Nutrition mediates the expression of cultivar–farmer conflict in a fungus-growing

Jonathan Z. Shika,b,1, Ernesto B. Gomezb, Pepijn W. Kooija,c, Juan C. Santosd, William T. Wcislob, and Jacobus J. Boomsmaa

aCentre for Social Evolution, Department of Biology, University of Copenhagen, 2100 Copenhagen, Denmark; bSmithsonian Tropical Research Institute, Apartado 0843-03092, Balboa, Ancon, Republic of Panama; cJodrell Laboratory, Comparative Plant and Fungal Biology, Royal Botanic Gardens, Kew, Richmond TW9 3DS, United Kingdom; and dDepartment of Biology, Brigham Young University, Provo, UT 84602

Edited by Bert Hölldobler, Arizona State University, Tempe, AZ, and approved July 8, 2016 (received for review April 16, 2016)

Attine evolved farming 55–60 My before humans. Although whose sexual fruiting bodies represent considerable investments. evolutionarily derived leafcutter ants achieved industrial-scale farm- An illustrative example is offered by the fungus-growing attine ing, extant species from basal attine genera continue to farm loosely ants, whose dispersing queens transmit symbionts vertically and domesticated fungal cultivars capable of pursuing independent re- whose farming workers therefore actively suppress wasteful for- productive interests. We used feeding experiments with the basal mation of inedible mushrooms (16, 17). Another example is pro- attine to test whether reproductive allocation vided by two independently derived lineages of fungus-growing Termitomyces conflicts between farmers and cultivars constrain crop yield, possibly that both evolved vertical transmission of explaining why their has remained limited in scale and cultivars while terminating mushroom production (18). However, productivity. Stoichiometric and geometric framework approaches although such correlated adaptive states are consistent with the- oretical predictions, and have been tested and confirmed in some showed that carbohydrate-rich substrates maximize growth of both plant–microbe symbioses (6, 7), no experimental work has tested edible hyphae and inedible mushrooms, but that modest protein whether allocation conflicts affect the ecological dynamics of an- provisioning can suppress mushroom formation. Worker foraging imal ectosymbioses in similar ways. was consistent with maximizing long-term cultivar performance: Here, we present such a test, using the common and widely ant farmers could neither increase carbohydrate provisioning with- distributed (Argentina to northern Mexico) basal fungus-growing EVOLUTION out cultivars allocating the excess toward mushroom production, nor ant Mycocepurus smithii as a model (19). This ant species is no- increase protein provisioning without compromising somatic cultivar table for being parthenogenetic (19), allowing controlled experi- growth. Our results confirm that phylogenetically basal attine farm- mentation with genetically uniform worker ants. It also rears an ing has been very successful over evolutionary time, but that unusual variety of fungal symbiont clones that are not irreversibly unresolved host–symbiont conflict may have precluded these wild- domesticated and apparently continue to exchange genes with type symbioses from rising to ecological dominance. That status was free-living relatives (16, 20). Cultivars have thus retained their full achieved by the evolutionarily derived leafcutter ants following full sexual potential so that ant farmers have explicit potential conflicts of a coevolving cultivar 30–35 Mya after the first with their crops over mushroom production that may damage attine ants committed to farming. symbiotic performance when expressed. The phylogenetically basal “lower attine” lineages to which social evolution | crop domestication | symbiosis | geometric framework | M. smithii belongs evolved more than 50 Mya when the ants Mycocepurus smithii Significance esources derived from mutualists often enhance the ecological Rdominance of partnered species (1, 2), but can also lead to Early subsistence farming implied significant physiological chal- conflicts between them (3, 4). Such conflicts typically stem from lenges for Neolithic farmers until they genetically isolated their diverging reproductive interests between hosts and symbionts, which crops through artificial selection and polyploidization. The attine may arise for two reasons. First, there is a fundamental conflict over ants faced analogous challenges when they adopted fungus symbiont mixing because any host compartment that is accessible farming 55–60 Mya. Whereas evolutionarily derived attine line- for multiple symbiont strains incurs a risk of symbionts competing ages irreversibly domesticated cultivars approximately 25 Mya for host resources (5) or of one of them free-riding on the services and ultimately realized industrial-scale farming, basal lineages of the other (6, 7), to the detriment of the host. The most harmo- retained small-scale farming, diversified, and now coexist with nious mutualisms are therefore expected to always involve a single advanced fungus-farmers in most New World tropical ecosys- strain of clonal symbiont per individual host. Typical examples are tems. We show that management of independent sexual re- the mitochondria and plastids that were domesticated by early eu- production in cultivars constrained farming productivity, echoing karyote protists (8, 9), and nutritional symbionts of insects with early human farming of unspecialized low-productivity crops. Loss of cultivar gene exchange with nondomesticated relatives specialized diets that are clonal within host individuals but geneti- – cally variable across hosts (10–12). Second, when mutualisms evolve likely reduced host symbiont conflict over reproduction, foster- ing the rise of ecologically dominant ant-agriculture. vertical symbiont transmission, hosts are selected to suppress ten-

