The early life of a leaf‐cutter ant colony constrains symbiont vertical transmission and favors horizontal transmission

Abstract Colonial organisms host a large diversity of symbionts (collectively, parasites, mutualists, and commensals) that use vertical transmission (from parent colony to offspring colony) and/or horizontal transmission to disperse between host colonies. The early life of some colonies, characterized by the dispersal and establishment of solitary individuals, may constrain vertical transmission and favor horizontal transmission between large established colonies. We explore this possibility with the miniature cockroach Attaphila fungicola, a symbiont of leaf‐cutter ants and the mutualist fungal gardens they cultivate. The early life of a leaf‐cutter colony is characterized by the dispersal of a female alate (winged “queen”) carrying a fungal pellet, and the subsequent establishment of a foundress (workerless “queen”) raising her incipient fungal garden and colony. Roaches hitchhike on female alates during leaf‐cutter nuptial flights, which strongly suggests that roaches are vertically transmitted to foundresses and their incipient colonies; however, weak compatibility between roaches and incipient gardens may constrain roach vertical transmission. Reciprocally, opportunities for horizontal transmission between large established colonies with abundant fungal gardens may weaken selection against roach‐induced harm (virulence) of incipient gardens. We use a laboratory experiment, behavioral observations, field surveys, and a transmission model to estimate the effect roaches have on the survivorship of incipient gardens and the frequency of roach vertical transmission. Contrary to traditional assumptions, our results indicate that roaches harm incipient gardens and predominantly use horizontal transmission between established leaf‐cutter colonies. Ultimately, “costs of generalism” associated with infecting disparate stages of a host's lifecycle (e.g., incipient vs. established colonies) may constrain the vertical transmission of roaches and a broad range of symbionts.

mature colonies can become enormous, colony lifecycles often commence with tiny solitary forms (Hölldobler & Wilson, 2009;Marti et al., 2015;Martin et al., 2018;Yang, 2007), and these humble beginnings can play a crucial role in the relationship between colonies and their symbionts. In particular, the early lives of colonies can influence patterns of between-colony transmission, constraining vertical transmission and favoring horizontal transmission. For example, theory predicts that beneficial symbionts (zooxanthellae) of stony corals should be vertically transmitted from parent to offspring corals (Bennett & Moran, 2015;Bull et al., 1991;Hartmann et al., 2017Hartmann et al., , 2019Herre et al., 1999); however, the same symbionts that benefit large established corals can harm their dispersing larvae (Hartmann et al., 2017(Hartmann et al., , 2019. For some corals, these larva-specific costs of infection favor coral acquisition of symbionts from the environment (horizontal acquisition) rather than from parents (vertical transmission), deviating from predictions that do not account for coral lifecycle heterogeneity (Hartmann et al., 2017(Hartmann et al., , 2019. For symbionts of eusocial insects, the solitary early life of a colony may present similar challenges to vertical transmission (herein, transmission from parent colony to incipient daughter colony), and accounting for colony lifecycle heterogeneity could dramatically alter predictions of transmission dynamics. Many colonies of ants, bees, wasps, and termites begin with just one or a few individuals and expand into colonies of thousands or millions of members, with resources growing in kind from meager to abundant (Hölldobler & Wilson, 2009;Tschinkel, 2011;Wheeler, 1910;Wilson, 1985). Leaf-cutter ants in the genus Atta exemplify this transformation. They begin with one or a few gynes (female reproductives) each carrying a pellet of mutualist fungus and develop into complex insect societies maintaining abundant fungal gardens (Forti et al., 2017;Marti et al., 2015). Notably, the gyne is the only ant that passes through the entire colony lifecycle, including the solitary incipient stage. She changes in form and function from a winged gyne dispersing from her parent colony on a nuptial flight (female alate) into a wingless, workerless gyne raising an incipient colony (foundress), and finally, after the eclosure of her first brood of workers, she becomes the queen of the colony's ergonomic and mature stages (i.e., growth and reproductive stages, respectively) (Fernández-Marín & Wcislo, 2005;Marti et al., 2015;Wilson, 1985).
