sizes between LITERATURE CITED oil specialists and oil generalists. However, we have range sizes Aguiar, C. M. L., and M. C. Gaglianone. 2003. Nesting biology of Centris for only 23 species and for few sister species pairs (Fig. 3). Geo- (Centris) aenea (Hymenoptera, Apidae, Centridini). Rev. Bras. Zool. 20:601–606. graphic range size may depend on too many other factors, besides Barringer, K. 1981. A taxonomic revision of Angelonia (Scrophulariaceae). the number of oil sources that a bee is able to use, and such a small Ph.D. dissertation, University of Connecticut, Storrs, CT. number of comparisons may be insufficient to detect a possible Bell, C. D., D. E. Soltis, and P. S. Soltis. 2010. The age and diversification of correlation between geographic range and oil-foraging breath. the angiosperms re-revisited. Am. J. Bot. 97:1296–1303. Cappellari, S. C., M. A. Haleem, A. J. Marsaioli, R. Tidon, and B. B. Simpson. 2011. Pterandra pyroidea: a case of pollination shift within Neotropical Malpighiaceae. Ann. Bot. 107:1323–1334. Conclusions Chauveau, O., L. Eggers, T. T. Souza-Chies, and S. Nadot. 2012. Oil-producing Although the Neotropical oil flower/Centris/Epicharis “mutual- flowers within the Iridoideae (Iridaceae): evolutionary trends in the flow- istic system” has long been known to result from convergent ers of the New World genera. Ann. Bot. 110:713–729. Cosacov, A., A. N. Sersic,´ V. Sosa, J. A. De-Nova, S. Nylinder, and A. A. evolution (Vogel 1974, 1988), the details have never been recon- Cocucci. 2009. New insights into the phylogenetic relationships, char- structed using dated phylogenies for at least the main partners acter evolution, and phytogeographic patterns of Calceolaria (Calceo- as done here. Our results show that the buildup of oil plant/oil lariaceae). Am. J. Bot. 96:2240–2255. bee interactions involved repeated instances of transitions from Darriba, D., G. L. Taboada, R. Doallo, and D. Posada. 2012. jModelTest 2: more models, new heuristics and parallel computing. Nat. Methods 9: Malpighiaceae to Calceolariaceae, Iridaceae, Krameriaceae, Or- 772. chidaceae, Plantaginaceae, and Solanaceae, and from flowers with Davis, C. C., and W. R. Anderson. 2010. A complete generic phylogenetic epithelial oil glands to those with fields of oil hairs, canalized tree of Malpighiaceae inferred from nucleotide sequence data and mor- by concomitant changes in bees’ oil-collecting structures from phology. Am. J. Bot. 97:2031–2048. pointed setae on two pairs of legs to soft, branched setae on just Davis, C. C., C. D. Bell, S. Mathews, and M. J. Donoghue. 2002. Laurasian migration explains Gondwanan disjunctions: evidence from Malpighi- the first pair of legs. The chemical composition of the few floral aceae. Proc. Natl. Acad. Sci. USA 99:6833–6837. oils that have been analyzed is simple, usually involving one ma- Davis, C. C., H. Schaefer, Z. Xi, D. A. Baum, M. J. Donoghue, and L. J. jor and few minor free fatty acids (Simpson et al. 1977; Vinson Harmon. 2014. Long-term morphological stasis maintained by a plant- et al. 1996, 1997; Cappellari et al. 2011), and oil-collecting is pollinator mutualism. Proc. Natl. Acad. Sci. USA 116:5914–5919. Dillon, M.O., T.Y. Tu, L. Xie, V. Quipuscoa-Silvestre, and J. Wen. 2009. Bio- triggered when the chemoreceptors on the tarsi of oil bees come geographic diversification in Nolana (Solanaceae), a ubiquitous member in contact with an oily surface (Dotterl¨ and Schaffler¨ 2007). This of the Atacama and Peruvian deserts along the western coast of South may make switching between oil flowers relatively easy. America. J. Syst. Evol. 47:457–476.

