Partner-specific induction of Spodoptera frugiperda immune genes in response to the entomopathogenic nematobacterial complex Steinernema carpocapsae-Xenorhabdus nematophila

The Steinernema carpocapsae-Xenorhabdus nematophila association is a nematobacterial complex (NBC) used in biological control of insect crop pests. The ability of this dual pathogen to infest and kill an insect strongly depends on the dialogue between the host’s immune system and each partner of the complex. Even though this dialogue has been extensively studied from the two partners’ points of view in several insect models, still little is known about the structure and the molecular aspects of the insects’ immune response to the dual infection. Here, we used the lepidopteran pest Spodoptera frugiperda as a model to analyze the respective impact of each NBC partner in the spatiotemporal immune responses that are induced after infestation. To this aim, we first analyzed the expression variations of the insect’s immune genes in the fat bodies and hemocytes of infested larvae by using previously obtained RNAseq data. We then selected representative immune genes for RT-qPCR investigations of the temporal variations of their expressions after infestation and of their induction levels after independent injections of each partner. We found that the fat body and the hemocytes both produce potent and stable immune responses to the infestation by the NBC, which correspond to combinations of bacterium- and nematode-induced ones. Consistent with the nature of each pathogen, we showed that X. nematophila mainly induces genes classically involved in antibacterial responses, whereas S. carpocapsae is responsible for the induction of lectins and of genes expected to be involved in melanization and cellular encapsulation. In addition, we found that two clusters of unknown genes dramatically induced by the NBC also present partner-specific induction profiles, which paves the way for their functional characterization. Finally, we discuss putative relationships between the variations of the expression of some immune genes and the NBC’s immunosuppressive strategies. Author summary Entomopathogenic nematodes (EPNs) are living in the soil and prey upon insect larvae. They enter the insect by the natural orifices, and reach the hemocoel through the intestinal epithelium. There, they release their symbiotic bacteria that will develop within the insect and eventually kill it. Nematodes can then feed and reproduce on the insect cadaver. By using transcriptomic approaches, we previously showed that Lepidoptera larvae (caterpillars of the fall armyworm Spodoptera frugiperda) produce a strong immune response in reaction to infestation by EPNs. However, we do not know if this immune reaction is triggered by the nematode itself -Steinernema carpacapsae - or its symbiotic bacteria - Xenorhabdus nematophila. To answer this question, we present in this work a careful annotation of immunity genes in S. frugiperda and surveyed their activation by quantitative PCR in reaction to an injection of the bacteria alone, the axenic nematode or the associated complex. We found that the immune genes are selectively activated by either the bacteria or the nematode and we discuss the implication of which pathway are involved in the defense against various pathogens. We also show that a cluster of newly discovered genes, present only in Lepidoptera, is activated by the nematode only and could represent nematicide genes.

members in the regulation of anti-nematode immunity (46,47). 98 In order to improve our understanding of the dialogue that takes place between this NBC and its host, we  These responses were found to be stable over the time post-infestation and to consist in combinations of X. 114 nematophila-induced and S. carpocapsae-induced responses in each tissue. The X. nematophila-induced 115 responses mainly correspond to genes that are classically involved in antibacterial immunity, whereas the 116 S. carpocapsae-induced ones mainly include lectins and genes potentially involved in melanization and 117 encapsulation. In addition, our RT-qPCR experiments show that two previously identified candidate 6 118 clusters of uncharacterized genes (48) also present partner-specific induction profiles. Our hypothesis is 119 that they may correspond to new types of anti-nematode and antibacterial immune factors found in 120 Spodoptera genus and lepidopteran species, respectively.

