In multispecies interactions, ecological circumstances may select for similar evolutionary responses in diverse species, resulting in recurrent convergent evolution [1]. Although convergent adaptation can be demonstrated by identifying functionally relevant phenotypes in different species, phylogenetic information improves our ability to make inferences about convergent evolution. A prominent example of such functionally convergent phenotype is found in the association between ants and plants [2][3][4], whereby many ant lineages have independently evolved specialized arboreal nesting habits [4,5]. Obvious instances of convergent evolution among ant-plant interactions can be identified through morphological and behavioural traits. Plants bearing domatia, extrafloral nectaries and food bodies have evolved independently multiple times in association with distantly related ants [3,4]. Once interdependencies between symbionts have been established, subsequent diversification within lineages, ant or plant, can lead to multiple species exhibiting symbiotic associations resulting from shared ancestry rather than convergence. Thus, morphologically and behaviourally similar sets of ant specialists, e.g. multiple species in the same closely related lineages, are assumed to result from a single evolutionary transition to specialization (e.g. within certain clades of Cecropia-symbiotic Azteca [6], Vachellia-inhabiting Pseudomyrmex [7]). In one case, however, recent studies have revealed an unexpected instance of convergence among closely related plant-ant species. The mutualistic ants associated with the iconic Pseudomyrmex-Vachellia system were discovered to comprise two clades separated by non-mutualistic species [8,9]. This independent evolution of two clades of Vachellia ants is striking because the specialized nest site selection and aggressive defence of nest space by these two Pseudomyrmex subgroups are lacking in the generalist species that separate them [9]. In this case, the two clades were morphologically distinguishable, but still considered to be sister clades prior to the molecular results [7]. Given this finding, other examples of convergence among closely related plant-ant species may exist in nature, but to reveal such cases, it is necessary to integrate natural history information with a robust molecular phylogenetic hypothesis [1,3,8,9].
Here, we report an even more extreme example of convergent evolution in an ant-plant system. The formicine genus Myrmelachista is a Neotropical lineage of exclusively arboreal ants. Myrmelachista and its sister taxon Brachymyrmex form a clade sister to all other formicine ants, and this clade is estimated to have originated in the Neotropics between 72.2 and 80 Ma [10]. Currently, 56 species and 13 subspecies of Myrmelachista are recognized [11], with species occurring from central Mexico to southern Argentina. The species exhibit a full range of arboreal nesting habits that we separate into two broad categories: (i) dead-stem nesters with generalized foragers [12,13] and live-stem nesters opportunistically nesting in a variety of plant species [12][13][14] or (ii) specialized live-stem nesters (=obligately symbiotic), nesting only in live stems of a narrow set of plant species, with most individuals of the plant species inhabited by the specialist ants [13,[15][16][17][18][19] (figure 1). Nesting in dead stems and nesting opportunistically in live stems appear to be part of a continuous trait space, and some species can be found nesting in either dead or live stems. In contrast, the ant-plant specialists are sharply differentiated in their distinct use of a restricted set of plant species (figure 1).
In Myrmelachista, the known specialized associations occur in two regions. In the northern portion of Amazonia in South America, M. schumanni is a specialist inhabitant of Duroia hirsuta (Rubiaceae) and Tococa (Melastomataceae), systematically poisoning all plants in the vicinity and creating monodominant vegetation patches called 'devil's gardens' [16,17,20,21]. In the devil's garden system, the host plants have conspicuous preformed domatia. In Middle America, seven species have specialized associations with small-sized plants of Mespilodaphne, Ocotea (Lauraceae) and Guarea (Meliaceae) dispersed in the understory [13,15,19]. The host plants involved in the Middle American associations lack conspicuous domatia, or if present, they are subtle [19]. One plant trait that correlates with the Middle American association is dwarfism; all associated plants are treelets that flower and fruit in the understory or forest gaps, while most of their congeners are large canopy trees. In neither of the specialized associations is there evidence for food bodies or extrafloral nectaries. Both associations are thought to be mutualisms [15,16,18,19,22], with either evidence or speculation that the ants: (i) provide defence against herbivores [15,16], (ii) remove spores, lichens, epiphylls and debris from leaf surfaces [18,19], (iii) protect from encroaching plants [16,17,21] and (iv) provide beneficial nutrients from food items brought into stems [22]. In return, the host plant provides living space for the ants.