dencies of their symbiont to continue investing in traits favoring Author contributions: J.Z.S., W.T.W., and J.J.B. designed research; J.Z.S. and E.B.G. per- horizontal transmission, because to quote Axelrod and Hamilton, formed research; J.Z.S. and P.W.K. contributed new reagents/analytic tools; J.Z.S., P.W.K., “...[a]nysymbiontthatstill has a transmission ‘horizontally’...would J.C.S., and J.J.B. analyzed data; and J.Z.S. and J.J.B. wrote the paper. be expected to shift from mutualism to parasitism...” (13). Such The authors declare no conflict of interest. selfish symbiont traits that use host resources to enhance independent This article is a PNAS Direct Submission. reproductive success without offering the host any returns (14) have Freely available online through the PNAS open access option. been likened to parasitic virulence (5), highlighting that mutualisms Data deposition: The sequences reported in this paper have been deposited in the Gen- are forms of reciprocal exploitation unless reproductive interests are Bank database. For a list of accession numbers, see Table S9. completely aligned (9, 15). 1To whom correspondence should be addressed. Email: [email protected]. – Host symbiont conflicts over horizontal symbiont transmission This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. are particularly acute when symbionts are multicellular eukaryotes 1073/pnas.1606128113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1606128113 PNAS Early Edition | 1of6 Downloaded by guest on September 26, 2021 specialized in farming gardens of basidiomycete fungi provisioned cultivar performance in vitro? (ii)Doantfarmersprioritize with nutritionally poor forest-floor detritus (21–23). This commit- harvesting proteins and carbohydrates based on their distinct ment to farming later led to the emergence of the “higher attines,” effects on somatic and reproductive growth of cultivars? (iii)Do including the and leafcutter ants that rear truly naturally available foraging targets allow farming ants to nutri- domesticated and coevolving cultivars on fresh plant-material sub- tionally regulate garden growth under field conditions? strates (22–25). However, it has remained underappreciated that (iv) Do farming ants deviate from maximal foraging expecta- relatively unproductive farming in the lower attines has otherwise tions for somatic cultivar growth to avoid overt reproductive been very successful in coping with the nutritional challenges of an conflict with their cultivars? The results that we obtained show exclusive fungal diet, in maintaining homeostatic growth conditions that there is a continuous threat of host–symbiont conflict over for small gardens, and in controlling fungal pathogens and social resource allocation toward horizontal symbiont transmission, – parasites (26 28). The continued survival and diversification of consistent with fundamental constraints in crop-yield pro- these small-scale farming mutualisms over evolutionary time thus ductivity that may have precluded the evolution of larger-scale suggests that these ants have found ways to control potential con- farming in phylogenetically basal attine ants. flicts emanating from the unavoidable sexual inclinations of their symbiotic cultivars. Results Because growth and reproduction in free-living basidiomycete Question 1: What Nutritional Blends Maximize Fungal Cultivar Performance fungi depend on access to specific amounts of protein and car- M. smithii in Vitro? We mapped somatic (hyphal) and reproductive (mushroom) bohydrates (29, 30), we explored whether and how M. smithii workers manage potential conflicts over cultivar reproduction by growth of fungal Petri dish cultures isolated from a provisioning fungus gardens with nutritional blends that promote colony across a landscape of 36 agar-based diets varying in P:C ratio (1:9–9:1) and P + C concentration (8 g/L–60 g/L) (diets hyphal growth while discouraging mushroom formation. We SI Methods performed geometric framework experiments (31) with a range modified from ref. 32) ( and Table S1). Fungal so- of nutritionally defined protein:carbohydrate (P:C) diets to matic growth was maximized by a high-carbohydrate blend (1:3 – generate a nutritional requirements map for fungal cultivars P:C 1:9 P:C) and constrained by protein in excess of 20 g/L that could be compared with nutritional foraging strategies of (Fig. 1A and Tables S2–S4), indicating that ant farmers can ant farmers. The graphical tools provided by the geometric maximize crop growth rates by harvesting carbohydrate-biased framework enabled us to quantify the amounts and blends of substrate. The likelihood of mushroom production was also protein and carbohydrates that lead to the expression of host– highestatP:Cratiosfrom1:3–1:9, but became already inhibited symbiont conflicts, and thus visualize nutritional trade-offs that by absolute protein amounts >10 g/L (Fig. 1B and Table S2–S4). could select for foraging behaviors to promote alternative cul- Thus, the greater protein tolerance of hyphal growth (<20 g/L) tivar growth trajectories. relative to mushroom production (<10 g/L) appears to provide a These experiments allowed us to address four fundamen- means by which ant farmers can nutritionally resolve potential tal questions: (i) What nutritional P:C blends maximize fungal allocation conflicts with their crops.

AB lahpyH htworg ,aera( )²mm egatnecreP fo setalp gnicudorp smoorhsum

60 60

1:9 1:6 1:3 1:2 1:9 1:6 1:3 1:2

50 50

40 40 250

1:1 1:1 30 30 200 2:1 30 2:1 0 3:1 3:1 Carbohydrates (g/L) 20 20

05

150 10 6:1 10 20 0 6:1 40 300 9:1 9:1

10 0 0 0 102030405060 0 102030405060 Protein (g/L) Protein (g/L)

Fig. 1. Mapping fungal cultivar performance across a landscape of 36 experimentally defined artificial media varying in total (grams per liter) amounts and relative (P:C ratio) amounts of protein and carbohydrates. (A) High-carbohydrate, low-protein nutritional blends maximized both fungal somatic growth rate (hyphal area in square millimeters after 32 d), and (B) the probability for plated cultivars to produce mushrooms (percent of surviving plates with mushrooms after 80 d). Dark blue indicates the lowest values for the two response variables and higher values are represented by increasingly red isoclines peaking at >300 mm2 for somatic growth area and >60% for plates with mushrooms. Fungal isolates (black circles) were inoculated on media spanning nine P:C ratios (1:9, 1:6, 1:3, 1:2, 1:1, 2:1, 3:1, 6:1, 9:1, gray lines extending from the origin) and four protein plus carbohydrate concentrations (8, 20, 40, 60 g/L). The response surface regressions producing the color-gradients were significant in both panels (P = 0.0001) (Tables S2–S4).

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1606128113 Shik et al. Downloaded by guest on September 26, 2021 ABs Total diet harvest (mg) Fig. 2. Experimental foraging preferences of farm- ing ants provisioning their cultivars. (A) A diet ex- 300 a 120 periment providing fixed nutritional rails of 1:6, 1:3, 1:1 c 1:6 1:3 200 1:1, 3:1, and 6:1 extending from the origin (31) 100 b showed how workers prioritized specific protein and 100 c carbohydrate mixtures over 22 d (colored dots). The 80 dashed curve connecting diet harvest values reflects 0 the overall nutritional compromises, showing that 60 1:6 1:3 1:1 3:1 6:1 workers prefer underharvesting carbohydrates rather Diet treatment Diet P:C than exceeding tightly regulated low levels of protein CD harvest. (B–D) Cumulative amounts of total diet, 40 1:6 120 120 a protein, and carbohydrates harvested during the P:C 3:1 1:3 diet experiment. Colonies consistently harvested low Carbohydrate harvest (mg) 20 1:1 80 80 6:1 3:1 amounts of protein and variable amounts of carbo- b 6:1 40 40 hydrates. Different letters indicate pairwise differ- 0 ences that were significant at P < 0.05. All error bars 020406080100 120 0 0 1:6 1:11:3 6:13:1 1:31:6 3:11:1 6:1 are SEs around the observed means. Diet harvest Protein harvest (mg) Diet treatment statistics are provided in Table S6.

Question 2: Do Ant Farmers Prioritize Harvesting Proteins and Carbo- substrates while inhibiting mushroom production via carefully hydrates Based on Their Distinct Effects on Somatic and Reproductive dosed frass provisioning. Growth of Fungal Cultivars? To understand how foraging ants navigate the trade-offs suggested by the in vitro cultivar growth Question 4: Do Farming Ants Deviate from Maximal Foraging Expec- results, we performed controlled feeding experiments with entire tations for Somatic Cultivar Growth to Avoid Overt Reproductive laboratory colonies to test whether M. smithii foragers collec- Conflict with Their Cultivars? To answer this question, we ex- tively blend protein and carbohydrates in amounts and ratios that amined how nutritionally variable substrates harvested by ants promote somatic growth while inhibiting mushroom production. impacted the performance of the entire farming symbiosis. We An experiment offering colonies single P:C ratio diets of 1:6, 1:3,

first established that a higher proportion of protein in synthetic EVOLUTION + 1:1, 3:1, or 6:1 (always at a 20 g/L protein carbohydrate con- P:C diets gave a proportional increase in nitrogen content of the centration) (Table S5) established that colonies tightly regulate 2 hyphal material on which the ants fed (χ 4 = 12.45, P = 0.014) protein harvest at low levels, while allowing for substantial var- (Fig. 4A). Second, these higher protein concentrations in diet iation in carbohydrate harvest across diets (Fig. 2 and Table S6). substrate and fungal food severely reduced performance of By strategically prioritizing protein regulation, ant workers ef- Mycocepurus farmers because protein-enriched crops had a fectively starved their cultivars of carbohydrates when offered χ2 = P = B ≥ higher likelihood of total failure ( 4 14.04, 0.007) (Fig. 4 diets with P:C ratios 1:1, as higher intake rates would imply C A and , Fig. S1,andTable S6), as well as higher worker mor- levels of protein harvesting that reduce yield (Fig. 1 ). At the D other end of the scale, the ant workers observed a rather con- tality (Fig. 3 and Table S6) and reduced brood mass of ant stant upper limit of carbohydrate harvesting despite the potential colonies (Table S6). Third, these results were unlikely to be af- for obtaining conflict-burdened cultivar growth benefits (Fig. 1A) fected by the type of cultivar because we did not detect a phylo- on increasingly carbohydrate-rich 1:3 and 1:6 diets (Fig. 2D). genetic signal in somatic growth rate variation across agar plates from nine randomly selected cultivars reared by M. smithii K = P = Question 3: Do Naturally Available Foraging Targets Allow Farming Ants colonies from the same population ( 0.419; 0.845) to Nutritionally Regulate Garden Growth Under Field Conditions? To (Fig. 5). Mycocepurus link nutritional decisions in laboratory colonies to food particles Overall, these results confirm that the farming harvested under field conditions, we performed extensive field system is generally adapted to provisioning fungus gardens with observations of M. smithii foragers in a lowland Panamanian low-protein substrate and that these adaptations have made the rainforest (Table S7). In terms of substrate types, M. smithii workers harvested mostly insect frass, wood fragments, and small seeds (Fig. 3). In terms of the underlying nutritional decisions, low foraging rates (on average four workers returning to their nests Substrate harvested in the field

with substrate per hour) resulted in small hourly amounts of wood substrate harvest (1.3-mg dry mass) (Table S7)withlow(<2%) average nitrogen content per substrate fragment (Table S8). From seed this finding, we inferred that Mycocepurus fungus gardens were Insect frass

