Reciprocal nutritional provisioning between leafcutter ants and their fungal cultivar mediates performance of symbiotic farming systems

Optimized food acquisition is challenging because foraged diet items are chemically complex and often nutritionally imbalanced. These challenges are likely magnified when foraged foods are used to provision others (e.g., offspring, nestmates, symbionts) with different nutritional requirements.
We used a theoretical framework of nutritional niches to study these provisioning challenges in leafcutter ants that cultivate a fungal symbiont with nutrients derived from freshly foraged plant fragments. While the leaf-cutting behaviours of free-ranging foragers are well studied, little is known about how colonies use these plant fragments to produce their fungal crop within underground nest chambers.
For instance, gardener ants are known to convert vegetation into a nutritional mulch that they plant on the fungus garden. However, it remains poorly understood how the ants use this mulch to target the specific nutritional needs of their fungal crop, and whether the cultivar signals if provisioned mulch meets its nutritional needs. Towards answers, we performed three experiments to assess the precision and specificity of nutritional regulation in farming systems of the Panamanian leafcutter ant Acromyrmex echinatior.
A laboratory feeding experiment with nutritionally defined diets showed that ant farmers collect a specific intake target for protein and carbohydrates and then linked strict protein regulation by foragers to the cultivar’s fundamental niche for protein. 
An in vitro experiment with the fungal cultivar in isolation did not detect a signal of protein stress that could be used by the ants to regulate their provisioning behaviour, but it did identify an elevated fatty acid that may reinforce optimal nutritional provisioning if detected by gardening ants.
A feeding experiment with isotopically labelled diets then revealed nutrient-specific and caste-specific allocation timelines, with nitrogen being assimilated into the cultivar’s nutritional rewards before being exclusively consumed by developing brood. In turn, these combined results help resolve the integrated behaviours that give rise to resilient leafcutter farming productivity. 
These results show how nutritional niches can help disentangle reciprocal provisioning dynamics between symbionts while providing a framework to explore the nutritional transactions that mediate symbiotic stability (e.g., sanctioning, screening, policing).

Methods
Do ants differentially regulate nutrients?
We established ten queenless A. echinatior subcolonies (hereafter colonies) from five parent colonies that were collected in Soberanía Park (Panama) and maintained at the University of Copenhagen in a climate-controlled room (25°C, 70–75% RH) (Table S1). We used a diet-choice experiment to test if ants regulate intake targets for protein (P) and/or carbohydrates (C). Each day for ten days, colonies were provided two pre-weighed nutritionally-defined agar-based diets (ca. 15 g) in separate Petri dishes containing protein:carbohydrate (P:C) ratios of either 1:6 or 2:1 and P+C dilutions of 30 g/L (modified from Dussutour & Simpson, 2008; Shik et al., 2018) (Fig. 1a, Table S2). These ratios and the concentration of nutrients were chosen based on published results showing the fungal cultivar’s needs for these macronutrients and concentrations of these macronutrients in typically foraged plant fragments (Shik et al., 2021). Cumulative diet intake was calculated as dry diet mass collected by foragers each day, using control diet dry:wet ratios to estimate initial diet dry mass (as per Kay et al., 2012; Krabbe et al., 2019). P and C intake levels were calculated using diet recipes to determine their dry weight fractions. 

We performed Bartlett’s homogeneity of variance test to test for differences in the regulation of P and C intake, a Wilcoxon signed-rank test to test for differences in diet (1:6 vs. 2:1) and nutrient (cumulative protein vs. cumulative carbohydrate) intake across colonies, and a one-sample t-test to test whether the P:C ratio foraged by colonies differed from the P:C ratios provided by each diet. These and all subsequent statistical analyses were performed using R 4.2.1 (R Core team, 2022). Additional methodological and analytical details for this experiment and all described below are provided in Supplementary Appendix S1.

Do colonies regulate nutrients relative to cultivar needs?
We tested whether nutritional foraging choices of A. echinatior colonies are related to the FNN dimensions of their fungal cultivar. As part of a recently published comparative study, an in vitro experiment was used to measure the FNN for hyphal growth across ratios and concentrations of P and C for the L. gongylophorus fungal cultivar isolated from three colonies of A. echinatior (Shik et al., 2021). In that experiment, the three cultivar isolates were confined to Petri dishes with 36 diet treatments (P:C ratios: 1:9,1:6, 1:3, 1:2, 1:1, 2:1, 3:1, 6:1, 9:1; P+C dilutions: 8, 20, 40, 60 g/L) and eight replicates per diet treatment (N = 864 total Petri dishes). After a period of growth, two cultivar performance traits were measured (growth area (mm2), and staphyla density (number of staphylae per mm2)). Here, we replot the resulting cultivar FNN heatmaps in a new way and use them to complement the colony-level feeding experiments described above (Appendix S1). 

Heatmaps were generated using the fields package in R studio (Nychka et al., 2015) where blue areas indicate low performance and yellow areas indicate high performance. Performance isoclines were generated using non-parametric thin-plate splines. Least-square regressions were used to assess the significance of the linear and quadratic terms (and their linear interaction) of variation in the dependent variables (growth area, staphyla density) across the P and C diet treatments. This analysis was performed at the level of each isolate and on the mean values across isolates used to generate the figure (Table S3).