Regarding the ant partners, the opportunist and specialist species of the Middle American radiation of Myrmelachista display a clear phenotypic distinction in worker colour, worker behaviour and queen behaviour. The worker caste of opportunist species varies in colour from black to dusky orange-brown, with the exception of a poorly known species that has orange workers (M. meganaranja Longino, 2006) [13]. In contrast, all specialist species have light orange to yellow workers (electronic supplementary material, figure S1). In addition, the specialist species show a high degree of morphological uniformity, with workers of different species essentially indistinguishable. With respect to worker behaviour, the opportunist species tend to occupy only a part of the host plant and lack patrolling behaviour. In contrast, specialist Myrmelachista occupy the entire host plant and often patrol new growth surfaces. The specialist species often harbour hemipteran partners (i.e. coccids and pseudococcids) in unique xylem pockets that extend from the nest lumen to the cambium. We have not observed these pockets in the nests of opportunist species. Finally, colony-founding queens of opportunist species presumably have relatively generalized searching behaviours, selecting a variety of plant cavities. Foundresses of specialists restrict their searching to particular species of host plants.
For Middle American Myrmelachista, where there are multiple, similar-looking specialist species present, it has generally been thought these species form a clade and that specialization arose once in the common ancestor [13]. This hypothesis is supported by the strong morphological and behavioural uniformity of specialists. However, there has never been an explicit phylogenetic reconstruction of the genus, and relationships among species remain speculative. In this study, we explore the evolution of nesting behaviour and this unique plant-ant specialization in Myrmelachista for the first time. We employ a phylogenomic approach using ultraconserved element (UCE) sequence capture [23,24] to estimate the phylogenetic relationships within the genus. Our taxon sampling covers the full morphological, behavioural and geographical range of Myrmelachista, with a particularly comprehensive sampling of the Middle American species. Our initial goal was to examine whether the South American and Middle American cases of myrmecophytism were the result of a single evolutionary specialization or independent events. We discovered not only the independent evolution of plant specialists in South and Middle America but also the unexpected result that the Middle American specialists were two independent clades, separated by generalist ancestors and arising nearly simultaneously.
Using a dataset comprised 1946 UCE loci and 45 terminals, we recovered a robust phylogeny of Myrmelachista ants and found little topological conflict across analyses (figure 2; electronic supplementary material, figures S2-S8). We found strong support for five main clades: a Patagonian clade, two Neotropical clades, a Devil's Garden clade and a Middle American clade. The Patagonian clade is sister to all remaining Myrmelachista. It is restricted to the southern portion of South America and contains several species. Those species whose natural history is known are opportunistic nesters. Neotropical clade 1 and Neotropical clade 2 are found throughout the Neotropics and comprise a diverse assemblage of opportunistic nesters. The Devil's Garden clade occurs in northern South America, well separated from the Middle American specialists, and contains M. schumanni. This clade may contain only the single-species M. schumanni or a complex of closely related species, all of which are devil's garden ants (electronic supplementary material, table S1). Finally, the Middle American radiation is a large clade that occurs in northwest South America, the Caribbean and throughout the Middle American Corridor (from Panama to Mexico). Unexpectedly, specialist plant-ant species in the Middle American clade were found to occur as two independent clades separated by opportunists (figure 2). Although partially overlapping in range, one specialist clade is centred on the Caribbean slope, from Mexico to Costa Rica (hereafter Caribbean clade), and the other specialist clade is centred on the Pacific slope of Costa Rica (hereafter Pacific clade). A constraint analysis comparing the main topology (figure 2) to a tree in which the two specialist clades were forced to be monophyletic firmly rejected the alternative hypothesis (p-value of approximately unbiased (AU) test AU = 0.00137; electronic supplementary material, table S2).
