Reproductive activation in honeybee (Apis mellifera) workers protects against abiotic and biotic stress

Abstract

Social insect reproductives exhibit exceptional longevity instead of the classic trade-off between somatic maintenance and reproduction. Even normally sterile workers experience a significant increase in life expectancy when they assume a reproductive role. The mechanisms that enable the positive relation between the antagonistic demands of reproduction and somatic maintenance are unclear. To isolate the effect of reproductive activation, honeybee workers were induced to activate their ovaries. These reproductively activated workers were compared to controls for survival and gene expression patterns after exposure to Israeli Acute Paralysis Virus or the oxidative stressor paraquat. Reproductive activation increased survival, indicating better immunity and oxidative stress resistance. After qPCR analysis confirmed our experimental treatments at the physiological level, whole transcriptome analysis revealed that paraquat treatment significantly changed the expression of 1277 genes in the control workers but only two genes in reproductively activated workers, indicating that reproductive activation preemptively protects against oxidative stress. Significant overlap between genes that were upregulated by reproductive activation and in response to paraquat included prominent members of signalling pathways and anti-oxidants known to affect ageing. Thus, while our results confirm a central role of vitellogenin, they also point to other mechanisms to explain the molecular basis of the lack of a cost of reproduction and the exceptional longevity of social insect reproductives. Thus, socially induced reproductive activation preemptively protects honeybee workers against stressors, explaining their longevity.

This article is part of the theme issue ‘Ageing and sociality: why, when and how does sociality change ageing patterns?'

1. Background

A pronounced life-history trade-off between reproduction and longevity characterizes most organisms [1,2], which can readily be demonstrated in artificial selection experiments [3]. Major hormone and intra-cellular signalling pathways govern the trade-off and thus involve several ‘ageing' genes that can extend lifespan when mutated [4,5]. These regulatory switches between alternative ‘survival' and ‘growth-and-reproduction' life-history syndromes coordinate numerous physiological mechanisms, including stress-responses, cellular homeostasis and metabolism, immunity and growth. Ageing is intimately linked to both oxidative stress [6] and immunoscenescence [7]. Specific trade-offs thus exist between reproduction and immunity [8] and reproduction and stress resistance [9,10] that may underlie the general trade-off between reproduction and longevity.

However, in social insects, the trade-off between reproduction and longevity does not exist [11]. Typically, the reproductive castes of social ant, wasp, termite and bee species outlive their non-reproductive kin by an order of magnitude and exhibit exceptional longevity relative to solitary insect species, with lifespans of several decades [12,13]. Many social insect queens exhibit exceptional longevity concomitant with exceptional reproductive output, reversing the traditional negative relation between reproduction and survival [14,15]. The mechanisms that cause the fundamentally reversed relationship are still unclear [16]. The endocrine control of reproduction in social insects may have been reorganized, removing the antagonistic switch between reproduction and self-maintenance [17]. For example, the egg-yolk protein vitellogenin pleiotropically affects oxidative stress resistance, immunity and life history in honeybees [18–20]. Honeybee queens may be protected from oxidative stress and immune challenges [15,21]. Alternatively, queens may upregulate immune and anti-oxidant genes despite their reproductive role [22]. Ultimately, social resource transfer may allow the reproductives of social insects to evade the cost of reproduction [23,24]. Their intrinsic survival advantage even depends on reproductive activation [25] and complements the effects of their protected environment and social facilitation of longevity [15].

The life-extending effect of reproductive activation in social insects is further supported by comparisons of non-reproductive and reproductive workers. In many species, the worker caste is functionally sterile but workers can activate their ovaries when not inhibited by the presence of queens and/or brood [26]. Reproductive workers significantly outlive their non-reproductive nest-mates [27,28]. This longevity advantage may be owing to the fact that the reproductive workers represent a non-random sample of the general worker population [13] or be related directly to reproductive activation and hence the differences between queen and normal worker physiology. Reproductive activation may be necessary for the longevity of social insect queens [25], and reproductively activated workers thus represent a good model to investigate the proximate factors of the extended lifespan of social insect queens without the confounding effects of alternative developmental pathways and other differences between castes [29]. However, such studies have so far only yielded limited mechanistic insights [30].

Here, we investigate stress resistance as a specific mechanism that may underlie the lifespan extension in reproductive social insects and ageing plasticity in general. We selected inoculation with Israeli Acute Paralysis virus (IAPV) and injection with the oxidative stressor paraquat as representative biotic and abiotic stressors. IAPV was chosen for the following reasons: (i) It is practically relevant as it has been linked to collapses of honeybee colonies [31]. (ii) It leads to acute, lethal infections of individual honeybees, facilitating measurements of resistance. (iii) IAPV is not as widespread as other viruses, which allows for a clear distinction between experimentally inoculated and control bees. The herbicide paraquat was chosen, even though it is of less practical relevance, because it is commonly used to induce oxidative stress experimentally in a wide variety of organisms [32] and established protocols for honeybees exist [20]. We compare the survival of reproductively activated and control workers of the European honeybee (Apis mellifera L) after IAPV and paraquat exposure and characterize gene expression patterns that accompany the observed survival differences to test the hypotheses that reproductive activation enhances the survival of stressors and that the enhanced survival is owing to defense mechanisms that are induced upon reproductive activation.

