Nutritional upgrading for omnivorous carpenter ants by the endosymbiont Blochmannia

Feeding experiments with chemically defined diets

Six colony fragments from each of eight colonies were fed different diets over a 12-week period and monitored for numbers of pupae raised from a total of 45 eggs and 45 first instar larvae supplied to each group.

Production of pupae was pooled for each set of colony fragments fed the same diet (n = 8), because colony identity did not significantly affect the number of pupae raised (ANOVA: p = 0.199, F_7,40 = 1.488). Numbers of pupae raised per colony fragment differed significantly by diet (ANOVA: p < 0.001, F_5,42 = 13.338). A posthoc (Tukey-HSD) test revealed that worker groups fed diet A (with essential amino acids; AA) raised more pupae than did control groups (diet A mean ± SD: 40 ± 9.02, controls: 29.63 ± 15.17 pupae), albeit not significantly more (Figure 1). With essential AA present in the diet, treatment with the antibiotic Rifampicin to reduce endosymbiont numbers produced no significant effect (diet AR: 24.5 ± 8.8 pupae), just as did omission of essential AA from the diet (diet B: 30.13 ± 14.6 pupae). However, worker groups treated with antibiotics and kept on a diet lacking essential AA raised significantly fewer pupae than did all other groups (diet BR mean ± SD: 4 ± 5.45 pupae) except for workers fed sucrose solution only; the latter produced pupal numbers statistically indistinguishable from those of BR groups (diet S: 8.25 ± 8.66 pupae) (Figure 1).

Over the 12-week experimental period, worker mortality differed significantly between feeding groups (ANOVA: p < 0.001, F_5,42 = 8.459) (Figure 2). Significantly fewer individuals died in control and sugar-only groups than in groups treated with antibiotic (Tukey-HSD test; diet C: 20.38 ± 6.5 dead workers from 150 and diet S: 21.75 ± 11.18, versus diet AR: 63.13 ± 25.51, and diet BR: 57 ± 18.63 dead workers) or those whose diet lacked essential AA (diet B: 54.38 ± 24.67 dead workers). Worker groups fed diet A (with essential AA) showed intermediate levels of mortality (37 ± 18.63 dead workers) (Figure 2). Mortality was not source colony-specific (ANOVA: p = 0.304, F_7,40 = 1.21) and not significantly correlated with the ability of worker groups to raise brood (Pearson's Product Moment Correlation Coefficient r = -0.158, n = 48, p = 0.284). Thus, brood production per worker group depended solely on diet, irrespective of the number of workers dying per group during the experimental period.

Brood may be eaten by workers either when damaged or when food is too limited [29]. For each worker group, we added the number of pupae produced and removed by us to the number of larvae still present, and subtracted this from 90 to quantify the number of brood items lost during trials. After 12 weeks, groups differed significantly in number of brood items missing from the total of 90 items supplied during the experiment (ANOVA: p < 0.001, F_5,42 = 6.287). Missing brood were fewest in worker groups fed with diet B (without essential AA), intermediate in groups fed with the control diet, diet A (with essential AA), or sucrose (Tukey-HSD posthoc test), and greatest in groups fed diets AR and BR (artificial diet with and without essential AA and treated with antibiotic) (Figure 3).

While total production of pupae did not differ significantly between diets A, AR, B, and C, we compared rates of pupal production over time for these diet categories using a logistic trend analysis. Colonies producing fewer than 20 pupae over the experimental period (one control group and one group fed diet AR) were outliers and thus omitted from the analysis. Production rates were standardized by dividing the number of pupae produced for each point of time by the final number of pupae in each group (Figure 4), thus yielding values between 0 (no pupae yet at the beginning of the experiment) and 1 (average maximum number of pupae produced per diet). These standardized production rates were fitted to a logistic function with two parameters (intercept and slope). The 95% confidence intervals were overlapping for all intercepts (fitted to zero), but for the slope, the confidence interval for diet AR did not overlap with those of the remaining groups. The shallower slope for AR (95% confidence limits (10.9; 20.0), versus diet A (3.515; 6.92), diet B (3.778; 7.121), and diet C (4.57; 9.561)) indicates longer development times for groups fed diet AR. Worker groups on diets A and B were able to raise 50% of the pupae within 22 to 27 days, whereas control groups and groups on diet AR need approximately 31 days to raise 50% of the maximum number of pupae (Figure 4). However, in control groups, the first quarter of all pupae produced is reached faster than in groups fed with diet AR. All groups produced brood faster in the first 6 weeks of the experimental period than thereafter.

Reduction of Blochmannia by antibiotic treatment

Reduction of Blochmannia numbers in midgut bacteriocytes after treatment with antibiotics was monitored using two different methods: real-time quantitative PCR and Fluorescent in situ hybridization (FISH). As qRT-PCR may also detect residual DNA from dead bacteria, FISH was performed to evaluate the amount of living bacteria by targeting the ribosomal RNA. Both techniques revealed significant differences among diets in endosymbiont numbers after the 12-week experimental period (FISH, Figure 5; qRT-PCR: Figure 6). The unspecific eubacterial 16s rDNA probe and the Blochmannia-specific probe yielded congruent pictures of the bacteriocytes. Quantitative real-time PCR revealed that workers differed significantly in their content of Blochmannia measured as pg of the gene tufB encoding an elongation factor (ANOVA: Welch- test of square-root transformed data: p < 0.001, F_5,42 = 5.77). Comparisons among groups (Tamhane posthoc test) revealed that numbers of endosymbionts were significantly reduced under diet AR (median: 0.094 pg tufB/worker) compared to those for all other diets (median: for diet A: 0.695, diet B: 0.280, diet C: 0.366, sucrose: 0.339 pg tufB/worker) except BR (median: 0.216 pg tufB/worker), the other diet with antibiotics. Numbers of Blochmannia in BR feeding groups were lower than those in all groups on diets lacking antibiotics, but significantly so only compared to diet A that had the highest tufB content of all groups.

Feeding experiments with ¹⁵N-labelled urea

To investigate whether the urease of Blochmannia is functional and facilitates N-incorporation into the amino acid pool of the host, we fed ¹⁵N-urea to the ants. Analysis of ¹⁵N-enrichment of free amino acids in the hemolymph of workers by gas chromatography-mass spectrometry (GC-MS) revealed significant labelling of several amino acids that could be unequivocally identified (see Methods section for details). Except for the branched-chain, essential amino acids valine, leucine and isoleucine, all amino acids analysed (including the non-essential amino acids alanine, glycine, proline, aspartic acid and glutamic acid as well as the essential amino acids methionine and phenylalanine) showed significantly elevated levels of heavy isotope incorporation; i.e. the naturally occurring level of the heavy isotope of the respective amino acid (M+1) showed no overlap with the mean level and range of standard error in experimental groups (Table 1). Labelling of hemolymph amino acids depends both on the pool size of pre-formed, unlabelled amino acids and turnover as well as the ¹⁵N-incorporation rate from ¹⁵N-urea. Therefore, amino acid pools became differentially labelled with ¹⁵N. Increase in levels of ¹⁵N-labelled amino acids was stronger for non-essential amino acids (between 4 mol % for glycine to 9 mol % for alanine) than for the essential amino acids methionine (3 mol %) and phenylalanine (3 mol %).

Examination of gut microflora of C. floridanus by community fingerprinting

Bacterial community fingerprinting by TGGE revealed no gut microbiota other than Blochmannia in dissected guts. Additional bands that were detected in whole ant extracts in some instances were only found on thorax controls but never in gut preparations. Cultivation attempts from C. floridanus midgut contents yielded a low number (1–20 clones per plate) of Serratia marcescens (accession number DQ117670), a known facultative insect pathogen, in two out of three examined ant colonies. FISH analysis of C. floridanus midguts with Blochmannia-specific and universal 16S rDNA probes was congruent and did not reveal any secondary microbiota.

Camponotus floridanus: sampling sites and culture

Founding queens of Camponotus floridanus were collected (by AE) in Florida in June 2003 and 2004 near Tarpon Springs, Pinellas County (28°8'55''N, 82°45'29''W). Colonies were kept in plastic containers (20 × 20 × 10 cm) with plaster nests in a climate chamber at Wuerzburg University (constant temperature of 25°C, 12 h light per day), and were fed twice a week with cockroaches (Nauphoeta cinerea), Bhatkar agar [36] and honey water (1:1) ad libitum. Prior to their use in experiments, colonies had reached a size of 2000 to 10000 workers.