provisioned with approximately 0.12-mg crude protein per hour, flower which translates into an average daily harvest of 1.4 mg–2.4 mg M. smithii worker other depending on circadian foraging patterns (SI Methods). Field provisioning rates thus provide cultivars with protein amounts similar to the mixtures that produced the red and yellow Fungus garden areas in the left-hand parts of the heat maps that we obtained in the laboratory (SI Methods), which is intriguing because these represent regions in the P:C acquisition landscape that are com- patible with simultaneous maximization of somatic cultivar growth (Fig. 1A) and reproduction via mushrooms (Fig. 1B). However, foraging Mycocepurus ants also have opportunities to navigate within this P:C acquisition landscape because the typical insect frass piece exhibited 2.5-fold more protein content than small Fig. 3. Natural foraging preferences of farming ants provisioning their wood fragments (Table S8), and frass pieces also exhibited cultivars. In the field, M. smithii workers harvest mostly insect frass (52%), substantial variation in protein content (0.67%–26.98%) (Table wood pieces (29%), and small seeds (14%) as natural substrate for the S8). This nutritional variation may provide foraging Mycocepurus fungal cultivars they maintain in hanging gardens and consume as their ants with a mechanism for enhancing hyphal growth with wood sole food source.

Shik et al. PNAS Early Edition | 3of6 Downloaded by guest on September 26, 2021 A 1.0 B 100 0.8 80 0.6 0.4 60 Diet P:C 0.2 1:6 0.0 40 1:3 1:1 -0.2 intact gardens

Change in fungal %N 20 3:1 -0.4 Percent colonies with Fig. 4. Garden nutrition, performance of cultivars, 6:1 and worker mortality in the diet experiments of Fig. 2. -0.6 0 1:6 1:3 1:1 3:1 6:1 0 5 201510 52 (A) The %N of fungal crops (and thus their protein Diet treatment Day content) increased when workers harvested protein- rich substrate, and (B and C) this was associated with CD1:6 1:3 1:1 3:1 6:1 a higher likelihood of crop failure, and (D) with 160 1:6 6:1 higher rates of worker mortality (see text for statis- tics). The pictures of experimental colonies (viewed 120 from above) in C contrast a 1:6 P:C diet [thriving fungus garden hung by workers from the lids of the 80 Petri dish, as is typical for healthy field gardens hanging from the ceiling of nest cavities (green im- 40 Worker number Worker age; see also Fig. 2A)], and a 6:1 P:C diet [dead 0 fungus garden discarded from the nest in an exter- healthy fungus crop failed fungus crop 020 200200200200 nal trash pile and with all brood having been dis- Day carded as well (red image)].

symbiosis protein-averse. However, despite this general aversion, productivity constraints because the carbohydrate-rich, protein- workers were inclined to harvest more protein and less carbohy- poor substrates that maximize growth rates of edible hyphae also drate than what could have maximized somatic hyphal growth. enabled the production of costly mushrooms when intake rates These patterns of restrained nutrient mixing appear consistent were high. with the necessity for farming ants to curtail potential host–sym- During the first 30–35 My of attine evolution, the ants pro- biont conflict over independent symbiont reproduction and with duced a monophyletic radiation of specialized fungus farmers. a productivity cost for the entire farming symbiosis to maintain However, their fungal cultivars were not genetically isolated stable mutualistic cooperation. from free-living fungi and could therefore not coevolve to any significant degree with their farmers as happened during the last Discussion approximately 25 My, after the ancestor of the higher attine ants The Questions Asked and the Answers Obtained. We used a state-of- came to initiate a new radiation based on a truly domesticated the-art nutritional geometry approach to understand how foragers cultivar lineage (16, 23–25, 35–38). Our study illustrates how this of fungus-growing ants adjust garden provisioning to suppress in- early interaction asymmetry between mutualistic partners has dependent reproduction by their cultivars in ways that also com- likely constrained ecological performance, which may help ex- promise symbiotic productivity. Our results expand the existing set plain both the low rates of fungus garden metabolism (39) and of geometric framework approaches (31–33) by: (i) providing a map the typically low rates of crop provisioning (four workers per nest of the nutrient requirements of fungal cultivars reared without per hour) (Table S7) in lower attine ants relative to evolution- farming ants (Fig. 1); (ii) using this map to track collective foraging arily derived Atta leafcutter ant colonies, where teeming foraging and provisioning decisions of ant farmers in laboratory experiments offering specific nutritional blends (Fig. 2), and in the field (Fig. 3); and (iii) evaluating the negative consequences of improperly biased nutrition for the combined performance of the Mycocepurus farming symbiosis (Fig. 4). 99 The results that we obtained yielded relatively clear answers to the four questions we posed. First, the cultivars of M. smithii thrivedwhenprovisionedwithcarbohydrate-biased substrate and suffered with increasing protein concentrations. Second, ant for- 62 aging preferences largely matched fungal preferences with workers harvesting substantial amounts of carbohydrates when they could 57 do so without simultaneously overharvesting protein from the same source, a strategy that avoided wasteful mushroom production 100 without excessive costs. Third, this behavior appeared to match 63 general adaptations to low-protein availability of saprophytic leaf- litter fungi on the forest floor, as expected because the cultivars of M. smithii remain closely related to free living fungi (16, 20, 34). At the same time, naturally available foraging targets, such as insect 04200 00 frass and dead plant fragments, are sufficiently variable in P:C ratios to allow differential garden provisioning, as we simulated Growth rate (mm²) in our garden-provisioning experiments. Fourth, differential Fig. 5. Somatic growth across M. smithii cultivar lineages. A sample of provisioning and nutrition of ants and cultivars does appear to cultivars from nine sympatric colonies of M. smithii exhibited substantial allow fine-tuned garden provisioning so that the expression of – genetic diversity without differences in growth rate (hyphal area in square host symbiont conflict over cultivar reproduction can be avoi- millimeters after 30 d + SE) in a phylogenetically controlled analysis (Blom- ded, because hyphal growth remained sustainable under higher berg’s K = 0.419; P = 0.845) (for details, see SI Methods and Table S9). The protein levels than mushroom formation. However, it also tree is a majority rule consensus maximum-likelihood chronogram with appeared that Mycocepurus ant farmers face fundamental nonparametric bootstrap support values.

