Romiguier et al. sequence 65 genomes to produce a phylogenetic tree of all ant subfamilies, with the subterranean leptanillomorph clade as the sister group of all other ants. A natural selection hotspot is detected during the emergence of formicoid clade, which is an ant clade grouping the most complex forms of sociality.
The evolution of eusociality has allowed ants to become one of the most conspicuous and ecologically dominant groups of organisms in the world. A large majority of the current 14,000 ant species belong to the formicoids, 1 a clade of nine subfamilies that exhibit the most extreme forms of reproductive division of labor, large colony size, 2 worker polymorphism, 3 and extended queen longevity. 4 The eight remaining non-formicoid subfamilies are less well studied, with few genomes having been sequenced so far and unclear phylogenetic relationships. 5 By sequencing 65 genomes, we provide a robust phylogeny of the 17 ant subfamilies, retrieving high support to the controversial leptanillomorph clade (Leptanillinae and Martialinae) as the sister group to all other extant ants. Moreover, our genomic analyses revealed that the emergence of the formicoids was accompanied by an elevated number of positive selection events. Importantly, the top three gene functions under selection are linked to key features of complex eusociality, with histone acetylation being implicated in caste differentiation, gene silencing by RNA in worker sterility, and autophagy in longevity. These results show that the key pathways associated with eusociality have been under strong selection during the Cretaceous, suggesting that the molecular foundations of complex eusociality may have evolved rapidly in less than 20 Ma.
To build a comprehensive phylogenetic tree including representatives of all extant ant subfamilies, we conducted two main types of analyses on the 4,300,911 amino acids from 4,151 single-copy protein-coding genes that we generated from 83 species (including 9 hymenopteran outgroups and 65 newly sequenced genomes; STAR Methods). First, we performed supermatrix approaches, where all the genes were concatenated for estimating a single species tree. Second, we performed supertree approaches, where the species trees were estimated from all gene trees. The two best resulting trees from each approach are summarized in Figure 1. Each node of the tree is supported with maximal support by the supermatrix analyses, and nearly all nodes (78/82) are congruently supported by both the supermatrix and supertree approaches. Importantly, the deepest and most important nodes of the ant phylogeny, including every relationship among the 17 ant subfamilies are maximally supported by both the supermatrix and supertree approaches.
Our results confirm that the so-called poneroid subfamilies (Ponerinae, Paraponerinae, Agroecomyrmecinae, Proceratinae, Apomyrminae, and Amblyoponinae) are monophyletic, [6][7][8][9] rather than paraphyletic. [10][11][12] Both the supermatrix and supertree provide maximum support for the poneroid monophyly. The relationships among poneroid subfamilies are also all congruently supported by both approaches (Figure 1), including for the Paraponerinae (one living species) and Agroecomyrmecinae (two living species), which are inherently difficult to relate with other subfamilies as their deep phylogenetic divergence result in long branches. An analysis of ultraconserved elements (UCE) markers for an increased dataset of 166 taxa (STAR Methods) retrieved the same subfamily relationships (Figure S1), further supporting the view that our phylogeny (Figure 1) is robust.
These analyses are important because the rooting of the ant phylogeny has been a controversial issue since the discovery of Martialis heureka, an extremely rare and morphologically divergent ant species that was initially inferred as the sister group of all other ants. 10 Some studies suggested that Leptanillinae is the subfamily sister to all other ants, 11,13 whereas a recent study suggested that Leptanillinae and Martialinae may form a monophyletic group. 9 Our analyses support this last hypothesis, with Leptanillinae + Martialinae forming a clade (hereafter referred to as the leptanillomorph clade) that is the sister group to all other ants (Figure 1). This conclusion was also supported by two further analyses controlling for outgroup composition, which has been suggested to affect the rooting of the ant tree. 9 First, we built an alternative supermatrix of 2,343 genes (983,951 amino acids) containing many outgroups, with 115 non-ant aculeate species borrowed from published transcriptomes. 14 These analyses revealed that the rooting of the ant phylogeny and subfamily relationships The topology is inferred according to the main supermatrix analysis (1,692,052 amino acid sites after cleaning; maximum likelihood; IQ-TREE PMSF C20 profile mixture model; Figure S1D); ultrafast bootstrap support is displayed first when node support is not maximal. Node support of the main supertree analyses (from gene trees of the 1,552 alignments of more than 500 amino acid sites; gene trees inferred with model search in IQ-TREE, and species tree with ASTRAL; Fig- ure S1E) is displayed second when not maximal (dash when node shows incongruence with the supermatrix analysis). Time divergence has been estimated using chronos with 12 calibration nodes (Table S2). Ant images from Alex Wild; used with permission. See also Figure S1 and Tables S1 andS2.
