Social environment affects the transcriptomic response to bacteria in ant queens

Abstract Social insects have evolved enormous capacities to collectively build nests and defend their colonies against both predators and pathogens. The latter is achieved by a combination of individual immune responses and sophisticated collective behavioral and organizational disease defenses, that is, social immunity. We investigated how the presence or absence of these social defense lines affects individual‐level immunity in ant queens after bacterial infection. To this end, we injected queens of the ant Linepithema humile with a mix of gram+ and gram− bacteria or a control solution, reared them either with workers or alone and analyzed their gene expression patterns at 2, 4, 8, and 12 hr post‐injection, using RNA‐seq. This allowed us to test for the effect of bacterial infection, social context, as well as the interaction between the two over the course of infection and raising of an immune response. We found that social isolation per se affected queen gene expression for metabolism genes, but not for immune genes. When infected, queens reared with and without workers up‐regulated similar numbers of innate immune genes revealing activation of Toll and Imd signaling pathways and melanization. Interestingly, however, they mostly regulated different genes along the pathways and showed a different pattern of overall gene up‐regulation or down‐regulation. Hence, we can conclude that the absence of workers does not compromise the onset of an individual immune response by the queens, but that the social environment impacts the route of the individual innate immune responses.

behavior and physiological immune system. Yet, in order to keep infections at bay, additional colony-level defenses have evolved that consist of collectively performed hygiene behaviors and organizational defenses, together forming the "social immunity" of the colony (Cremer, Armitage, & Schmid-Hempel, 2007;Evans & Spivak, 2010).
Proximately, these social immunity behaviors occur in response to pathogen exposure, and it was shown that individual infection after bacterial injection in honeybees affects the behavior of their nestmates already 6 hr after injection so that the bacteria-injected individuals are subject to increased allogrooming and aggression (Richard, Holt, & Grozinger, 2012). This suggests that nestmates can sense immune response, and in honeybees and ants, this has recently been shown to be mediated by cuticular hydrocarbons (Hernández López, Riessberger-Gallé, Crailsheim, & Schuehly, 2017;Pull et al., 2018)-important cues for chemical communication in insects (Howard & Blomquist, 2005).
Ultimately, all social immunity measures help to keep the colony free from disease, and to-in particular-prevent disease spread to the most valuable colony members, the reproductive queens .
It has been suggested that these social immune measures may interfere with the evolution as well as expression of individual immunity (Barribeau et al., 2015;Evanset al., 2006;Viljakainen et al., 2009), possibly reducing the need for individual immune responses.
Most studies have focused on the analysis of the genome and immune components, where it was found that social insects have neither strongly reduced nor enlarged immune repertoires (Barribeau et al., 2015;Simola et al., 2013), and all major insect immune pathways being represented (Toll, Imd, JAK/STAT, JNK). Recent work has shown that social, colony-level pathogen defenses affect the functionality of individual-level immune responses. In ants and honeybees, exposure to resin, which has antibacterial properties and which these insects use as a nest building material, leads to decreased investment in physiological immune response (Borba, Klyczek, Mogen, & Spivak, 2015;Castella, Chapuisat, Moret, & Christe, 2008;Simone et al., 2009).
The physiological immune defenses in insects comprise cellular and humoral responses, the former including phagocytosis of small microorganisms and encapsulation of larger parasites and the latter composed of several signaling pathways that culminate in the production of antimicrobial peptides and other effector molecules (Ferrandon, Imler, Hetru, & Hoffmann, 2007). The core genes encoding for these immune system components are retained across several insect orders (Viljakainen, 2015). Moreover, the immune responses are interconnected with stress responses, which in insects have an immune-enhancing effect via stress hormones releasing energy for both stress and immune responses (Adamo, 2017). This interconnection may be particularly relevant for our study where we test the effect of worker presence or absence in the context of infection, as it is known that social isolation may induce stress and interfere with disease defense abilities in insects (Boulay, Quagebeur, Godzinska, & Lenoir, 1999;Kohlmeier, Holländer, & Meunier, 2016;Koto, Mersch, Hollis, & Keller, 2015).
While previous work has focused mostly on worker-worker interactions, we here test how ant queens, the most important individuals of the colony, modulate their individual immune defenses after bacterial infection when they have access to social immunity or not (presence or absence of workers). We control for the fact of social isolation by also determining gene expression patterns of uninfected queens when alone or in the presence of their workers.
We used queens of the Argentine ant Linepithema humile (Figure 1) that we injected with a combination of gram+ and gram− bacteria or sterile saline solution and then kept in either isolation or with workers. After injection, changes in gene expression patterns were analyzed at four time points using RNA-seq: shortly after injection (2 hr), during the development of the immune response (4 and 8 hr post-injection), and when the immune response was | 11033 VILJAKAINEN Et AL. and tested for an interaction between the two, in particular if the response to bacterial infection differed between the two rearing conditions.