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1606128113 Shik et al. Downloaded by guest on September 26, 2021 highways mobilize thousands of workers to carry leaf fragments Table S1). We placed 5-mm-diameter plugs of pure fungal culture in 60 × to their nests. Although these productivity constraints may well 15-mm Petri dishes containing 12 mL of 36 sterile synthetic agar-based diets. have prevented the lower attines from obtaining an ecological We sealed the dishes with parafilm (diet × dilution: n = 12), checked them footprint in Neotropical ecosystems comparable to the leafcutter every few days for contamination, and removed infected agar from some ants, the substantial radiation of farming strategies and species plates while discarding plates if overrun with contaminants. After 32 d we photographed noncontaminated plates (n = 360) and es- diversity represented by the extant lower attines also suggests 2 that these early farmer lineages gained control over the expres- timated fungal expansion rate (fungal area after 32 d, mm ) with ImageJ sion of selfish symbiont traits soon after becoming farmers. (NIH Image; v1.49g) (as per ref. 32). After 80 d we photographed all instances of mushroom growth on the same plates (n = 47), and used percent surviving The stoichiometric and geometric framework approaches de- plates with mushrooms for each diet treatment as the dependent variable in veloped for this study highlight a mix of evolutionary adaptations subsequent analyses. We used least-square regressions with both linear and and constraints associated with ant subsistence farming. Our quadratic terms to evaluate how fungal expansion rates and percentages of present study clarifies some of the nutritional mechanisms by surviving fungus plates with mushrooms varied across the 36 diet protein and which lower attine ants may have tamed their cultivars as long as carbohydrate combinations (32). Nonparametric thin-plate splines that do not gene flow with free-living fungal relatives precluded proper do- constrain the shape of the response surface were determined with the “fields” mestication. However, the precise mechanisms by which M. smithii package v.2.14.0 in R and used to map fungus growth and mushroom pro- ant workers regulate garden nutrition and suppress mushroom duction across the 36 diets (32, 45, 46). production need further study, both in the laboratory where we We measured growth rate over 30 d on PDA of pure fungal cultures isolated have far from exhausted the opportunities offered by the geo- from nine M. smithii colonies (n = 5 dishes per colony) using the protocols metric framework approach, and in the field, where further studies described above. To test for fungal strain growth differences, we isolated DNA could manipulate foraging on frass and dead plant material to from these fungal samples with 10% (wt/vol) Chelex extractions (47) followed explore health consequences for natural M. smithii colonies. by PCR amplification of the conserved nuclear large subunit (LSU) rRNA. The resulting PCR products were sequenced at Beijing Genomics Institute (BGI) Remarkable Parallels with the Challenges of Human Subsistence Europe, producing sequences that were deposited in GenBank (see Table S9 for corresponding accession numbers). We used these sequences to construct a Farming. The earliest human subsistence farmers faced severe chronogram using the function phylosig from the ape package in R v3.0–8to physiological challenges (40, 41), in part because plant-based test for a phylogenetic signal in fungal growth rates across colonies using agricultural diets provided more carbohydrates and less protein Blomberg’s K (48). than the lean game that dominated diets of ancestral hunter- We extended the geometric framework logic described for fungal cultivars gatherers (31). Modern humans appear to retain these “outdated”

