C. fellah queens had been collected following a mating flight in Tel Aviv, Israel, in 2007. The colonies had been reared at 27°C below a 12-h/12-h gentle/darkish cycle, supplied water and sugar answer advert libitum, and fed weekly with flies and a synthetic ant meals [34]. For 1 yr previous to the experiment (C. fellah staff typically don’t stay longer than a yr [20]), all newly eclosed staff had been paint-marked weekly with a colour code indicating their week of delivery (Fig 1). C. fellah queens mate solely as soon as, so inside colonies, all staff are full sisters [35].

Every colony was created by randomly subsampling roughly 100 staff (roughly 20% of the employee inhabitants) from lab inventory colonies, together with the queen and roughly 20% of the brood. The colonies had been housed in double-chamber setups, with a nest field (170 × 123 mm) saved in fixed darkness linked by way of a plastic tube (inside diameter 19 mm) to a foraging area (170 × 123 mm) below a 12-h/12-h gentle/darkish cycle. Distinctive 1.4 mm² matrix barcodes from the ARTag library [36] had been fastened to the thorax of every ant utilizing SAUER pores and skin adhesive (Fig 1). These tags weigh a fraction of what ants can carry, and their presence doesn’t alter habits [37].

Colonies had been repeatedly video recorded for 7 days at 6 frames per second. The monitoring system saves video recordsdata and the place and orientation of every tag in every body. Full technical specs and supply code can be found at: https://github.com/formicidae-tracker.

Employees that had been detected in fewer frames than 2 commonplace deviations beneath the colony imply (5 or 6 staff per colony) had been excluded from the evaluation (“01-MergeData.R”). Every employee’s bodily setting was quantified by discretizing the sector into an equilateral grid of hexagonal cells utilizing the imply ant physique size as the interior cell diameter and by counting the variety of frames by which every employee was detected in every hexagon (“spatial constancy.ipynb”) [38]. To quantify every employee’s social setting, we inferred pairwise social interactions from the monitoring information as in [10]. Head and physique areas had been annotated for every ant, and interactions had been outlined as occurring when the top areas of two ants overlapped. The whole variety of pairwise interactions was used to create weighted social networks (Fig 2; “interplay community.ipynb”). This proximity-based definition of social interactions captures trophallaxis (roughly 16% of interactions), grooming (roughly 8% of interactions), and antennation occasions (roughly 35% of interactions), in addition to events when 2 people transfer previous each other or pause subsequent to at least one one other (roughly 42% of interactions). This excessive price of “false constructive” interactions has no impact on the information evaluation as a result of the frequency of pairwise false positives was nicely correlated with the frequency of actual pairwise interactions, and absolutely the numbers should not necessary as a result of interplay counts are successfully normalized inside colonies throughout evaluation. As an example this level, we artificially elevated the proportion of false constructive interactions by evaluating the pairwise interplay counts contemplating solely the overlap of head areas (blue in Fig 1B) and the interplay counts when contemplating additionally the overlap of physique areas (inexperienced in Fig 1B), which will increase the common variety of pairwise interactions from 96 to 260, and which massively will increase the false constructive price (e.g., many cases of body-to-body overlap). The R-squared worth between the pairwise interplay scores measured in these 2 methods was 0.812.

To quantitatively characterize particular person place within the social community (“social maturity”), we used the mushy group detection FacetNet (https://c4science.ch/supply/facet_unil) [20,39,40]. This strategy permits for overlapping social communities by outputting a steady quantity within the vary 0 to 1 denoting the extent to which a given node belongs to a given group (i.e., the place a given employee is positioned between the nurse and forager communities). For this evaluation, the variety of communities was fastened at two primarily based on earlier analyses of C. fellah social networks [20]. In step with this classification, and in keeping with earlier outcomes, there was a U-shaped distribution of social maturity scores, with most staff being deeply embedded in a single or the opposite group (Fig D in S1 Textual content; “02-Fundamental.R”).

These manually annotated behaviors weren’t normalized by detection frequency as a result of when a person’s identification was not detected within the chosen body, we adopted the person by the movies till the identification was resolved. Therefore, foraging was sure by 0 and 1, whereas all different behaviors had been sure by 0 and 80, although with completely different distributions. To present equal weight to all behaviors throughout dimensionality discount, every habits was normalized between 0 and 1 inside colonies (“02-Fundamental.R”). PCA of the normalized behavioral information yielded equally V-shaped plots throughout the 4 colonies independently, and when information had been pooled throughout colonies (Fig E in S1 Textual content; “02-Fundamental.R”). For the correlational analyses, together with differential expression evaluation, PC1 of behavioral area was calculated individually for staff from every colony. For the machine studying–primarily based evaluation PC1 of behavioral area was calculated utilizing people from all colonies collectively. The efficiency of the 6 annotated behaviors mapped onto the social networks in a constant method (Fig F in S1 Textual content; “02-Fundamental.R”).

The transcriptomic reads had been mapped to the C. fellah reference genome (BioProject: PRJNA901066) with STAR v2.7.8a and counted with FeatureCounts utilizing default parameters in any respect steps (mapping on the gene, not the transcript stage) [41,42] (“Mapping.sh”; “Counting.sh”). After mapping and counting, we obtained 38.7 ± 8.3 million (imply ± SD) reads per particular person.

Earlier than working the differential expression evaluation, we filtered out genes with <100 reads throughout all samples (i.e., 1,667 out of 14,664 genes). We used DESeq2 [43] to determine genes that had been differentially expressed by every of age, social maturity, and PC1 of the behavioral, microbiota, and bodily setting information, in addition to for every aforementioned variable when controlling for one another variable, for every of the 4 colonies individually (“03-GeneExpression.R”). In DESeq2, important differential expression was assessed with a Wald check, and the Benjamini and Hochberg technique was used to acquire a number of testing adjusted p-values. Genes had been thought-about as differentially expressed when the adjusted p-value was <0.05.

Employees had been floor sterilized by dipping in 95% ethanol, soaking for 1 minute in 5% bleach, then rinsing with sterilized water. The abdomens had been then eliminated and crushed in PowerBead tubes, and DNA was extracted utilizing the DNeasy PowerSoil package following the producer’s protocol (https://www.qiagen.com/de/assets/obtain.aspx?id=91cf8513-a8ec-4f45–921e-8938c3a5490c&lang=en). For every batch of DNA extractions, we additionally carried out clean DNA extractions by which no tissue was added, to manage for potential contaminants within the reagents. A mock group composed of 16S rRNA gene plasmids [10] was processed and sequenced alongside the experimental samples to verify for biases launched throughout PCR and sequencing. To characterize the microbiota, we amplified the V4 hypervariable area of the 16S rRNA gene following the Illumina 16S metagenomic sequencing preparation information with minor modifications (https://help.illumina.com/paperwork/documentation/chemistry_documentation/16s/16s-metagenomic-library–prep-guide-15044223-b.pdf) utilizing primers 515F-Nex (TCGTCGGCAGCGTCAGATGTGTATAAGAGACACCGCGGTAA) and 806R-Nex (GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGGACTACHVGGGTWTCTAAT), which include the adapter sequences for Nextera XT indexes and the primers for the V4 area of the 16S rRNA gene [44]. Briefly, we carried out PCR amplifications in a complete quantity of 25 μl, utilizing 2.5 μl template DNA, 12.5 μl of Invitrogen Platinum SuperFi DNA Polymerase Grasp Combine, 5 μl MilliQ water, and a couple of.5 μl of every primer (5 μM). PCR circumstances had been as follows: 98°C for 30 s adopted by 25 cycles of 98°C for 10 s, 55°C for 20 s, and 72°C for 20 s, and by a last extension step at 72°C for five min. Amplifications had been verified by 2% agarose gel electrophoresis. The PCR merchandise had been then purified with Clear NGS purification beads (CleanNA) in a 1:0.8 ratio of PCR product to beads and eluted in 27.5 μl Tris (10 mM, pH 8.5). A second PCR step was carried out to append dual-indexes to every pattern utilizing the Nextera XT index package (Illumina). Second-step PCR amplifications had been carried out in a complete quantity of 25 μl utilizing 2.5 μl of the merchandise from the primary PCR, 12.5 μl of Invitrogen Platinum SuperFi DNA Polymerase Grasp Combine, 5 μl MilliQ water, and a couple of.5 μl of every Nextera XT index primer. Thermal cycle circumstances had been as follows: a primary denaturation step at 95°C for 3 min adopted by 8 cycles at 95°C for 30 s, 55°C for 30 s, 72°C for 30 s, and a last extension step at 72°C for five min. We once more purified the PCR merchandise utilizing Clear NGS purification beads in a 1:1.12 ratio of PCR product to beads and eluted them in 27.5 μl Tris (10 mM, pH 8.5). The amplicon concentrations had been quantified by PicoGreen, and amplicons had been then pooled in equimolar focus except the unfavorable management and clean DNA extractions, which had been diluted 10×. These controls are meant to point out any background noise and contamination, so the identical sequencing depth is pointless. We verified that the ultimate pool was of the appropriate measurement utilizing a Fragment Analyzer (Superior Analytical) and carried out sequencing on an Illumina MiSeq sequencer on the Genomic Know-how Facility of the College of Lausanne, producing 2 × 250 bp reads.

