Latent gene network expression underlies partial re-evolution of a polyphenic trait in the worker caste of ants

  1. Angelly Vasquez-Correa
  2. Johanna Arnet
  3. Travis Chen
  4. Ehab Abouheif  Is a corresponding author
  1. Center for Evolutionary and Organismal Biology, Women’s Hospital, School of Medicine, Zhejiang University, China
  2. McGill University, Department of Biology, Canada

eLife Assessment

This important study explores whether complex structures that are lost during evolution can re-evolve, which is a long-standing debate in evolutionary and developmental biology. The authors demonstrate that re-evolution can occur if the gene regulatory network that underlies the development of complex traits is maintained. The evidence supporting its conclusions is convincing and the work will be of interest to those studying the evolution and development of complex traits.

https://doi.org/10.7554/eLife.110148.3.sa0

Abstract

Polyphenisms–where alternative phenotypes develop from a single genome in response to environmental cues–are not only widespread in nature, but also occur at multiple levels of biological organization, from cells to individuals to societies. Polyphenism is thought to promote phenotypic diversification through the gain, loss, and re-evolution of alternative phenotypes. After the origin of a polyphenism, one of the alternative phenotypes often retains the developmental capacity to produce the ancestral trait, thereby permitting the other to evolve rapidly. Yet, little is known about the developmental processes underlying the re-evolution of polyphenic traits, and how they may produce phenotypic diversification. Here, we address this question by focusing on the caste polyphenism in ant societies, which produces a winged queen caste and a wingless worker caste in a single colony in response to environmental cues. We show, in a hyperdiverse group of ants, that a caste-specific trait called the ocelli (three simple eyes on the dorsal head) is always present across queen castes but was lost and partially re-evolved multiple times, giving rise to novel patterns (one ocelli) in the worker castes. Surprisingly, we discovered that a hidden (latent) expression of the ocelli gene regulatory network in worker castes that lost ocelli underlies the partial re-evolution of ocelli in this group. We therefore propose that latent developmental potentials may generally persist across polyphenic systems, including ant castes, and may facilitate the partial re-evolution of novel phenotypic patterns.

eLife digest

The re-emergence of traits lost over evolutionary time has long fascinated biologists and sparked debate. Yet we still do not fully understand the developmental and evolutionary mechanisms that allow lost traits to reappear. This question is particularly relevant in species that exhibit polyphenism – the ability of the same genome to produce different forms in response to environmental signals.

Ants are a striking example of polyphenism in action. Within a single colony, the same genome can produce morphologically distinct queens, males and workers. One of the most remarkable polyphenic traits is the presence of ocelli, three simple light-sensing eyes on the top of the head. Queens and males consistently possess ocelli, but in workers they have been repeatedly lost and gained during evolution. Across different species, workers may have one, two or all three ocelli, suggesting that the trait has evolved independently multiple times.

This remarkable evolutionary flexibility led Vasquez-Correa et al. to propose that the developmental programme for forming ocelli remains dormant in workers, making it easier for the trait to reappear. To test this idea, the researchers reconstructed the evolutionary history of worker ocelli across the hyperdiverse ant subfamily Formicinae.

Their ancestral state reconstruction revealed a single evolutionary gain of worker ocelli, followed by multiple losses and independent reappearances. To understand how this repeated re-evolution occurred, the researchers examined ant head development, focusing on the eye-antennal disc. This tissue gives rise to the eyes, antennae, ocelli and head capsule.

Using hybridization chain reaction, a technique that visualizes gene activity in fixed tissues, they tracked the expression of genes involved in ocellus development throughout development. Remarkably, even in workers that lacked ocelli, key developmental genes were briefly active, suggesting that the underlying gene regulatory network remained intact. This retained developmental programme appears to have enabled the repeated evolution of ocelli in Formicinae workers and soldiers.

Polyphenism is widespread in nature, occurring at many levels of biological organization, from individual cells to complex societies. The findings of Vasquez-Correa et al. suggest that dormant developmental programmes may commonly persist after traits are lost, providing hidden evolutionary potential. If similar latent gene activity exists in other organisms, it could help explain how lost traits repeatedly re-evolve in new forms, offering fresh insights into the origins of evolutionary novelty.

Introduction

Polyphenism is a form of developmental plasticity where alternative phenotypes develop from a single genome in response to environmental cues (Nijhout, 2003). It is a phylogenetically widespread feature of plants and animals that has evolved at different levels of biological organization (Hanna et al., 2024; West-Eberhard, 1989; West-Eberhard, 2003). For example, at the population level, the mouth form polyphenism in nematode worms produces alternative big tooth (omnivorous) or small tooth (bactivorous) mouth forms that develop in response to pheromones, crowding, salt concentration, temperature, and culturing substrate (Bento et al., 2010; Bose et al., 2012; Ragsdale et al., 2013; Werner et al., 2018). At the colony-level, caste polyphenism in eusocial insects produces morphologically differentiated queen and worker castes that develop in response to temperature and nutrition (Chandra et al., 2018; Evans and Wheeler, 2001; Korb, 2025; Rajakumar et al., 2024). And finally, at the cellular-level, polyphenism occurs within a multicellular individual, where a single genome gives rise to differentiated cell-types during development, such as between germline and somatic cells, in response to internal cues like morphogen gradients (Brunet and King, 2017; Davison and Michod, 2021; Devlin et al., 2023). Polyphenism has been proposed to promote, at the macroevolutionary scale, phenotypic diversification through the gain, loss, and re-evolution of alternative phenotypes (West-Eberhard, 2003). This is based on the idea that, once a polyphenic trait originates, one of the alternative morphs retains the capacity to produce the trait in the genome while the other is freer to evolve. This hypothesis has received support from a comparative study of mouth form polyphenism across 90 species of nematode worms showing a phylogenetic association between the gain and loss of alternative mouth form phenotypes and the phenotypic diversification of mouth parts (Susoy et al., 2015). Another supporting example is fat synthesis in parasitic wasps, which revealed an association between developmental plasticity and the loss and subsequent re-evolution of fat synthesis in one species (Peters et al., 2017; Visser et al., 2010; Visser et al., 2021). Yet, the underlying developmental and genetic processes facilitating the gain, loss, and re-evolution of polyphenic traits remain poorly understood. (Forni et al., 2026; Sommer, 2020; West-Eberhard, 2003).