| Samples
Workers and queens of the Argentine ant L. humile were collected from the European main supercolony in Castell d'Aro, Spain, in April 2011 and kept in artificial nests in climate chambers (Sanyo) set to 27°C for 14 hr of light and 21°C for 10 hr of dark. The ants were fed with honey and cockroaches three times per week. Approximately 3 weeks prior to the experiments, small sub-colonies originating from two stock colonies and each consisting of a single queen and 10 workers were placed into petri dishes (diameter 9 cm) with a plastered ground and fed with 10% sugar water.

| Bacteria used for infections
We used the gram-positive bacterium Staphylococcus aureus and the gram-negative bacterium Serratia marcescens in combination for infecting the queens with the aim to induce gene expression of both Toll and Imd innate immune signaling pathways, since in Drosophila, gram-positive bacteria are known to induce mainly the Toll pathway and gram-negative bacteria the Imd pathway (Ferrandon et al., 2007). The bacteria were grown overnight in LB medium (Merck).
The bacterial suspension was centrifuged, and the pellet was washed three times in sterile saline solution (hereafter called Ringer) prepared following the protocol described in Aubert & Richard (2008).
The pellet from the final wash was suspended in Ringer. For the injections, bacterial suspensions were diluted, bacterial cells counted using Neubauer counting chamber, and S. marcescens and S. aureus dilutions mixed to get a solution representing both bacterial species in equal quantity.

| Injections and social environment
Linepithema humile queens were studied for effects on genomewide expression patterns at four time points (2, 4, 8, and 12 hr) after bacterial versus control injections in the presence or absence of five workers in a full factorial design. All the injections were made at the same time of day (in the morning) within a time window of 3 hr. Sample information is summarized in Table 1. The queens were randomly assigned for either bacterial or Ringer injection and were first transferred to small petri dishes on ice to cold-immobilize them for injection. Microinjections were performed using Picoliter Injector PLI-100 Plus (Harvard Apparatus) set at 10 psi for 1 s using spiked glass needles with inner diameter of 11.9 μm (Biomedical Instruments), resulting in an injection volume of about 65 nl. This volume was injected twice between the second and third tergite of the abdomen, containing approx. 1,300 bacterial cells (50:50 mix of S. marcescens and S. aureus). The controls were injected twice with 65 nl of sterile Ringer. After injection, the queens were transferred back to their original petri dish nests and kept together with five workers (social environment, the queens referred to as "social queens" hereafter) or reared alone by removing the workers (the queens referred to as "isolated queens" hereafter) at constant room temperature (22°C) and with 10% sugar water ad libitum. Each treatment at each time point was repeated three times. At 2,4,8,, the ants were frozen in liquid nitrogen and kept in −80°C freezer until RNA extraction. The wholebody samples were disrupted and homogenized in TissueLyser II (Qiagen) using stainless steel beads (5 mm diameter). Total RNA was extracted using RNeasy Micro Kit (Qiagen) following the protocol provided with the kit and including DNA removal using RNase-free DNase I. RNA was quantified using Agilent 2100 Bioanalyzer, and the samples were sent to BGI Tech Solutions (Hong Kong) for library preparation (Illumina TruSeq RNA Sample Prep Kit) and mRNA sequencing (100 bp paired-end reads) with Illumina HiSeq2000.

| Bioinformatic analyses
The filtering of raw sequence data was performed by BGI and included adapter removal, removal of reads with more than 10% of undetermined bases, and removal of reads with more than 50% of low quality bases (Q < 10). Quality controlled clean data obtained from BGI were used for further analyses. The clean reads were mapped to the L. humile reference genome (GCF_000217595.1) using STAR v.2.4.1b (Dobin & Gingeras, 2015 Benjamini and Hochberg (1995). In this study, we used a false discovery rate (FDR) <10%.
Insects, including L. humile, are known to harbor RNA viruses (Gruber et al., 2017;Shi et al., 2016) which may have an effect on host gene expression (Doublet et al., 2017;Gerth & Hurst, 2017 faa" (accessed 7 January 2018) using BLASTX 2.6.0+ with an evalue threshold of 10 −4 . Contigs that matched insect viruses and that had a query coverage of at least 400 amino acids were used in the following steps. The unmapped reads from each sample were mapped against the selected blast-annotated virus contigs using default settings in BWA-MEM v.0.7.17 , and the mapped reads were counted using samtools v1.4    Figure 2).

| Viral load of queens
Eight RNA viruses were identified in the Trinity-assembled contigs of reads that could not be mapped to the L. humile genome (Viljakainen, Holmberg, Abril, & Jurvansuu, 2018 a priori viral load of the ants, which ranged from low to high levels. Importantly, we found that the gene expression profiles of the ants were affected by viral load (Figure 3), so that we controlled for viral load in the analysis of differential gene expression.

