Pathogen‐induced maternal effects result in enhanced immune responsiveness across generations

Abstract Parental investment theory postulates that adults can accurately perceive cues from their surroundings, anticipate the needs of future offspring based on those cues, and selectively allocate nongenetic resources to their progeny. Such context‐dependent parental contributions can result in phenotypically variable offspring. Consistent with these predictions, we show that bacterially exposed Manduca sexta mothers oviposited significantly more variable embryos (as measured by mass, volume, hatching time, and hatching success) relative to naïve and control mothers. By using an in vivo “clearance of infection” assay, we also show that challenged larvae born to heat‐killed‐ or live‐Serratia‐injected mothers, supported lower microbial loads and cleared the infection faster than progeny of control mothers. Our data support the notion that mothers can anticipate the future pathogenic risks and immunological needs of their unborn offspring, providing progeny with enhanced immune protection likely through transgenerational immune priming. Although the inclusion of live Serratia into oocytes does not appear to be the mechanism by which mothers confer protection to their young, other mechanisms, including epigenetic modifications in the progeny due to maternal pathogenic stress, may be at play. The adaptive nature of maternal effects in the face of pathogenic stress provides insights into parental investment, resource allocation, and life‐history theories and highlights the significant role that pathogen‐induced maternal effects play as generators and modulators of evolutionary change.

Consistent with the predictions made by parental investment theory, the empirical research on maternal effects, and the facts that immune responses are energetically costly (Bonduriansky, Runagall-McNaull, & Crean, 2016;Schmid-Hempel, 2003) and that energy is both limited and limiting, we hypothesized that Manduca sexta mothers who experienced pathogenic insults prior to reproduction would make differential contributions toward their future progeny, resulting in phenotypically variable brood (McGinley, Temme, & Geber, 1987;Schmid-Hempel, 2003;West-Eberhard, 1989). Furthermore, we also hypothesized that such contributions (including immune-related products) could be incorporated into eggs resulting in more immunecompetent offspring than progeny of mothers who did not experience such pathogenic pressures. Here, we report on the increased volumetric variability of embryos when mothers were exposed to pathogenic bacteria during or soon after oogenesis. Moreover, maternal treatment impacted both embryonic developmental timing and hatching success as well as immune responsiveness of first instar larvae born to mothers treated with heat-killed or live bacteria, relative to larvae from control mothers.

| Bacterial and insect cultures, and pupal injections
The Gram-negative bacterium, Serratia marcescens, was chosen to elicit a maternal immune response because it is an ecologically relevant pathogen, commonly found on foliage and in soil (Sikorowski, Lawrence, & Inglis, 2001). For further details on the choice of pathogen and the rationale behind our experimental design, see Appendix S1. Manduca sexta fertilized eggs (i.e., embryos) and larvae were originally obtained from Carolina Biological Supply (Burlington, NC).
Larvae were reared on standard artificial diet (Bell & Joachim, 1976) at 25°C under a 16-hr:8-hr light/dark cycle and were injected as pupae 2 days prior to their expected date of eclosion. A subset of female pupae were randomly weighed, and all female pupae were allocated to one of the following four different treatments ( Figure 1): naïve (unmanipulated), saline injection (injected with 10 μl of sterile Burns-Tracey saline [BTS]), heat-killed bacterial injection (injected with 10 μl of a 10 8 /ml heat-killed S. marcescens in BTS), and live-bacteria F I G U R E 1 Schematic of research protocol injection (injected with a nonlethal total dose of 4,000 live S. marcescens bacterial cells suspended in 10 μl of BTS). Although heat-killed-Serratia-treated mothers received a larger dose of immune elicitors relative to the live-Serratia-injected mothers, such dose increased the probability that the elicited physiological responses in the mothers had "trickle-down" consequential effects across generations without the added negative impact of disease itself. The maternal naïve treatment served as an important second control to identify the effects of maternal stress associated with cuticular puncturing. Male pupae always remained untreated (naïve). Before injection, all females' abdominal side was swabbed with 70% ethanol and injections were performed by inserting the sterile needle of a 10μl Hamilton syringe between the fourth and fifth ventral abdominal segments of the pupa. Following injection, the area was swabbed with 70% ethanol for a second time, and immediately after, male and female pupae were placed at the base of flying cages. Although a slight possibility existed that alcohol swabbing did not completely sterilize the cuticle before injections, great care was given to ensure that no contaminant bacteria other than our Serratia strain was injected into the pupa. For studies in which identification of both parents was required (morphometric experiments [mass and volume data], see below), individual treated females were mated with individual untreated males in cages (30 × 30 × 60 cm). For all additional experiments, groups of 4-6 identically treated females were mated with similar numbers of untreated males in larger breeding cages (40 × 40 × 75 cm). Although the parentage in these cages could not be determined, such breeding groups resulted in a more efficient production of known same-aged embryos. Upon eclosion, parents were provided with 10% sucrose as a food source. The mated moths oviposited on 30 mm foam plugs infused with tobacco extract and suspended from the cage ceiling.
Embryos were then collected and allocated to the different experiments below.

