DISCUSSION

We describe a robust pattern where individual and superorganism (i.e. ant colony) growth increased with dietary protein, both as a function of the amount provided and its quantity relative to carbohydrates. This pattern was supported across two independent experiments on two common ant species in central Europe and the midwestern United States. These results support our hypothesis that colony size and the size of individuals within colonies are protein limited. Availability of protein likely constrains colony growth and is also a possible regulator of size and caste determination within colonies.
We also show evidence of nutrient trade-offs for colony growth and survival, as well as for individual size and lipid storage. A visual schematic of nutrient trade-offs is depicted in Fig. 1, such that for a particular trait there is a single optimal P:C. Trade-offs exist when different traits have different optima, namely, when the lines representing trait values over differing P:C values intersect. The lines for both colony (growth vs. survival) and individual traits (size vs. lipid content) intersect in our data (Fig. 9). The area of intersection represents the optimal investment, assuming that each of the traits has equal fitness effect. For both individual and colony-level traits, for both species, the intersection is at the level of the intermediate P:C used in this study. Interestingly, we may have detected the optimum P:C for both colony survival and growth; that is, intermediate P:C levels have the highest value of both survival and growth. On the other hand, the maximum values for the individual-level traits of size and lipid content are at the extreme P:C levels used in this study – the largest workers were produced at the highest relative P and the fattest workers at the highest relative C. The existence of trade-offs at the individual level suggest that colonies can manipulate the nutrients available to individuals within a nest in response to a changing external environment. For example, in response to environmental stress, colonies may modify the nutrients available to individuals within the nest, biasing provisioning toward carbohydrates over protein. At the colony level, a nutritionally mediated trade-off between growth and survival implies that selection may optimize nutrient collection to benefit the life-history trait most beneficial at a given time in the life cycle (including having variable strategies for nutrient collection).

Development of ant colonies

Individual workers and colonies (superorganisms) increased in size with increasing amounts of protein in the diet, and as a function of how much food was available, and yet there was no interaction between the two (Figs. 4-7, 9). However, the high mortality exhibited by L. neoniger colonies under one of the experimental diets (L) is consistent with other studies that find either lesser stress resistance or nitrogen toxicity in high protein/low carbohydrate diets (Fig. 4)(Cook, Eubanks, Gold, & Behmer, 2010; Dussutour & Simpson, 2009, 2012). We hypothesize that the high colony mortality was due to a decrease in lipid stores, as individual lipid content is a buffer for colony survival (Dussutour, Poissonnier, Buhl, & Simpson, 2016; Christopher R Smith, 2007). The lower survival in L. neoniger , but not L. niger , with the highest protein (L) diet may be due to different overwintering conditions for each species. L. neoniger was overwintered at a higher temperature compared to L. niger (although the temperature transition was the same for both species) and thus workers may have had higher metabolic rates and depleted greater amounts of their lipid reserves, leading to higher worker and colony mortality. In agreement with our data, lipid storage in other insects is increased with increased carbohydrate content of diet (Dussutour et al., 2016; Warbrick-Smith, Behmer, Lee, Raubenheimer, & Simpson, 2006). Note, the higher temperature used in L. neoniger was due to the concurrent overwintering of multiple, ecologically different, species.
Other studies have demonstrated that a change in the macronutrient ratios, toward a higher protein content, has toxic effects (Dussutour & Simpson, 2009, 2012; Harrison, Woods, & Roberts, 2012; Simpson & Raubenheimer, 2009). However, the P:C ratios in our study were all relatively high compared to studies documenting toxicity – increased protein content in our study was generally correlated with both increased worker number and worker size. A decrease in colony growth with even higher levels of protein was found in our preliminary study, which may have been due to a toxicity effect. Together, our preliminary data and the data presented in this paper suggest that optimal growth is achieved through a trade-off between sufficient carbohydrates for lipid storage on the one hand, and enough protein for growth on the other (but less than will cause toxicity). In both Lasius species, this optimum appears to be at a P:C of 1:4 to 1:8. These results are consistent with previous studies using similar diets, though those studies measured mortality rather than reproduction - mortality was lowest between 1:3 and 1:5 in Rhytidoponera sp., L. nigerand Solenipsis invicta (Cook et al., 2010; Dussutour & Simpson, 2009, 2012).
Colony growth was non-linear with respect to the amount of food provided. Colonies fed once per week did not grow any better than those fed twice weekly, suggesting that we found a maximum growth rate for colonies on the provided diets. Doubling food availability essentially saturated their ability to turn those nutrients into new ants. Colonies fed only once every two weeks, however, had decreased growth. Therefore, our levels of nutrient provisioning were appropriate to assess nutrient limitation as a function of both the amount and P:C ratio of provided diets.

