RESULTS

Colony Survival and Growth

The first census of colony growth was in August 2017 and the last in April 2018 (months spent overwintered were not included in the analysis). Colony survival was affected only by diet treatment inL. neoniger (G = 9.19, df = 3, P = 0.03, Fig. 4a), but neither species showed an effect of feeding frequency on survival. Only the diet low in carbohydrates (L) showed decreased survival relative to all other treatments (P < 0.05 in all comparisons after FDR correction). For both species, each fixed effect (diet P:C, feeding frequency and time) in the mixed-effects model was statistically significant, as were the interactions Time*Diet and Time*Frequency (Fig. 5, Table S2). The latter two-way significant interactions suggest that the dynamics of diet P:C and feeding frequency changed with time, some of which were likely due to differences becoming more pronounced over time. Additionally, species-specific effects on growth were apparent from the non-recovery of L. neoniger colonies post-winter, especially due to differential survival (above).
In the analysis of overall growth, diet P:C and feeding frequency were statistically significant for both species, but there were no significant interactions (Fig. 4b, Table S3). The high carbohydrate diet (H) produced the lowest growth, differing from the low carbohydrate (L,P < 0.05) and medium-high carbohydrate (MH, P< 0.005) diets in post-hoc comparisons. The frequency that colonies were fed was highly significant, with only colonies fed every other week suffering reduced growth relative to the other two treatment levels (P < 0.001 in both post-hoc comparisons). Colonies of L. niger grew larger than colonies of L. neoniger , seemingly owing to differences in overwintering and recovery from winter

Worker Phenotypes

Differences among treatments in worker phenotypes are summarized in Fig. 6 for both species and full results are given in Table S4. All measured traits were significantly affected by diet P:C and feeding frequency separately in L. niger , but only head width exhibited a significant diet P:C x Frequency interaction. In L. neoniger , worker lean mass and lipid content were affected by diet P:C only, whereas head width was affected by feeding frequency only. L. neoniger worker dry mass was the only trait affected by both diet P:C and feeding frequency, and no interactions were significant for all the measured traits of this particular species.
Overall, increasing protein content in diet increased head width, lean and dry mass (Fig. 6). But as noted above, the effect was the opposite for worker lipid content. Although L. neoniger head width was not affected by diet P:C, the directional difference in means is consistent with the results obtained for lean and dry mass (i.e., more proteinaceous diets correspond to larger workers).
In addition, while feeding frequency only had a statistically significant effect on L. neoniger worker dry mass (both higher frequency treatments were marginally different than the lower frequency,P = 0.06 and P = 0.08), the direction of the effect was consistent across all dependent variables assessed for both species (Fig. 6) - more frequent feedings (i.e., greater food availability) led to larger and fatter individuals.

Stable Isotope Pulse Experiment

There was a statistically significant effect of caste by sample time for atomic enrichment of both 15N and13C (K-W test, χ2 > 12, d.f. = 3, P < 0.005 for both 15N and13C). It can be noticed that incoming nutrients are first ingested by workers, with worker atomic per cent of15N and 13C peaking at 24h. The 24h workers had nearly double the atomic enrichment of 13C compared to all other treatments and had five-fold higher enrichment of15N (Fig. 7). Workers were statistically different (P < 0.05) from larvae at both 24 and 96-hour collections for 15N, but the difference was not statistically significant in the 24-96 hour comparison for13C. No other pairwise comparisons were statistically significant. We had predicted that we would see a movement of more nitrogen to larvae relative to carbon, but this was not evident in the data. However, if anything the opposite appears more likely. Larval atomic per cent 15N remained constant over both sample times while the mean for atomic per cent 13C increased over sampling points (this result is trending toward being statistically significant, P = 0.11). Larval C:N was much higher than that of adult workers, 7.24 ± 0.18 vs. 5.00 ± 0.11 (mean ± SE, F1,39 = 108.9, P < 0.001), likely due to the high lipid content of larvae. Thus, despite a high protein need relative to workers for growth, the carbohydrate needs of larvae were substantial.

Diet preference

About 20 baiting trials were conducted for each species in the field, but several time intervals and bait stations had to be discarded due to intrusion by other species or a lack of Lasius presence. L. niger clearly tended to avoid the diet with the highest protein content (termed “L” in this study because of the low carbohydrate content) but did not differentiate among the remaining diets (Fig. 8), although pair-wise comparisons revealed that ML and MH were significantly preferred over L (P < 0.05). On the other hand, L. neoniger exhibited an overwhelming preference for the ML diet, with diet being highly significant in the mixed-effects model (χ2 = 52.01, P < 0.001). This trial was also replicated with two L. neoniger (ca. 2 years) lab colonies, and the result was qualitatively similar (data not shown).
In order to understand whether the preference for ML was driven by the numerator (carbohydrates) or the denominator (protein) in the ratio, we offered new baits with both macronutrients to L. neoniger field colonies, yet in series where only one varied. We hypothesized that the preference for ML was a composite of a combined preference for carbohydrates and proteins, especially because their response to the gradient in ratios was non-linear. When carbohydrates were held constant the ants had a strong preference for lower levels of protein; the number of ants present with the least amount of protein was double that compared to the most protein. The intermediate level of protein received an intermediate level of visitation. The lab results were consistent with the field data. The effect of diet in these trials was again highly statistically significant (χ2 = 29.09, P< 0.001). When protein was held constant, the ants had an increasing preference for more carbohydrates, with the preference increasing with increasing levels of carbohydrates (χ2= 13.47, P = 0.001) (Fig. S1).