Here, we address this question by focusing on caste polyphenism in the eusocial colonies of ants, which consists of a morphological division of labor between a winged reproductive queen caste and wingless non-reproductive worker caste in almost all 16,962 valid described ant species (AntWeb, 2026; Ward, 2014). The differential expression of polyphenic traits, such as wings, that develop in queens but not workers, are called ‘caste-specific’ traits (Miura, 2005). It has been shown that polyphenic traits, including caste polyphenism in ants, are produced during development by the differential expression of highly conserved gene regulatory networks (GRN) in response to environmental cues (Abouheif and Wray, 2002; Béhague et al., 2018; Casasa et al., 2021; Davidson et al., 2023; Lenuzzi et al., 2023; Rajakumar et al., 2024; Vizueta et al., 2025). However, how the expression of these GRNs influences the evolution of caste-specific traits in ants remains unknown.

Here we focus on the ocelli, which are three small single-lens eyes on the dorsal head of most flying insects. Ocelli complement the function of the compound eyes by mediating orientation using polarized light and in the synchronization of daily activity (Buschbeck and Bok, 2023; Krapp, 2009). We investigate the evolution of ocelli in a hyperdiverse subfamily of ants (Formicinae), where they are universally present in the winged reproductive caste (queens and males) as three large ocelli that aid in mating flights and dispersal (Moser et al., 2004; Narendra et al., 2016). In contrast, ocelli in the wingless worker caste are evolutionarily labile, showing dramatic variation across species in the presence/absence or number of ocelli in the worker caste (Johnson and Rutowski, 2022; Narendra et al., 2016; Narendra and Ribi, 2017; Schwarz et al., 2011; Figure 1A). In some species, adult workers completely lack ocelli, such as in Camponotus floridanus, while in others they are present and vary in number–there are species whose adult workers have all three ocelli or just one single ocellus, and these can be present in all or in only a subset of workers (Figure 1A). For example, workers in Cataglyphis bicolor have all three ocelli, which function in light sensing and navigation, acting as a celestial compass that provides crucial directional information (Fent and Wehner, 1985; Figure 1A). In contrast, all workers of Polyrachis bihamata have just a single medial ocellus (Hung, 1967; Figure 1A), and in workers of Dinomyrmex gigas, a single medial ocellus evolved only in a subset of individuals in the worker caste called ‘soldiers’ (or major workers) with large heads, but are absent in other individuals called ‘minor workers’ with small heads (AntWeb, 2026; Figure 1A). How this dramatic variation in ocelli in the worker caste of formicine ants has evolved remains poorly understood.

The presence and absence of ocelli in queens and workers across ants.

(A) Ocelli develop in the winged reproductive castes across species of Formicinae ants exemplified by C. floridanus, C. bicolor, P. bihamata, and D. gigas (white dashed circles). Ocelli in the worker caste of formicine ants (white dashed circles) are evolutionarily labile, showing no ocelli (C. floridanus), all three ocelli in all individuals of the worker caste (C. bicolor), or only one ocellus in only the soldiers (P. bihamata and D. gigas). (B) Schematic representation of the ocelli GRN in Drosophila melanogaster adapted from Figure 5A from Jean-Guillaume and Kumar, 2022. The genes investigated in this study are highlighted in orange. Arrowheads indicate activation, and bars indicate repression. Queens are scaled to 1 mm, and workers and soldiers are scaled according to the queen of each species. Asterisks indicate that C. floridanus was used for gene expression studies. Photos from Antweb (AntWeb, 2026).

This evolutionary lability of ocelli across the worker castes of formicine ants also provides an opportunity to understand how the GRN underlying development of ocelli influenced the evolution of this caste-specific trait. Ants are holometabolous insects, in which adult body parts develop from imaginal discs, semi-independent clusters of cells in the larvae (Held, 2002; Koch and Abouheif, 2019). In the fruit fly D. melanogaster, ocelli develop from the eye-antenna imaginal disc located at the ventral region of the head capsule. The eye-antenna imaginal disc also gives rise to the head capsule, eye, antenna, and maxillary palps (Held, 2002). D. melanogaster is the only insect where the GRN underlying ocelli development has been well characterized at the third larval stage (Blanco et al., 2009; Domínguez-Cejudo and Casares, 2015; Sabat et al., 2017; Figure 1B). The gene orthodenticle (otd; formerly known as ocelli-less) is a selector gene that is necessary for specifying the ocellar region in the developing head capsule. otd expression in the ocellar domain, together with other genes like hedgehog (hh), initiates the development of the ocellar region and the three ocelli (two lateral ocelli and one medial ocellus) in the eye-antennal disc. The activation of these genes regulates the expression of downstream genes, such as otd regulating the expression of defective proventriculus (dve; Blanco et al., 2009; Jean-Guillaume and Kumar, 2022; Yorimitsu et al., 2011), whereas hedgehog (hh) activates a portion of the retinal determination network, such as eyes absent (eya), twin of eyeless (toy), sine oculis (so), and atonal (ato). These retinal determination genes have been shown to be regulated by independent regulatory enhancers from the compound eye (Blanco et al., 2009; Blanco et al., 2010; Jean-Guillaume and Kumar, 2022; Figure 1B). Because the eye-antennae disc and the ocelli GRN have only been well characterized in D. melanogaster, it remains unknown whether they are conserved in ants.

To understand how the GRN underlying ocelli development may have influenced the evolution of this caste-specific trait, we first inferred the evolutionary history of ocelli in adult workers across the Formicinae using ancestral state reconstruction. We then characterized the eye-antennae disc in ants using three genes, eyeless (ey), distal-less (dll), and otd-1, which are known to mark the eyes (ey), the antenna (dll), and the head capsule and ocelli (otd-1). This characterization allowed us to investigate the expression of five key genes in the ocelli GRN, otd-1, hh, toy, eya, and so, during development of the winged and wingless castes across two formicine species.