At the scale of the colony, changes in form and function result from the codependent processes of fungal garden cultivation  and sociogenesis, "the process by which colony members undergo changes in caste, behavior, and physical location incident to colonial development," (Wilson, 1985(Wilson, , pp. 1489) creating a dynamic within-nest environment for symbionts. During the lifecycle of a colony, these "guests" can experience changes in their host colony's size, resources, defenses, nest architecture, interaction networks, and other qualities that affect symbiont fitness (Cremer & Sixt, 2009;Hughes et al., 2008;Parmentier, 2020;Rynkiewicz et al., 2015;Tschinkel, 1993;Woodard et al., 2013). As such, colony stage-dependent variation, including traits specific to the early life of a colony (Moreira et al., 2019), should be consequential for symbiont ecology and evolution.
Moser reports that "the roach inhabits the fungus gardens of most nests," but does not provide a specific estimate of prevalence (1964, pp. 1048). At our field sites in Austin, TX, mature leaf-cutter colonies with chronic roach infections (>5 years) survive and reproduce apparently unimpaired, so we use the neutral term "symbiont" instead of "parasite" to describe them (Phillips, 2021); however, the effect of roaches on incipient colony survival is unknown.
During the mass upheaval of a colony nuptial flight, roaches hitchhike on a small proportion of their host colony's dispersing female alates (<7%), and typically each "infected" female alate bears a single phoretic female roach (Moser, 1967a;Phillips, 2021;Phillips et al., 2017;Waller & Moser, 1990). Hitchhiking (i.e., co-dispersal) on female alates has traditionally been interpreted as a behavior that initiates vertical transmission, likely because it suggests roaches remain with female alates as they become foundresses, and that roaches then co-establish with foundresses and their incipient colonies (Djernaes et al., 2020;Moser, 1967aMoser, , 1967b; however, there is no evidence that roaches persist as infections through the incipient stage of colony development. Furthermore, recent findings indicate that hitchhiking roaches can abandon female alates after nuptial flights and subsequently ride leaves carried by foragers into the nests of established colonies (Phillips, 2021). Accordingly, hitchhiking on female alates may facilitate a complex mode of horizontal transmission between established colonies ("female alate-vectored transmission," Phillips, 2021) rather than vertical transmission to incipient colonies. In other words, roach co-dispersal with female alates can be uncoupled from roach co-establishment with foundresses and incipient colonies, and it remains unclear how frequently co-establishment and thus vertical transmission occurs (Figure 1).
The low host quality of incipient colonies (extremely high mortality, low tolerance for disturbance, meager incipient gardens) may limit roach co-establishment and constrain vertical transmission, favoring routes of horizontal transmission that bypass incipient colonies (direct or female alate-vectored transmission between established colonies) (Moser, 1964;Phillips, 2021). Reciprocally, if roaches rarely or never use routes of vertical transmission that pass through incipient colonies, selection on roaches to avoid overexploiting and damaging incipient gardens should be weak (weak incipient garden compatibility). Alternatively, if roaches rely heavily on vertical transmission for dispersing between nests, they should be under strong selection to successfully co-establish with incipient colonies, and to minimize harm and possibly provide benefits to incipient gardens (strong incipient garden compatibility) (Combes, 2001;Genkai-Kato & Yamamura, 1999;Iritani et al., 2019;Lipsitch et al., 1996). To test whether roaches exhibit strong or weak incipient garden compatibility, and whether roaches primarily use vertical or horizontal transmission, we use a laboratory experiment to estimate the effect individual roaches have on the survivorship of low-volume fungal gardens in artificial foundress chambers, and we use field surveys and a between-colony transmission model to estimate the contribution of vertical transmission to roach prevalence among mature leaf-cutter colonies. A. texana colonies were infected with roaches (73% mature colony "infection" prevalence: 3/5 mature colonies with roaches at BFL, 5/6 mature colonies with roaches at Hornsby Bend). In total, 420 roaches were collected from 7,791 female alates (an average of 5.5% of female alates from colonies with roaches had a single roach attached). Roach prevalence per infected colony ranged from 2.2% to 6.8% of female alates bearing a single hitchhiking roach. Data have been deposited at the Dryad Data Repository  https://doi.org/10.5061/dryad.8sf7m 0cnt). Specimens of A. fungicola collected and not lost or destroyed during these and other experiments are accessioned (Accession number: UTIC255785) at the Insect Collection of University of Texas at Austin (https://biodi versi ty.utexas.edu/colle ctions).