122
Hemocytes' and fat body's immune responses 123 In order to get an accurate picture of the S. frugiperda transcriptional immune responses to the NBC 124 infestation, we first used a previously published list of immune genes identified by sequence homology in 125 the S. frugiperda genome (49). We then looked at their expression variations in the fat body and in the 126 hemocytes (S1A Table) and we completed the repertoire with additional putative immune genes that we 127 directly identified from our RNAseq data (S1B Table). In total, we present the annotation of 226 immune 128 or putative immune genes of which 132 were significantly modulated at 15 h post-infestation (hpi) (Sleuth, 129 p-value < 0.01; |Beta| > 1; all count values > 5 in at least one condition) in one or both tissues (Fig 1). 130 Among them, 62 were involved in antimicrobial responses (Fig 1A), 18 were related to melanization (Fig   131   1B), 23 were involved in cellular responses ( Fig 1C) and the 29 remaining genes were grouped in a category 132 called "diverse" due to pleiotropic or poorly characterized functions (Fig 1D). 133 Antimicrobial responses. In the antimicrobial response category, 58 genes were found to be 134 upregulated in at least one of the two tissues ( Fig 1A). The signaling genes encoded 3 and 8 members of 135 the Imd and Toll pathways, respectively, as well as 5 short catalytic peptidoglycan recognition proteins 136 (PGRP-S), which are probably involved in the regulation of these pathways by peptidoglycan degradation 137 (50, 51) ( Fig 1A). Four other genes were considered as involved in recognition. They encoded Gram 138 negative binding proteins (GNBPs), which have been reported to recognize peptidoglycans or β-glucans 139 and participate in the further activation of the Toll pathway (22) (Fig 1A). Finally, the effector genes 140 encoded 33 antimicrobial peptides (AMPs) belonging to all the S. frugiperda's AMP families (49) plus 4 141 lysozymes and lysozyme-like proteins (LLPs) (Fig 1A). Depending on their families and on the insect 142 species, AMPs can present varied activity spectra, ranging from antiviral or antibacterial activities to anti-143 fungal and anti-parasitic ones (52). Varied activity spectra have also been found for several insects' 144 lysozymes and LLPs (53-57). Interestingly, all of the categories and subcategories cited above were 145 represented in the two tissues, indicating that their antimicrobial responses are diversified and that the 146 factors responsible for their disappearance in the hemolymph (24, 41) probably act at a post-transcriptional 147 level. About a half of the genes presented similar and significant induction profiles in the hemocytes and 148 in the fat body. This is for instance the case of the usually anti-Gram negative bacteria attacin, cecropin and 149 gloverin AMPs (52), which were all highly induced in the two tissues (Fig 1A), suggesting they both 150 respond to the bacterial partner X. nematophila. On the other hand, all the induced GNBP, lysozyme and 151 LLP genes were found to be either significantly induced in the hemocytes or in the fat body, and in the 7 152 AMP category, tissue-specificities were observed for diapausin, defensin-like and most moricin genes (Fig   153   1A).

154
Only 8 antimicrobial response genes were found to be significantly downregulated ( Fig 1A). Interestingly, 155 4 were involved in the Imd pathway whereas the 4 remaining ones were dispersed between the AMP, GNBP 156 and lysozyme categories (Fig 1A). The Imd pathway downregulated factors included sickie and the akirin 157 in the hemocytes and SMARCC2 and BAP60 in the fat body ( Fig 1A). In D. melanogaster, Sickie 158 participates in the activation of Relish, the transcription factor of the Imd pathway (58) and the akirin acts 159 together with the Brahma chromatin-remodeling complex, containing BAP60 and SMARCC2, as cofactor 160 of Relish to induce the expression of AMP genes (59). Given the potent induction of anti-Gram negative 161 bacteria immune responses in the two tissues, the down-regulation of these genes could be attributed to 162 immune regulations. However, it has been shown that in the close species S. exigua, live X. nematophila 163 reduces the expression of several AMP genes, including attacin, cecropin and gloverin (42, 60, 61). It would 164 thus be of particular interest to determine whether the observed down-regulations are related to this 165 immunosuppressive effect.

166
To summarize, the antimicrobial responses are potent and diversified in the two tissues, with a common 167 induction of genes that probably respond to X. nematophila. Yet unexplained tissue-specific responses were 168 observed and the results show a down-regulation of Imd pathway members that could be related to a  Melanization. In the melanization category, 16 genes were found to be upregulated in at least one of the 173 two tissues (Fig 1B). These genes firstly encoded 6 serine proteases ( Fig 1B) that were considered as 174 members of the prophenoloxidase (proPO) system. The proPO system is an extracellular proteolytic 175 cascade ending in the maturation of the proPO zymogen into PO, which initiates the melanization process 176 (62). Among the upregulated serine proteases, PPAE2 is the only one that is known to take part in proPO 177 processing whereas the other proteases were included in this category because of their characteristic CLIP  Punch-like (64, 65) as well as 4 genes, Reeler-1 and 3 Hdd23 homologs, that are involved in melanization 182 and nodule formation in other models (66, 67) ( Fig 1B). Despite of tissue-specific induction patterns, serine 183 proteases and serpins were found in the two tissues (Fig 1B), suggesting that both participate in the 184 stimulation of the proPO system, which is consistent with results obtained in other interaction models (68-185 70). However, with the exception of the DDC, all the melanization enzymes as well as the nodulation-8 186 related genes were specifically induced in the hemocytes (Fig 1B), which is consistent with the very 187 localized nature of this immune response (65) that is mainly mediated by hemocyte subtypes.

188
Finally, only 2 genes, PPAE1 and Yellow-like 2, were found to be significantly down-regulated in this 189 category ( Fig 1B). Both were specifically repressed in the hemocytes, which could be due to functional 190 interferences with their upregulated homologs (PPAE2 and Yellow-like 1).

191
In summary, our results suggest that both the hemocytes and the fat body participate in induction and  like) (85) were also found upregulated ( Fig 1C). Only 2 genes (Ced-6-like, Rbsn-5-like) of the cellular 216 responses category were found to be upregulated in the fat body ( Fig 1C) and both encoded intracellular 217 proteins that are probably not related to immunity in this tissue.

218
All the 4 down-regulated putative cellular immunity-related genes were specifically modulated in the 219 hemocytes ( Fig 1C). They encoded 2 Rho GTPase Activating Proteins (RhoGAP-like), a scavenger 220 receptor similar to the D. melanogaster Croquemort receptor (SR-B3) and a homolog of the D. In agreement with these assertions, we found that 2 of these genes 246 were down-regulated in the fat body, but all 4 genes were upregulated in the hemocytes (Fig 1D). Two 247 other signaling genes were found to be specifically overexpressed in the hemocytes. The first one is a 248 homolog of the Litopenaeus vannamei (Decapoda: Penaeidae) leucine-rich repeat flightless-I-interacting 249 protein 2 (LRRFIP2-like) (Fig 1D), which has been shown to upregulate AMP expression in L. vannemei 250 as well as in D. melanogaster (98). On the other hand, 3 signaling genes were found to be strictly down-251 regulated ( Fig 1D). Interestingly, these genes included a member of the TGF-β pathway (BAMBI-like) in all but one of these IMPI homologs were found to be specifically upregulated in the hemocytes, a tissue-279 specificity that had not been highlighted in previous reports (121, 122).

280
Finally, the remaining genes of unknown function encoded Spod-x-tox, a protein without antimicrobial 281 activity which contains tandem repeats of defensin-like motifs (123), 3 REPAT genes, which are known to 282 be induced in the midgut after exposure to toxins, viruses and intestinal microbiota perturbations in the 283 close species S. exigua (124-126), and Hdd1, which is induced in response to bacteria and peptidoglycan 284 in the lepidopteran models Hyphantria cunea and Bombyx mori (127, 128) ( Fig 1D). 285 In summary, we found an important additional mobilization of several relevant candidate immune genes, In order to put the S. frugiperda immune responses in relation with the infectious process, we then described 292 their temporal dynamics in each analyzed immunocompetent tissue. To this aim, we monitored with RT-293 qPCR experiments the induction levels of selected representative immune genes from 5 hpi, the mean time 294 at which nematodes release X. nematophila in the hemocoel, to 20 hpi, which is about 9 hours before the 295 first insect deaths (S1 Fig). 296 In the hemocytes, the selected genes included 15 genes of the antimicrobial response, 2 genes involved in 297 melanization, 5 cellular response genes, 2 lectins and one IMPI-like gene. At 5 hpi, only 2 genes, encoding 298 a lebocin antibacterial (52) AMP (Lebocin 2) and the negative regulator Pirk of the Imd pathway (129), 299 were found to be significantly upregulated. However, most of the selected genes that are strongly induced 300 at later time points also presented positive log2 fold changes at this time point (Fig 2A). From 10 to 20 hpi, 301 all selected genes but few exceptions (cecropin D, Tg-like and Integrin -like) due to biological variability 302 were significantly upregulated at each time point (Fig 2A). Clustering analyses based on Pearson 303 coefficients however revealed 3 distinct clusters of covariations. The first one contained 13 genes belonging 304 to all the categories cited above and corresponded to very stable induction patterns (Fig 2A). The second 305 one, which contained 8 genes involved antimicrobial and cellular responses plus the selected C-type lectin 306 (CLECT (ccBV)), corresponded to slightly increasing patterns (Fig 2A). Finally, the third one, which 307 contained the Relish and Pelle members of the Imd and Toll pathways (22), an integrin and the DDC 308 melanization enzyme (130) genes, corresponded to slightly decreasing patterns (Fig 2A). 309 In the fat body, the selected genes included 15 genes of the antimicrobial response, 2 genes involved in 310 melanization, one galectin gene (Galectin 1) and an IMPI-like gene (IMPI-like 3). At 5 hpi, all 7 selected 311 AMPs, PGRP-S1 and Galectin 1 were found to be upregulated ( Fig 2B). All these genes were among the 312 most strongly overexpressed at later time points. Such as in the hemocytes, most of the selected genes were 313 then significantly upregulated from 10 to 20 hpi ( Fig 2B). In this tissue, the genes only subdivided into two 314 main covariation clusters: a cluster of genes with stable induction patterns and a cluster of genes with 315 increasing induction patterns. The first cluster contained 10 genes of which 8 were involved in antimicrobial 316 responses, one encoded a melanization-related serine protease (Snake-like 2) and one encoded the Galectin 317 1 (Fig 2B). The second cluster contained 9 genes, of which 7 were involved in antimicrobial responses, one 318 encoded the DDC melanization enzyme (130) and the last one encoded the IMPI-like 3 (Fig 2B). 319 Altogether, the results obtained for the two tissues show that most of the transcriptional immune responses 320 induced at 15 hpi take place between 0 and 10 hpi, which is comparable to timings observed in other 321 interaction models (131)(132)(133). The results also indicate that these responses are globally stable across the 322 time post-infestation despite some distinct gene induction patterns in each category of response.

323
Interestingly, while we were hoping to discriminate between an early response, probably activated by the 324 nematode presence, and a later response, probably reacting to bacterial growth, we did not find any clear  In the hemocytes, 14 genes presented higher induction levels in response to X. nematophila than in response 340 to the axenic nematode (Fig 3). In the antimicrobial category, they included the negative regulator Pirk of 341 the Imd pathway (129), all the selected attacin, cecropin, gloverin, lebocin and gallerimycin AMPs, the 2 342 selected PGRP-S, and also probably the Imd pathway transcription factor Relish (22) (Fig 3A). As indicated 343 above, the Imd pathway, as well as the attacin, cecropin and gloverin AMP families, are known to take part  Once again, all of these genes are susceptible to play a part in an immune response to a pathogenic 351 bacterium even though most of them could act on diverse types of invaders. Surprisingly, we found that X. 352 nematophila strongly over-induces the transglutaminase (Tg-like) putative clotting factor (85) (Fig 3C).

353
This result could suggest that the bacterium is actually the main responsible for tissue damages at this time 354 point and/or that Tg-like expression is induced in response to bacteria. Importantly, this result is in 355 agreement with the study of Yadav and colleagues (43), who showed that the D. melanogaster Fondue 356 clotting factor was induced after infestation by the NBC but not after infestation by axenic nematodes.

357
Remarkably, most of the genes that were mostly induced by X. nematophila presented higher induction 358 values in response to the bacterium alone than in response to the whole NBC. However, this observation 359 cannot be directly interpreted as an antagonistic effect of the nematode partner since it could be due to 13 360 changes in the relative proportions of each hemocyte subtype, which would not necessarily reflect absolute 361 variations in their numbers. In addition, the nematode partner specifically induced the overexpression of 362 the selected C-type lectin (CLECT (ccBV)) and was probably the main inducer of the Galectin 1, the 363 tetraspanin D76 homolog (Tsp-like 3) and the selected diapausin AMP (Diapausin 5) (Fig 3A, 3C and 3D). 364 As mentioned before, the M. sexta tetraspanin D76 is known to take part in encapsulation (83)  factor (Kr-like factor 1), were similarly induced by each of the three pathogens (Fig 3A, 3B and 3C), 371 suggesting that these responses are induced by the 2 partners without any additive effect.

372
In the fat body, statistical analysis of the results firstly revealed that the induction levels of Pirk as well as 373 of the selected cecropin and gloverin AMPs were significantly lower in response to the axenic nematode 374 than in response to the NBC and to X. nematophila (Fig 4A), suggesting the bacterial partner is the main 375 responsible for their inductions. In addition, despite non-significant statistics, the results for the selected 376 attacin AMP, PGRP-S6 and GNBP3 showed similar induction patterns ( Fig 4A). As for the hemocytes, the 377 induction patterns of Pirk and of the attacin, cecropin and gloverin AMPs suggest that the fat body's 378 antimicrobial response to X. nematophila is well adapted to the type of pathogen that is met. On the 379 contrary, the induction levels of the melanization-related serine protease (Snake-like 2) was significantly 380 lower in response to X. nematophila than in response to the NBC and to the axenic nematode (Fig 4B), 381 suggesting that the nematode partner is the main responsible for its induction. Similar induction patterns 382 were obtained for the Toll pathway members Toll and Cactus (22) as well as for Galectin 1 (Fig 4A and  gallerimycin AMP, PGRP-S1 and the DDC melanization enzyme, which presented a lesser induction when 392 each NBC partner was injected alone (Fig 4A and 4B). These results suggest synergistic effects of the 393 nematode and of the bacterium on the induction of these genes.

394
In summary, we found in the 2 tissues that most of the selected genes presented partner-specific induction  Table). 416 In order to learn more about their putative functions, we decided to analyse, as we did for the known 417 immune genes, their induction patterns across the time post-infestation and in response to each NBC partner 418 in the corresponding tissues. In both cases, we found that the induction dynamics of the genes were very 419 similar to those of immune genes, with an upregulation that becomes significant at 5 or 10 hpi and with 420 globally stable induction patterns from 10 to 20 hpi (Fig 5A and 5B). 421 In the case of the GBH cluster, the results that we got for the 2 NBC-responsive genes (GBH1 and GBH3) 422 in the hemocytes indicate that they are significantly less induced after axenic nematode injection than after 423 NBC and X. nematophila injections, suggesting that the bacterium is the main responsible for their up-424 regulation (Fig 5C). We could hypothesize an acquisition by horizontal gene transfer from bacteria of the 425 GBH genes. In this case, their putative involvement in the antibacterial immune response would be 426 particularly interesting, since bacterial genes hijacking for immune purpose has only been reported once in 427 metazoans, in the tick Ixodes scapularis (136). Such a hypothesis however requires functional confirmation.

428
In the case of the Unk cluster, we found that the 4 most induced genes in the fat body (Unk2 to 5) are all 429 strongly and similarly induced by the NBC and by the axenic nematode whereas they are not induced by 430 X. nematophila (Fig 5D). The results are very similar for the least expressed Unk gene (Unk1), for which 15 431 we only found a significant induction for the injection of axenic nematodes (Fig 5D). This partner-specific 432 induction pattern suggests the Unk genes are involved in specific aspects of the insect responses to the 433 infestation. In addition, the putative involvement of the Unk genes in the response towards the nematode 434 partner seems to be in agreement with their early mobilization during the infectious process and with their 435 overexpression in the midgut, which is the entry site of the nematode. In our previous study, we had combinations of X. nematophila-and S. carpocapsae-induced responses that seem to be well adapted to the 449 nature of each partner (Fig 6). Continuing this work with more functional and mechanistic approaches is now required to get an accurate 466 picture of the molecular dialogue between the NBC and the immune system. In the longer term, such 467 approaches could help to identify the precise causes of the immune system's failure against this NBC and 468 thus the conditions that are required for an adequate use of this NBC against insect pests.   PBS and dried on paper towel before being placed in 12-well plates. The pathogens efficacies were checked 534 by monitoring 12 control and 12 infected larvae's survival for 72 h after infestation or after injection.

535
Production and storage of bacterial symbionts 536 X. nematophila strain F1 isolated from nematobacterial complexes strain SK27 was conserved at -80°C.

537
Within 3 weeks before each experiment, they were grown for 48 h at 28°C on NBTA with erythromycin 538 (15 µg/mL). The colonies were then conserved at 15°C and used for overnight culture at 28°C in 5 mL   The primers (S3 Table) were designed with the Primer3Web tool (148). Their efficiency was estimated by     Acknowledgments 629 We thank the quarantine insect platform (PIQ), member of the Vectopole Sud network, for providing the 630 infrastructure needed for pest insect experimentations. We are also grateful to Clotilde Gibard and Gaëtan 631 Clabots for maintaining the insect collections of the DGIMI laboratory in Montpellier. This work was 632 supported by grants from the French Institut National de la Recherche Agronomique.