2. Methods

(a). Colony set-up and sample collection

Experimental colonies were established in May 2017 by shaking 600 honeybees of mixed ages and colony origins from the University of North Carolina Greensboro (UNCG) apiary into three-frame mating nuc hives that were maintained until the following August. Four queenright colonies were each headed by an unrelated queen of unknown age. Four queenless colonies were established with only workers and some capped brood and were replenished with small cohorts of newly emerged workers regularly to compensate for worker mortality and keep colony size and age structure similar among treatments. All colonies were monitored for visible signs of parasites and pathogens, and powdered sugar treatments and manual collections were applied to limit Varroa mites and small hive beetles, respectively. Adequate food stores were maintained in each colony throughout the experiment through supplemental feeding.

Frames with ready-to-emerge workers were collected from an array of stock colonies from the UNCG apiary. These frames were kept in an emergence incubator at 33°C and 50% relative humidity overnight. Newly emerged workers were carefully brushed from the frames and marked with cohort-specific enamel paint (Testors, Vernon Hills, IL) on their thorax. Multiple cohorts were introduced and the IAPV study was performed in June/July, while the paraquat study was completed in July/August. For each study, one cohort that survived best to permit adequate sampling at 15 and 25 days of age was selected. These two sampling ages were chosen to represent typical in-hive (pre-foraging) and outside (foraging) life history stages that differ drastically in vitellogenin [33] and mortality [34]. To assess the effect of social condition on worker physiology, ten random, apparently healthy workers from each of the four experimental groups (2 social conditions × 2 ages) were collected, immediately stored in microcentrifuge tubes with 400 µl of RNAlater™, and transferred to the laboratory for storage at −80°C until further analysis. About 120 remaining workers (free of any visible disease symptoms) of each experimental group were transferred to the laboratory and randomly assigned to be stressed or receive a sham control treatment.

(b). Stress application and survival analysis

Individual workers were briefly anesthetized on ice before treatment application. For the IAPV study, the dorsal thoracic hairs of the workers were shaved off, followed by an application of either 2 µl of water (sham control) or 2 µl of an IAPV solution containing 5.2 × 10⁷ copies [35]. For the paraquat study, workers were injected either with 1 µl of PBS solution containing paraquat (Sigma-Aldrich, St. Louis, MO) at a concentration of 150 µg g⁻¹ [20] or with 1 µl of PBS as vehicle control. Individuals that did not recover from anesthetization after treatment were excluded from the study.

Each treatment and control group was split into three replicates containing 10 to 15 individuals that were housed together in a sterile plastic cup. Survival was monitored every 24 h for 5 days while these groups were maintained under optimal incubator conditions (33°C, and 50% humidity) with ad lib water and sucrose candy. Pollen or other protein sources were not offered because the potential detrimental effects (introduction of pathogens or toxins, and risk of unsanitary conditions) were considered to outweigh nutritional benefits on immunity and stress resistance [36,37] during the short-term acute survival studies. Dead bees were removed from the cages daily and mortality was monitored. Survival rates were compared for the paraquat and IAPV stress experiments separately with Cox proportional hazard regression, followed by log-rank tests in SPSS v. 21 (IBM). A subsample of living workers (n = 10 per age, reproductive condition and treatment group) was collected from each group across cups after 24 h and 120 h for the paraquat study and after 24 h for the IAPV study (high mortality did not allow for a 120 h collection) for a total of 280 post-treatment samples that were individually frozen and later combined with the 40 pre-treatment samples for subsequent ovary dissections and molecular analyses (Experimental design: figure 1).

Figure 1.

Experimental design. Workers from queenright (QR) and queenless (QL) social conditions were sampled at 15 and 25 days of age. Before treatment applications, samples were taken from each group (Pre-Treat). Remaining workers were subjected to either stress treatment (para = paraquat injection, IAPV = Israeli Acute Paralysis Virus inoculation) or the corresponding control treatment (PBS = PBS injection, Shav = Thorax shaving). These workers were monitored for survival and live samples were collected after 24 hours and 120 hours. The samples were dissected to test ovary status (symbolized by an ‘O' at the top right corner of each sample box), studied for the expression of several target genes with qPCR (symbolized by red text) and subjected to Taq-seq for transcriptional profiling (thick-outlined boxes). For each sampling group (box), 10 individual workers were collected. (Online version in colour.)