Divergence dating analyses revealed that crown-group Myrmelachista evolved approximately 29 Ma (95% highest posterior density (HPD) 21.35-37.71 Myr) during the late-Oligocene (figure 3). The Middle American radiation evolved approximately 12 Ma (HPD 9.52-16.67 Myr). The three specialist clades were estimated to be reasonably young and similar in age, originating in the Pliocene. The crown-group ages were 3.83 Ma (HPD 2.61-5.30 Myr) for the Devil's Garden clade, 3.02 Ma (HPD 2.18-3.97 Myr) for the Caribbean clade and 3.32 Ma (HPD 2.44-4.30 Myr) for the Pacific clade of the Middle American specialists.
We examined the evolution of nesting phenotype by performing multiple ancestral state reconstructions (ASR) analyses using maximum parsimony (MP), maximum likelihood (ML) and Bayesian approaches. Analyses overwhelmingly supported the independent evolution of at least three instances of live-stem specialization from opportunist ancestors (figure 3; electronic supplementary material, tables S3-S5).
The MP analyses recovered specialization occurring independently in the Caribbean and Pacific clades as the most parsimonious scenario with different models returning the shortest total transition cost by either recovering as or having M. cooperi coded as an opportunist species (for the hidden model) (electronic supplementary material, table S3 and figure S10).
For ML analyses, the ARD (all rates different among state transitions) models favoured the ancestor for the Caribbean and Pacific clades likely being a specialist (i.e. reversal favoured over convergence) (electronic supplementary material, table S3 and figures S11-S13). Similar log-likelihood values and AIC (Akaike information criterion) scores for the ML analyses were obtained with M. cooperi coded as opportunist and the models with equal rates among state transitions (ER) (recovering independent evolution of Middle American specialists) and ARD (electronic supplementary material, tables S4 and S5). However, the ER model was preferred over the ARD model after a log-likelihood ratio test (p = 0.04; electronic supplementary material, table S4).
Our reversible-jump Markov Chain Monte Carlo (rjMCMC) analyses also supported the scenario of independently evolved specialists. Interestingly, outputs recovered the ancestor of the Pacific and Caribbean clades likely being a specialist, supporting a scenario of reversal to the opportunist condition for the ancestor of the Pacific clade and subsequent regaining of the specialist condition for one Pacific subclade (containing M. osa, M. haberi, M. lauropacifica and M. RSP027, a new specialist species from Costa Rica) (electronic supplementary material, tables S3 and S6 and figure S14).
Owing to its peculiar morphology and natural history, additional experiments could confirm M. cooperi as a social parasite of specialist Myrmelachista. Similar results from our ASR analyses for all methods were obtained when M. cooperi was either coded as opportunist or unknown for nesting habit (electronic supplementary material, table S3). Overall, our results indicate that the Middle American Myrmelachista clade radiated rapidly (figure 3) and generated not only opportunist species but also two likely independent clades of remarkably similar specialist species that arose nearly simultaneously-and perhaps allopatrically-on the Caribbean and Pacific sides of Middle America.