Feeding experiments (antibiotic treatment accompanied by chemically defined diet)

From each of eight large queenright colonies of C. floridanus, six groups of 150 randomly picked minor workers were transferred independently into separate artificial nests. Each of the six worker groups was fed three times per week, but on different diets:

• Control: Bhatkar agar, cockroaches and honey water ad libitum.

• A: a cube (0.5 cm sidelength) of chemically defined diet containing all trace metals, vitamins, growth factors and amino acids (AA) essential for the ants (for exact composition and preparation of the diet see [28].

• AR: same as A but with 100μl of a solution containing 2% of the antibiotic Rifampicin mixed into the food every other week.

• B: chemically defined diet like A except that AA essential for the host were omitted and compensated by higher amounts of non-essential AA, so that the total amount of AA in the diet was not changed.

• BR: chemically defined B-diet mixed with 100μl of a 2% Rifampicin-solution every other week.

• S: water containing sucrose ad libitum (20% w/v) only.

Each worker group was provided with 15 eggs and 15 larvae (1 mm length) at the beginning of the experiment, as well as after 4 and 8 weeks (45 eggs and 45 larvae in total). Over a period of 12 weeks the success of the workers in raising pupae from the early stage brood was monitored. Worker groups were checked three times a week for new pupae as well as for dead workers, and both were removed to prevent supplementation of the worker-diet through cannibalism. As dead workers may already have been disassembled or cannibalized by nestmates, they were counted by the number of head capsules found in the nest. Dead workers were not replaced, and thus the number of workers within each group declined over the experimental period. At the end of the experimental period, all brood members still present in the colony were counted in order to calculate the amount of brood lost during the experimental period. When food is limited or brood items are damaged, brood may be eaten by workers [29,31] and thus the nutritional value of these items would be recycled.

Real time qPCR

DNA was isolated from whole gasters (including guts and bacteriocytes) of two workers from each group of the six different feeding regimes (see Feeding experiments section). Whole ants were cooled on ice first, and then gasters were cut off with fine scissors. Gasters were homogenized after shock-freezing with liquid nitrogen in an Eppendorf Cap. DNA extraction was performed with a commercial kit (Gentra Systems Puregene^®, Minneapolis, MN, USA) according to the manufacturer's recommendations. The DNA pellet was resuspended in 50μl double distilled water. For each reaction of real time qRT-PCR, 100 ng total DNA were used as template, and a fragment of approx. 150 bp of tufB (see primer sequence in [19]) was quantified; this exemplary housekeeping gene of B. floridanus encodes an elongation factor. C(t) values of the samples were transformed to pg of tufB by using 5, 10, 40 and 100 pg of a purified tufB PCR product obtained with the same primers as a standard template in each run. For each sample, two independent qRT-PCRs were performed using a DNA Engine Opticon System (MJ Research, Waltham, MA, USA) and the ABsolute QPCR SYBR Green Mix (ABgene, Hamburg, Germany). All reactions were performed according to the manufacturers' instructions. The pg tufB measured by qRT-PCR were square-root transformed before the statistical analysis.

Fluorescent In Situ hybridisation (FISH)

To determine the success of the antibiotic treatment, bacteriocytes were visualized by FISH with oligonucleotide probes Eub338 [37]), targeting a conserved region of the eubacterial 16S rRNA, and with Bfl172 (5'-CCTATCTGGGTTCATCCAATGGCATAAGGC-3'), targeting a 16S rRNA region specific for B. floridanus and the closely related B. rufipes. Probes were labelled with the fluorescent dyes Cy3 or FITC at the 5' end by the manufacturer (Metabion, Planegg-Martinsried, Germany). Probe Bfl172 was designed using the software Primrose [38]. Specificity of the probes was checked with RDP II Probe Match [39].