Results

Partial reversion to a single ocellus occurs three times independently within the tribe Camponotini (Formicinae)

We first performed an ancestral state reconstruction to infer the evolutionary history of worker ocelli across the subfamily Formicinae. Our ancestral state reconstruction inferred a single re-gain of worker ocelli at the base or early within the Formicinae (Figure 2). Subsequent to the gain of ocelli early in the evolution of the Formicinae, our analysis inferred a single, well-supported loss of worker ocelli at the base of the tribe Camponotini (Figure 2). Following this single loss, we inferred three (well-supported) independent and partial re-evolution to a single medial ocellus in three different genera: Camponotus gibbinotus, P. bihamata, and D. gigas (Figure 2). In P. bihamata, the single ocellus occurs in all workers in the colony, whereas in C. gibbinotus and D. gigas the single ocellus occurs only in the large-headed soldiers (Figure 1). We therefore investigated the developmental role of the ocelli GRN underlying these independent partial reversions to a single medial ocellus in this tribe of ants.

Ancestral state reconstruction of ocelli reveals three well-supported reversions of ocelli in the worker caste in the tribe Camponotini.

Maximum clade credibility tree of formicine ants from Blaimer et al., 2015. Ancestral state reconstruction for the presence (blue-colored circles) and absence (red-colored circles) of ocelli based on stochastic character mapping. Each pie chart for the nodes represents the posterior probabilities, scaled by the weight of evidence for each model. Species used to analyze ocelli GRN expression are highlighted in gray, and the species that re-evolved one ocellus is indicated as partial re-evolution. Three tribes within the Formicinae are marked (Melophorini, Plagiolepidini, Lasiini, Myrmelachistini and Camponotini) by arrows.

A fate map characterizing the development of the head capsule, antennae, eyes, and ocelli within the eye-antenna disc in ants

To understand whether the GRN underlying the development of ocelli influenced the reversions of this caste-specific trait in workers, we first had to characterize the development of the eye-antenna disc in ants. In D. melanogaster, the eye, antennae, maxillary palps, ocelli, and head capsule develop from the eye-antenna disc, which is segregated into regions marked by the expression of highly conserved developmental genes (Haynie and Bryant, 1986; Held, 2002). In the Florida carpenter ant C. floridanus, we found that, similar to D. melanogaster, expression of otd-1 marks the precursor regions of the head capsule and ocelli (Figure 3—figure supplement 1), ey marks the precursor regions of the eyes, and dll marks the precursor regions of the antennae (Figure 3). During the first larval instar, expression of otd-1 emerges primarily in the head capsule in the middle part of the disc between the antenna and compound eye (Figure 3B). In contrast, dll and eya expression delineate the precursor regions of the antenna and eye (Figure 3C and D). During the second and third larval instar, otd-1 is expressed in the developing head capsule and ocelli in the medial region of the disc (Figure 3G and L), while dll is confined to the antenna and ey to the compound eye region (Figure 3H, I, M and N). Finally, during the fourth (final) larval instar, a developmental threshold mediated by juvenile hormone acts as a switch point to determine whether larvae will develop either into a minor worker or soldier (MacMillan et al., 2025). Once larvae have been determined, expression patterns of otd-1, dll, and ey in worker-destined larvae (Figure 3P–T) or soldier-destined larvae (Figure 3U–Y) remain expressed in the same regions as in the second and third instars (Figure 3Q–T, V–Y). Together, our characterization shows that the eye-antennae disc and the regional identities within it, including the precursor region of the head capsule, ocelli, eyes, and antennae, are conserved in ants relative to Drosophila.

Figure 3 with 1 supplement see all
Characterizing development of the eye-antenna imaginal disc in worker castes of C. floridanus using orthodenticle-1 (otd-1), distal-less (dll), and eyeless (ey) gene expression to mark the developing head capsule and ocelli, antenna, and eyes.

Fluorescent images in panels A, F, K, P, U represent the development of the entire eye-antenna imaginal disc marked with the nuclear stain DAPI across all four larval stages, where the head capsule region is labeled as ‘Hc’, the antennal region is labeled as ‘An,’ and the eye region is labeled as ‘Eye’. Panels B, G, L, Q, and V represent the development of the head capsule (Hc) marked by the genes orthodenticle-1 (otd-1 in magenta); panels C, H, M, R, and W represent the antennal region marked by distal-less (dll) in green color; and panels D, I, N, S, and X represent the eyes (Eye) is eyeless (ey) (yellow); (A–E) First instar, images are to scale (white bar in A). Note: the green or yellow staining outside of the structures highlighted by the white arrows in panels C, D, and E is background noise. (F–J) Second instar, images are to scale (white bar in F), (K–O) third instar, images are to scale (white bar in K), (P–T) fourth instar worker-destined larvae, images are to scale (white bar in P). (U–Y) Fourth instar soldier-destined larvae, images are to scale (white bar in U). Scale bars = 100 µm.

Expression of the ocelli GRN is conserved in winged reproductive castes but is latent in species whose adult workers completely lack ocelli

We next asked whether the ocelli GRN is conserved in the winged reproductive caste (males) relative to Drosophila and whether it is expressed in workers that entirely lack ocelli as adults. We address these questions using two species C. floridanus and Polyrachis rastellata. We chose these two species because C. floridanus is closely related to C. gibbinotus and P. rastellata is closely related to P. bihamata, which are two of the species that our ancestral state reconstruction inferred independent partial reversions to single medial ocellus in the worker caste (Figure 2). In the winged male caste of C. floridanus, we found that otd-1, eya, and so are expressed where the three ocelli will develop, while toy and hh are expressed in the inter-ocellar region (the tissue that separates the three ocelli) (see white arrowheads in Figure 4A–B’ and Figure 5A–C’). Because these genes are similarly expressed within the ocellar region within the eye-antennal disc of D. melanogaster, we infer that the ocelli GRN is conserved in the winged reproductive castes in ants.