(c). Ovary studies

The ovaries of the pre-treatment and all 24 h post-treatment workers were assessed, while the 120 h samples were omitted owing to concerns that stress treatment may impact ovary status [38]. Samples that had initially been frozen without RNAlater (Thermo Scientific, Wilmington, DE) were submerged in 1 ml of RNAlater ICE™ (Thermo Scientific, Wilmington, DE) for more than 24 h before dissection. The abdomen of each bee was separated from the remainder of the body and dissected in a chilled RNAlater ICE^TM solution according to previous established procedures [39]. Left and right ovaries were independently scored on a four-point scale according to previous studies [39]. The average ovary development of workers was calculated between the left and right ovaries and compared between the queenright and queenless conditions separately at 15 and 25 days of age with two independent Mann–Whitney U tests.

(d). RNA extraction

RNA was extracted from abdominal tissues of control and reproductively activated workers from the paraquat and IAPV experiments (figure 1). Ten workers of each group before treatment, 24 h after treatment (paraquat injection or IAPV inoculation) and sham treatment (PBS injection or shaving), and 120 h after paraquat injection and PBS injection were analysed using a standard Trizol™ (Life Technologies, Carlsbad, CA, USA) protocol. Briefly, the abdomens without their ovaries (which had been separated to score the ovary activation of individuals) were placed on dry ice and then crushed to powder with a disposable pestle. The pulverized tissues were mixed with 1 ml of Trizol™. Chloroform (200 µl) was added and mixed with the sample before centrifugation for 15 min at 12 000 rcf. The supernatant was isolated and RNA was precipitated with 0.5 ml isopropanol on ice for 15 min, followed by centrifugation for 10 min at 12 000 rcf. The resulting RNA pellet was washed with 1 ml of 75% molecular grade ethanol and air dried. The dried RNA was resuspended in 100 µl of molecular grade water and stored at −80°C for future processing.

(e). Quantitative real-time polymerase chain reaction

After RNA quantification with a Nanodrop 1000™ spectrophotometer (Thermo Scientific, Wilmington, DE), samples were treated with 1 unit of DNAse (Thermo Scientific, Wilmington, DE) and diluted with molecular grade water to a concentration of 20 ng µl⁻¹. The diluted RNA was used to synthesize cDNA using the SensiFAST™ cDNA Synthesis kit (Bioline, Taunton, MA) according to the manufacturer's recommendations. The reactions, using a mixture of random hexamer and oligo dT primers, were carried out in a thermocycler (Mastercycler, Eppendorf, Germany) with the following program: 25°C for 10 min, 42°C for 15 min, and 85°C for 5 min.

Quantitative real-time polymerase chain reaction (qPCR) assays were carried out on a StepOnePlusTM cycler (Applied Biosystems, Foster City, CA) with SensiFAST SYBR Green™ kits (Bioline, Taunton, MA), modified slightly from previously established protocols [35]. The following targets were selected: Vitellogenin, the immune genes Dicer-like, Defensin, Apidaecin and Argonaute-2, the oxidative stress defense genes Glutathione-s-transferase D1, CuZn Superoxide dismutase and Catalase, and the three viruses IAPV, DWV-A and DWV-B (electronic supplementary material, file S1). The total reaction volume was 18 µl with a final primer concentration of 0.4 µM and 2 ul of 1 : 10 diluted cDNA. Two negative controls (no target and no reverse transcriptase) were included for each gene. After an initial denaturation step of 3 min at 95°C, 40 cycles of 95°C for 20 s, 60°C for 30 s and 72°C for 1 min were performed with quantification of fluorescence during the 72°C step [31]. Cycle threshold (C_T) values were determined with a single gene-specific fluorescence threshold across all reactions, averaging the C_T of two technical replicates. Gene expression was quantified as ΔC_T sensu [40] by subtracting the mean C_T of two reference genes, actin and RPS5 [41] from the C_T of the target gene. Glutathione-S-transferase did not amplify in any samples and was therefore not analysed. Levels of IAPV, DWV-A and DWV-B were determined by absolute quantification [41]. Differences of ΔC_T of genes and absolute levels of viruses were evaluated with general linear models.

(f). Transcriptome analysis with Tag-Seq

Aliquots of the extracted total RNA from the 25-day-old workers in the paraquat experiment (QR no treatment: n = 9, QR post PBS injection: n = 9, QR paraquat injection: n = 9, QL no treatment: n = 6, QL PBS injection: n = 7 and QL Paraquat: n = 6) were sent to the Genomic sequencing and analysis facility (GSAF) of the University of Texas at Austin for library preparation and sequencing, following a modified version of the original Tag-Seq protocol [42,43]. After library preparation and indexing for multiplex sequencing, samples were sequenced in three lanes of the HiSeq 4000 Illumina platform. Tag-Seq reduces the transcriptome complexity by restricting the library to the mRNA sequences near the 3′ end poly-A tail. It therefore affords higher multiplexing and sequencing depths but cannot detect non-polyadenylated RNA or alternative splicing.