| Morphometric analyses of embryos
Embryonic mass and volume measurements were used as proxy for embryo quality (Rossiter, 1991). While embryonic mass data were collected across a total of 61 mothers (naïve = 15, saline = 18, heatkilled Serratia = 16, live Serratia = 12), embryonic volume was collected from a total of 59 mothers, mostly the same mothers from which we collected embryonic mass (naïve = 13, saline = 16, heatkilled Serratia = 18, live Serratia = 12). Mothers usually oviposit for up to 3 days posteclosion, and all of their embryos were collected throughout this time period. Thus, our protocol captured the entire reproductive output of each female/maternal treatment. Each embryo was individually weighed and photographed under a 4× magnification (SPOT Q camera). Given that embryos are spherical, we used the cross-sectional area of the embryo image to calculate its volume by using the formula of a sphere V = 4/3(πr 3 ). Out of a total of 3,720 collected embryos, we had both mass and volume data for 2,250 embryos across all four maternal treatments (naïve = 499, saline = 540, heat-killed-Serratia = 377 and live-Serratia injected = 834).

| In vivo bacterial clearance assay: microinjections, bleeding, and enumeration of recovered bacteria
A clearance of infection assay was developed in which we not only determined the ability of larvae to reduce/eliminate (i.e., clear) a Serratia infection in vivo, but also established the time course for eradicating of such infection. Embryos from saline (n = 90), heat-killed (n = 63) and live-Serratia (n = 75) maternal treatments were placed on insect diet shortly after oviposition. These embryos and their mothers were a different subset than those used for the morphometric analyses described above. Because the aim of the study was to test whether maternal exposure to heat-killed or live bacteria resulted in more immune competent progeny relative to the offspring of saline-injected mothers, this experiment did not include larvae of naïve mothers. Two days posthatching, the larvae were cold-immobilized, swabbed with 70% ethanol, and then injected with a live Serratia (a total of 228 larvae across the three treatments) through the ventral intersegmental membranes with a pulled glass capillary tube (Fig. S1). The challenge dose/ larva was 1.5 × 10 5 live Serratia cells suspended in 1 μl of BTS, administered using a microinjection apparatus (Picospritzer). Additionally, to test whether our Manduca laboratory colony was naturally infected with Serratia, first instar larvae from naïve mothers were also injected with 1 μl of sterile BTS lacking Serratia (n = 40). To visually confirm that both the bacterial and control saline solutions successfully reached the hemocoel, both solutions were colored with sterile red food coloring (1 μl of dye/40 μl of BTS). The food coloring was metabolized within the first 24 hr with no apparent negative effects (Rosengaus, Malak, & McKintosh, 2013). Subsequently, larvae were placed in labeled cups containing nutrient medium. A subset of these insects were bled either immediately (providing a baseline Serratia re-

| Developmental milestones and hatching success of progeny as a function of maternal treatment
To determine the impact that maternal treatment had on developmental milestones of offspring (e.g., hatching time and their hatching success), we followed the development of a total of an additional set of 3,275 embryos (n = 834 embryos from naïve mothers, 604 from saline-injected mothers, 973 embryos from heat-killed-bacteriainjected mothers and 864 from live-bacteria-injected mothers). This set of embryos was different from those used in the morphometric and clearance of bacteria assays. Within the first 24 hr after the onset of oviposition, each embryo was individually coded and maintained on Manduca diet in 96-well cell culture plates. These embryos were checked daily for a maximum of 7 days, and the number of days elapsed from oviposition to hatching and their hatching success were quantified.