| Incipient garden survivorship experiment
We collected paired female alate ants and roaches for use in the experiment, where alates collected from the field already had attached A. fungicola roaches. Using naturally paired ants and roaches ensured that both species came from the same natal colony, and thus controlled for potential intercolony differences (e.g., chemical profiles). We removed the wings of the female alates and placed the de-winged alate (herein, "foundress") and her attached roach in a 5 cm diameter container ("foundress chamber") with 20 mg of fungal garden ("incipient garden") from a laboratory colony. Notably, 20 mg is larger than the inoculum of fungus that foundresses initially regurgitate when founding a new colony under natural conditions (Marti et al., 2015). Long-term survival in the laboratory of foundresses provided only with their inoculum is extremely rare, complicating experimentation and highlighting the extreme fragility of incipient colonies.
Experiments were conducted under laboratory conditions described in Phillips et al. (2017), with all replicates kept at room temperature (22-24°C). Chambers were checked for 1 min every 24 hr in low-light conditions to determine mortality of fungal gardens, foundresses, and roaches. The fungal garden was marked as effectively dead if it was dismantled and scattered in decaying clumps in upper and/or lower corners of the chamber and if the foundress did not tend any portion of the garden for at least 30 s ("uncaring" foundresses), or if the foundress was dead (without a caretaker, the fungal garden is effectively dead). Alternatively, the fungal garden was marked as living if the foundress tended a contiguous portion of the garden for at least 30s (i.e., the foundress' head and F I G U R E 1 An adult female Attaphila fungicola roach on part of an established leaf-cutter fungal garden mandibles maintained a position facing and over the garden, typically manipulating and antennating it: see end of Video S4, e.g., of foundress tending behavior). Additionally, 30 min following initial set-up, each foundress chamber was observed for 3 min to determine if the inhabiting roach disturbed the incipient garden. Garden disturbance was scored if roach contact caused any observable movement, physical dislocation, or fragmentation of the garden, and subsequently, we categorized gardens as either "disturbed" (one or more observations of roach disturbance: see Videos S1-S3, e.g., of disturbance) or "undisturbed" (no observations of roach disturbance).
To test the effect of A. fungicola on incipient garden mortality, we used a mixed-model Cox proportional hazards model with rightcensored daily mortality as the response variable and treatment (presence or absence of A. fungicola) as the explanatory variable.
Nest location was nested within flight date as a random effect to account for variation in survivorship between flight dates and nests.
Survival analyses were run using version 2.2 of the coxme package and the survival R package (Therneau, 2015). Additionally, we used a chi-square test to determine whether there was a difference in the proportion of dead foundresses between the roach-present and roach-absent treatments. All analyses used R version 4.0.1 (R Core Team, 2020).

| Between-colony transmission model
To explore the maximum contribution of vertical transmission to roach infection prevalence among mature colonies, we develop a simple model that assumes exclusive vertical transmission. We use this model to estimate mature colony infection prevalence (V) from foundress infection prevalence (J) and the effect of roach presence on the likelihood of a foundress reaching the mature colony stage (δ).
See Figure 2 for a diagram of the model, and Table 1 for parameter symbols and definitions.
Parameter δ is the net effect of roach presence on foundress and incipient colony survivorship. A value of δ < 1 indicates that the roach is harmful to foundresses and their incipient colonies, δ = 1 that the roach is neutral, and δ > 1 that the roach is beneficial. To estimate δ from our incipient garden survivorship experiment, we use the inverse of a hazard ratio calculated from our survivorship analysis. This is a dimensionless measure of the effect roaches have on incipient garden survivorship.
Our field estimate of foundress infection prevalence, J, is not directly based on foundress infection prevalence (i.e., co-establishing roaches) because roaches have not been observed in incipient colonies in our study region (Phillips et al., 2017) and we are not aware of estimates of foundress infection prevalence in any other region and for any other Attaphila species. This makes our best direct estimate F I G U R E 2 Diagram of model. See Table 1 for definitions of symbols.
of J zero and would indicate that vertical transmission does not occur. To account for the possibility that we have not observed these rare events, including the possible deposition of roach ootheca (egg cases) on alates or in foundress chambers, we estimate the maximum potential value of J from the maximum proportion of female alates with hitchhiking roaches collected from a single mature A. texana colony. This is likely a highly conservative estimate of J given that roaches co-establishing with foundresses seem much rarer than roaches co-dispersing with female alates (see Section 4). The maximum prevalence of roaches on a single nest mound in Austin, TX that has been recorded is 0.07 (On 15 May 2016, 50/719 female alates collected from a single nest mound surface preparing for nuptial flights at Brackenridge Field Laboratory). As we describe in "Model Results" below, using this conservative estimate of J = 0.07 helps estimate a conservative maximum possible contribution of roach vertical transmission to mature colony infection prevalence (V).