to the entire farming symbiosis. We first cataloged the substrate harvested by EVOLUTION nutritional adaptations, regulating the intake of protein more M. smithii workers during approximately 20 h of observations at 22 colonies strongly than carbohydrates, and thus overeating sugars, starch, and in Soberanía Park from November 2013 to December 2013 (Table S7). We lipids to reach set targets for protein consumption when navigating then collected additional substrate for analyses of elemental N (%N) from today’s nutritional landscape that has become even more biased October 2014 to December 2014. This process allowed us to estimate the toward carbohydrates (42). The extant relatives of the earliest small- percent crude protein in harvested substrate (Table S8), assuming that scale fungus-farming ants in the present study echo this “leveraging” proteins contain, on average, 16% N (49). We also harvested entire of protein intake (Fig. 2), although the mechanisms likely differ M. smithii colonies from May 2014 to June 2014 and established them in because attine ants appear to have an upper protein tolerance set by plastic containers with ad libitum water and ground polenta as forage fungal symbionts adapted to low protein conditions (Figs. 2 and 4). for garden maintenance. M. smithii workers also appeared to regulate protein harvest more We performed a P:C diet laboratory feeding experiment within the tightly, and at lower levels, than nonfarming ant species (33, 43, 44), geometric framework (32, 34, 50) to analyze the foraging strategies by which may present additional physiological challenges because the which workers prioritize protein and carbohydrates when harvesting nutri- ants need substantial amounts of protein to fuel colony growth. tionally defined agar-based substrates. The diets contained 20 g/L macro- We have shown that Mycocepurus ant farmers could neither in- nutrients and known P:C ratios modified from the literature (50) (Table S5). We used a no-choice experiment confining colonies to a single P:C diet (1:6, crease carbohydrate provisioning without cultivars allocating the ex- 1:3, 1:1, 3:1, or 6:1), simulating field conditions with a constant P:C ratio of cess toward mushroom production, nor increase protein provisioning substrate. We replaced diets every day and estimated cumulative protein without compromising somatic cultivar growth. The ants appear to and carbohydrate harvest of colonies from dietary P:C ratios and dry:wet navigate nutritional intake trade-offs in a prudent and possibly fit- mass ratios of control diet pieces (SI Methods). ness-maximizing manner, just as human subsistence farmers have We used general linear mixed models (GLM) to test for diet treatment always tried to do. The phylogenetically basal forms of attine farming effects on: cumulative diet harvest (total diet, protein, and carbohydrates), have remained very successful over evolutionary time, despite not final worker number, final colony mass (adult ants + brood + fungus gar- reaching ecological dominance, quite possibly because they evolved dens), and final brood mass (SI Methods, Fig. S1, and Table S6). We included in an empty niche and could gradually respond to selection for the initial colony mass as a covariate and colony ID as a random factor because highest possible productivity. When higher attine ants eventually the largest of the 20 collected colonies was initially divided into five sub- emerged, they gradually moved into another, functionally herbivo- colonies and distributed across diet treatments, yielding 25 experimental rous niche, so did not compete for the same resource base, which colonies. We estimated initial worker number, retrospectively, from dead may have contributed to the coexistence of ancestral subsistence workers collected during the experiment and final survivors. Changes in colony farming and industrial-scale farming of attine lineages in many demography were observed over 29 d, whereas diet harvest was measured neotropical ecosystems. In contrast, the cultural evolution of human over 22 d because crop failure on high-protein diets precluded measurements agriculture proceeded rapidly and modern industrial-scale farming toward the end of the experiment. We analyzed the proportion of colonies with failed crops at the end of the experiment using a GLM with a binomial almost inevitably replaces subsistence farming because the two ag- distribution with a logit link function, and diet protein availability as ordinal ricultural practices compete for the same land, water, and fertilizers. variable. Fungal samples were collected from each colony for %N analysis on the first and last days of the experiment, and a Kruskal–Wallis test was used to Methods examine changes in fungal crop %N for each dietary P:C treatment. We used a geometric framework approach to assess the nutritional re- Nutritional substrate selected by workers could potentially offer a biased quirements for maximizing hyphal growth and mushroom production in pure estimate of fungal resource provisioning if workers used substantial clonal cultivar inoculates isolated from a large colony of M. smithii (177198) amounts of carbohydrates (but not proteins) to fuel their own metabolic collected from the rainforest at Soberanía National Park, Panama (N 9.11489, demands. We tested this conjecture by comparing the carbon mass in

W 79.69784), and grown on Petri dishes filled with a standard potato dextrose harvested sucrose with estimates of carbon mass respired as CO2 by workers agar (PDA) medium. We offered nine different P:C ratios (1:9, 1:6, 1:3, 1:2, 1:1, in the 1:6 and 6:1 P:C diet treatments over 22 d. Mass-balance calculations 2:1, 3:1, 6:1, 9:1) and four concentrations (8, 20, 40, and 60 g/L), yielding 36 (detailed in SI Methods), based on metabolic data from a previous study (39), replicated culture conditions in which we estimated hyphal growth rates and indicated that worker maintenance metabolism required a small fraction of probabilities of mushroom production (modified from ref. 32) (SI Methods and the harvested sucrose carbon mass, from 2.17 ± 1.00% to 5.96 ± 4.25% in the

Shik et al. PNAS Early Edition | 5of6 Downloaded by guest on September 26, 2021 high-carbohydrate 1:6 P:C and low-carbohydrate 6:1 P:C diets, respectively, on earlier drafts. The Autoridad Nacional del Ambiente y el Mar provided per- suggesting that the observed intake targets were not greatly influenced by mission to sample ants in Panama and export samples to Denmark. Elemental worker energy demands. This inference is consistent with farming ants rep- analyses were performed in the Turner Laboratory at the Smithsonian Tropical Research Institute. J.Z.S. was supported by a Postdoctoral Fellowship from the resenting a very small part of a colony-farm’s total biomass and energy con- Smithsonian Institution Competitive Grants Program (to W.T.W., J.J.B., and J.Z.S.) sumption (39). and by a Marie Curie International Incoming Fellowship (Grant 327940 Integrat- ing Social Evolution and Metabolic Ecology) hosted by J.J.B. J.C.S. thanks Jack W. ACKNOWLEDGMENTS. We thank M. Smeekes for field assistance; D. Donoso for Sites Jr. (Brigham Young University) and the National Science Foundation for verifying Mycocepurus smithii identifications; A. Dussutour and D. Nash for sta- financial support as a postdoctoral fellow. Additional support was provided tistical guidance and for providing M. smithii images; S. Wilder for advice about by general research funds from the Smithsonian Tropical Research Institute (to diets; and L. Ugelvig, N. Sanders, and M. Poulsen for offering insightful comments W.T.W.) and Danish National Research Foundation Grant DNRF57 (to J.J.B.).