1842 EVOLUTION JULY 2015 GAIN AND LOSS OF SPECIALIZATION IN TWO OIL-BEE LINEAGES, CENTRIS AND EPICHARIS (APIDAE)

Dotterl,¨ S., and I. Schaffler.¨ 2007. Flower scent of floral oil-producing Lysi- Rasmussen, C., and J. M. Olesen. 2000. Oil-flowers and oil-collecting bees. machia punctata as attractant for the oil-bee Macropis fulvipes.J.Chem. Pp. 23–31 in pp. 161–178. Ø. Totland, W.S. Armbruster, C. Fenster, Ecol. 33:441–445. U. Molau, L.A. Nilsson, J.M. Olesen, J. Ollerton, M. Philipp and J. Drummond, A. J., M. A. Suchard, D. Xie, and A. Rambaut. 2012. Bayesian Agren,˚ ed. Scandinavian Association for Pollination ecoloy honours phylogenetics with BEAUti and the BEAST 1.7. Mol. Biol. Evol. Knut Faegri. Norwegian Academy of Science & Letters 39, Oslo. 29:1969–1973. Renner, S. S., and H. Schaefer. 2010. The evolution and loss of oil-offering Engel, M. S. 2006. A giant honey bee from the middle Miocene of Japan flowers: new insights from dated phylogenies for angiosperms and bees. (Hymenoptera: Apidae). Am. Museum Novit. 3504:1–12. Philos. Trans. R. Soc. Lond. B Biol. Sci. 365:423–435. Goldblatt, P., A. Rodriguez, M. P. Powell, T. J. Davies, J. C. Manning, M. Sersic,´ A. N., and A. A. Cocucci. 1999. An unusual kind of nectary in the oil vander Bank, and V. Savolainen. 2008. Iridaceae “Out of Australasia”? flowers of Monttea. Flora 194:393–404. Phylogenetic tree, biogeography, and divergence time based on plastid Sigrist, M. R., and M. Sazima. 2004. Pollination and reproductive biology of DNA sequences. Syst. Bot. 33:495–508. twelve species of neotropical Malpighiaceae: stigma morphology and Janz, N., and S. Nylin. 2008. The oscillation hypothesis of host-plant range and its implications for the breeding system. Ann. Bot. 94:33–41. speciation. Pp. 203–215 in K. J. Tilmon, ed. Specialization, speciation, Silvestro, D., and I. Michalak. 2012. RaxmlGUI: a graphical front-end for and radiation: the evolutionary biology of herbivorous insects. Univ. of RAxML. Org. Divers. Evol. 12:335–337. California, Berkeley, C A. Simpson, B. B. 1989. Krameriaceae Flora Neotropica Monograph 49. The Janz, N., S. Nylin, and N. Wahlberg. 2006. Diversity begets diversity: host New York Botanical Garden, NY. Pp. 1–109. expansions and the diversification of plant-feeding insects. BMC Evol. Simpson, B. B., and J. L. Neff. 1981. Floral rewards: alternatives to pollen Biol. 6:4. and nectar. Ann. Missouri Bot. Gard. 68:301–322. Kay, K. M., J. B. Whittall, and S. A. Hodges. 2006. A survey of nuclear ribo- Simpson, B. B., J. L. Neff, and D. Seigler. 1977. Krameria, free fatty acids somal internal transcribed spacer substitution rates across angiosperms: and oil-collecting bees. Nature 267:150–151. an approximate molecular clock with life history effects. BMC Evol. Simpson, B. B., A. Weeks, D. M. Helfgott, and L. L. Larkin. 2004. Species Biol. 6:36. relationships in Krameria (Krameriaceae) based on ITS sequences and Lovo, J., R. C. Winkworth, and R. Mello-Silva. 2012. New insights into morphology: implications for character utility and biogeography. Syst. Trimezieae (Iridaceae) phylogenetic tree: what do molecular data tell Bot. 29:97–108. us? Ann. Bot. 110:689–702. Snelling, R. R. 1974. Notes on the distribution and taxonomy of some North Machado, I. C. 2004. Oil-collecting bees and related plants: a review of the American Centris (Hymenoptera: Anthophoridae). Contributions in Sci- studies in the last twenty years and case histories of plants occurring in ence. Vol. 259. Natural History Museum, Los Angeles Co unty, CA. Pp. NE Brazil. Pp. 255–280 in B. M. Freitas and J. O. P. Pereira, eds. Solitary 1–41. bees: conservation, rearing and management for pollination. Imprensa ———. 1984. Studies on the taxonomy and distribution of American Cen- Universitaria,´ Fortaleza, Brazil. tridine bees (Hymenoptera: Anthophoridae). Contributions in Science. Madison, W. P., and D. R. Madison. 2009. Mesquite: a modular system for Vol. 347. Natural History Museum, Los Angeles Co unty, CA. Pp. evolutionary analysis. Version 2.7. http://mesquiteproject.org. 1–69. Martins, A. C., A. J. C. Aguiar, and I. Alves-dos-Santos. 2013. Interaction Stamatakis, A. 2006. RAxML-VI-HPC: maximum likelihood-based phyloge- between oil-collecting bees and seven species of Plantaginaceae. Flora netic analyses with thousands of taxa and mixed models. Bioinformatics 208:401–411. 22:2688–2690. Martins, A. C., G. A. R. Melo, and S. S. Renner. 2014a. The corbiculate bees Stewart, A. J. A., T. M. Bantock, B. C. Beckmann, M. S. Botham, D. arose from New World oil-collecting bees: implications for the origin of Hubble, and D. B. Roy. 2015. The role of ecological interactions in pollen baskets. Mol. Phylogenet. Evol. 80:88–94. determining species ranges and range changes. Biol. J. Linn. Soc. DOI: Martins, A. C., M. D. Scherz, and S. S. Renner. 2014b. Several origins of 10.1111/bij.12543. floral oil in the Angelonieae, a Southern Hemisphere disjunct clade of Tate, J. A., M. C. Acosta, J. Mcdill, E. A. Moscone, B. B. Simpson, and A. A. Plantaginaceae. Am. J. Bot. 101:2113–2120. Cocucci. 2009. Phylogenetic tree and character evolution in Nierember- Molau, U. 1988. Scrophulariaceae - Part I. Calceolarieae. Flora Neotrop. gia (Solanaceae): molecular, morphological, and cytogenetic evidence. 47:0–41. Syst. Bot. 34:198–206. Neff, J. L., and B. B. Simpson. 1981. Oil-collecting structures in the An- Taylor, D. W., and W. L. Crepet. 1987. Fossil floral evidence of Malpighiaceae thophoridae (Hymenoptera): morphology, function, and use in system- and an early plant-pollinator relationship. Am. J. Bot. 74:274–286. atics. J. Kansas Entomol. Soc. 54:95–123. Vinson, S. B., G. W. Frankie, and H. J. Williams. 1996. Chemical ecology Nylin, S., J. Slove, and N. Janz. 2013. Host plant utilization, host range of bees of the genus Centris (Hymenoptera: Apidae). Florida Entomol. oscillations and diversification in nymphalid butterflies: a phylogenetic 79:109–129. investigation. Evolution 68:105–24. Vinson, S. B., H. J. Williams, G. W. Frankie, and G. Shrum. 1997. Floral lipid Nylinder, S., U. Swenson, C. Persson, S. B. Janssens, and B. Oxelman. 2012. chemistry of Byrsonima crassifolia (Malpighiaceae) and a use of floral A dated species tree approach to the trans-Pacific disjunction of the lipids by Centris Bees (Hymenoptera: Apidae). Biotropica 29:76–83. genus Jovellana (Calceolariaceae, Lamiales). Taxon 61:381–391. Vivallo, F. 2013. Revision of the bee subgenus Centris (Wagenknechtia) Palmer, J. D. 1991. Plastid chromosome: structure and evolution. Pp. 5–53 Moure, 1950 (Hymenoptera: Apidae: Centridini). Zootaxa 3683:501– in L. Bogorad and I. K. Vasil, eds. The molecular biology of plastids. 537. Academic Press, San Diego, CA. Vivallo, F. and G. A. R Melo. 2009. Taxonomy and geographic distribution Rambaut, A. 2014. FigTree: tree figure drawing tool. Institute of Evolutionary of the species of the hyptidis group (Hymenoptera: Apidae: Centri- Biology, University of Edinburgh, Edinburgh, U . K. dini), with description of a new species from central Brazil. Zootaxa Rambaut, A., M. A. Suchard, and A. J. Drummond. 2014. Tracer v1.6, 2003– 2075:33−44. 2013: MCMC trace analysis tool. Institute of Evolutionary Biology, Vogel, S. 1974. Olblummen¨ und olsammelnde¨ Bienen. Trop. Subtrop. University of Edinburgh, Edinburgh, U. K. Pflanzenwelt 7:1–267.