Mature colonies with roaches (M i ) are generated in our model by the development of foundresses with roaches (F i ) that survive to colony maturity with their roach infection intact. This is determined by the rate that foundresses with roaches reach the mature colony stage (s i ) and the persistence of inherited roaches through colony development (z). If we assume mature colonies with roaches die at rate q, the change of mature colonies with roaches over time is as follows: Mature colonies without roaches (M u ) are generated by foundresses without roaches that survive to colony maturity (F u ), determined by the rate that foundresses without roaches reach the mature colony stage (s u ), and by the rate that foundresses with roaches (F i ) lose their inherited roaches and reach the mature colony stage (s i (1 − z)).
We assume that foundresses with roaches that lose them during colony development (e.g., the roaches die) are as likely to reach colony maturity as foundresses that maintain roaches through colony development (s i ). In other words, we assume that the likelihood of foundresses reaching the mature colony stage is independent of the duration of roach infections. This assumption is consistent with our experimental results, which indicate A. fungicola has a rapid effect on low-volume fungal garden survivorship (Figure 3). We also assume that mature colonies with roaches die at the same rate as mature colonies without roaches (q). This assumption is based on observations of similar nest surface frequencies of A. fungicola (i.e., similar proportions of female alates with hitchhiking roaches during nuptial flight preparations of a given infected nest: roughly 2%-7%) over a span of 5 years with no apparent reduction in colony health or size of nuptial flights. This assumption is also consistent with the general prediction that symbionts of large, long-lived colonies are likely to evolve relatively low virulence (Hughes et al., 2008).  Second, we estimate V max under the conservative assumption that roaches have no effect on incipient fungal garden survivorship (neutral, δ = 1). By "conservative assumption," we specifically mean an assumption that selects parameter values deviating from more realistic values (i.e., values based on experiments, surveys, or natural history observations) in a way that maximizes model estimates of V.
For all estimates, we make the conservative assumption that roaches are never lost or cleared after occupying an incipient colony's foundress chamber (z max = 1). In nature, z < 1 is certainly more accurate. Foundresses, for instance, have been observed attacking and killing roaches (Phillips et al., 2017). As discussed above, our es- generally absent during nuptial flights (Phillips et al., 2017;Waller & Moser, 1990). It has been proposed that A. fungicola are parthenogenetic in Louisiana (Waller & Moser, 1990) where no male A. fungicola have been collected from A. texana colonies.

| Model results
By incorporating our conservative values of z and J (z = 1, J = 0.07) and our laboratory estimate of δ (δ = 0.3) into Equation 6, we calculate that V = 0.02 (2% mature colony infection prevalence). Under these same conditions, but assuming roaches have no effect on the likelihood of a foundress reaching the mature colony stage (δ = 1), we calculate that V = 0.07 (7% mature colony infection prevalence).
Our field survey estimate of mature colony infection prevalence is

| D ISCUSS I ON
Symbionts inherited by host propagules (e.g., plant seeds, coral larvae, ant gynes) must co-disperse and co-establish with propagules for vertical transmission to be successful (Bibian et al., 2016). As a consequence, accounting for challenges that symbionts face during both host dispersal and establishment can help identify constraints on vertical transmission.
The propagules of leaf-cutter ant colonies are gynes and the mutualist fungus they carry and care for, and the early life of a colony is marked by their dispersal (winged female alates carrying fungal pellets) and establishment (workerless foundresses raising incipient fungal gardens) (Helms, 2018;Marti et al., 2015;Moser, 1967a).
During leaf-cutter nuptial flights, the symbiotic cockroach A. fungicola hitchhikes on female alates (co-dispersal), a behavior that strongly suggests roaches are vertically transmitted to incipient colonies (Moser, 1967a;Waller & Moser, 1990); however, roach coestablishment with foundresses may be limited by weak compatibility with incipient gardens. The collective results of our experiment, behavioral observations, field surveys, and model indicate that roaches are weakly compatible with incipient gardens, that they at most rarely use vertical transmission, and that they primarily use horizontal transmission between established colonies.
Given the extreme fragility of incipient fungal gardens, we would expect selection for compatibility with incipient gardens to be strong for any vertically transmitted symbiont and for such symbionts to avoid harming or to even benefit incipient gardens during co-establishment (Fries & Camazine, 2001;Genkai-Kato & Yamamura, 1999;Herre et al., 1999;Lipsitch et al., 1996). In contrast, our results suggest roaches have evolved fixed responses to robust gardens rather than plastic behaviors that can be attuned to delicate gardens. In artificial foundress chambers, we observed roaches feeding on and rubbing against gardens (the latter may help the roaches acquire a colony's chemical profile) (Nehring et al., 2016) (see Videos S1-S3), behaviors that are likely harmless to established gardens but could be catastrophic to incipient gardens and responsible for their accelerated failure (Figure 3). Also, roaches appeared to stress foundresses, consistent with observations from a previous study (Phillips et al., 2017). By antagonizing a foundress, a roach could indirectly cause significant damage to the garden (Moreira et al., 2019) (see Video S4, e.g., of a foundress turning away from her garden while grooming a roach off of her body; note, these interactions were observed but not scored as disturbances in our "incipient garden survivorship experiment" because they do not involve direct contact between roach and garden). Lastly, roaches might act as vectors of "hyperphoretic spores" and microbial garden diseases that kill incipient gardens (Di Prisco et al., 2011;Hughes et al., 2004;Moreira et al., 2019;Moser & Blomquist, 2011).

| Mixed-mode transmission between colonies
Attaphila fungicola vertical transmission may occur rarely, with populations of roaches using both vertical and horizontal transmission (i.e., mixed-mode transmission) (Antonovics et al., 2017;Ebert, 2013). The only field observation that ostensibly describes roaches co-establishing with incipient colonies notes A. fungicola in "new burrows made by [A. texana] queens" (Moser, 1967a, pp. 304). Other field observations suggest co-establishment and vertical transmission are rare. Roaches have not been collected from A. texana foundress chambers in central Texas (Phillips et al., 2017),

nor in incipient nests of its sister species Atta mexicana in Organ
Pipe Cactus National Monument, AZ, and attempts to have these roaches co-establish with foundresses have been unsuccessful (Phillips et al., 2017;pers. communication Alex Mintzer). Also, in northern Mexico, individuals of an unidentified species of Attaphila were observed running around independently of nearby A. mexicana foundresses searching for nest sites (Sánchez-Peña, 2005) suggesting that these roaches had abandoned foundresses before vertical transmission could be completed (i.e., roach co-dispersal uncoupled from co-establishment).
If Attaphila roaches exhibit mixed-mode transmission, perhaps encounters with foundress predators (e.g., armadillos, grackles, myrmecologists) influence a roach's decision to either take a chance remaining with a foundress likely to die (vertical transmission) or abandon the foundress and risk seeking an established colony (female alate-vectored horizontal transmission). In north Texas, a roach jumped off of a foundress seeking a nest site and disappeared into the grass, an escape apparently prompted by a researcher's collection of the foundress (U. G. Mueller, personal communication). Also, it is possible that roaches deposit ootheca (egg cases) on female alates or with foundresses before abandoning them; however, in field experiments where roaches were released while attached to female alates and foundresses, this was not observed (Phillips, 2021). In another study, roaches deposited ootheca within a few days of being collected with female alates during nuptial flight preparations, suggesting ootheca deposition would not have occurred during the nuptial flight itself (Waller & Moser, 1990). Even in the unlikely scenario that every hitchhiking roach attaches an ootheca to its co-dispersing female alate, or deposits an ootheca during co-establishment with a foundress, and assuming that ootheca and potentially accompanying adult roaches are harmless to incipient gardens, our model predicts vertical transmission would still occur infrequently (under these conservative conditions, our model predicts that the maximum proportion of mature colonies infected through vertical transmission is 10%; see Section 3.1).