Latent expression of otd-1 and hh genes in the ocelli GRN in workers and soldiers of C. floridanus during the 4th larval instars.

Expression of orthodenticle-1 (otd-1) is yellow, and hedgehog (hh) is magenta. Early 4th instar; (A, B) males (C, D) soldiers and (E, F) workers. Late 4th instar: (A,’ B’) males (C’, D’) soldiers and (E’, F’) workers. Images are not to scale.

Latent expression of toy, eya, and so within the ocelli GRN in the developing workers and soldiers of C. floridanus.

Expression of toy (green), eya (yellow), and so (magenta). Early 4th instar; (A–C) males, (D–F) soldiers, and (G–I) workers. Late 4th larvae stage; (A’–C’) males, (D’–F’), soldiers, and (G’–I’). Images are not to scale.

Surprisingly, we discovered that the ocelli GRN remains latently expressed in minor worker- and soldier-destined larvae of C. floridanus, which completely lack ocelli as adults. In soldier-destined larvae, all five genes remain expressed in the ocellar region within the eye-antennal disc at the beginning of the last larval instar (Figures 4C, D, 5C–F). By the end of this instar, otd-1 and hh remain expressed (Figure 4C’ and D’), but toy, eya, and so are either down-regulated or absent relative to their expression in the compound eye region (Figure 5D’–F’). In minor worker-destined larvae, three of the five genes (otd-1, hh, eya) remain expressed in the ocellar region during the early part of the last larval instar (Figures 4E, F, 5H), whereas toy and so are expressed in the compound eye region but absent (interrupted) in the ocellar region (Figure 5G and I). Furthermore, in P. rastellata, whose worker caste is composed of similarly sized individuals with no subcastes, we found that otd-1, eya, and so remain expressed in the ocellar region within the eye-antennal disc during the early part of the last larval instar (Figure 6). Together, our results show that despite the absence of ocelli in adult workers for millions of years, the expression of the ocelli GRN remains latent during larval development.

Latent expression of the ocelli GRN in the eye-antennal disc in worker larvae of Polyrachis rastellata at 4th larval stages.

Expression of selected genes otd-1 (yellow), eya (yellow), and so (magenta) at early 4th instar larvae. Images are not to scale.

Finally, we performed Scanning Electron Microscopy (SEM) in C. floridanus male, soldier, and minor worker pupae. In minor worker and soldier pupae, we discovered the existence of rudimentary ocelli that appear at the beginning of pupal development, continue to be elaborated, and then are eliminated before they molt into adult workers (Figure 7). These ocelli rudiments are highly reduced relative to the fully functional ocelli found in male pupae. Finally, the pattern and timing of development of ocelli rudiments in the minor worker and soldier pupae coincide with the spatial expression and timing of interruption of the latent expression of the ocelli GRN. In soldier-destined larvae, ocelli GRN expression is interrupted later in development than in minor worker-destined larvae, and consequently, the ocelli rudiments in soldier pupae continue to develop longer and are more elaborated relative to those in minor worker pupae (Figure 7B–D, F–H). Therefore, expression of the latent ocelli GRN in the eye-antennae disc results in the development of ocelli rudiments in worker pupae and are then eliminated in adult workers.

Development of rudimentary ocelli in worker and soldier pupae of C. floridanus.

(A) SEM showing ocelli development in males at mid-stages of pupal development. SEM showing development of rudimentary ocelli on (B–E) soldiers and (F–I) minor workers during early (day 6–7), mid (day 16–17), and late (19-20) pupal development. These ocelli rudiments disappear prior to adult stage (E–I).

Discussion

Our developmental and evolutionary data provide evidence that the latent expression of genes in workers lacking ocelli as adults is part of a latently expressed ocelli GRN, which likely facilitated at least three independent evolutionary reversions of this trait in the worker caste of species within the Camponotini clade. The latent expression patterns of genes in the ocelli region of developing workers lacking ocelli as adults are the same as in males that will develop fully functional ocelli but are only interrupted late in larval development. Furthermore, the timing and pattern of these late interruptions coincide with the degree of development of rudimentary ocelli in minor worker and soldier pupae before they disappear in adults (Figure 7). This indicates that, although the expression of these genes is latent, they still retain the capacity to produce rudimentary ocelli in the pupal stage before they disappear in adults. And finally, although eya, toy, and so, are part of both the ocelli and compound eye GRNs in Drosophila, the ocelli GRN has its own distinct identity and the genes within this GRN have distinct regulatory elements and are selectively regulated (Jean-Guillaume and Kumar, 2022; Zimmerman et al., 2000). In ants, our data show that the selector gene for compound eye development in insects (eyeless / Pax-6) is expressed in the compound eyes and not ocelli, and the selector gene for ocelli (otd-1) is expressed in the ocelli and not compound eyes. We further show that the latent expression of toy, eye, and so are downregulated or absent in the ocellar region, but at the same time, are strongly. expressed in the regions of the compound eyes. Therefore, the mutually exclusive expression of the selector genes eyeless / Pax-6 in the compound eyes and otd-1 in ocelli suggests that, like in Drosophila, the compound eye and ocelli GRNs have distinct identities, ultimately leading to differential expression of downstream genes and production of different cell types; compound eyes are produced from multiple imaging-forming facets, while ocelli are produced from a single lens (Buschbeck and Bok, 2023; Jean-Guillaume and Kumar, 2022; Mishra et al., 2021). Altogether, our data show that the expression of these genes in workers lacking ocelli as adults is part of a latently expressed ocelli GRN.