| Effect of social context
We first investigated the effect of social isolation per se by analyzing differentially expressed genes in Ringer-injected queens that were either  (Figure 3, Table 3, and Appendix: Table A1). The total number of DEGs across all time points was 134 with 82 up-regulated and 52 down-regulated genes. GO enrichment analysis showed enrichment of biological processes "single-organism metabolic process," "carbohydrate phosphorylation," and "cellular glucose homeostasis" in the up-regulated genes at 12 hpi (Appendix: Table   A2). We hence found that social context affected queen energy metabolism, but had no effect on immune gene expression.

| Effect of bacterial infection depending on social context
We found that the effect of bacterial infection depended strongly on the social context the queens were reared at, even if the overall  Table 4).
The majority of the DEGs were not directly related to immune response, the core immune genes representing only 5% and 9% of the regulated genes in social and isolated queens, respectively. To get insight on the affected biological processes and molecular functions, GO enrichment analysis was carried out for treatment contrasts with F I G U R E 3 Differentially expressed genes induced by (a) isolation in Ringer-injected control queens without viral load as a cofactor, (b) isolation in Ringer-injected control queens and viral load taken into account as a cofactor, (c) bacterial injection in queens at social environment without viral load as a cofactor, (d) bacterial injection in queens at social environment and viral load taken into account as a cofactor, (e) bacterial injection in isolated queens without viral load as a cofactor, (f) bacterial injection in isolated queens and viral load taken into account as a cofactor at least 10 DEGs. The up-regulated genes of social queens at 4 hpi showed enrichment of proteolysis and serine-type endopeptidase inhibitor activity (Appendix : Table A2). In the isolated queens, downregulated genes at 2 hpi showed enrichment of proteolysis, molybdopterin cofactor biosynthetic process, and serine protease inhibitor activity (Appendix: Table A2). The genes categorized as being involved in proteolysis and having serine protease inhibitor activity were largely the same genes as in the social queens. Oxidation-reduction process was enriched in the down-regulated genes of isolated queens at 12 hpi including gene encoding for phenoloxidase (LOC105668871) and several cytochrome P450 protein-coding genes.

| Activation of Toll signaling pathway
The expression of immune genes indicated activation of the signaling pathway Toll (Table 4). In social queens, five Toll pathway genes were up-regulated. These included beta-1,3-glucan-binding protein ( Two genes encoding Toll receptors were up-regulated, LOC105678817 and LOC105678648, both similar to Drosophila Toll (NP_524518, 29% identity, 70% coverage and 30% identity, 48% coverage, respectively).
In the isolated queens, seven Toll pathway genes were up-regulated and two down-regulated (Table 4). Three genes encoding for serine proteases, all involved in the activation of the Toll pathway in a similar way described above for modSP and snake, were up-regulated: two limulus clotting factor C-like (LOC105673362 and LOC105673363) similar to Drosophila modSP (NP_536776, 27% identity, 97% coverage and 31% identity, 94% coverage, respectively) and additionally, serine protease gd-like (LOC105671866) similar to gastrulation-defective in Drosophila (NP_001303552, 29% identity, 84% coverage) that presumably activates serine protease snake (Rose et al., 2003). One gene encoding Toll-like protein (LOC105678784) similar to Drosophila Toll (NP_524518, 36% identity, 90% coverage), which is a transmembrane receptor, was up-regulated, and two Toll-like protein-coding genes were down-regulated (LOC105678817 and LOC105678912).
Two genes downstream of Toll receptor were up-regulated, cactus-1a (LOC105678482) and cactus-1b (LOC105678483), both similar to Drosophila cactus (NP_476943, 43% identity, 50% coverage and 38% identity, 76% coverage, respectively) which is an inhibitor of NF-κB transcription factor Dorsal that positively regulates the transcription of antimicrobial peptides (AMPs; Ferrandon et al., 2007). Notably, our analysis only revealed two Toll pathway genes overlapping between the social and isolated queens, limulus clotting factor C-like (LOC105673362) and protein Toll-like (LOC105678817), of which the latter was significantly up-regulated in the social queens while downregulated in the isolated queens.

| Activation of Imd signaling pathway
Genes along the Imd pathway were not as widely represented among the DEGs as the Toll pathway genes (Table 4). Down-regulated in both social and isolated queens was a gene encoding for peptidoglycan-recognition protein SC2 (PGRP-SC2, LOC105675773), which is a negative regulator of the Imd pathway (Bischoff et al., 2006). In addition, isolated queens showed up-regulation of a gene encoding for the NF-κB-like transcription factor Relish (LOC105668729) (Ferrandon et al., 2007) and for uncharacterized protein (LOC105678813) similar to Drosophila poor Imd response upon knock-in (NP_001286686, 39% identity, 24% coverage), which, again, is a negative regulator of the Imd pathway (Kleino et al., 2008).