Overall, the disparity between high mature colony prevalence and low female alate and foundress prevalence of roaches suggests vertical transmission is rare-unless roaches are somehow beneficial to incipient colonies and colonies with roaches disproportionately reach maturity. Our experiment indicates no such mutualism occurs (Figure 3). Infrequent vertical transmission could still play an important role in roach population dynamics and evolution, and vertical transmission might occur at higher frequencies in areas where the density of established leaf-cutter colonies is low and there are fewer opportunities for horizontal transmission (e.g., range frontiers) (Mueller et al., 2011). Also, Attaphila individuals, "strains" and species could vary in their compatibility with incipient colonies, creating within-species and between-species variation in the frequency of vertical transmission. A comparative analysis of transmission strategies among Attaphila might reveal conditions that facilitate vertical transmission, but we know little about the life histories of most species or how many species exist (Bohn et al., 2021;Bolivar, 1901;Brossut, 1976;Djernaes et al., 2020;Nehring et al., 2016;Rodriguez et al., 2013;Sánchez-Peña, 2005;Wheeler, 1900).

| Potential roach strategies for mitigating virulence during co-establishment
Virulence (i.e., symbiont-induced harm to a host) can be adaptive or nonadaptive for symbionts (Bull, 1994;Leggett et al., 2013). A common model of adaptive virulence frames it as a property emerging from the trade-off between transmission period and transmission rate: Increasing within-host reproduction is costly because it increases virulence and reduces symbiont transmission period (i.e., kills the host faster), but beneficial because it increases symbiont transmission rate (i.e., rate of infection of new hosts) (Bull, 1994;Day, 2003). In contrast, the roach-induced harm observed in our experiment probably represents nonadaptive virulence (Figure 3). A Texas leaf-cutter foundress raises her incipient colony in a closedoff ("claustral") underground chamber (Marti et al., 2015), the same small space a vertically transmitted roach would presumably occupy during co-establishment. If a roach contributes to the death of an incipient garden, it likely seals the fate of itself and its progeny in a shared grave with garden and foundress.
Nonadaptive virulence can be described as "virulence of no selective value per se…a coincidental byproduct of [symbiont] evolution in a different host species" (Bull, 1994(Bull, , pp. 1424(Bull, -1425. As this suggests, a major cause of nonadaptive virulence is infecting the "wrong" host, a host that a symbiont has not co-evolved with and may not be compatible with (i.e., a host outside of the symbiont's host range) (Bull, 1994;Combes, 2001;Leggett et al., 2013). Although the "wrong" host often refers to an incompatible host species or strain, here we use it to refer to a potentially incompatible colony lifecycle stage, the incipient colony, which for Attaphila represents a radically different host environment than an established colony. Under this premise, roach behaviors that harm incipient colonies could arise as a byproduct of roach co-evolution with established colonies. For example, if roaches have evolved an adaptive attraction to fungal gardens in the garden-rich environment of established colonies, the same attraction may be nonadaptively virulent when expressed in the garden-poor environment of incipient colonies. As discussed above, roaches might be able to mitigate this harm by adjusting their behavior during co-establishment to avoid incipient gardens (i.e., behavioral plasticity, Leggett et al., 2013), or by exclusively using "behavior-less" ootheca to co-establish with incipient colonies (i.e., ontogenetic niche shift, ten Brink & de Roos, 2017;Werner & Gilliam, 1984). Additionally, roaches may be able to mitigate harm by targeting leaf-cutter co-foundresses instead of solitary foundresses. Leafcutter foundresses can join together to start a new colony, and these co-foundress collectives exhibit higher survivorship and produce larger incipient gardens than solitary foundresses (Cahan & Julian, 1999;Mintzer, 1987). As a consequence, co-foundresses and their incipient gardens could exhibit a greater tolerance for roaches (Ayres & Schneider, 2012;Cremer et al., 2018;Pull et al., 2013), increasing the likelihood of both roach and incipient colony survival; however, roaches have been observed abandoning co-foundresses during the excavation of new colonies (Phillips, 2021;Phillips et al., 2021), suggesting that if roaches do infect incipient colonies, targeting co-foundresses may not be a preferred strategy.
In general, abundant resources during co-establishment should reduce the risk of symbiont overexploitation and catastrophic damage. Consider the early life of an ant-plant-homopteran mutualism, one in which a sap-sucking scale insect (the homopteran) co-establishes with an ant foundress on a myrmecophytic tree (Gaume et al., 1998). Although scale insects can be vectors of disease (Brown, 2016), and infestations can damage host plants (Golan et al., 2015), the sap-sucking of one or a few scale insects during co-establishment is unlikely to mortally wound a tree and doom the tripartite symbiosis. Now imagine if leaf-cutter foundresses initiated colonies with tree-sized fungal gardens instead of seed-sized fungal gardens. Presumably a roach in this scenario would be innocuous during co-establishment regardless of its behavior, and vertical transmission would not be constrained by incipient gardens.