Lasius sitiens EX1585 (USA) Brachymyrmex Myrmelachista heeri EXRP182 (Costa Rica) nebulosus EXRP181 (Honduras) depilis cf. aphidicola EXRP168 (Colombia) admotus EXRP163 (Brazil) EXRP169 (USA) pilipes EXRP174 (Brazil) hoffmanni EXRP103 (Chile) mayri EXRP170 (Argentina) RSP020 EXRP137 (Brazil) RSP024 EXRP97 (Brazil) nodigera EXRP100 (Brazil) bruchi EXRP177 (Argentina) bambusarum EXRP175 (Brazil) mexicana EXRP58 (Costa Rica) reclusi EXRP50 (Colombia) catharinae EXRP180 (Brazil) schumanni EXRP48 (Bolivia) schumanni EXRP68 (Ecuador) schumanni EXRP77 (Guyana) RSP063 EXRP21 (Guatemala) RSP065 EXRP179 (Brazil) lauroatlantica EXRP4 (Costa Rica) nigrocotea EXRP29 (Costa Rica) flavoguarea EXRP28 (Costa Rica) flavocotea EXRP71 (Costa Rica) RSP057 EXRP63 (Mexico) RSP027 EXRP42 (Costa Rica) osa EXRP65 (Costa Rica) haberi EXRP1 (Costa Rica) 0.02 substitutions/site Myrmelachista clades Patagonian Generalist/opportunistic Nesting habit Live-stem specialist Neotropical 1 Neotropical 2 Middle Am. Radiation Devil's Garden 46.6/53 100/75 98.2/98 93.3/91 100/82 75.8/80 52/64 97.5/94 cf. zeledoni EXRP44 (Costa Rica) zeledoni EXRP7 (Costa Rica) lauropacifica EXRP74 (Costa Rica) cooperi EXRP17 (Costa Rica) gagates EXRP124 (Dominican Republic) plebecula EXRP30 (Costa Rica) skwarrae EXRP22 (Mexico) brevicornis EXRP49 (Bolivia) meganaranja EXRP64 (Costa Rica) longiceps EXRP41 (Costa Rica) longiceps EXRP43 (Costa Rica) RSP011 EXRP166 (Colombia) RSP014 EXRP56 (Venezuela) RSP056 EXRP6 (Panama) RSP026 EX819 (Costa Rica)
We recovered a robust phylogenetic hypothesis for Myrmelachista, with most nodes receiving maximum support and, more importantly, exhibiting little topological conflict across analyses. We found that specialized ant-plant associations occurred independently at least three times in the genus. The South American devil's garden ants are distantly related to the other specialists and thus clearly evolved independently from the Middle American specialist species-plus, devil's garden species are phenotypically distinct from the other specialists (e.g. worker coloration and overall body shape, antennomere count for female castes, host plant use). For the Middle American specialists, we discovered that what was hypothesized to be a single radiation of ant-plant mutualists is actually two independent evolutionary events, occurring nearly simultaneously and with strikingly parallel morphological and behavioural features. One specialist clade is restricted to the Pacific slope of Costa Rica.
The other specialist clade occurs mainly on the Caribbean slope, from southern Mexico to Costa Rica. Most of the Caribbean clade's diversity is in Costa Rica, with the species found in southern Mexico representing a relatively recent split from one of the Costa Rican species. This pattern suggests that both specialist clades originated in Costa Rica and the Caribbean clade later dispersed northwards.
The evolution of Middle American Myrmelachista specialists is comparable to the Pseudomyrmex-Vachellia system [9], which also involves convergent evolution of host plant specialization. Both systems are relatively young, but differ in the degree of morphological divergence, the timing of events and the degree of sympatry. The two acacia ant clades are morphologically distinguishable, while the two Myrmelachista clades are not. The evolution of the two acacia ant clades was not simultaneous. One lineage arose in the Late Miocene in northern Middle America, and the other lineage probably originated in southern Middle America about 3 Myr later, apparently colonizing the pre-existing mutualism [9]. In contrast, the Myrmelachista clades arose simultaneously.
The two Pseudomyrmex lineages are largely sympatric, while the two Myrmelachista clades are mostly non-overlapping, with only a small zone of contact. The one exception is morphospecies Myrmelachista RSP027 (a Pacific clade Myrmelachista Brachymyrmex 85.66 29.08 3.83 3.02 3.32 Patagonian Neotropical 1 Neotropical 2 Middle Am. Radiation Devil's Garden Probability of being a specialist Cretaceous 120 110 100 90 80 70 60 50 40 30 20 10 0 Quat. Pleistocene Pliocene Miocene Oligocene Eocene Paleocene Upper Lower Paleogene opportunist specialist length = 50.724 Neogene Lasius sitiens (USA) heeri (Costa Rica) nebulosus (Honduras) depilis (USA) cf. aphidicola (Colombia) admotus (Brazil) pilipes (Brazil) hoffmanni (Argentina) mayri (Argentina) nodigera (Brazil) RSP020 (Brazil) RSP024 (Brazil) bruchi (Argentina) bambusarum (Brazil) mexicana (Costa Rica) reclusi (Colombia) catharinae (Brazil) RSP063 (Guatemala) RSP065 (Brazil) cf. zeledoni (Costa Rica) zeledoni (Costa Rica) cooperi (Costa Rica) gagates (Dominican Republic) plebecula (Costa Rica) skwarrae (Mexico) brevicornis (Bolivia) RSP011 (Colombia) RSP014 (Venezuela) RSP056 (Panama) RSP026 (Costa Rica) lauroatlantica (Costa Rica) nigrocotea (Costa Rica) flavoguarea (Costa Rica) flavocotea (Costa Rica) RSP057 (Mexico) meganaranja (Costa Rica) longiceps (Costa Rica) longiceps (Costa Rica) RSP027 (Costa Rica) osa (Costa Rica) haberi (Costa Rica) lauropacifica (Costa Rica) schumanni (Bolivia) schumanni (Guyana) schumanni (Ecuador) species inhabiting the lower cloud forest envelope that can be found all the way to the Caribbean slope of the Cordillera Volcánica Central).