EVOLUTION JULY 2015 1843 ALINE C. MARTINS ET AL.

———. 1988. Die Olblumensymbiosen:¨ Parallelismus und andere Aspekte Zanella, F. C. V. 2002. Sistematica,´ filogenia e distribuic¸ao˜ geografica´ das ihrer Entwicklung in Raum und Zeit. Z. Zool. Syst. Evol. 26:341– especies´ sul-americanas de Centris (Paracentris) Cameron, 1903 e de 362. Centris (Penthemisia) Moure, 1950, incluindo uma analise´ filogenetica´ Xi, Z., B. R. Ruhfel, H. Schaefer, A. M. Amorim, M. Sugumaran, K. J. do “grupo Centris” sensu Ayala 1998 (Hymenoptera, Apoidea, Centri- Wurdack, P. K. Endress, M. L. Matthews, P. F. Stevens, S. Mathews, dini). Rev. Bras. Entomol. 46:435–488. et al. 2012. Phylogenomics and a posteriori data partitioning resolve the Cretaceous angiosperm radiation Malpighiales. Proc. Natl. Acad. Sci. Associate Editor: M. Friedman USA 109:17519–24. Handling Editor: R. Shaw

Supporting Information Additional Supporting Information may be found in the online version of this article at the publisher’s website:

Figure S1. Maximum-likelihood tree for Calceolaria rooted on Kohleria (Gesneriaceae) based on a matrix of 103 taxa and 1544 aligned nucleotides. Figure S2. Maximum-likelihood tree for Sysirinchieae, Trimezieae, and Tigridieae rooted on Iris based on a matrix of 100 taxa and 5905 aligned nucleotides. Figure S3. Maximum-likelihood tree for Krameriaceae and Nierembergia (Solanaceae). Figure S4. Chronogram for Calceolariaceae rooted on Kohleria (Gesneriaceae) obtained under a Bayesian relaxed clock model from the same matrix as used in Figure S1. Figure S5. Chronogram for Sysirinchieae, Trimezieae, and Tigridieae obtained under a Bayesian relaxed clock model from the same matrix as used in Figure S2. Figure S6. Chronograms for Nierembergia and Krameria. Table S1. Bee species and angiosperm families visited for oil collecting, with relevant references. Table S2. Range sizes for 23 bee species, used in the statistical tests for a correlation between range size and oil-host specialization.

1844 EVOLUTION JULY 2015