Several hypotheses may explain how the ocelli GRN came to be latently expressed and maintained in developing workers that lack ocelli as adults. Perhaps the most simplistic hypothesis proposes that the presence of functional ocelli in adult queen and male castes maintains the ocelli GRN intact in the genome by keeping it under positive natural selection. This hypothesis assumes that this, as a side consequence, leads to expression of the ocelli GRN in the worker castes lacking ocelli. However, caste determination between queens and workers occurs through the action of a developmental threshold or switch, where a continuous environmental cue is translated into discrete phenotypic outcomes (Abouheif, 2021; MacMillan et al., 2025; Qiu et al., 2022; Rajakumar et al., 2024; Schultner et al., 2023). Once caste determination has occurred, the genome is expressed differentially during the developmental trajectories of queens and workers (Abouheif, 2021; Abouheif and Wray, 2002; Barkdull and Moreau, 2023; Béhague et al., 2018; Chandra et al., 2018; Khila and Abouheif, 2010; Qiu et al., 2022; Vizueta et al., 2025). These trajectories are decoupled, and consequently, can evolve largely independently (Abouheif, 2021; Powell et al., 2020; Vong et al., 2025). The dramatic variation in the number, size, and presence/absence of ocelli in worker castes across the Formicinae (see Figure 1) supports the largely independent evolution of ocelli in workers from those in queens, which always develop three ocelli. Furthermore, we observe similar patterns of variation in the wings and ovaries between queen and worker castes across ant species (Cronin et al., 2013; Khila and Abouheif, 2010; Monnin and Peeters, 2008; Rajakumar et al., 2018; Rajakumar et al., 2012). Therefore, while the presence of ocelli in males and queens maintains the ocelli GRN in the genome and creates a potential for expression of this GRN in developing workers lacking ocelli, this cannot solely explain how this latent expression became actualized (released and fixed) and what maintained it over millions of years.

One hypothesis for how this ocelli GRN became latently expressed and has been maintained in workers lacking ocelli is pleiotropy, which potentially results from the multiple roles that genes within the ocelli GRN play within the same imaginal disc (the eye-antennal disc). For instance, otd-1 and hh also determine the regional identity of the head capsule, while hh, toy, eya, and so also play key roles in compound eye development in specifying structures such as optic lobes, cone differentiation, and rhabdomeres development (Blanco et al., 2009; Domínguez-Cejudo and Casares, 2015; Jean-Guillaume and Kumar, 2022). Another example is the highly conserved developmental regulatory gene sonic hedgehog (shh), which plays a key role in limb development across animals. The vestigialization of hindlimbs in snakes and therefore the re-evolution of hindlimbs in extinct species across the phylogeny of the group is thought to be maintained during development through pleiotropic enhancers that drive shh expression. This means that the same enhancer (ZRS) that drives shh expression in the external genitalia also drives it (pleiotropically) in the developing limb region of snakes (Leal and Cohn, 2018). Future studies should attempt to explore whether expression of these conserved genes in multiple regions of the eye-antennal disc is driven by shared enhancers. Alternatively, we cannot rule out the hypothesis that the ocelli GRN has been co-opted to play a novel, yet currently unknown, function during worker larval development. Recent discoveries on the evolution of the wing GRN in ants provide support for this hypothesis. The wings, another nearly universal caste-specific trait in ants, develop in the reproductive male and queen caste, but are halted in the worker caste in response to environmental cues (Abouheif and Wray, 2002). The wing GRN is also found to be latently expressed in the wingless worker caste of ants and was thought to be functionless (Abouheif and Wray, 2002). However, it was recently discovered that the latent expression of this wing GRN in wingless worker caste acquired a novel function to generate big-headed soldiers in the hyperdiverse ant genus Pheidole (Rajakumar et al., 2018; Rajakumar et al., 2012).

Finally, our inference that this latent expression of the ocelli GRN in workers facilitated the partial reversion to a single medial ocellus is supported by: (1) the close phylogenetic relationship between species that lack ocelli in adult workers but retain a latent expression of the ocelli GRN (C. floridanus and P. rastellata), and those that underwent a partial phylogenetic reversion to a single medial ocellus (C. gibbinotus, D. gigas, and P. bihamata); (2) the presence of a developmental capacity or potential of the latent ocelli GRN expression to produce rudimentary ocelli in the pupal stage of C. floridanus workers that completely lack ocelli as adults; (3) the ability to experimentally induce only one, only two, or all three ocelli in similarly sized adult workers that normally lack them by applying high doses of Juvenile Hormone (JH) to worker-destined larvae in the ant Monomorium pharaonis (Li et al., 2024); and finally (4) in nature, the rare induction of a single medial ocellus by mermithid parasites in soldiers of Pheidole pallidula that typically lack ocelli in natural colonies (Laciny et al., 2019; Passera, 1976). The natural or experimental induction of worker individuals with only a single medial ocellus in different ant species also supports the inference that the single medial ocellus can be developmentally dissociated from the other two lateral ocelli. This suggests that reversion can facilitate the appearance of novel patterns of ocelli development in the workers if selected for.

In C. floridanus and P. rastellata, there is a latent ocelli GRN expression for all 3 ocelli, providing a springboard to facilitate the partial phylogenetic reversion to a single medial ocellus in C. gibbinotus, P. bihamata, and D. gigas. In the genus Polyrhachis, however, some species in the same subgenus as P. bihamata, such as P. bellicosa, have three ocelli. Because the phylogenetic relationships within this subgenus have yet to be resolved, the independent reversion of ocelli in the ancestor of this subgenus may have resulted either in a single ocellus as reflected in P. bihamata or in three ocelli as reflected in P. bellicosa, and two ocelli were subsequently lost giving rise to the single medial ocellus in P. bihamata (Hung, 1967). These possibilities further reinforce the different evolutionary pathways by which this latent potential may facilitate novelty after reversion.

Future functional, genomic, and comparative analyses of the ocelli GRN between larval stages, individuals within the worker caste, and species will ultimately reveal the architecture of the ocelli GRN and whether its underlying enhancers and promoters are modular or pleiotropic. This, in combination with manipulations of insect hormones, such as JH and ecdysone, would also elucidate whether variation in the size, presence/absence, number of ocelli is regulated by continuous or switch-like developmental mechanisms. And finally, determination of the organismal and ecological function of the ocelli will be important to understand the adaptive significance of the latent expression of the ocelli GRN at both the individual and colony-level.