Sequencing data were processed on the Longleaf cluster through UNC Chapel Hill. Custom perl and bash scripts used for previous Tag-Seq projects (https://github.com/z0on/tag-based_RNAseq; [42]) were implemented in this environment. Briefly, reads were concatenated across lanes, filtered for quality for a minimum q-score of 33 in fastx_clipper, mapped to the honeybee transcriptome (Amel_4.5), and counts were generated per gene in SAMtools [44]. Differential gene expression analyses were carried out in in DESeq2 in the R statistical environment [45]. Low-expression genes (those with fewer than three counts in more than 90% of samples) were removed from the dataset, leaving 6710 expressed genes. The gene counts were normalized and log-transformed using a regularized log transform with the command rlog() in DESeq2 to determine the significance of differentially expressed genes (DEGs) based on an adjusted p < 0.05. Three comparisons of experimental groups were performed: (i) untreated workers from queenless or queenright colonies to test the effect of reproductive activation; (ii) queenright workers injected with paraquat or PBS to test the effect of oxidative stress in normal control workers and (iii) queenless workers injected with paraquat or PBS to test the effect of oxidative stress in reproductively active workers. Directional overlap analyses between the DEG lists were performed with hypergeometric tests. Up- and downregulated DEGs identified from the queenless versus queenright workers (i) were compared with DEGs in response to paraquat in queenright workers (ii) and with DEGs identified by Galbraith et al. [46] in response to IAPV infection.

Functional enrichment analyses of gene ontology (GO) terms were performed on the expressed genes in the three experimental comparisons. To include all data irrespective of significance threshold, the GO_MWU package in R was used (https://github.com/z0on/GO_MWU), based on the signed p-values of the DESeq2 analysis and GO terminology from one-to-one orthologues found in Drosophila melanogaster through HymenopteraMine [47]. For each GO term, a rank-sum test of the associated genes was performed to determine whether their rank distribution was significantly skewed toward up- or downregulation. For significant GO terms, the number of DEGs among the total number of genes in each GO term was also computed and listed [48].

3. Results

(a). Social manipulation induces reproductive activation in workers

Removing the queen and most of the brood from our experimental colonies to induce worker reproductive activation increased ovary activation scores (figure 2). While the effect was not significant after 15 days (Mann–Whitney U = 1,167.5, n = 100, p = 0.460), the 25-day-old workers exhibited significantly higher ovary activation scores (U = 325.5, n = 100, p < 0.001) under queenless than under queenright conditions.

Figure 2.

Worker ovary activation. Removal of the queen and most of the brood as our queenless treatment compared to the unmanipulated queenright treatment resulted in increased ovary activation scores. The average activation scores of the right and left ovary of individual workers were not significantly different between queenless and queenright colony conditions after 15 days, but 25-day-old workers in queenless conditions had significantly more activated ovaries than their counterparts in queenright conditions. Median is shown ± quartiles, range and outliers. (Online version in colour.)

(b). Social manipulation increases survival of stressors

Reproductively activated and sterile control workers were subjected to paraquat and IAPV treatments paired with corresponding controls to test the prediction that reproductive activation increases survival after stress exposure. In support of this prediction, the effect of injecting paraquat compared to PBS was dependent on reproductive activation: paraquat injection significantly decreased survival relative to PBS injection in control workers (χ² = 25.3, n = 120, p < 0.001) but not in reproductively activated workers (χ² = 0.4, n = 106, p = 0.522). At 15 days of age, paraquat-injected control workers survived on average 76.8 ± 4.0 (s.e.) hours, significantly lower than the 104.8 ± 3.7 h of their PBS-injected counterparts and the 113.5 ± 2.2 and 105.2 ± 4.9 h of the reproductively activated workers injected with paraquat and PBS, respectively (all: p < 0.001; figure 3a). The survival of the reproductively activated 25-day-old workers was not significantly different between PBS-injected (108.0 ± 4.4 h) and paraquat-injected (113.1 ± 2.8 h) groups (χ² = 0.1, n = 100, p = 0.720; figure 3a), while workers of the control group did not survive in sufficient numbers to age 25 days for the equivalent test to be conducted.

Figure 3.

Cage survival of experimental worker groups. (a) Paraquat experiment: while paraquat injection (Paraq) decreased survival compared to PBS injection (PBS) in control workers from a queenright colony (Contr), neither age group (15- or 25-day olds) of reproductively activated workers (Repro) from a queenless colony experienced a significant decrease in survival. (b) IAPV experiment: virus inoculation (IAPV) decreased survival compared to the sham treatment for all conditions, but the effect was less pronounced and thus survival higher in reproductively activated workers (Repro) than in non-reproductive control workers (Contr). The relatively high mortality of the 15-day-old sham-treated controls suggests that shaving off the thoracic hairs followed by a water application can reduce survivorship of caged honeybee workers on its own.