Midguts of C. floridanus workers from control groups and groups fed with diet A (containing all essential AA) and treated with antibiotics (2% Rifampicin w/w mixed into the food) were dissected in sterile PBS (pH 7.2). Gut contents were removed prior to fixation on microscopic slides for 2 h in 4% (w/v) paraformaldehyde in PBS. Samples were dehydrated by subsequent 3-min washes in 50%, 70% and 100% ethanol. After air-drying, they were covered with 40μl of hybridisation buffer (35% formamide, 900 mM NaCl, 20 mM Tris/HCl pH 7.5, 0.2% SDS and 2 ng/μl of each fluorescently labelled oligonucleotide probe) and incubated for 2 h at 46°C in a humid chamber in the dark. After hybridisation, the slides were washed for 30 min in 70 mM NaCl, 20 mM Tris/HCl pH 7.5, 5 mM EDTA, 0.01% SDS at 48°C. Samples were rinsed in sterile distilled water, air-dried, mounted in 2 μl VectaShield^® (Vector Laboratories Inc., Burlingame, CA, USA) and analyzed with a Leica DMR microscope (Leica Microsystems, Wetzlar, Germany). Photographs were taken with a RT Slider digital camera (Diagnostic Instruments Inc., Sterling Heights, MI, USA) and the Meta Imaging Series 5.0 (Universal Imaging, Downington, PA, USA).

Feeding experiments with ¹⁵N-labelled urea

From four different colonies of C. floridanus, two groups of 30 workers each (10 minors and 20 majors) were isolated and placed into artificial nests. After a hunger period of 3 days, during which all groups had access to water ad libitum but were not fed, control groups were fed honey water (50% w/v) containing 1% urea (Roth), and experimental groups were fed with honey water (50% w/v) with 1% ¹⁵N-labelled urea (Sigma Aldrich, Taufkirchen, Germany). Workers were anesthetized using CO₂ 8 h after first access to the honey water. Each worker's gaster was removed to prevent contamination of the hemolymph sample with crop contents. The worker's head was then severed from the alitrunk and hemolymph was collected with a 2 μl microcapillary tube from the posterior opening of the head capsule, as well as from the anterior opening of the alitrunk. From each group of workers, approximately 6 μl of hemolymph were collected and frozen (-20°C) immediately after collection.

Amino acid analysis from ant hemolymph by gas chromatography-mass spectrometry (GC-MS)

Hemolymph (10 μl) was diluted with 50 μl of double distilled water, centrifuged (1000 g for 30 sec), and the supernatant was saved. The pellet was extracted twice with 50 μl of double distilled water, and the combined water phases (150 μl) were treated with 30 μl of methanol, 20 μl of pyridine and 30 μl of N-propyl chloroformate. After a 5-min incubation time, the mixture was extracted twice with 500 μl of chloroform. The organic phases were taken to dryness under a stream of nitrogen. The residue (N-propyloxy-carbonylamide, methyl ester derivatives of amino acids) was reconstituted in hexane for GC-MS analysis. Some of the amino acids could not be efficiently derivatised and therefore escaped detection. Amino acids that could not be measured were all amino acids with additional amide (Asn, Gln), hydroxy (Thr, Ser, Tyr), thiol (Cys) and amine (Arg, Lys) groups in the side chains as well as His and Trp.

GC-MS was performed using a Varian 3400 gas chromatograph interfaced to a Finnigan MAT quadrupole SSQ 700 mass spectrometer. Amino acid derivatives were analyzed on a 25 m, 0.25 μm film thickness Optima-5 column (Macherey and Nagel, Düren, Germany). The injector of the splitless mode was set at 300°C. The GC column temperature was programmed from 80°C to 300°C at 10°C/min. The MS source was set at 150 °C and the electron energy at 70 eV. Compounds were analysed in the positive chemical ionisation mode using isobutane as reactand gas.

Bacterial community fingerprint of C. floridanus gut-microflora

To assess the bacterial gut microbiota in C. floridanus the guts of workers from four different colonies were dissected in isolation buffer (35 mM TrisCl, pH 7.6, 25 mM KCl, 250 mM sucrose). Five mid- and hindguts, midgut contents and thoraxes respectively, were pooled for DNA extraction from each colony, followed by universal 16S rDNA PCR and temperature gradient gel electrophoresis (TGGE) as previously described [13].

For analysis of cultivable bacterial species guts from workers of three colonies were homogenized in 50 μl Luria Bertani (LB) medium, plated on LB agar plates and incubated overnight to several days at 30°C and 37°C.

Statistical analysis

Statistical analyses were performed either with Statistica 7.0 or with SPSS 12.0 (SPSS GmbH). The logistic trend analysis to compare development time between feeding regimes was performed with S-Plus (SAS Institute).