More broadly, our findings suggest that the ancestral and latent GRN expression (also known as ancestral developmental potential) we observed may generally underlie polyphenic systems, including caste-specific traits in ants and other eusocial organisms. We therefore propose that ancestral developmental potentials facilitate the re-evolution of polyphenic traits (West-Eberhard, 2003), and when these potentials facilitate only the partial re-evolution of alternative phenotypes, novel phenotypic patterns appear. We hope our findings not only inspire future work testing these proposals in polyphenic organisms, but also in non-polyphenic ones, where the polyphenism occurs at the cellular level but not at the level of the whole organism. Here, the cellular polyphenism produces alternative cell-types from a single genome in response to internal cues, such as morphogen gradients within the organism. If alternative cell types retain homologs or serial homologs of specific traits (Jackman et al., 2025; Lynch, 2023), then this raises the possibility that ancestral and latent developmental potentials may generally facilitate re-evolution of alternative cell types in multicellular organisms.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Gene (Camponotus floridanus)eyeless
distal-less
hedgehog
eye absent
sine oculis
twin of eyeless
orthodenticle
GenBankXM_025414466
XM_025412727.1
XM_011262474.3
XM_011262474.3
XM_011252868.3
XM_011268499.3
XM_020028684.2
Commercial assay or kitHybridization Chain Reaction in situ hybridization probesMolecular Instruments
Software, algorithmR software
Fiji
Photoshop
Illustrator
Tracer
Geneious
MEGA 12 alpa






Ant maintenance and collection

Request a detailed protocol

Colonies of C. floridanus were collected at Gainesville (Florida, USA), and P. rastellata were collected at Mae Tang (Chiang Mai, Thailand), under the following import permit numbers from the Canadian Food Inspection Agency: Florida (P-2016–02921, P-2018–00809) and Thailand (Ants from Asia, P-2019–00011). Ants were maintained in plastic boxes with glass test tubes filled with water-constrained cotton wool. They were fed mealworms and the Bhatkar–Whitcomb diet (Bhatkar and Whitcomb, 1970). Colonies were maintained at 25 °C with 60% humidity in complete darkness. All colonies were housed in growth chambers at McGill University’s Phytotron Facility under Plant Pest Containment Level 1 certification numbers PC-2016–057 and PC-2018–265 from the Canadian Food Inspection Agency. None of the species collected are endangered as determined by their absence on the IUCN Red List of Threatened Species (https://www.iucnredlist.org/, and search Formicidae) or were collected on protected lands.

Larvae fixation and in-situ HCR

Request a detailed protocol

Gene sequences were obtained from NCBI GenBank database (Sayers et al., 2022) using genome BLAST against the assembled C. floridanus genome: eyeless (ey; XM_025414466), distal-less (dll; XM_025412727.1), hedgehog (hh; XM_011262474.3), eye absent (eya; XM_025414466), sine oculis (so; XM_011252868.3), and twin of eyeless (toy; XM_011268499.3) genes. For otd, two paralogs of the gene were found in ants (XM_020028684.2 and XM_025415314.1), which is a result of a gene duplication event that has also been reported in wasps, bees, and beetles (Lynch et al., 2006; The Honeybee Genome Sequencing Consortium, 2006; Zattara et al., 2017). The two otd paralogs sequences were then aligned by multiple sequence alignment using all of the known orthodenticle related sequences in insects: D. melanogaster (NM_001369965.1), A. mellifera: otd-1 (XM_026446161.1), otd-2 (XM_006571236.3), Nasonia vitripennis: otd-1 (XM_008212114.4), otd-2 (XM_031926951.2), and T. castaneum otd-1 (XM_008192467.2), otd-2 (XM_008192470.2), and A. pisum (XP_008180802.1). To determine the otd-1 paralog to D. melanogaster (otd-1), a maximum likelihood gene tree was inferred using genetic distance model Hasegawa Kishino Yano (HKY) and 500 bootstrap replicates as incorporated in MEGA12 alpha (Kumar et al., 2024). Probes corresponding to all genes were chosen for the hybridization chain reaction experiments using the fluorescence Hairpins (B1 546, B2 488, B3 647) synthesized by Molecular Instruments.

First, second, third, and fourth larval instars of soldier and minor worker-destined larvae and fourth instar of male destined larvae of C. floridanus and worker larvae of P. rastellata were collected and subsequently fixed in a PEM 4% formaldehyde solution for 2 hr at room temperature. Fixed samples were then dehydrated progressively in methanol baths (25%, 50%, 75% methanol for 15 min each, and 100% overnight at 4 °C) and stored in 100% methanol at −30 °C until use. All gene expression analyses were conducted by in situ Hybridization Chain Reaction (HCR), following the protocol for HCR (v3.0 protocol; Schwarzkopf et al., 2021). After the tissue was pre-hybridized in a prewarmed Probe Hybridization Buffer (Molecular Instruments) for 30 min at 37 °C and incubated with HCR probes in a Probe Hybridization Buffer overnight at 37 °C. Tissues were washed the next day in a prewarmed Probe Wash Buffer four times, 15 min each, and washed in 5 X SSCT (UltraPure 20XSSC Buffer, Invitrogen, diluted in water) three times for 5 min at room temperature. Tissues were pre-amplified in Amplification Buffer (Molecular Instruments) for 30 min at room temperature and incubated with snap-cooled HCR hairpins in Amplification Buffer overnight at room temperature. Tissues were then washed with 5 X SSCT at room temperature twice for 5 min, for 30 min, and once for 5 min before being mounted on glycerol-DAPI 80%.

Microscopy

Request a detailed protocol

Confocal imaging was used to describe gene expression using a Leica SP8 confocal microscope. Fiji (Schindelin et al., 2012) was used for image processing. Scanning electron microscopy (SEM) was done on a Hitachi TM3030 Scanning Electron Microscope.