In the virus stress experiment, workers were topically inoculated with IAPV solution or subjected to a sham inoculation without virus. IAPV inoculation significantly decreased survival relative to sham treatment in 15-day-old control workers (χ² = 22.4, n = 83, p < 0.001) and to a lesser extent also in their reproductively activated counterparts (χ² = 13.1, n = 72, p < 0.001) and reproductively activated 25-day-old workers (χ² = 12.9, n = 70, p < 0.001), while insufficient numbers of non-reproductive control workers survived to age 25 days for an equivalent IAPV survival test to be conducted. Overall, the 25-day-old sham-treated, reproductively activated workers lived longest (more than 120 h), followed by 15-day-old sham treated (113.6 ± 4.4 h), 25-day-old IAPV inoculated (111.5 ± 2.6 h) and 15-day-old IAPV inoculated (93.1 ± 4.4 h) reproductively activated workers. The 15-day-old control workers survived 99.9 ± 4.3 h after sham treatment and 75.0 ± 3.8 h after IAPV inoculation (figure 3b). Thus, the survival of IAPV inoculation was significantly lower in non-reproductive control workers than the reproductively activated 15-day-old (Fisher's exact p < 0.001) and 25-day-old (p < 0.001) workers.

(c). Social manipulation and treatment affect specific transcripts

To verify reproductive activation and stress treatment effects at the physiological level and to test the hypothesis that the increased stress resistance of reproductively activated workers is owing to the upregulation of specific immune and anti-oxidant genes, pre- and post-treatment samples from all experimental groups were compared with qPCR (electronic supplementary material, file S2). Increased vitellogenin expression in queenless compared to queenright control workers confirmed reproductive activation of the queenless workers (table 1). The effectiveness of IAPV inoculation was also confirmed by higher IAPV quantities in the inoculated than the sham-treated workers (table 1). Gene expression differences in response to reproductive activation and/or stress application were identified in some immune and anti-oxidant genes according to our hypothesis, while other genes did not vary in expression (table 1; electronic supplementary material, file S2), leading to the subsequent study of all transcripts via Taq-seq analysis in the paraquat stress experiment.

Table 1.

Experimental effects^a on relative expression of specific molecular targets.

target	reproductive activation effectsa (control = QR, reproductively activated = QL)	treatment effectsa in control and reproductively activated workers after 24 and 120 h
paraquat experiment
vitellogenin	pre-treatment: QL > QR: F1,18 = 5.9, p = 0.026 24 h after PBS: QL > QR: F1,18 = 14.1, p = 0.001 24 h after paraquat: QL > QR: F1,17 = 12.0, p = 0.003	—
toll-6	24 h after PBS: QR > QL: F1,17 = 17.1, p = 0.001	controls: 24 h after PBS > 24 h after paraquat: F1,17 = 14.0, p = 0.002
Apidaecin	—	controls: 24 h after paraquat > 24 h after PBS: F1,14 = 8.5, p = 0.011
Argonaute-2	120 h after PBS: QL > QR: F1,18 = 6.4, p = 0.021	reproductively activated: 120 h after PBS > 120 h after paraquat: F1,18 = 6.7, p = 0.019
Catalase	24 h after paraquat: QR > QL: F1,17 = 5.8, p = 0.027	controls: 120 h after paraquat > 120 h after PBS: F1,17 = 4.7, p = 0.044
IAPV experiment
vitellogenin	pre-treatment: QL > QR: F1,18 = 5.9, p = 0.026 24 h after sham: QL > QR: F1,16 = 124.1, p < 0.001 24 h after IAPV: QL > QR: F1,18 = 10.0, p = 0.005	—
toll-6	24 h after sham: QR > QL: F1,17 = 11.1, p = 0.004 24 h after IAPV: QR > QL: F1,18 = 4.9, p = 0.040	—
Argonaute-2	—	reproductively activated: 24 h after IAPV > 24 h after sham: F1,18 = 7.2, p = 0.015
IAPV	—	controls: 24 h after IAPV > 24 h after sham: F1,16 = 30.0, p < 0.001 reproductively activated: 24 h after IAPV > 24 h after sham: F1,18 = 20.2, p < 0.001

^aOnly significant (p < 0.05) effects are shown.