We obtained a complete of 11,552,825 uncooked sequences, from the 289 belly samples, 4 unfavorable PCR controls, 4 mock group samples and 16 clean DNA extractions. Uncooked sequencing information high quality filtered with Trimmomatic [45] utilizing LEADING:3, TRAILING:3, SLIDINGWINDOW:4:15, and MINLEN:180. The standard-filtered information had been analyzed with the Divisive Amplicon Denoising Algorithm 2 (DADA2) pipeline (“dada2” package deal model 1.20.0 in R) [46]. All features had been run utilizing the beneficial parameters (https://benjjneb.github.io/dada2/tutorial.html) besides that we set randomize = TRUE and nbases = 3e8 on the learnErrors step, and pool = TRUE in the course of the pattern inference step. The SILVA database v.138 was used for taxonomy assignments of the recognized ASVs. We eliminated any ASV categorized as mitochondria, chloroplast, or Eukaryota (“phyloseq” package deal v1.36.0 [47], “subset taxa” perform). We then used each the “prevalence” and “frequency” strategies (technique = “both”) within the R package deal “decontam” v. 1.12.0 [48] to determine and take away contaminants launched throughout moist lab procedures, utilizing the unfavorable PCR controls and the clean samples as reference, which allowed us to determine 14 such ASVs (Renibacterium sp., Ralstonia sp., Microbacterium sp., Leifsonia sp., Cutibacterium sp., Enhydrobacter sp., Gordonia sp., 2 Sphingomonas ASVs, 3 Methylobacterium-Methylorubrum ASVs, a Comamonadaceae, and a Finegoldia). The ultimate information set consisted of 10,459,506 reads belonging to 79 ASVs (“Microbiota ABC.csv”; Fig G in S1 Textual content).

Within the analyses, we excluded Blochmannia, which is an obligate intracellular endosymbiont of Camponotus ants offering important amino acids and with a possible position in nitrogen recycling [49,50]. This bacterium was current in all people (“02-Fundamental.R”).

Euclidean distance was used to calculate pairwise in habits, mind gene expression, intestine microbiota profile (utilizing relative abundances of various ASVs), and bodily setting. To generate the structure for every layer within the multiplex community, we used the R package deal “multinet” [51]. Intralayer edges join nodes whose similarity/interplay frequency was above the higher quartile, whereas interlayer edges join like nodes (i.e., node i in layer a with node i in layer b). We used default values for all parameters aside from gravity (set to 1) and iterations (set to 1,000). The networks had been plotted utilizing python package deal “mnet.” To check the correlation between layers, every layer was decreased in complexity to be univariate. Social maturity was used for the social layer and age is de facto univariate. For the opposite 4 layers, we used PCAs (“02-Fundamental.R”; “04-InterlayerCorr.R”). We extracted PC1 of the microbiota, habits, and bodily environmental information. We extracted PC2 of the gene expression information as a result of all organic variables had been finest correlated with this principal element, whereas PC1 was finest correlated with extraction batches from the molecular lab (Fig H in S1 Textual content).