Does the cultivar signal its nutritional needs?
We used an untargeted approach to test for changes in profiles of 37 fatty acids putatively involved in signalling (Table S4) when fungal cultivars from five A. echinatior colonies were confined to Petri dishes with C-biased (1:3) and P-biased (3:1) P:C media (N = 45 Petri dishes; n = 3 plates per treatment per colony). Based on the FNN results, we selected P:C diets at the 8 g/L P + C dilution using the diet recipe from Crumière et al. (2021). This analysis included a test for the presence of linoleic acid (9,12 octadecadienoic acid (Z,Z)-, methyl ester, Table S4) due to its previous identification as an important fatty acid metabolized by L. gongylophorus and its association with behavioural modification of leafcutter ants (Khadempour et al., 2021). After 42 days of in vitro growth, fungal tissue was collected from the surface of growth media, flash frozen in liquid nitrogen and stored at -80 ºC for lipid extractions and analysis with gas chromatography mass-spectrometry. Values were adjusted relative to results from a control media treatment cultured in the same way but without being inoculated with the fungal cultivar. 

To test for P:C diet effects on lipid profiles, we performed a Principal Coordinate Analysis (PCoA) using the vegan package version 2.6-2 (Oksanen et al., 2022) with Bray-Curtis dissimilarity and a PERMANOVA in the vegan package with colony of origin included as a random effect to test for statistical significance. Candidate compounds were identified using a random forest analysis, which is a supervised classification and variable selection method for multivariate data, in the randomForest package V4.7-1.1 (Liaw & Wiener, 2022).

Do worker castes play distinct roles in nutritional regulation?
The function of three putative adult worker castes in A. echinatior colonies remains uncertain (Bot & Boomsma, 1996; Larsen et al., 2014) but knowledge of task discretization among these castes can help interpret whether measures of nutritional regulation described in sections 2.5 and 2.6 (see below) are sufficiently precise. We thus combined morphological, behavioural, and isotopic data to test for differences among minor, media and major workers. We first collected 10 workers of the body size categories representing each putative caste from each of six colonies (N = 180 workers) and measured their head widths (mm) and dry weight (mg). We then used a mixed model log-log scaling approach with colony identity as a random factor to test whether caste exhibit continuous variation in head with (HW) with body mass (M). If caste regressions exhibit different slopes, this provides evidence for ‘break points’ in body shape that define ‘physical caste’ systems (Oster & Wilson, 1979). We next used a modified ethogram approach (Santos et al., 2005; Calheiros et al., 2019) to identify and quantify three fungus-gardening behaviours in standardized nest fragments (fungus inspection, fungus licking, fungus reconstruction). We tested for differences among the three body size categories in the frequency of task performance using mixed models with colony identity as a random factor followed by posthoc tests (Table S5). 

Does the cultivar exhibit nutrient-specific assimilation and allocation?
We performed an isotopic tracer experiment on Day 0 of the diet choice experiment described above (section 2.1) where the P:C diets were enriched with known amounts of the heavy isotopes of carbon (13C-enriched glucose) and nitrogen (15N-enriched ammonium nitrate) (Table S2). Just before providing colonies with enriched diets on Day 0, we collected samples from each colony to quantify baseline enrichment of 13C and 15N (i.e., natural abundance). Following the heavy isotope ‘pulse’, we thereafter provided colonies ‘unenriched’ P:C diets and then collected samples for isotopic analyses on Days 1, 2, 5 and 10 (as per Shik et al., 2018).

On each sampling day, we collected samples of fungus gardens from each colony at three vertical layers (top, middle, bottom), and from two tissue types (hyphae (all three levels), staphylae (middle layer)). At each sampling day, we also sampled adult ants by pooling individuals of three putative castes from each colony (n = 5 minor; n = 5 media; n = 3 major). We separately analysed ant body segments to distinguish between assimilation (head-thorax) and transport/nestmate-dissemination/cultivar-provisioning (abdomen) (as per Shik et al., 2018).

How does the fungal cultivar regulate nutritional provisioning of its workforce?  
The isotopic data were used to test for caste-specific timelines of nutrient ingestion and allocation. Developing ants (larvae) are also called a digestive caste that, among other functions, likely mediates nutrient exchange within fungus gardens (Erthal et al., 2007). We thus also analysed enrichment timelines in larvae and pupae collected from the middle layer of the fungus garden of each colony on each sampling day. We further tested for nutrient-specific colony waste disposal by collecting trash samples from each colony every day. All samples were dried at 50°C for ≥ 24 hours and were then crushed, homogenized, and added in ca. 0.15 mg aliquots into pre-weighed tin capsules. Isotopic analyses of 15N/14N and 13C/12C were conducted using an Isoprime isotope mass spectrometer (Cheadle Hulme, UK) coupled to a Eurovector CN analyser (Pavia, Italy). For subsequent mixed-model statistical analyses, we calculated excess μg 15N and μg 13C per g dry mass of each sample (hereafter 15N and 13C as per Shik et al., 2018) (Table S6).