|Home | About | Journals | Submit | Contact Us | Français|
Expression of several genes possibly involved in the symbiotic relationship between the obligate intracellular endosymbiont Blochmannia floridanus and its ant host Camponotus floridanus was investigated at different developmental stages of the host by real-time quantitative PCR. These included a set of genes related to nitrogen metabolism (ureC, ureF, glnA, and speB) as well as genes involved in the synthesis of the aromatic amino acid tyrosine (tyrA, aspC, and hisC). The overall transcriptional activity of Blochmannia was found to be quite low during early developmental stages and to increase steadily with host age. However, a concerted peak of gene expression related to nitrogen recycling could be detected around the entire process of pupation, while expression of biosynthesis pathways for aromatic amino acids was elevated only during a short phase in pupation. These data suggest an important role of certain metabolic functions for the symbiotic interactions of the bacteria and an individual host organism in early phases of development. General relevance of Blochmannia for its ant host was tested in fostering experiments with worker groups of Camponotus floridanus, and their success in raising pupae from first-instar larvae was used as a fitness measure. Groups treated with antibiotics had a significantly reduced success in raising the brood in comparison to untreated control groups, indicating that the symbiosis is relevant for the development of the entire colony.
Blochmannia species are obligate intracellular endosymbionts of ants of the genus Camponotus and closely related genera (20, 21). The genus Camponotus comprises more than 1,000 species, which are ubiquitously present in most terrestrial habitats (5). So far, Blochmannia spp. could be detected in every Camponotus species (>30) investigated, indicating an important role of the bacteria for the biology of the host animals (8, 18, 20). The bacterial endosymbionts reside in specialized cells, the bacteriocytes, which are intercalated between enterocytes of the midgut tissue of the host. In females, the bacteria also occur in the oocytes, leading to a vertical transmission of the bacteria (6, 13, 19, 20). Blochmannia is closely related to other insect endosymbionts, like Wigglesworthia glossinidia, the endosymbiont of the tsetse fly Glossina brevipalpis, and Buchnera aphidicola, the endosymbiont of aphids (1, 2, 3, 9, 11, 22). Phylogenetically, these bacteria belong to the Enterobacteriaceae and have a common ancestor with Escherichia coli. Aphids and tsetse flies are characterized by their strict food specialization to diets limited in essential nutrients. Their symbioses are clearly based on food supplement, the endosymbionts providing their hosts with essential metabolites, like amino acids and vitamins, respectively (10, 16). However, many members of the genus Camponotus are considered to be omnivorous, and accordingly, a nutritional basis for this symbiosis is not obvious at first glance.
The genome of Blochmannia floridanus has recently been sequenced. With only 705 kbp in size and 583 protein-encoding genes, it is extremely reduced, a feature typical of the genomes of the insect endosymbionts characterized so far (1, 9, 11, 22, 23, 24). B. floridanus still encodes basic metabolic functions required for energy metabolism or replication to maintain life in the specialized environment of the eukaryotic host cell. Additionally, a set of genes which seems to be related to the symbiotic function of the bacteria is present (11). For example, the Blochmannia genome encodes most biosynthetic pathways for amino acids essential for the host, with only the arginine biosynthesis pathway missing. In contrast, many of the nonessential amino acid pathways are missing. Among the few nonessential amino acid pathways retained in Blochmannia, the presence of the tyrosine biosynthesis pathway is striking and suggests that the biosynthesis of aromatic amino acids may be an important aspect of the symbiotic relationship between Blochmannia and its host (27). Most interestingly, a complete urease gene cluster is present in the B. floridanus genome. Urease activity results in the production of CO2 and NH3, the latter of which is a potent cell toxin, mainly due to its detrimental effect on membrane potential and ion transport (7).