1. Douglas AE (2010) The Symbiotic Habit (Princeton Univ Press, Princeton, NJ). 31. Simpson SJ, Raubenheimer D (2012) The Nature of Nutrition: A Unifying Framework 2. Bruno JF, Stachowicz JJ, Bertness MD (2003) Inclusion of facilitation into ecological from Animal Adaptation to Human Obesity (Princeton Univ Press, Princeton, NJ). theory. Trends Ecol Evol 18(3):119–125. 32. Dussutour A, Latty T, Beekman M, Simpson SJ (2010) Amoeboid organism solves 3. Jandér CK, Herre EA, Simms EL, Irwin R (2012) Precision of host sanctions in the fig complex nutritional challenges. Proc Natl Acad Sci USA 107(10):4607–4611. tree-fig wasp mutualism: Consequences for uncooperative symbionts. Ecol Lett 15(12): 33. Dussutour A, Simpson SJ (2009) Communal nutrition in ants. Curr Biol 19(9):740–744. 1362–1369. 34. Wardle DA (2002) Communities and Ecosystems: Linking the Aboveground and 4. Hoeksema JD, Schwartz MW (2003) Expanding comparative-advantage biological Belowground Components (Princeton Univ Press, Princeton, NJ). market models: contingency of mutualism on partners’ resource requirements and 35. De Fine Licht HH, Schiøtt M, Mueller UG, Boomsma JJ (2010) Evolutionary transitions acquisition trade-offs. Proc Biol Sci 270(1518):913–919. in enzyme activity of ant fungus gardens. Evolution 64(7):2055–2069. 5. Frank SA (1996) Host-symbiont conflict over the mixing of symbiotic lineages. Proc 36. De Fine Licht HH, Boomsma JJ, Tunlid A (2014) Symbiotic adaptations in the fungal Biol Sci 263(1368):339–344. cultivar of leaf-cutting ants. Nat Commun 5:5675. 6. Kiers ET, et al. (2011) Reciprocal rewards stabilize cooperation in the mycorrhizal 37. Kooij PW, Aanen DK, Schiøtt M, Boomsma JJ (2015) Evolutionarily advanced ant symbiosis. Science 333(6044):880–882. farmers rear polyploid fungal crops. J Evol Biol 28(11):1911–1924. 7. Bever JD, Richardson SC, Lawrence BM, Holmes J, Watson M (2009) Preferential al- 38. Schiøtt M, De Fine Licht HH, Lange L, Boomsma JJ (2008) Towards a molecular un- location to beneficial symbiont with spatial structure maintains mycorrhizal mutual- derstanding of symbiont function: Identification of a fungal gene for the degradation – ism. Ecol Lett 12(1):13 21. of xylan in the fungus gardens of leaf-cutting ants. BMC Microbiol 8:40. 8. Bourke AFG (2011) Principles of Social Evolution (Oxford Univ Press, New York). 39. Shik JZ, et al. (2014) Metabolism and the rise of fungus cultivation by ants. Am Nat 9. Cosmides LM, Tooby J (1981) Cytoplasmic inheritance and intragenomic conflict. 184(3):364–373. – J Theor Biol 89(1):83 129. 40. Piperno D, Pearsall DM (1998) The Origins of Agriculture in the Lowland Neotropics 10. Aanen DK, et al. (2009) High symbiont relatedness stabilizes mutualistic cooperation (Academic, London). – in fungus-growing termites. Science 326(5956):1103 1106. 41. Mummert A, Esche E, Robinson J, Armelagos GJ (2011) Stature and robusticity during 11. McCutcheon JP, Moran NA (2011) Extreme genome reduction in symbiotic bacteria. the agricultural transition: Evidence from the bioarchaeological record. Econ Hum – Nat Rev Microbiol 10(1):13 26. Biol 9(3):284–301. 12. Hosokawa T, Koga R, Kikuchi Y, Meng XY, Fukatsu T (2010) Wolbachia as a bacter- 42. Raubenheimer D, Machovsky-Capuska GE, Gosby AK, Simpson S (2015) Nutritional – iocyte-associated nutritional mutualist. Proc Natl Acad Sci USA 107(2):769 774. ecology of obesity: From humans to companion animals. Br J Nutr 113(Suppl): 13. Axelrod R, Hamilton WD (2002) The evolution of co-operation. Narrow Roads of Gene S26–S39. Land, Volume 2: Evolution of Sex, ed Hamilton WD (Oxford Univ Press, Oxford), p 155. 43. Cook SC, Eubanks MD, Gold RE, Behmer ST (2010) Colony-level macronutrient regu- 14. Law R, Lewis DH (1983) Biotic environments and the maintenance of sex—Some ev- lation in ants: mechanisms, hoarding and associated costs. Anim Behav 79(2):429–437. idence from mutualistic symbioses. Biol J Linn Soc Lond 20(3):249–276. 44. Kay AD, Shik JZ, Van Alst A, Kaspari M (2012) Diet composition does not affect ant 15. Herre EA, Knowlton N, Mueller UG, Rehner SA (1999) The evolution of mutualisms: colony tempo. Funct Ecol 26(2):317–323. Exploring the paths between conflict and cooperation. Trends Ecol Evol 14(2):49–53. 45. Lee KP, et al. (2008) Lifespan and reproduction in Drosophila: New insights from 16. Mueller UG (2002) Ant versus fungus versus mutualism: Ant-cultivar conflict and the nutritional geometry. Proc Natl Acad Sci USA 105(7):2498–2503. deconstruction of the attine ant-fungus symbiosis. Am Nat 160(Suppl 4):S67–S98. 