Evolution of ocelli in the formicine clade

Request a detailed protocol

The evolution of ocelli on workers across the subfamily Formicinae was inferred using ancestral reconstruction (ASE) for discrete traits incorporated in the R package Phytools 4.3.3 (Revell, 2024). The ASE analysis was based on the UCE70 phylogeny for the clade Formicinae published by Blaimer et al., 2015. The species Camponotus floridanus and P. bihamata were added manually to the phylogeny. To determine the presence or absence of ocelli in workers, photographs of the studied species from the database AntWeb, Version 8.114 were used (AntWeb, 2026). The observations from the database were contrasted with published information from the literature (Appendix 1—table 2). Ocelli were classified as present in the worker caste if individuals exhibit any of the three ocelli (two lateral and one medial ocellus). In the case of the presence of worker polymorphism, ocelli were classified as present if any one of the three ocelli was present within any of the worker subcastes. Whereas the absence was the complete lack of ocelli across workers and soldiers.

Four separate models of ocelli evolution were tested for each character in phytools: Equal rate ‘ER’, all transitions rate different ‘ARD’, and an irreversible model allowing only transitions between presence and absence, and another irreversible model allowing only transitions between absence and presence. We compared the fit of our models by computing Akaike information criterion (AIC) and Akaike weights and conducting pairwise likelihood ratio tests. The new function incorporated in phytools 4.3.3, simmap, was used to generate stochastic character maps under each of the four models tested (Appendix 1—table 1). The stochastic mapping that resulted from the stochastic simulation represented the frequencies that are equal to the weight of evidence supported by each model (Revell, 2024).

Appendix 1

Appendix 1—table 1
Model selection under maximum likelihood estimation implemented in phytools (Revell, 2024).

The results are ordered by decreasing Akaike Weights (w).

Modellog(L)d.f.AICweight
Equal Rates–41.808109185.61621730.70944317
Unequal Rates–41.720603287.4412050.28485677
Irreversible: Absence to Presence–46.813104195.62620810.00475637
Irreversible: Presence to Absence–48.430553198.86110650.00094368
Appendix 1—table 2
Database with the references used in this study for presence and number of ocellus (1) and absence (0) of ocellus.
SubfamilyGenusSpeciesOcelli CountReferences
outgroupAcanthoponeraminor0AntWeb.org
FormicinaeAcropygaacutiventris0AntWeb.org
outgroupAneuretussimoni0AntWeb.org
FormicinaeAnoplolepiscustodiens0AntWeb.org
FormicinaeAnoplolepisgracilipes0AntWeb.org
FormicinaeAphomomyrmexafer3AntWeb.org
FormicinaeBajcaridristheryi3Santschi, 1936
outgroupBrachymyrmexdepilis0AntWeb.org
FormicinaeCalomyrmexalbertisi0AntWeb.org
FormicinaeCalomyrmexlaevissimus0AntWeb.org
FormicinaeCamponotusfloridanus0AntWeb.com
FormicinaeCamponotusgibbinotus1AntWeb.org
FormicinaeCamponotushyatti0MacKay and Mackay, 2002
FormicinaeCamponotusmaritimus0Ward, 2005
FormicinaeColobopsissaundersi0AntWeb.org
FormicinaeColobopsisvitiensis0Mann, 1920
FormicinaeCataglyphiscursor3AntWeb.org
outgroupCladomyrmapetalae0Agosti, 1991
outgroupDolichoderuspustulatus0AntWeb.org
FormicinaeDinomyrmexgigas1AntWeb.org
FormicinaeEchinoplaaustralis0AntWeb.org
FormicinaeEuprenolepisprocera0Lapolla, 2009
FormicinaeFormicamoki3Cole, 1943
FormicinaeFormicaneogagates3AntWeb.org
FormicinaeGigantiopsdestructor3Smith, 1858
FormicinaeIberoformicasubrufa3Antwiki- genus
FormicinaeLasiophanesatriventris3AntWeb.org
FormicinaeLasiuscalifornicus0AntWeb.org
FormicinaeLasiusniger3AntWeb.org
FormicinaeLepisiotacanescens3Sharaf et al., 2020
outgroupManicabradleyi0AntWeb.org
outgroupMyrmeciapyriformis3AntWeb.org
FormicinaeMyrmecocystusflaviceps3AntWeb.org
FormicinaeMyrmecorhynchusemeryi3
(soldiers and media,
absent in minors
)
Wheeler, 1917
outgroupMyrmelachistaflavocotea0AntWeb.org
FormicinaeMyrmoterasiriodum3AntWeb.org
outgroupNothomyrmeciamacrops0AntWeb.org
FormicinaeNotoncuscapitatus3AntWeb.org
FormicinaeNotostigmacarazzii3AntWeb.org;
Emery, 1911
FormicinaeNylanderiadodo0Lapolla et al., 2011
FormicinaeNylanderiahystrix3Kallal and LaPolla, 2012
FormicinaeOecophyllalonginoda0AntWeb.org
FormicinaeOecophyllasmaragdina0Cole and Jones, 1948
FormicinaeOpisthopsisrespiciens0AntWeb.org
FormicinaeParaparatrechinaglabra3AntWeb.org
FormicinaeParaparatrechinaoceanica0AntWeb.org
FormicinaeParatrechinaantsingy3LaPolla and Fisher, 2014
FormicinaeParatrechinalongicornis3AntWeb.org
FormicinaeParatrechinazanjensis3LaPolla et al., 2013
FormicinaePetalomyrmexphylax3Snelling, 1979
FormicinaePlagiolepisalluaudi0AntWeb.org
FormicinaePolyergusbreviceps3AntWeb.org; Smith, 1947
FormicinaePolyrhachisdecumbens0Kohout, 2006
FormicinaePrenolepisemmae3AntWeb.org
FormicinaePrenolepisimparis0Williams and Lapolla, 2016
FormicinaeProformicamongolica3AntWeb.org
FormicinaeProlasiusconvexus3McAreavey, 1957
FormicinaePseudolasiusaustralis0Emery, 1925
FormicinaePseudonotoncushirsutus3Shattuck and O’Reilly, 2013
outgroupRhytidoponerachalybaea0AntWeb.org
FormicinaeRossomyrmexanatolicus3AntWeb.org
FormicinaeSantschiellakohli3Forel, 1893
FormicinaeTeratomyrmexgreavesi3Shattuck and O’Reilly, 2013
outgroupTetraponerarufonigra3Ward, 2001
FormicinaeZataniaalbimaculata3AntWeb.org

Data availability

All data reported in this paper are provided in Appendix 1—table 2. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the corresponding author upon request.