There are some similarities between the two systems. Both systems appeared to have evolved in the context of niche shifts. For Pseudomyrmex, the mutualism was initiated with a shift from closed to open habitats, where plants were more prone to herbivore pressure [9]. For Middle American Myrmelachista, the niche shift was probably linked to elevational changes. The opportunist relatives of the specialist clades are found in montane rainforests and cloud forests. The specialists are restricted to middle and lower elevations. The evolution of mutualistic Myrmelachista may be linked to downslope colonization in a lowland environment where the generalist arboreal niche is already packed with hostile arboreal ant competitors (e.g. Crematogaster, Azteca, Camponotus). The acacia system was a move to a high-energy system, requiring highly energetic rewards for conspicuous and pugnacious ants. The Myrmelachista system was the opposite, moving to a low-energy system and remaining 'under the radar' to other arboreal ants, with few overt rewards and relatively inconspicuous ant defenders. The extreme uniformity of traits across the mutualistic instances in the Myrmelachista system, an evolutionary déjà vu, may indicate that there are limited, almost deterministic solutions to the challenges posed by the environment.
Host plants in prominent examples of ant-plant symbioses typically bear prominent traits to attract ant partners [25], such as conspicuous preformed domatia and extrafloral nectaries [5]. These are lacking in the Myrmelachista system, or at most, there is a slight swelling of shoot tips that facilitates colonization [18,19]. The factors predisposing the development of mutualism are thus unclear. However, some plant traits correlate with the mutualism, even among unrelated plant hosts. One of the most prominent is dwarfism. Host plants of specialist Myrmelachista are small treelets, rarely more than 4 m tall, flowering and fruiting in the understory of small forest gaps. In contrast, the congeners not involved in symbiotic interactions with Myrmelachista are all tall canopy trees. This pattern holds for hosts in the Lauraceae (Mespilodaphne and Ocotea) and Meliaceae (Guarea). Dwarfism may have evolved independently of ant presence and been a pre-existing resource for ant colonization, or dwarfism might result from the mutualistic interaction. A potential dissolution of the mutualism might occur when host plants achieve a certain size, and plant growth might need to match colony growth to maintain the interaction [19]. Therefore, plants may thrive by limiting the nest space ants can occupy and effectively protect. Another potential explanation is that ants act as stressors for plant development, directly preventing growth. Being in a small understory treelet may have facilitated the radiation of specialist Myrmelachista. A second characteristic that Myrmelachista host plants share is a thin cortical layer and soft pith, traits favouring entry by stem-nesting ants [26]. Similar traits are observed in other mutualistic systems involving formicine ants (Cladomyrma in Southeast Asia [27]; Petalomyrmex in Africa [28]). In those systems involving other formicines, plant partners seem to be recent lineages [5]. Divergence timing estimates for the host plants in the Myrmelachista system are yet to be explored. However, molecular evidence shows that Guarea tafae-malekui (host for Myrmelachista flavoguarea) is confined to Costa Rica and has arisen recently from the South American Guarea stock [29] and that the genus Mespilodaphne is a short-branched lineage within the Ocotea complex [30].
Throughout Costa Rica in the wet forest understory, one will find one or more species of host treelets, nearly 100% of whose stems will be occupied by tiny yellow ants. These ants look the same and act the same. Longino distinguished seven species based on queen colour, size and head shape [13]. Worker differences are extremely subtle to non-existent. Thus, to find these species divided among two independent clades in our UCE phylogeny was completely unexpected and begs the question 'what factors, both ecological and genomic, could produce such extreme similarity between two independent clades'? A potential explanation for the convergence itself is that the evolutionary pathway from opportunist to specialist is deterministic, with the evolution of host plant specialization being a similar outcome. Deep homology (i.e. the homology of complex regulatory circuitry inherited from a common ancestor) [31,32] may be providing the foundation for independently evolved traits, such as loss of integumental pigmentation in workers and host-searching behaviour in founding queens.
However, the deterministic nature of specialization alone cannot explain the convergence in the timing of the origination of the two clades of specialists in the Middle American radiation. Therefore, one could speculate that the presence of common selective agents precipitating specialization occurred simultaneously on the two slopes of Costa Rica, e.g. environmental factors, plant host availability and niche competition. It is possible that simultaneous ecological conditions for mutualism may occur in different regions within the range of Myrmelachista, such that a specialist niche is frequently available but may be dynamic, shifting over space and time. A combination of precursor material for convergence and constrained pathways to mutualism may allow specialists to evolve repeatedly, similarly filling the niche each time it appears [33].
Another explanation is that occasional introgression with opportunist species could transport the genomic pieces linked to the specific lifestyle across different lineages, similar to what is observed with Heliconius butterflies, in which genes for adaptive wing patterns cross species boundaries [34,35]. In this scenario, rather than a sporadic gene exchange, specialist species of Myrmelachista could form an admixed 'population', with mass hybridization generating the persistence of the specialist phenotype. For pairs of closely related species, we might expect the repeatability of genomic evolution after hybridization events [36,37]. The genomic landscape (i.e. composition, conserved recombination maps, shared mechanisms of selection) of specialist Myrmelachista could, therefore, allow for the specialist phenotype to be 'floating' above species boundaries.
Additionally, young clades might experience comparatively faster speciation and extinction rates [38]. The specialization in Myrmelachista could be an evolutionary dead end, frequently occurring but rapidly going extinct-such that there may have been a long history of specialists stretching far back in time. While specialization in narrow ecological niches may allow species to survive in competitive or unique environments (e.g. [39,40]), it may also imply that specialized taxa may be doomed to extinction before they can transition to a more stable state over evolutionary time [41]. The growing empirical evidence available from clades with detailed phylogenetic and ecological information suggests contrasting outcomes, with niche specialization pushing clades towards the edge of extinction (e.g. [42,43]) or not (e.g. [40,44,45]).
One ASR (electronic supplementary material, figure S12) favoured reversal over convergence. Instead of recent convergence, a single old lineage of specialists may have experienced multiple instances of reversal to the niche of opportunist nesting (see electronic supplementary material, tables S3-S5 and figures S11-S14). Are convergence of specialization and numerous reversals to the opportunistic condition equally probable? Although our results support the former, in processes similar to those described above, specialist species could harbour older genetic elements that allow the reappearance of opportunist lineages. This is similar to what was recently observed in fruit flies, with hybrids rapidly regressing to one of the parental species, becoming nearly indistinguishable in morphology and behaviour [46]. The loss of specialization in Myrmelachista, therefore, could be linked with an apparent opportunist phenotype having evolved mechanisms to inhibit specialization.
While it is well established that ant-plant mutualisms have evolved multiple times independently across different ant lineages, it is uncommon to find several originations of the behaviour within individual ant genera. Our results reveal that plant-ant specialization has evolved at least three times in the unique Myrmelachista system. Furthermore, this 'under the radar' system includes an instance of remarkable convergent evolution. The déjà vu within Middle American Myrmelachista shows that independently evolved ant-plant mutualisms can be strikingly similar in tempo and mode of evolution. The cryptic nature of the Myrmelachista-plant mutualisms in the Middle American clade makes the study of its symbiotic origins challenging, but a comparison of independent origins could reveal the degree of convergence versus divergence in the underpinning mechanisms. Questions about the selective forces driving and maintaining these symbiotic interactions are open for further studies. Therefore, exploring the genomic signatures of ant-plant mutualistic interactions should be the next step, revealing mechanisms of selection and functional space [47].
Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Rodolfo da Silva Probst (probstrodolfo@gmail.com).
(ii) Data, code and materials Raw Illumina reads and contigs representing UCE loci have been deposited at Dryad (https://doi.org/10.5061/dryad.4j0zpc8m9) [48]. A complete list of NCBI (National Center for Biotechonology Information) accession numbers and further details of samples and voucher specimens can also be accessed at Dryad, along with all UCE matrices, Trinity contigs, tree files, unfiltered UCE alignments and additional data analysis files (partitioning schemes and log files) [48]. The Phyluce package and associated programs can be downloaded from https://www.ultraconserved.org/#software or its GitHub repository [49]. The ant-specific baits used for UCE enrichments are available from Arbor Biosciences (arborbiosci.com/genomics/targeted-sequencing/mybaits/mybaits-expert/mybaits-expert-uce/). The UCE bait sequence file is available at Figshare (https://figshare.com/arti- cles/dataset/Hymenoptera_UCE_and_Exon_bait_sets_from_Branstetter_et_al_2017_/4630375/1).
Our sample set comprised 45 specimens. Thirty-eight were species of Myrmelachista: one from the Caribbean, 21 from Middle America and 16 from South America. The remaining seven specimens were outgroup taxa, selected based on a phylogeny of the Formicinae by Blaimer et al. [10]. We employed a phylogenomic approach, combining target enrichment of UCEs with multiplexed, next-generation sequencing [23,24]. All molecular work was performed following a phylogenomic pipeline combining target enrichment of UCEs with multiplexed, next-generation sequencing described in Branstetter et al. [24] and included DNA extraction, library preparation, UCE enrichment and sample pooling. Sequencing was performed on an Illumina HiSeq 2500 at the University of Utah genomics core facility or on a HiSeq X at Novogene Corporation, Inc. For UCE enrichment, we used an ant-customized bait set ('ant-specific hym-v2') targeting 2524 UCE loci across Hymenoptera [24].
We carried out multiple phylogenetic reconstructions on the final matrix (1946 UCE loci for a 75% complete matrix) using the likelihood-based program IQ-Tree v2.0.4 [50] to explore the effects of data partitioning, including entropy-based Sliding-Window Site Characteristics (SWSC-EN) partitioning [51], genealogical concordance, topological structure and sensitivity analyses (additional information in the electronic supplementary material). We used the coalescent program Astral-III [52] to account for gene tree discordance arising from independent lineage sorting while inferring a species tree. We inferred a dated time tree for Myrmelachista and its species using BEAST2 v2.6.3 [53] and MCMCTree v4.9j, part of the PAML package [54].
All the phylogenetic analyses were conducted using either the CIPRES platform [55] or the Center for High Performance Computing at the University of Utah.
In this study, we explored ASR for nesting habit for our Myrmelachista dataset using three distinct approaches and several different programs: MP [56], ML [56][57][58] and rjMCMC [59] (see electronic supplementary material, "Supplementary Text" for a complete description of methods). Ancestral trait estimation for nesting habit was based on known natural history data (collection data, specimen labels or literature information) for all Myrmelachista taxa included in our dataset. Terminals were trimmed to avoid duplicate specialists that could influence analyses outputs. We classified nesting habit as a discrete binary character: (0) opportunistic nester and (1) specialized live-stem nester. For the input tree, we selected the ML tree inferred using the SWSC-EN partitioning method owing to its improved sensibility to UCE mutation rates when compared with concatenated or partitioned by locus [51]. Therefore, results and interpretation are dependent on this specific topology-which could potentially extend to ASR outputs. For an extensive explanation of model approaches for each reconstruction analysis, please refer to the electronic supplemental material.
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20241214 Downloaded from https://royalsocietypublishing.org/ on 12 July 2024
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20241214 Downloaded from https://royalsocietypublishing.org/ on 12 July 2024
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 291: 20241214 Downloaded from https://royalsocietypublishing.org/ on 12 July 2024