| Maternal translocation of bacteria during oogenesis
We investigated whether live Serratia (and/or its constituents) were incorporated into the embryo via transovarian transmission, a phenomenon reported in both Galleria mellonella and Tribolium castaneum (Freitak et al., 2014;Knorr et al., 2015). To this end, embryos (n = 18 from up to seven different live-bacteria-injected mothers) were frozen, and 20 μm cryosections were cut and collected onto positively charged Thermo Scientific™ Shandon™ ColorFrost™ Plus Slides. The resulting 16-20 sections/embryo permitted observation of the entire embryo. Further description of this protocol along with the preparation of positive controls is available in Appendix S1.

| Statistical analyses
Given that the main aim of our work was to test whether immune elicitation (by components of the heat-killed Serratia or live bacteria) affected progeny's phenotype, analyses focused on how measures of embryo quality and immune responsiveness varied between each of the maternal treatments (naïve, heat-killed, and live Serratia) and the reference saline-injected maternal treatment. Neither embryo mass (all Shapiro-Wilk statistics >.92 and their corresponding significance p < .0001 across the four maternal treatments), embryo volume (all Shapiro-Wilk Statistic >.87 all p < .0001), nor numbers of recovered bacteria (all Shapiro-Wilk Statistics >.55, all p < .0001) were normally distributed despite attempts to normalize the data through log 10 , log n , and arcsin transformations. Hence, data were analyzed using several nonparametric tests.
Embryonic mass and volume data across the four maternal treatments were subjected to a Levene test for homogeneity of variance. Overall differences in the median embryonic mass and volume were also analyzed using Kruskal-Wallis (KW) tests across the four treatments with subsequent Mann-Whitney (MW) tests between treatments and a Bonferroni correction due to multiple pairwise comparisons (setting a conservative significance threshold of p ≤ .01).
Unfortunately, this latter analysis could not control for the effect of maternal identity nor for the fact that multiple embryos originated from the same mother. Hence, we generated two different linear mixed-effect models based on 2,250 embryos for which we had both mass and volume data. The first model tested the effect of maternal treatment (fixed categorical variable) on embryo mass while including maternal identity as a random-effect term and embryo volume as a fixed covariate. The second model similarly tested the effect of maternal treatment on embryo volume while controlling for embryo mass (fixed covariate) and while accounting for the nonindependence of the observations and residuals obtained from the same mothers (random effect).
Both models focused on main effects and interactions between the maternal treatment and embryo volume or maternal treatment and embryo mass, respectively. Pairwise comparisons between maternal treatments were also generated as part of these linear mixed-effect models while using a Bonferroni correction.
Differences in the rates of Serratia recovery as a function of maternal treatment (while controlling for time of bleeding) were also analyzed with a GLM that included maternal treatment as a categorical

| Morphometric analyses of embryos
The variance in embryonic mass (Levene statistic = 9.7, df 1 = 3, df 2 = 2,246, p < .0001; Figure 2a) and embryonic volume (Levene statistic = 54.4, df 1 = 3, df 2 = 2,246, p < .0001; Figure 2b) across the four maternal treatments was not homogeneous. Yet, embryonic mass data clustered more tightly around the median than embryonic volume, which exhibited broader dispersion (Figure 2a Figure 2b). Further detailed analyses showed that embryos larger than 3 mm 3 (Figure 2b) were not all progeny of a single large mother. These more voluminous embryos were progeny of between 12% to 16% of the females across the various maternal treatments. These same mothers (not necessarily the heaviest) also produced a range of smaller embryos, including embryos that were as small as 1.0-1.6 mm 3 (e.g., in the heat-killed Serratia treatment). after a Bonferroni correction, see Table S1 for these and additional pairwise comparisons between all maternal treatments, Appendix S1).
There was a significant difference in the interaction between maternal treatment and embryo volume (F = 31.6, df numerator = 3, df denominator = 2,242, p ≤ .0001).
With respect to embryo volume, a second linear mixed-effect model showed that both maternal treatment and embryo mass sig-  Table   S2. There was no significant interaction between maternal treatment and embryo mass (F = 1.5, df numerator = 3, df denominator = 2,242, p ≤ .2).