| Costs of generalism may constrain vertical transmission
Vertical transmission from parent to daughter incipient colonies requires both roach encounters with and compatibility with incipient colonies (Combes, 2001). Hitchhiking on female alates (co-dispersal) facilitates encounters with incipient colonies because roaches simply have to remain with female alates as they transition into foundresses. Indeed, these easy encounters seem to be the basis for assuming hitchhiking is a first step in vertical transmission, and that co-dispersal is tightly linked to co-establishment; however, vertical transmission also requires compatibility with both incipient colonies and established colonies, while horizontal transmission requires compatibility with only established colonies. In the first case (vertical transmission), a roach must be a "generalist" of host colony lifecycle stages, while in the latter case (horizontal transmission), a roach can be a "specialist" of just established colonies. As a consequence, costs of generalism may ultimately constrain vertical transmission, not horizontal transmission, and attenuate the link between roach co-dispersal and co-establishment.
Some authors have divided costs of generalism that constrain symbiont compatibility with distinct hosts (i.e., restrict host range) into two categories: ecological costs and evolutionary costs (Benmayor et al., 2009;Leggett et al., 2013). Symbionts are susceptible to ecological costs when their potential hosts vary in quality, a scenario "analogous to that assumed in optimal foraging theory, where patches vary in quality" (Benmayor et al., 2009, pp. 764). In this context, ecological costs for symbionts are opportunity costs that result from infecting lower-quality hosts instead of higher quality hosts (Bull, 2006;Heineman et al., 2008). Incipient leaf-cutter colonies are intrinsically lower-quality hosts than established colonies, possessing much higher mortality rates, fewer resources, and a lower tolerance for disturbance than established colonies. As such, we would expect vertical transmission from parent to daughter incipient colonies, but not horizontal transmission between established colonies, to impose strong ecological costs of generalism on roaches. Evolutionary costs of generalism associated with roach vertical transmission are less clear. Evolutionary costs arise if a symbiont's ability to infect one kind of host (the "novel host") is associated with reduced performance in another kind of host (the "original host") (Benmayor et al., 2009;Leggett et al., 2013). These costs could result from antagonistic pleiotropy between roach traits enhancing performance in incipient colonies (e.g., avoidance of fungal garden) and traits enhancing performance in established colonies (e.g., attraction to fungal garden).
Although many insect societies found colonies with just one or a few individuals ("independent founders"), some found colonies with a large number of individuals ("dependent founders") (Cronin et al., 2013;LeBrun et al., 2013;Vargo & Porter, 1989). Army ants reproduce through a process called "colony budding," in which a new queen accompanied by a large group of nestmates break off from their parent colony to form a new colony (Cronin et al., 2013).
Budding allows an army ant colony to effectively skip the incipient stage and begin its life as an established colony (Cronin et al., 2013;Denny et al., 2004;Kronauer et al., 2010;Rettenmeyer et al., 2011). argue that army ant colonies should be more susceptible than independent founders (e.g., leaf-cutters) to inheriting colony symbionts such as phoretic mites (Berghoff et al., 2009) and socially transmitted microbes (Łukasik et al., 2017).
If leaf-cutter colonies were to reproduce through colony budding as army ants do, how would this affect Attaphila transmission? The vertical transmission of Attaphila might be less constrained by costs of generalism, and daughter colonies would likely inherit roaches more frequently.

| CON CLUDING REMARK S
Colonies are lifecycles, and many begin with just one or a few individuals. From the perspective of a colony symbiont, the solitary early life of a colony represents a radically different host environment than that of a large established colony. Compared to established colonies, incipient colonies possess few resources, succumb easily to disturbance, and suffer high rates of mortality. All else being equal, incipient colonies are lower-quality hosts than established colonies, and infecting the former instead of the latter can be costly. Across a broad range of host-symbiont systems, these costs may constrain routes of vertical transmission that pass through incipient colonies and favor routes of horizontal transmission that bypass them.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interests.