(d). Reproductive activation buffers oxidative stress effects on entire transcriptome

Based on the survival and qPCR results, we selected one experimental paradigm to characterize the entire transcriptomic consequences. Tag-Seq [43] was used for the first time in honeybees to compare the effects of reproductive activation and paraquat stress in reproductively activated and in non-reproductive workers on the entire transcriptome with large sample sizes (46 total samples). The comparison of the transcriptomes of queenless to queenright 25-day-old workers prior to treatments with DESeq2 involved 6710 genes with data and 2070 of these genes were differentially expressed (DEGs): reproductive activation upregulated 1006 and downregulated 1064 genes (figure 4), with, respectively, 820 and 871 of these genes having a known function (electronic supplementary material, file S3). Queenright control workers exhibited differential expression of 1277 genes after paraquat versus PBS injection, with 590 upregulated (481 with known function) and 687 downregulated (482 with known function) genes in the paraquat relative to the PBS group (electronic supplementary, material, file S3). In sharp contrast, only two differentially expressed genes were discovered when comparing paraquat- and PBS-injected workers that were reproductively activated by queenless conditions (figure 4).

Figure 4.

Transcriptome responses to social condition and paraquat stress. Taq-seq comparison between control and reproductively activated workers revealed about 25% differential gene expression (top). When non-reproductive control worker transcriptomes were compared between paraquat- and PBS-injected groups, approximately 20% of genes were significantly up- or downregulated (bottom left). Differentially expressed genes (DEGs) significantly overlapped between reproductive activation (top) and paraquat response in control workers (bottom left). Workers that had been reproductively activated by queenless conditions only showed significant expression differences between paraquat- and PBS-injected groups for two genes (bottom right). In all plots, the y-axis depicts log-fold change of gene expression in treatment versus control samples, the neutral line is indicated in red, and a log-fold change of 0.5 in either direction is given in blue. The x-axis indicates the mean of normalized read counts. Red dots indicate genes that exhibit significant expression differences. (Online version in colour.)

Quantitatively, reproductive activation led to more pronounced gene expression differences (ranging from 267-fold downregulation to 198-fold upregulation) than paraquat exposure in queenright, non-reproductive workers (35-fold downregulation to 24-fold upregulation). Directional overlap analyses between the DEGs indicated significant overlap between genes that were upregulated by reproductive activation and paraquat exposure in queenright workers (110 common genes corresponding to a 1.7 × enrichment, p < 0.001), while genes downregulated by reproductive activation and paraquat injection showed no significant overlap (p = 0.377) with only 36 genes in common (electronic supplementary material, file S3). Significant overlap was also found with the IAPV-responsive genes identified by Galbraith et al. [46] among the upregulated genes, with 91 shared genes (equalling a 2.7-fold overrepresentation, p < 0.001) and a weaker but significant overlap among the downregulated genes (14 shared genes, equalling a 1.7-fold overrepresentation, p = 0.035).

Extraction of overall functional transcriptome differences between experimental groups was performed with GO-MWU analysis, based on expression differences of all genes quantified as signed p-values from DESeq2. Numerous enriched GO terms in up- and downregulated genes were identified (figure 5). Most of the highly significant GO terms in the comparison of reproductively activated and sterile workers before stress treatment were the result of upregulated genes and related to many homeostatic functions, including ‘regulation of hormone levels' and ‘defense response', and many metabolic functions, most prominently lipid metabolism. GO terms extracted from the gene expression differences between paraquat- and PBS-injected workers were mostly linked to downregulated genes and included many nervous system functions, while upregulated genes were overrepresented in categories that included ‘oxidation–reduction process', ‘electron transport chain' and others related to energy metabolism.

Figure 5.

Comprehensive gene ontology analyses of worker bee transcriptomes. GO analyses of the gene expression differences between workers that were reproductively activated under queenless versus sterile workers under queenright conditions (a) and between PBS- and paraquat-injected sterile workers (b) were analysed by GO-MWU tests [48]. Significant GO terms are adaptively clustered and colours indicate which terms are enrichment with upregulated (red) or downregulated (blue) genes. Size and darkness of the text symbolize the significance of each term and the numbers in front of each term indicate the proportion of DEGs among all genes associated with that term. (Online version in colour.)

4. Discussion

The exceptional longevity of social insect reproductives remains an intriguing case of comparative ageing research with general implications for understanding the plasticity and mechanisms of ageing [11]. Comparisons of non-reproductive and reproductive castes, young and old reproductives, or a combination thereof are beginning to emerge that suggest some involvement of common stress defense mechanism such as the upregulation of stress resistance genes [22,30,49]. However, results vary among studies and causes and consequences are difficult to disentangle without experimental manipulations. Furthermore, it has remained unresolved how much of the longevity increases in social insect reproductives is owing to adoption of ancestral mechanisms versus taxon-specific, potentially novel processes.