| Developmental milestones and hatching success of progeny as a function of maternal treatment
After controlling for the effects of embryonic mass and volume (maternal identity and mass were not available), the time course of hatch-

| Clearance of infection assay
Accurate enumeration of recovered bacteria from the larvae's hemocoel required first a positive identification of Serratia. The use of Serratia-specific primers corroborated that counting only the pink/red bacteria would represent reliable recovery rates of Serratia used in the larval challenges. None of the DNA from white CFUs growing in our plates was amplified by the Serratia-specific primer (Fig. S4). Positive controls for the PCR assays included one pink CFU grown from our experimental frozen bacterial stock. We also confirmed that our laboratory Manduca colony was not naturally infected with this microbe as hemolymph of saline-injected larvae (who were progeny of naïve mothers) never showed positive growth for Serratia.
Results from our GLM test helped assess the impact that maternal treatment had on the larvae's ability to clear bacterial infection.  Table S3 for additional pairwise comparisons).

| Maternal translocation of Serratia
Of the total 18 immunofluorescently stained samples, none showed positive fluorescence for S. marcescens in any region of the embryo (Fig. S6). We are confident that our protocol would have detected high concentrations of S. marcescens based on the strong signal seen in our positive controls. However, we realize that minute amounts of S. marcescens may have escaped detection.

| DISCUSSION
Manduca sexta has been a long-standing test organism for the study of insect immunity (Kanost & Blissard, 2015;Kanost, Jiang, & Yu, 2004;Zhen, Najar, Wang, Roe, & Jiang, 2008 and references therein). We feel confident that both cuticular injury and, particularly, the exposure to immune elicitors or live Serratia set the stage for physiological trade-offs between a mother's somatic/immune maintenance and her reproductive system (see Appendix S1 for analyses on impacts of maternal treatment on maternal biological measures). Using this tractable model species, we focused on whether pathogen-induced maternal effects resulted in increased phenotypic variability of their progeny in terms of embryological mass, volume, developmental milestones as well as immune competency in first instar larvae.

| Morphometric analyses of embryos
Although overall, embryonic mass and volume were both influenced by maternal treatment, embryo mass appears to be a more constrained physical attribute than embryo volume (Figure 2a,b). The fact that these two morphometric measures of embryo quality respond differently to maternal pathogenic stress suggests that, just like in the gypsy moth (Rossiter, 1991), mass of Manduca embryos may not be as plastic and, hence, may not be a reliable measure of differences in maternal provisioning, embryonic metabolic state (Maino, Kearney, Nisbet, & Kooijman, 2014), developmental transformation , embryo quality, and, ultimately, fitness. Interestingly, the mass and volume of embryos oviposited by saline-injected mothers (even after controlling for maternal identity and after accounting for the fact that multiple embryos originated from the same mother) were significantly different from embryos of naïve mothers (Figure 2b, Tables S1 and S2 pairwise comparisons). This suggests that the mere act of aseptically puncturing the cuticle of the maternal pupal case elicits cross-generational phenotypic effects. Cuticle abrasion and wounding are known to elicit immune responses (Brey et al., 1993;Johnston & Rolff, 2013) (Razzell, Wood, & Martin, 2011). Thus, recognizing that