46. Nychka D, Furrer R, Sain S (2015) Fields: Tools for Spatial Data. R package version 17. Poulsen M, Boomsma JJ (2005) Mutualistic fungi control crop diversity in fungus- 8.2-1. Available at https://cran.r-project.org/web/packages/fields/index.html. Accessed growing ants. Science 307(5710):741–744. May 22, 2015. 18. Aanen DK, et al. (2002) The evolution of fungus-growing termites and their mutu- 47. Walsh PS, Metzger DA, Higuchi R (1991) Chelex 100 as a medium for simple extraction alistic fungal symbionts. Proc Natl Acad Sci USA 99(23):14887–14892. of DNA for PCR-based typing from forensic material. Biotechniques 10(4):506–513. 19. Rabeling C, et al. (2011) Cryptic sexual populations account for genetic diversity and 48. Garland T, Jr, Bennett AF, Rezende EL (2005) Phylogenetic approaches in comparative ecological success in a widely distributed, asexual fungus-growing ant. Proc Natl Acad physiology. J Exp Biol 208(Pt 16):3015–3035. Sci USA 108(30):12366–12371. 49. Rothman JM, Dierenfeld ES, Hintz HF, Pell AN (2008) Nutritional quality of gorilla 20. Mueller UG, Rehner SA, Schultz TR (1998) The evolution of agriculture in ants. Science diets: Consequences of age, sex, and season. Oecologia 155(1):111–122. 281(5385):2034–2038. 50. Dussutour A, Simpson SJ (2008) Description of a simple synthetic diet for studying 21. Leal IR, Oliveira PS (2000) Foraging ecology of attine ants in a Neotropical savanna: – Seasonal use of fungal substrate in the cerrado vegetation of Brazil. Insectes Soc nutritional responses in ants. Insectes Soc 55(3):329 333. 47(4):376–382. 51. Cazin J, Jr, Wiemer DF, Howard JJ (1989) Isolation, growth characteristics, and long- 22. De Fine Licht HH, Boomsma JJ (2010) Forage collection, substrate preparation, and term storage of fungi cultivated by attine ants. Appl Environ Microbiol 55(6): – diet composition in fungus-growing ants. Ecol Ent 35(3):259–269. 1346 1350. 23. Nygaard S, et al. (2016) Reciprocal genomic evolution in the ant-fungus agricultural 52. Liu K, Raghavan S, Nelesen S, Linder CR, Warnow T (2009) Rapid and accurate large- symbiosis. Nat Commun 7:12233. scale coestimation of sequence alignments and phylogenetic trees. Science 324(5934): – 24. Schultz TR, Brady SG (2008) Major evolutionary transitions in ant agriculture. Proc 1561 1564. Natl Acad Sci USA 105(14):5435–5440. 53. Posada D (2008) jModelTest: Phylogenetic model averaging. Mol Biol Evol 25(7): – 25. De Fine Licht HH, et al. (2013) Laccase detoxification mediates the nutritional alliance 1253 1256. between leaf-cutting ants and fungus-garden symbionts. Proc Natl Acad Sci USA 54. Zwickl D (2006) Genetic algorithm approaches for the phylogenetic analysis of large 110(2):583–587. biological sequence datasets under the maximum likelihood criterion. PhD disserta- 26. Currie CR, Mueller UG, Malloch D (1999) The agricultural pathology of ant fungus tion (The University of Texas at Austin, Austin, TX). gardens. Proc Natl Acad Sci USA 96(14):7998–8002. 55. Huelsenbeck JP, Ronquist F (2001) MRBAYES: Bayesian inference of phylogenetic 27. Hughes WO, Pagliarini R, Madsen HB, Dijkstra MB, Boomsma JJ (2008) Antimicrobial trees. Bioinformatics 17(8):754–755. defense shows an abrupt evolutionary transition in the fungus-growing ants. 56. Rambaut A, Drummond AJ (2007) Tracer v1.4. Available at beast.bio.ed.ac.uk/Tracer. Evolution 62(5):1252–1257. Accessed December 15, 2007. 28. Adams RMM, et al. (2013) Chemically armed mercenary ants protect fungus-farming 57. Shik JZ, Hou C, Kay A, Kaspari M, Gillooly JF (2012) Towards a general life-history societies. Proc Natl Acad Sci USA 110(39):15752–15757. model of the superorganism: predicting the survival, growth and reproduction of ant 29. Plunkett BE (1953) Nutritional and other aspects of fruit-body production in pure societies. Biol Lett 8(6):1059–1062. cultures of Collybia velutipes (Curt.) Fr. Ann Bot (Lond) 17(66):193–217. 58. Waters JS, Holbrook CT, Fewell JH, Harrison JF (2010) Allometric scaling of metabo- 30. James SW, McLaughlin DJ (1988) The influence of carbohydrate source and concentra- lism, growth, and activity in whole colonies of the seed- Pogonomyrmex tion and light on fruit body development in Clavicorona pyxidata. Mycol 80(1):89–98. californicus. Am Nat 176(4):501–510.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1606128113 Shik et al. Downloaded by guest on September 26, 2021