Here, we studied the link between stress resistance and reproduction directly. We manipulated honeybee workers to become reproductively activated, demonstrated that this reproductive activation increases survival of a virus and an oxidative stressor, and found that the better oxidative stress survival was linked to a buffering effect because transcriptome changes after paraquat exposure were negligible in reproductively activated individuals, in contrast with their non-reproductive counterparts. Exclusion of the ovaries in these transcriptome comparisons of the abdomen allowed us to bias our analyses towards tissues (e.g. fat body and midgut) and molecular processes with more direct functions in immunity and stress resistance, as opposed to the reproductive functions of the ovary. We identified significant overlap between genes upregulated by reproductive activation and in response to paraquat exposure, suggesting that the protection induced by reproductive activation preempts at least partially the actual stress response. Genes involved include some well-known public mechanisms of ageing and stress resistance but also some novel processes and unknown genes (see discussion below and electronic supplementary material, file S3). Gene ontology analyses suggest that metabolic adjustments and storage proteins play a key role. Thus, we tie the longevity advantage of insect reproductives to increased stress resistance and report that many ancestral mechanisms are involved.

The list of upregulated DEGs that are shared between reproductive activation and paraquat exposure included the G-protein coupled receptor Methuselah, a classic ageing gene that ties increased life expectancy to oxidative stress resistance [50]. The DEG lists also shared the adiponectin receptor involved in insulin signalling [51], the corazonin receptor that may be involved in nutritional and stress signalling [52], and several oxidoreductases, such as the oxidative stress-responsive peroxidase [53]. Glutathione S-transferase, Catalase and various other oxidoreductases were also upregulated by reproductive activation but not in response to paraquat, indicating that reproductive activation upregulates oxidative stress defense. This result in combination with the upregulation of other stress response genes, such as JNK and oxidative stress-induced growth inhibitor 2, in reproductively active workers indicates the possibility that reproductive activation in honeybees causes some oxidative stress [54], enhancing subsequent oxidative stress resistance and survival [55].

The 20 most upregulated genes in reproductively active versus non-reproductive workers did not belong to the canonical ageing or stress response pathways and their expression was not affected by paraquat. Nevertheless, they may have increased the bees' resistance to paraquat by buffering against oxidative stress in other ways. One such mechanism, the co-option of the egg-yolk protein vitellogenin to oxidative stress resistance, has already been identified in worker bees [20] and is also linked to the longevity of honeybee queens [17]. Vitellogenin is preferentially oxidized by paraquat [20] and among the most upregulated genes in our reproductively activated workers, in contrast with a previous study [56]. Thus, vitellogenin may contribute significantly to the increased stress resistance of our reproductively activated workers. However, the transcriptome analyses indicate more systemic changes that extend to other storage proteins, such as hexamerins 70a and 110, vigilin, and another very high-density lipoprotein. Although these proteins have not yet been investigated in detail, they may have acquired multi-functionality that could be similar to vitellogenin and thus serve similar buffering functions and prevent resource trade-offs [57,58]. These systemic changes are consistent with major hormone signalling responses to queenless conditions: ultraspiracle and the ecdyson receptor upregulation suggest enhanced ecdyson signalling, while juvenile hormone signalling may be decreased by upregulation of juvenile hormone esterase, which is positively linked to reproduction in honeybee workers and queens [59] and associated with physiological ageing and stress response in honeybee workers [60]. Further juvenile hormone involvement was suggested by the upregulation of krüppel homolog 1 [61] and takeout, a nutrient sensing, juvenile hormone binding hemolymph protein whose overexpression increases life expectancy in Drosophila [62].

No significant degree of overlap was found between the genes that were downregulated by reproductive activation and in response to paraquat, suggesting that downregulation of biological processes was quantitatively less important for the protection against stress by reproductive activation. Nevertheless, the 36 overlapping genes contained several interesting genes, such as Major Royal Jelly Protein 1 and 2, which have also been found upregulated in isolated ovaries of reproductive workers [59]. These proteins are major constituents of the brood food [63] and their downregulation in reproductively activated workers may indicate resource allocation towards the ovaries instead of the glandular production of brood food [64]. Paraquat stress in non-reproductive workers may lead to a depletion of freely available proteins and therefore diminish brood food production for nursing. In the reproductively activated workers only two genes were affected in response to paraquat, and the poorly characterized venom acid phosphatase was similarly downregulated by paraquat in non-reproductive workers. Genes that are downregulated in response to paraquat stress may indicate consequences of damage or resource reallocation instead of protective or repair mechanisms.