Colony and Individual Phenotypes

Our study demonstrated that average ant size increased with increasing protein (relative to carbohydrates) and the total amount of diet provided. While we did not examine gene expression in this study, it has been established that caste determination in some social insects is regulated behaviourally and morphologically by nutrient-sensing genes, including those involved in insulin and Tor signalling (Wheeler, Buck & Evans 2006; Patel et al. 2007; Toth et al. 2007; and reviewed in Chandra et al., 2018; Smith, Toth, Suarez, & Robinson, 2008; Toth & Robinson, 2007), and caste determination is largely explained by size variation (Trible & Kronauer, 2017). Thus, a universal feature of ‘organismality’ is that size scales with the nutritional environment during development because the translation of nutrients into growth is achieved through conserved processes at the cellular level. If there existed a super-superorganism then we would also expect that its size, along with the size of its constituent parts, is nutritionally regulated; note that supercolonies, as seen in some ant species (Holway, Lach, Suarez, Tsutsui, & Case, 2002), do not likely fit the evolutionary definition of ‘organism’.
While worker size (such as head width or lean mass) was regulated by the quantity of protein in the environment, individual lipid content was regulated by the amount of carbohydrates (Fig. 6). As noted above, there was thus a trade-off faced by organisms with regard to their nutritional choices for these two major macronutrients (Fig. 9). Solitary insects, including caterpillars and last instar grasshoppers, tend to maximize growth with P:C near 1:1 (reviewed in Behmer 2009), whereas flies performed better with a more carbohydrate bias (Lee et al., 2008; Young, Buckiewicz, & Long, 2018). It is clear that in these studies there are life-history trade-offs inherent in different P:C ratios, such as longevity (maximized with increasing decreasing protein) and egg-laying (higher at increased protein).
While data are currently limited across insect taxa, we hypothesize that superorganism growth is maximized at a higher carbohydrate diet compared to most solitary insects. This prediction is premised on the majority of the standing biomass of a superorganism being non-growing, non-reproducing workers. While larvae have a higher carbon to nitrogen (C:N) ratio than workers, this is largely due to their high lipid content. The colony’s germline, reproductive gynes and males, have a higher nitrogen content (i.e., much lower C:N ratio) (Schmidt et al., 2012; C R Smith et al., 2008; Chris R Smith & Suarez, 2010). A logical extension of this prediction is that social insect growth and development is less protein limited than solitary insects and this may have been an inherent benefit to social living. This hypothesis is in line with how metabolic rate scales with body size in insect organisms and superorganisms. On a per mass basis, metabolism scales constantly across solitary and social insects (Hou, Kaspari, Vander Zanden, & Gillooly, 2010). Thus, eusociality and increased colony size are selectively advantageous with regard to increased metabolic efficiency.

Nutrient flows within colonies and colony-level preferences

We predicted that larvae would receive incoming protein, disproportionately, compared to carbohydrates and that the opposite would be true for workers. Using stable isotope labelling of the most preferred and optimal diet of L. neoniger , fed as a pulse, we traced nitrogen (as a proxy for protein) and carbon (as a proxy for carbohydrates) through colonies over four days. As is necessarily the case, nutrients were transferred from workers to larvae, but the signature of excess isotope enrichment in workers was gone in four days. Contrary to our expectation, after four days there were no differences between workers and larvae for either nutrient, and there, if anything (the result is marginally statistically significant), was the opposite pattern of nutrient transfer than predicted with larvae having slightly more of an excess of carbon than workers, but not nitrogen. Both workers and larvae at four days still showed atomic excess of labelled isotopes of both elements compared to colonies fed unenriched diets (i.e., they still had labelled nitrogen in/on their body). It is possible that proteins were being stored in workers until larvae were hungry - the timeframe of our study was insufficient to examine differences between developmental castes in diet assimilation. Additional studies using the basic isotope pulse strategy employed here, but with more time-points and a higher resolution of sampling, will help disentangle some of the complex interactions within colonies that are difficult to infer from behaviour, such as the relative distribution of nutrients to different sub-groups within a nest, and how rapidly they are distributed and assimilated (Shik et al., 2018).
Workers are non-growing, and thus feed primarily on carbohydrates to fuel their metabolism. As discussed above, we show that colony growth is maximized at intermediate levels of protein in the diet. When given choices of the same diets used to assay colony growth, colony preference was not in alignment with the optimum for colony growth in L. niger – higher protein diets increased growth, but higher carbohydrate diets were preferred. In L. neoniger , on the other hand, colony preference was uncannily aligned with optimal colony growth (data were consistent between field and lab colonies). Further work done only withL. neoniger found that preference for protein and carbohydrates in the diet is a result of clear preferences for each of the varying macronutrients, and thus the ants are judging the ratio. When each protein and carbohydrates were manipulated individually, the ants had a clear preference for increasing carbohydrates and for decreasing amounts of protein. What caused a decrease in preference at the higher levels of carbohydrates in the first preference experiment on L. neonigeris unclear, but the effect persisted across multiple preparations of the diets and was consistent in the lab and field (Fig. 8, Fig. S1).
It is difficult to draw generalizable conclusions from a single (one-season) diet/bait preference assay, and it is also difficult to judge optimality in diet from growth only under laboratory conditions. That being said, increased carbohydrate preference by foragers may be an adaptive strategy because this prioritizes survival over growth, and ant colonies tend to be long-lived. For example, when starved, colonies will prioritize investment in growth over investment in reproduction (Smith, 2007), presumably because they will have future attempts at reproduction should they survive the current period of low resource levels. Preference/foraging, though, may not be optimal due to many types of constraints (Pyke, 1984), and studies on some ants have failed to tightly link forager preference with productivity/fitness (Seal & Tschinkel, 2007). Foragers are responsive to the presence of larvae in the nest and adjust the collection of resources to more protein in their presence (Cassill & Tschinkel, 1999; Dussutour & Simpson, 2009). Therefore, it is not as though overall colony preference, as expressed through foragers, is not regulated by feedback. Furthermore, ants are capable of filtering nutrients once foragers return to the colony. Although in non-social organisms this mechanism would be composed mainly by selective absorption of nutrients and excretion of excess, ants use the larvae as a protein stomach and the nutrients are distributed across nest members (Deby L Cassill, Butler, Vinson, & Wheeler, 2005; Sorenşen, Busch, & Vinson, 1985; E. O. Wilson, 1976), or to stores/trash enriched with protein (Dussutour & Simpson, 2009).