References

  1. Website
    1. AntWeb
    (2026) California Academy of Science
    Accessed April 23, 2026.
  2. Book
    1. Davison DR
    2. Michod RE
    (2021) Phenotypic plasticity and evolutionary transitions in individualit
    In: Pfennig DW, editors. Phenotypic Plasticity &Amp; Evolution. CRC Press. pp. 241–266.
    https://doi.org/10.1201/9780429343001-13
    1. Emery C
    (1925)
    Genera Insectorum
    Hymenoptera, fam. formicidae, subfam. Formicinae, Genera Insectorum, Louis Desmet-Verteneuil.
  3. Book
    1. Held L
    (2002) References
    In: Held L, editors. Imaginal Discs: The Genetic and Cellular Logic of Pattern Formation. Cambridge University Press. pp. 307–440.
    https://doi.org/10.1017/CBO9780511529733.018
    1. Hung ACF
    (1967)
    A revision of the ant genus polyrhachis at the subgeneric level (hymenoptera: formicidae)
    Transactions of the American Entomological Society 93:395–422.
    1. Kohout RJ
    (2006)
    Review of polyrhachis (cyrtomyrma) forel (hymenoptera: formicidae: formicinae) of Australia, borneo, new guinea and the solomon Islands with descriptions of new species
    Memoirs-Queensland Museum 52:87.
  4. Book
    1. MacKay WP
    2. Mackay E
    (2002)
    The Ants of New Mexico (Hymenoptera: Formicidae)
    Edwin Mellen Press.
    1. Santschi F
    (1936)
    Liste et descriptions de fourmis du Maroc
    Bulletin de La Société Des Sciences Naturelles Du Maroc 16:198–210.
  5. Book
    1. Smith F
    (1858)
    Catalogue of Hymenopterous Insects in the Collection of the British Museum: 6: Formicidae
    Order of the Trustees.
    1. Ward PS
    (2014) The phylogeny and evolution of ants
    Annual Review of Ecology, Evolution, and Systematics 45:23–43.
    https://doi.org/10.1146/annurev-ecolsys-120213-091824
    1. Wheeler W
    (1917)
    The Australian ant-genus Myrmecorhynchus (Ern. André) and its position in the subfamily Camponotinae
    Transactions of the Royal Society of South Australia 41:14–19.

Article and author information

Author details

  1. Angelly Vasquez-Correa

    1. Center for Evolutionary and Organismal Biology, Women’s Hospital, School of Medicine, Zhejiang University, Hangzhou, China
    2. McGill University, Department of Biology, Montreal, Canada
    Contribution
    Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1478-8682
  2. Johanna Arnet

    McGill University, Department of Biology, Montreal, Canada
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0009-0000-2478-9446
  3. Travis Chen

    1. Center for Evolutionary and Organismal Biology, Women’s Hospital, School of Medicine, Zhejiang University, Hangzhou, China
    2. McGill University, Department of Biology, Montreal, Canada
    Contribution
    Data curation, Formal analysis, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0009-0003-6789-1928
  4. Ehab Abouheif

    1. Center for Evolutionary and Organismal Biology, Women’s Hospital, School of Medicine, Zhejiang University, Hangzhou, China
    2. McGill University, Department of Biology, Montreal, Canada
    Contribution
    Conceptualization, Supervision, Funding acquisition, Validation, Investigation, Writing – original draft, Project administration, Writing – review and editing
    For correspondence
    abouheif@zju.edu.cn
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7651-1737

Funding

Natural Sciences and Engineering Research Council of Canada (Discovery)

  • Ehab Abouheif

Natural Sciences and Engineering Research Council of Canada (BESS-CREATE)

  • Angelly Vasquez-Correa

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Hermogenes Fernandez-Marín, who inspired this work by sharing his knowledge about ocelli in soldiers across the Atta species. We thank Lloyd Davis, Marc Seid, Rajendhran Rajakumar, and Shelly Berger for help with collecting Camponotus floridanus colonies. We thank Gregory A Wray, Mary Jane West-Eberhard, Friedrich Markus, Guilherme Gainett, Arjuna Rajakumar, and Rajendhran Rajakumar for discussions and/or comments on the manuscript, and Juan Carlos Penagos for input on phylogenetic analysis. We thank Erik Plante and undergraduates for help with feeding and maintaining lab colonies of Camponotus floridanus. Finally, we thank McGill University’s Integrated Quantitative Biology Initiative (IQBI) and Advanced Bioimaging Facility (ABIF) for imaging support. This work was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant to EA and by a Doctoral fellowship from NSERC BESS-CREATE program to AV-C.

Version history

  1. Preprint posted:
  2. Sent for peer review:
  3. Reviewed Preprint version 1:
  4. Reviewed Preprint version 2:
  5. Version of Record published:

Cite all versions

You can cite all versions using the DOI https://doi.org/10.7554/eLife.110148. This DOI represents all versions, and will always resolve to the latest one.

Copyright

© 2026, Vasquez-Correa et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 782
    views
  • 53
    downloads
  • 0
    citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Angelly Vasquez-Correa
  2. Johanna Arnet
  3. Travis Chen
  4. Ehab Abouheif
(2026)
Latent gene network expression underlies partial re-evolution of a polyphenic trait in the worker caste of ants
eLife 15:RP110148.
https://doi.org/10.7554/eLife.110148.3

Share this article

https://doi.org/10.7554/eLife.110148