The reproductively activated workers enjoyed a more than threefold paraquat survival advantage with more than 80% survival regardless of age, while more than 60% of the non-reproductive young control workers died after the paraquat injection. Survival of the reproductively activated workers was not significantly affected by paraquat. By contrast, IAPV affected non-reproductive and reproductively activated worker survival but the effect was less severe in reproductively activated workers and survival of the non-reproductive workers was lower. The more distinct protection against paraquat by reproductive activation was the reason to perform the transcriptome analysis on the paraquat experiment. However, the targeted qPCR analysis of candidate genes was performed on all experimental groups and showed similar effects of reproductive activation in both experiments. Thus, the analysis of reproductively activated versus non-reproductive workers may also be used to understand differences in constitutive immunity that could explain the better IAPV survival of reproductively activated workers. In this context, the upregulation of the following genes is of interest: aubergine, involved in the piwi pathway that confers antiviral resistance [65], the RNA-induced silencing complex component Argonaute-2 [66], the RNA regulator roquin-1 [67], and possibly the venom peptides mellitin and apamin that have antiviral activity [68]. None of these genes were identified as IAPV-responsive in adult workers [46] but the significant overlap of our DEG list and IAPV-responsive genes in general may further explain why reproductive activation protects against IAPV infection. The overlap included krüppel homolog 1 and the corazonin receptor, implicating again central regulatory pathways in the reversal of the trade-off between reproduction and survival [57].

Our gene ontology analyses based on the expression patterns of all genes supported the above interpretations. The enrichment of GO terms with either up- or downregulated genes is driven by the DEGs to varying degrees, but the GO-MWU includes all genes with valid data [48]. Thus, the significant enrichment of upregulated genes in metabolic-, defense response-, homeostasis- and hormone-related GO terms and of downregulated genes in RNA- and mitochondrial-related GO terms by reproductive activation corroborates and generalizes our above conclusions. The GO enrichment owing to paraquat effects in non-reproductive workers was also congruent with expectations because the most significant GO terms enriched for upregulated genes were ‘cellular catabolic process', ‘oxidation-reduction process' and energy-related. Most GO enrichments in response to paraquat were caused by downregulation, which is noteworthy because the number of up- and downregulated DEGs of known function was almost identical. Numerous downregulated GO terms were consistent with reduced investment in growth and reproduction in the paraquat stressed workers. However, it is important to remember that our analyses only included about half of all known genes [69] because the other genes were not sufficiently expressed. Moreover, many unknown genes were significantly upregulated in response to both, paraquat and reproductive activation. These genes may be particularly interesting because they have presumably evolved rapidly and may harbour new domains that confer unique functionality [70]. While our data suggest that regulatory changes have changed the relationship between the ancient processes of reproduction and detoxification, it cannot be excluded that such novel molecular mechanisms also play an important role.

Laying honeybee workers are longer lived and more stress-resistant than non-reproductive workers but de-queening colonies to reproductively activate workers is not a practical solution for preventing the ongoing declines in honeybee health [71]. Laying workers are in a terminal investment stage because they can only produce males, dooming the survival of the colony [72]. Terminal investment often favours reproduction over somatic maintenance [73], but reproductive workers do not compromise immunity or other survival mechanisms while investing in reproduction. Thus, the observed patterns may be more indicative of the general life-history evolution of social insects and explain their ageing patterns. Nevertheless, our characterization of transcriptomic changes that convey stress resistance in honeybee workers can be used as a foundation for identifying mechanisms that can be manipulated to improve honeybee health.

5. Conclusion

In summary, this study provides evidence for several mechanisms that can explain the exceptional longevity of social insect reproductives. The longevity is owing to a mix of extrinsic and intrinsic factors [15] and the intrinsic factors may be owing to reproductive activation [25,28]. Our study links enhanced survival of a biotic and an abiotic stressor to reproductive activation, which provides the experimental link between stress resistance and longevity [74] in social insects. The transcriptome data indicate that a mixture of known and potentially novel mechanisms are involved, all of which await further exploration in the light of social evolution.

Supporting information

Electronic supplementary material, S1: Primer sequences used for qPCR analysis.

Electronic supplementary material, S2: Boxplots of qPCR results.

Electronic supplementary material, S3: Excel file containing all DEGs identified by the Taq-seq analysis.

Data accessibility

The raw data of this study are available from the corresponding author upon reasonable request. Reads of the Tag-Seq experiment have been deposited at NCBI' Short Read Archive (Bioproject PRJNA661322).

Authors' contributions

A.K. designed and performed the experiments, analysed the data and wrote the first draft of the manuscript. J.H. performed and analysed the Taq-seq analysis and participated in writing the final draft of the manuscript. O.R. conceptualized the overall study, analysed data and wrote the final version of the manuscript.

Competing interests

The authors declare no competing interests.

Funding

The study was supported by the UNCG Biology Department (https://biology.uncg.edu/) to A.K. and the Ecology, Evolution and Behavior Graduate Program of UT Austin (https://cns.utexas.edu/eeb-graduate-program) to J.H. Further funding for the project was provided to O.R. by the US Department of Agriculture (Animal and Plant Health Inspection Service: 16-8130-0636-CA), the US Department of Defense (Army Research Office: W911NF1520045) and the National Institutes of Health (National Institute on Aging: R21AG046837).