| Melanization
Melanization-an active mechanism to encapsulate pathogens within the host-was induced in both social and isolated queens indicated by a number of up-regulated genes (

| Interaction effect of social isolation and bacterial injection
An interaction analysis of social isolation and bacterial injection at all the four time points showed regulation of 20 genes across all time points in the bacteria-injected isolated queens (Tables 3 and 5). These included down-regulation at 4 hpi of the Toll receptor activator-gene spätzle (LOC105678357), up-regulation at 12 hpi of the Imd pathway signaling gene imd (LOC105672003) and down-regulation, also at 12 hpi, of hemocyte protein-glutamine gamma-glutamyl transferaselike (LOC105670674) similar to transglutaminase (NP_609174, 37% identity, 90% coverage) in Drosophila that inhibits the Imd pathway transcription factor Relish (Maki, Shibata, & Kawabata, 2017).
Hence, in both presence and absence of workers, all important immune defense pathways were triggered in the queens, with the highest number in the Toll pathway.

| D ISCUSS I ON
In this study, we tested how the presence or absence of workers affects ant queen immune response after bacterial infection. When testing for the effect of our experimentally induced bacterial infections, we found that existing viral load had an effect on differential gene expression analyses and that virus presence and load should be taken into account in these types of analysis, as previously reported (Gerth & Hurst, 2017). To further control for the effect of social worker presence or absence per se in the absence of an infection, we first analyzed differential gene expression between Ringer-injected TA B L E 5 Differentially expressed genes in bacteria-injected isolated queens in interaction analysis of social environment and bacterial treatment with false discovery rate <10%. T2, T4, T8, and T12 indicate the post-injection time points queens reared in the social environment or alone. Despite our sample size being large enough to detect significant effects of rearing on queen metabolism, we could not detect any general effect on queen immune gene expression.
We then tested whether worker presence or absence interfered with the queens' individual immune response to bacterial infection over the course of infection. Overall, we found that injection of bacteria, over the four time points studied, affected the expression of similar numbers of genes in both social and isolated queens but interestingly, the social queens up-regulated the majority of the genes, whereas in the isolated queens, down-regulation was prevailing. This general down-regulation might be a consequence of social isolation, which has been shown to affect life-history traits by reducing longevity in workers of the ant Camponotus fellah (Boulay et al., 1999;Koto et al., 2015), yet did not compromise the innate immune response of bumblebees after pathogen challenge (Richter, Helbing, Erler, & Lattorff, 2012). In the group-living earwig Forficula auricularia, rearing individuals alone also lead to a transiently increased susceptibility after pathogen exposure shortly after isolation, yet an indistinguishable survival of individuals living isolated or in groups for longer periods (Kohlmeier et al., 2016), hence the effects of social isolation may be plastic, both varying over time and across species.
The isolated queens regulated slightly, but not significantly, higher number of the core immune genes than social queens (23 vs. 14, χ 2 = 3.38, df = 1, p = 0.07) and the immune gene expression in the isolated queens did not show the overall pattern of down-regulation observed in all of their DEGs: 65% of immune genes were up-regulated as opposed to 37% of all DEGs (χ 2 = 8.11, df = 1, p = 0.004). Therefore, even though gene expression in the isolated queens showed a trend of down-regulation, the activation of cellular and humoral immune cascades was comparable to the social queens.
An interesting observation was two enriched GO terms, serine protease inhibitor activity and proteolysis, both categories including approximately 20 genes, which were up-regulated at 4 hpi in the bacteria-injected social queens and down-regulated at 2 hpi in the bacteria-injected isolated queens. All except one of the genes in the serine protease inhibitor-category were long noncoding RNAs (lncRNA) which Blast2Go annotation found to contain a protease inhibitor domain suggesting they might regulate serine protease inhibitors (serpins). In insects, serpins are known to be involved in the regulation of immune signaling cascades, phagocytosis, and digestion (Gubb, Sanz-Parra, Barcena, Troxler, & Fullaondo, 2010), and the expression of serpin-related lncRNAs could be involved in the fine-tuning of various arms of immune response. It is striking that the gene expression patterns of these genes showed opposite directions in the two rearing conditions, highlighting the strong effect of social environment on the general response to bacterial injection.
As a conclusion, this study shows that ant queens were equally able to activate innate immune signaling cascades after bacterial injection when kept together with workers or when reared alone.
This reveals that pathogen-injected queens raise an induced immune response even in the presence of rearing workers, yet that worker presence interferes with which exact set of genes is regulated.
Hence, we could show that individual queen responses are not compromised, but modulated by their social context.