| Developmental milestones as a function of maternal treatment
Maternal treatment not only influenced physical attributes of the embryos but also impacted their developmental timing and hatching success. A higher proportion of progeny from live-bacteria-injected mothers hatched on the third day postoviposition relative to progeny from the other treatments. This consistent pattern across the 7-day census period translated into a 90% hatchability rate. Just as in the morphometric experiment, our hatchability experiment shows that exposure to live bacteria during oogenesis appears to hasten the development of embryos beyond the effects recorded for heat-killedbacteria-and saline-injected mothers. The fact that progeny of the latter two maternal treatments had faster development and higher hatchability than embryos from naïve mothers points to differential maternal contributions when facing general stress (e.g., cuticular wounding  (Kinsella, 1966;Pant, Kumar, & Dhar Singh, 1979)] which could result in faster development and higher hatchability. All the above questions remain to be answered.

| Clearance of infection assay
In There are several putative mechanisms, not necessarily mutually exclusive, that can explain the apparent heightened immune competency of progeny from heat-killed-and live-Serratia-injected mothers.
First, the translocation of bacteria and/or bacterial components from Serratia-treated mothers to oocytes could elicit the embryo's own immunological responses in a strain-specific manner (Freitak et al., 2014;Knorr et al., 2015). We found no evidence for this mechanism (Appendix S1). Second, mothers could incorporate prefabricated nutrients and immune-related compounds into the developing oocytes (i.e., antimicrobial peptides, mRNAs, RNA-binding proteins, enzymes), a strategy that could presumably protect progeny against microbial challenges until the offspring's own transcriptional machinery becomes active (Sysoev et al., 2016). These maternal "gifts" have been reported in both invertebrate and vertebrate species (Broggi, Soriguer, & Figuerola, 2016;Grindstaff, Brodie, & Ketterson, 2003;Hasselquist & Nilsson, 2009;Moret, 2006;Rossiter, 1991;Sadd & Schmid-Hempel, 2007;Seppola, Johnsen, Mennen, Myrnes, & Tveiten, 2009;Trauer-Kizilelma & Hilker, 2015a, 2015bZanchi et al., 2012). Finally, mothers could modulate the progeny's gene expression through DNA methylation and/or histone acetylation (Oldroyd et al., 2014) which could result in the lower microbial loads and earlier eradication of Serratia in offspring of heat-killed and live-injected Serratia. Notably, some of the above mechanisms proposed to explain maternal effects in the face of disease can also apply to instances where paternal effects occur (Rodgers, Morgan, Leu, & Bale, 2015). Fathers could transmit directly and/or indirectly (through the female) information to their unborn progeny in the form of epigenetic changes or immune-related compounds via their seminal fluids. Establishing whether maternal and paternal effects are additive or multiplicative may help identify environmental reasons underpinning phenotypic differences (i.e., plasticity) and their evolutionary consequences (Senner, Conklin, & Piersma, 2015;West-Eberhard, 1989).
Our findings demonstrate the significant role that pathogens play as both agents of selection and generators of phenotypic variability.
The recognition that environmental pressures (including pathogenic burden) can impact life-history traits across generations has represented a major paradigm shift in our understanding of the interdependence among environmental stress, physiological responses, inheritance as well as rates of evolutionary innovation and change (Houri-Ze'evi et al., 2016;West-Eberhard, 1989). Hence, this is a fertile area that merits further research.

ACKNOWLEDGMENTS
We thank Alexander Chancellor, Meghan Chin, Theodore Bowe, John

CONFLICT OF INTEREST
We declare we have no competing interests.

AUTHORS' CONTRIBUTIONS
RBR and WS conceived, designed, ran statistical analyses and coordinated the study. They also drafted the manuscript. NH performed both the clearance of bacteria assay and the histoimmunofluorescence experiments and with the help of CB, enumerated and verified the identity of the enumerated bacteria. JK, JM, and MZ raised, collected, and quantified M. sexta embryos used in this study and assisted in data analyses. JM and BG performed developmental and hatching success studies. All authors gave final approval for publication.

ETHICS
Manduca sexta were purchased from Carolina Biological Supply, and these research organisms do not require animal ethics approval. To avoid pain, animals were cold-immobilized before any injections were performed.

DATA ACCESSIBILITY
Data presented in this paper are available as an electronic Appendix S1.