Replication of the bacteria was found to be coordinated with host development and possibly regulated by the host (26). Replication takes place mainly during pupation and in young workers, which may indicate a particular importance of this symbiosis in distinct developmental stages of the host. In the present work, we investigated the differential expression of key enzymes involved in nitrogen metabolism by real-time quantitative PCR. These enzymes were chosen because the genome sequence of Blochmannia indicated their relevance for the symbiosis (11, 27). In particular, we investigated expression of genes related to nitrogen recycling (ureC, ureF, glnA, and speB) and to the synthesis of aromatic amino acids (tyrA, aspC, and hisC). ureC encodes a urease structural protein and ureF an accessory factor required for nickel incorporation in the active site of the urease, which is required for the activity of the enzyme. glnA encodes glutamine synthetase required for the incorporation of ammonium produced by urease in the amino acid biosynthesis pathways. speB encodes an enzyme of the arginase family which produces urea as an end product. In the case of the tyrosine biosynthesis pathway, we chose tyrA, which codes for prephenate dehydrogenase, while the aspC and hisC genes encode aminotransferases, which alternatively catalyze the final step of tyrosine synthesis. Expression of the bacterial genes was determined in eggs, larvae immediately before pupation, and pupae immediately after pupation, after metamorphosis, and immediately before eclosion, as well as in imagines immediately after eclosion (callows), several days old (taken from inside the nest), and older (foragers collected outside the nest). The tufB gene, encoding an elongation factor, was chosen as an exemplary housekeeping factor that should be expressed at a relatively constant rate and thus can be used as an internal standard of gene expression.
Finally, we discuss the possible role of the symbiotic interaction of Blochmannia with respect to the individual host animals and the entire colony. General relevance of Blochmannia for its host Camponotus floridanus at the colony level was tested with fostering experiments in parallel to the gene expression studies. If Blochmannia is required for normal development of individuals in the colony, then antibiotic treatment of workers that care for the brood may reduce colony fitness in that either fewer pupae develop from eggs or development of brood is slower.
Colonies of Camponotus floridanus were kept in artificial nests and cultivated in a climate chamber at the University of Würzburg at a constant temperature of 30°C, 50% humidity, and a 12-h day-night rhythm. Twice a week, they were fed with a combination of cockroaches (Nauphoeta cinerea), honey-water (50% honey, wt/wt) and Bhatkar agar (4).
Ants were taken from nine different developmental stages: eggs (E), larvae 2 to 3 mm in size (L1), larvae 3 to 4 mm in size (L2), pupae before metamorphosis (P1), pupae after metamorphosis (P2), pupae shortly before eclosion (P3), adults immediately after eclosion (before melanization was finished) (W1), adults estimated to be a few days old (taken from the nest) (W2), and older adults (foragers) (W3). Samples of each developmental stage were collected randomly from five different colonies and pooled before RNA isolation. Before RNA isolation, bacteria were extracted from their hosts as described earlier (26). Briefly, the abdomens of about 100 workers or pupae and whole animals of the remainder of the life stages were crushed in a glass homogenizer in isolation buffer (35 mM Tris-HCl, pH 7.6, 25 mM KCl, 250 mM sucrose). Host cell debris was removed by filtration through a nylon filter of 28 μm pore size. Bacteria were harvested by centrifugation, and the bacterial cell pellet was used immediately for RNA isolation. Total RNA of B. floridanus was isolated using an E.Z.N.A. bacterial RNA isolation kit (Peqlab, Erlangen, Germany). DNA contamination of the RNA preparations was removed by DNase I treatment (DNAfree; Ambion, United Kingdom), and the RNA yield was determined photometrically. On average, 20 μg of total RNA was obtained by this procedure. Contamination by eukaryotic RNA was evaluated by real-time quantitative PCR (RT-qPCR), amplifying a fragment of the host 18S rRNA. Levels were found to be quite similar in all samples and about equimolar to Blochmannia 16S rRNA as determined by primers specific for B. floridanus (21) (data not shown). One microgram of total RNA of each preparation was used for cDNA synthesis. cDNA was synthesized using a RevertAid first-strand cDNA kit (Fermentas, St. Leon Rot, Germany) and the reverse primers listed in Table Table1.1. All PCR primers were purchased from Sigma Genosys (Steinheim, Germany).
A fragment of about 150 bp of each of the rrs, tufB, ureC, ureF, glnA, speB, tyrA, aspC, and hisC genes of B. floridanus and the gene encoding the 18S rRNA of the host ant C. floridanus was amplified using the respective primers listed in Table Table1.1. The respective PCR products served as a standard for quantification. RT-qPCRs were performed using a DNA Engine Opticon system (MJ Research, Waltham, Mass.) and a qPCR core kit for SYBR green I (Eurogentec, Cologne, Germany). All reactions were performed according to the manufacturers' instructions. In each qPCR, 1 μl of the cDNA synthesis reaction was used as a template, corresponding to 50 ng of RNA template. Quantification was performed using at least two independent RNA preparations and repeated at least three times per replicate.
Thirteen founding queens of Camponotus floridanus were collected in the field in June 2003 in Tarpon Springs, Florida (collection by A. Endler and S. Diedering), and reared as described above for 18 months, until colonies had reached a size of approximately 600 to 1,000 workers. From each of the 13 colonies, two subcolonies were created by transferring 200 workers each into two separate artificial nests. The rest of the workers remained in the mother colony with the single egg-laying queen. Subsequently, the mother colony (including the queen and thus all brood produced by the queen) and one of the worker groups were treated with the antibiotic rifampin (C43H58N4O12; Serva Elektrophorese GmbH); the other worker group, used as a control, was not. The antibiotic was fed to the ants twice a week over a period of 13 weeks in a solution containing 1% rifampin, 49.5% water, and 49.5% honey (wt/wt) as described previously (19). Workers that died were not replaced with workers from the mother colony, and thus the numbers of workers in the subcolonies decreased steadily during the experimental period. Adult ants generally require a carbohydrate-rich diet only; however, their requirement for nitrogen-rich food rises strongly when they have to raise brood. Therefore, every week eggs and first-instar larvae were removed from each mother colony and divided equally between the two worker subcolonies, and the workers' success in raising pupae from the brood was used as a fitness measure. Pupae raised from eggs and larvae in the worker subcolonies were removed, counted, and stored.
In order to test whether the endosymbionts within the brood itself are sufficient for normal development, irrespective of the status of the endosymbionts within the workers that nurse them, we suspended the treatment with antibiotics after 13 weeks and removed all brood from the worker groups. The subcolonies were then kept an additional 4 weeks and were given eggs and small larvae collected from other lab colonies of C. floridanus that had not been treated previously with antibiotics. Again, pupae raised by the workers were collected and counted every week over a period of 4 weeks.
To test for possible toxic effects of rifampin on the workers themselves, the remaining workers within worker groups were counted at the end of the experiment after 17 weeks (13 weeks of treatment phase plus four subsequent weeks), and levels of mortality between antibiotic-treated subcolonies and controls were compared. Additionally, it was tested whether the number of remaining workers in a subcolony correlated with the number of raised pupae.
All statistical analyses were performed with Statistica 6.1. (Statsoft) and SPSS 12.02 (SPSS, Inc.).
Expression of several bacterial genes, including ureC, ureF, glnA, speB, tyrA, aspC, and hisC (see the introduction), was analyzed by real-time quantitative PCR in different developmental stages of the host ant Camponotus floridanus. Standardization of the absolute amount of cDNA detected in each sample to the amount of Blochmannia 16S rRNA yielded virtually the same expression pattern (data not shown). Expression of tufB and of the genes encoding metabolic proteins investigated showed a characteristic pattern, with significant differences in overall gene expression between different life stages of the host (F = 61.5, df = 8, P < 0.001; unifactorial analysis of variance [ANOVA] of log10-transformed data followed by a Scheffé post hoc test) (Fig. (Fig.1).1). We observed minimal overall bacterial gene expression in small larvae (L1) and in late pupal stages (samples P2 and P3), while in adult ants the bacterial gene expression increased significantly with age (samples W1 to W3) (Fig. (Fig.11).
In order to uncover differential gene expression in the bacteria, expression of the metabolic genes was standardized relative to the expression of the housekeeping tufB gene. The gene expression patterns showed significant differences between life stages (n = 341; F = 97.9, df = 8, P < 0.001; unifactorial ANOVA of log10-transformed gene expression relative to tufB), between genes within a life stage (n = 216; F = 180.8, df = 3, P < 0.001; unifactorial ANOVA of log10-transformed gene expression relative to tufB), and in the interaction between genes and life stage (F = 27.3, df = 24, P < 0.001; unifactorial ANOVA followed by Scheffé post hoc tests). The relative expression levels of ureC varied only threefold throughout the life span of the host (Fig. (Fig.2),2), and a slight increase of expression was observed only with late larvae and early pupae (samples L2 to P2). In contrast, the expression of ureF was barely detectable in eggs, where the expression of tufB exceeded expression of ureF about 100-fold, but ureF expression increased significantly in larvae (sample L), where it reached about 23% of tufB expression, and even exceeded tufB expression in later phases, reaching a maximum in late larvae. This late-larva phase also corresponds to the maximum level of ureC expression. A second peak of ureF expression could be observed in older workers (sample W3). In contrast, expression of ureF is reduced in older pupae and young workers, comprising the period of eclosion and sclerotization (samples P3 to W2).
The expression pattern of glnA strongly resembled that of ureF. The expression of glnA was significantly lower in eggs than in other life stages, with a tufB-to-glnA expression ratio of approximately 100:1. Relative glnA expression then increased to reach a maximum in middle-aged pupae (sample P2). A second minimal level of glnA expression was observed in young workers (sample W1), where tufB expression exceeded glnA expression about sevenfold. From that time point on, the expression of glnA increased steadily. In early developmental stages of the host (samples E and L1), ureF and glnA were expressed at a very low level, but their expression increased strongly in later stages (samples L2 to W3), with the expression of ureF exceeding the expression of glnA three- to sixfold.
The expression levels of the gene encoding the urea-producing enzyme SpeB were generally lower than those of ureF and glnA and did not differ significantly between life stages. A peak of speB expression was observed in young pupae (sample P1), while minimal gene expression was reached in old pupae immediately before eclosion (sample P3). From this time point on, expression increased steadily, reaching a second peak in old workers (sample W3). Overall, relative expression levels of ureC and ureF were most elevated in late larvae and early pupae, while maximal glnA and speB expression was observed in young and middle-aged pupae, respectively.
Expression of tyrosine biosynthesis genes in different developmental stages of the host was investigated. Again, expression was depicted relative to the expression of tufB. The gene expression pattern showed significant differences between life stages (n = 293; F = 33.2, df = 6, P < 0.001; unifactorial ANOVA of log10-transformed gene expression relative to tufB), between genes within a life stage (n = 125; F = 493.6, df = 2, P < 0.001; unifactorial ANOVA of log10-transformed gene expression relative to tufB), and in the interaction between genes and life stage (F = 15.4, df = 12, P < 0.001; unifactorial ANOVA followed by Scheffé post hoc tests). During most developmental stages of the host, the ratio of tyrA to tufB expression varied only slightly, between 0.2 and 0.4 (Fig. (Fig.3).3). In comparison to other life stages, the expression of tyrA was significantly elevated only in middle-aged pupae (sample P2), reaching a level almost similar to that of tufB. The expression of hisC was significantly lower in all life stages than that of tyrA and aspC but followed the same pattern, with a peak in middle-aged pupae. The expression of aspC was quite constant during most developmental stages, its level being similar to that of tyrA. Only in eggs (sample E) and again in middle-aged pupae (sample P2) was the expression of aspC significantly elevated, reaching about 80% and 130% of the tufB expression level, respectively. Overall, minimal tyrosine biosynthesis gene expression was observed in larvae, while each gene was expressed at its maximum level in middle-aged pupae. It should be noted that tyrosine synthesis is elevated only for a short phase, midpupation (sample P2), while the genes involved in nitrogen recycling are induced slightly earlier and for a more extended period of time.
To investigate a possible role of Blochmannia at the colony level, experiments were carried out with animals treated with rifampin, which, as previously described, causes a strong decrease of the bacterial load in treated animals (19). Confirming the previous results, rifampin treatment did not interfere significantly with the viability of the adult workers and no obvious toxic effect on these animals could be noted. In the feeding group treated with antibiotics, a mean number of 72.2 workers of the initial 200 workers per subcolony survived, in comparison to 66.4 of the untreated controls, during the time span of the experiments (n = 13; T = 35, P > 0.05; Wilcoxon's test for matched pairs). During the course of the experiment, subcolonies treated with antibiotics raised significantly fewer pupae from the first-instar larvae given to them than untreated subcolonies from the control group derived from the same mother colony (mean ± standard deviation [SD] for controls, 80.5 ± 16.3; mean ± SD for rifampin-treated subcolonies, 23.7 ± 6.4; n = 13; T = 0, P < 0.002; Wilcoxon's test for matched pairs) (Fig. (Fig.4).4). Since there is no correlation between the number of surviving ants and the number of raised pupae (product moment correlation coefficient, r = 1.1916 × 10−5; P > 0.05), the better success in raising brood cannot be attributed to the differences in the surviving numbers of workers.
In order to differentiate between the relevance of Blochmannia during larval development of each individual and a possible indirect effect of rifampin treatment on larval development due to a reduction in the workers' capability to raise brood, the fostering experiments were continued for 4 weeks after cessation of the 13-week antibiotic treatment. All subcolonies (controls and treatment group) were supplied with equal numbers of first-instar larvae taken from untreated colonies of C. floridanus. The numbers of pupae raised during this experiment were generally lower than those in the 13-week treatment phase, owing to the shorter period of time (Fig. (Fig.5).5). However, once more the success in raising brood was reduced in the rifampin-treated subcolonies in comparison to untreated controls, albeit not significantly (mean ± SD for untreated controls, 2.4 ± 2.7; mean ± SD for rifampin-treated subcolonies, 1.2 ± 1.2; n = 13).
The aim of this study was to uncover biological functions of the endosymbiotic bacteria of the carpenter ant Camponotus floridanus (i) by investigating the transcriptional activities of several bacterial genes that may be important during different stages of the host's life cycle and (ii) by monitoring development of brood in colonies treated with antibiotics.
The kinetics of expression of the investigated Blochmannia genes was analyzed by RT-qPCR. In general, relatively low bacterial gene expression levels were found in early phases of host development. Bacterial gene expression strongly increased in adult life stages of the host, with the highest transcriptional activity in workers of several months of age. This was very surprising, as previous studies have shown that the symbiosis degenerates in workers with increasing age, leading to a strong reduction in the number of Blochmannia genome complements present in the ant hosts (26). However, the high absolute gene expression level in the imagines points to an important function of the symbiosis also in the adult phase of the animals' life cycle (see below).
Expression analysis of the genes involved in nitrogen recycling revealed that the ureF gene is virtually not expressed in eggs but that its expression has two peaks later in the ants' ontogeny (in late larval stage to midpupal stage and in older workers). In contrast, ureC expression is almost constitutive throughout the life span of the animals. Thus, in spite of both genes being involved in the same enzymatic function, their expression patterns are uncoupled. The expression of ureF can exert a regulatory function, because urease requires the incorporation of nickel for activity, which is accomplished by the accessory factor UreF (7). The structure of the urease gene cluster in B. floridanus suggests the presence of at least two transcriptional units, one comprising the ureDABC genes and a second operon, consisting of the ureF and ureG genes, located immediately downstream. These two operons are separated by an intergenic region of roughly 200 bp likely to contain a promoter directing ureFG transcription (11). In support of this assumption, RT-qPCR analysis revealed less than 1‰ of cotranscription of ureF from the ureC promoter (data not shown). In Blochmannia pennsylvanicus, the intergenic region separating the putative urease operons is about twice the size of that in B. floridanus (9); sequence similarities were detected only in the region proximal to the ureF gene, which should contain the promoter, while no sequence similarities could be detected in the distal region (data not shown). Thus, the genomic organization of the urease gene cluster supports the finding of differential expression of urease genes and indicates that urease activity is modulated during the development of the animals. Interestingly, the expression of the glnA gene, encoding glutamine synthetase, follows the expression pattern of ureF. Glutamine is the main entrance gate of nitrogen into the cell metabolism. Increased glutamine synthetase activity may therefore contribute to the assimilation of nitrogen required for the biosynthesis of symbiosis-relevant building blocks and, concomitantly, may help in the detoxification of the potent cell toxin ammonium generated by the urease activity probably expressed at the same time point.
The detection of differential gene expression in Blochmannia is striking, since only very few regulatory genes have been retained in this reduced genome. Among these, there is the zur gene, encoding a transcriptional factor recognizing divalent cations as a signal that may be a sensor for the availability of nickel required to generate a functional urease. It is therefore tempting to speculate that Zur is involved in differential regulation of the urease gene clusters. In general, so far only very little information regarding gene regulation in such endosymbiotic bacteria is available. In B. aphidicola, which also retains only very few transcription factors, a very limited transcriptional response to environmental stresses such as heat shock and nutrient deprivation was observed. These results are in agreement with the hypothesis that, in line with genome reduction, the flexibility of the transcriptional response has also been reduced drastically (14, 25). In fact, in B. aphidicola, so far only in the case of the metE gene (required for methionine biosynthesis) was a differential upshift of expression observed under conditions of limited amino acid content in the diet. In line with this finding, among the few transcription factors retained is the merR gene, which therefore may be involved in the differential regulation of the metE gene (14).
Since the data presented here indicate a concerted peak of expression of key enzymes of nitrogen metabolism, these findings suggest a cooperation of urease and glutamine synthetase in order to feed ammonium into amino acid synthesis during brood development, especially in early pupation, where no food uptake coincides with high metabolic activity. On the other hand, in later life stages of the host the expression of ureF exceeds the expression of glnA and may therefore, by the massive production of toxic ammonium, contribute to the degeneration of the symbiosis described previously (26).
Tyrosine is a nonessential amino acid for insects (15). The tyrosine biosynthesis pathway is among the few pathways for nonessential amino acids which have been retained in the streamlined Blochmannia genome (27). Regarding the development of the holometabolous host animal, it is known that large amounts of aromatic amino acids are required for the sclerotization of the cuticle after each molt and especially immediately after eclosion of the imago (12). However, the solubility of these amino acids is very low and insects are known to store large amounts of soluble derivatives such as the respective glucosides or phosphates in their hemolymph to avoid shortcomings in the supply of these important compounds (15). Expression of the tyrosine biosynthesis genes tyrA, hisC, and aspC is significantly increased during a short phase in pupation. This may indicate that the bacteria contribute to production of aromatic amino acids in this developmental step and therefore to sclerotization of the cuticle following eclosion. Availability of tyrosine required for sclerotization plays a role in the ontogeny of individuals only during development from egg to imago but not in adult workers themselves since they do not molt any longer.
The fostering experiments conducted in parallel to the gene expression studies revealed that Blochmannia is relevant for colony fitness, since antibiotic treatment in order to reduce the number of symbionts led to a significant reduction in the worker group success in raising brood. Blochmannia could play an important role for its hosts on two different levels. First, the endosymbiont may be vital for the ontogeny from egg to imago. Second, the endosymbiont may also play a role at the colony level as food is exchanged between nest mates, including the brood, via trophallaxis and regurgitation. Thus, it may be sufficient when workers harbor the endosymbionts, and colonies may not suffer a fitness loss, even when the brood itself stems from an antibiotic-treated queen. The antibiotic treatment had no detectable negative effect on the adult workers themselves, as their mortality was not increased in comparison to that of controls. The lack of negative effects of lowering the bacterial load in adult animals by antibiotic treatment may indicate, therefore, that in this stage of the host's life cycle the bacteria are less important for the individual animal but they may be relevant at the colony level by providing high-quality food to the brood. The fostering experiments of ant colonies revealed that the ant host of Blochmannia can develop from egg to imago in spite of treatment with antibiotics that should lead to a strong reduction in the number of endosymbionts in the host (19). Eggs laid by queens that were treated with antibiotics and then raised by workers that were also treated still developed into healthy adults, albeit at a significantly reduced level in comparison to the level of brood raised by untreated worker groups. The lower number of pupae raised may be due to either slower development of the brood or higher mortality. Thus, it is likely that the nutrition provided by the antibiotic-treated workers is of lower quality, probably due to clearance of the endosymbiotic bacteria from their hosts. That provisioning with food by workers plays a significant role in brood development is supported by the result that after suspension of treatment with antibiotics, the group formerly treated still raised fewer pupae than the untreated controls. Arrest of development of the brood at an early larval stage was not observed, as the workers themselves may have contributed to brood development with their own body reserves. Additionally, dead brood may be used as a nitrogen source and cannibalized to feed the surviving larvae. Cannibalistic behavior that may buffer times of food shortage has previously been observed for C. floridanus (17).
In conclusion, the data presented in the manuscript indicate that the bacteria contribute to the life cycle of their host animals on two levels. First, the observed differences in gene expression of genes potentially relevant during host development indicate that Blochmannia has an important function mainly in the phase of pupation, since genes important for nitrogen recycling as well as for tyrosine biosynthesis are up-regulated at this stage. Second, the fostering experiments with antibiotic-treated workers suggest that the contribution of the bacteria also concerns the fitness of the host at the colony level, since the bacteria indirectly contribute to the maintenance of the brood development, probably by affecting the quality of the food provided by the workers to the larvae.
We thank Dagmar Beier for critically reading the manuscript and for discussions. We are grateful to Annett Endler and Jürgen Liebig for providing C. floridanus colonies and valuable comments on the biology of these ants.
This work was supported by the priority program of the DFG (SFB 567/C2).