are not affected by the inclusion of all these outgroups (Figure S1B). Second, using random combinations of outgroups (see STAR Methods for details), we always found the same rooting, with strong support for the leptanillomorph clade being sister to all other ants (bootstrap values: 100 for 109 trees and 99 for the remaining 6 trees). Because all species of the leptanillomorph clade are pale, blind, and have a similar hypogeic ecology, this suggests that some early ants may have escaped extinction by retreating to these stable subterranean habitats before other lineages diversified by developing novel morphological and behavioral adaptations. 10 According to our divergence date estimates, the common ancestor of the leptanillomorphs lived around the Jurassic-Cretaceous boundary (145 Ma) shortly after the common ancestor of all extant ants (150 Ma). This suggests the possibility that subterranean lifestyles existed in the ancestors of extant ants or, more likely, that a hypogeic lifestyle originated at an early stage in the history of leptanillomorphs. This result contrasts with the fossil evidence because the earliest-known fossilized crown ants were not specialized to subterranean habitats, and they come from Burmese amber deposits that are 99 Ma old. 15,16 Set against our divergence date estimates, this indicates a gap in the ant fossil record of 50 Ma, further emphasizing an existing discrepancy between fossils and molecular data when it comes to the question of ant origins. 17 This is reminiscent of the debate on the origin of placental mammals, which are estimated be in the middle Cretaceous by molecular data, whereas there is no fossil record before the K-T crisis. 18 It has been suggested that the lack of fossils may stem from the occurrence of only a few lineages of placental mammals and perhaps small populations sizes during the Cretaceous. 19 Similarly, it is possible that the abundance of crown ants was low at first and only sufficiently increased with the rise of angiosperm 6 to be represented in the fossil record. Alternatively, there may be methodological biases leading to overestimation of divergence dates 20,21 or incorrect phylogenetic placement of early ant fossils. 16 The pervasive positive selection is associated with the origin of the socially diverse formicoid clade To investigate the molecular changes associated with the evolution of complex eusociality, we conducted positive selection analyses on the 4,151 ortholog genes of the 75 ant genomes (STAR Methods). The percentage of genes under positive selection varied greatly among the 38 branches ranging from 0% to 2.6% in a single branch (Figure 2). Strikingly, the branch leading to the Colors indicate the percentage of genes significantly under positive selection (aBSREL analysis) on each branch. Colony size and polymorphism data represent the maximum observed value in the genus, and they have been extracted from the literature. 23 Significantly enriched functional categories (biological process) under positive selection are represented as a word cloud for the formicoid branch. The size of the font is proportional to the p value (Fisher's exact test; larger font indicating the most significant ones, from 0.011 to 0.047). Darker colors indicate the three functions, with the highest numbers of significant annotated genes under positive selection. See also Data S1.
formicoid clade stood out as a clear outlier with a 30-fold higher rate of positive selection compared with the average of other tree branches. There were 110 positively selected genes on the branch subtending the formicoid clade, whereas the average number of genes with positive selections was only 3.1 in other branches (maximum value, 20 genes). This finding is particularly remarkable, given that the genes considered in our analysis are highly conserved universal orthologs across Hymenoptera. 22 This indicates that extensive molecular changes in wellconserved core genes occurred along the branch giving rise to the formicoids. By contrast, there was no evidence of a further burst of positive selection later in the evolutionary history of the formicoids, including in the multiple branches leading to the most complex eusocial species (Figure 2). This suggests that most of the genetic innovations that are specific to complex eusociality in formicoids occurred in less than 20 Ma during the early Cretaceous (Figures 1 and2).
Functional enrichment analyses for the formicoid branch revealed that histone acetylation was the most significantly overrepresented function among the 110 positively selected genes. Histone acetylation is well known for controlling transcriptional activity, 24 reprograming the foraging behavior of the major worker caste into the minor worker caste, 25 colony activity rhythms, 26 and the longevity/fecundity trade-off in workers. 27 Histone acetylation is also involved in the caste determination of honeybees through the effect of royal jelly, 28 and it has been identified as a key caste-specific enhancer of transcription regulating the differential larval development of queens and workers. 29 Interestingly, our analyses revealed positive selection on histone acetyltransferase (Data S1), a gene previously linked to functions potentially relevant to eusociality, such as the regulation of worker polymorphism. 30 The second most significant function was autophagy. Autophagy has repeatedly been shown to be essential for queen lifespan extension 31,32 and the castespecific programed cell death responsible for the divergent ovary development in queen and worker honeybees. 33 Finally, the third most significant function was gene silencing by RNA (Figure 2). From our results, we retrieved the gene Tudor-SN, which is a candidate for controlling worker sterility in honeybees. 34 Altogether, these results reveal that the common ancestor of the formicoid ants underwent important genomic changes relative to the regulation of gene expression (e.g., histone acetylation and gene silencing RNA) and soma maintenance (e.g., autophagy).
These changes may have been important in allowing the evolution of extreme division of labor in formicoid clade, which is the ant clade comprising the vast majority of species exhibiting extreme forms of complex eusociality (e.g., maximum colony size of 3 million polymorphic workers in formicoid Dorylus species compared with a maximum of 50,000 monomorphic workers in some poneroid species of the genus Leptogenys 23 ). However, given that the nine formicoid subfamilies also display some species with less complex levels of eusociality, this implies that although the genomic changes that occurred during the early Cretaceous may have favored the emergence of extreme division of labor and more overtly complex forms of eusociality, they did not necessarily lead to such changes in social organization. Knowing which selective pressures triggered these dramatic molecular changes remains an open and intriguing question.
By providing genome-wide data for all ant subfamilies, this study infers the leptanillomorph clade as the sister clade to all other extant ants and clarifies controversial subfamily relationships that will be important for further comparative studies in ants. The comparative genome analysis also reveals important changes in key molecular pathways implicated in the differential gene expression of queens and workers. This burst of molecular innovations, which occurred over 20 Ma in the early Cretaceous, possibly played an important role in facilitating the evolution of complex eusociality, including the large colony sizes, extensive caste polymorphism, and extreme fecundity/longevity of queens that characterize multiple lineages of formicoids.
Detailed methods are provided in the online version of this paper and include the following: