Identification of the Gene Repertoire of the IMD Pathway and Expression of Antimicrobial Peptide Genes in Several Tissues and Hemolymph of the Cockroach Blattella germanica

Antimicrobial peptide (AMP) genes, triggered by Toll and IMD pathways, are essential components of the innate immune system in the German cockroach Blattella germanica. Besides their role in killing pathogenic bacteria, AMPs could be involved in controlling its symbiotic systems (endosymbiont and microbiota). We found that the IMD pathway was active in the adult female transcriptomes of six tissues (salivary glands, foregut, midgut, hindgut, Malpighian tubules and fat body) and hemolymph. Total expression of AMP genes was high in hemolymph and salivary glands and much lower in the other sample types. The expression of specific AMP genes was very heterogeneous among sample types. Two genes, defensin_g10 and drosomycin_g5, displayed relevant expression in the seven sample types, although higher in hemolymph. Other genes only displayed high expression in one tissue. Almost no expression of attacin-like and blattellicin genes was observed in any sample type, although some of them were among the genes with the highest expression in adult female whole bodies. The expression of AMP genes in salivary glands could help control pathogens ingested with food and even determine gut microbiota composition. The low expression levels in midgut and hindgut are probably related to the presence of beneficial microbiota. Furthermore, a reduction in the expression of AMP genes in fat body could be the way to prevent damage to the population of the endosymbiont Blattabacterium cuenoti within bacteriocytes.


Introduction
Most species of cockroaches (Blattodea) live in tropical/subtropical regions, and only a few interact with humans and may be considered pests [1]. The German cockroach Blattella germanica (Linnaeus) is a pest that affects human environments around the world. It impacts human health in several ways, including the transmission of pathogens or by disseminating antibiotic-resistant bacteria. How B. germanica, like other cockroaches, has evolved to become a human pest, adapted to unhealthy habitats, is of great interest. Genomics resources are providing molecular understanding of adaptation and immune response. The genomes of some species of cockroaches such as Periplaneta americana (Linnaeus) [2] or B. germanica [3], in combination with the analysis of transcriptomes [4], have made it possible to identify the expansion of several gene families encoding antimicrobial peptide (AMP) genes that would act on pathogenic bacteria.
However, cockroaches also interact with other non-pathogenic or beneficial bacteria. In fact, cockroaches are paradigmatic, as two symbiotic systems coexist in each individual: a bacterial endosymbiont (Blattabacterium cuenoti, hereinafter Blattabacterium) located in included 28 genes (see sequences in Supplementary Materials CDS_IMD_path.fasta and protein_IMD_path.fasta). The sequences of these genes were obtained from adult and nymph transcriptomes. To produce the complete CDS in cad and Dredd_2, it was necessary to obtain the last 3 -codons from the genome sequence [3]. Because of its length, a complete assembled CDS could not be obtained for Bruce, so the sequences included in Supplementary Materials derived from the genome annotation. The reported CDS of only three genes (akirin, ben and Npc2) were identical to the annotated CDS of the genome; in other cases, there were differences of at least one segment, due to the prediction of a different gene structure. The locus tag qualifiers of CDS with complete or partial coverage are indicated ( Table 1). The IKKbeta, IKKg and key genes were not identified in either the transcriptomes or the genome. NOT Kenny † The annotated CDS in the genome and those obtained from the transcriptomes differ in more than one exon in most cases. ‡ Because a complete transcript could not be obtained, the CDS was taken from the genome. § The last codons of the CDS were recovered from the genome sequence. # The transcript is located in two genome places with 100% identity and 100% coverage.

Expression of IMD Pathway Genes in Six Tissues and Hemolymph of B. germanica
We analyzed the levels of expression of 28 IMD pathway genes in female adult tissues (fat body, foregut, hindgut, Malpighian tubules, midgut and salivary glands) and hemolymph by estimating the means of the normalized expression (GeTMM) (Figure 1).

of 15
# The transcript is located in two genome places with 100% identity and 100% coverage.

Expression of IMD Pathway Genes in Six Tissues and Hemolymph of B. germanica
We analyzed the levels of expression of 28 IMD pathway genes in female adult tissues (fat body, foregut, hindgut, Malpighian tubules, midgut and salivary glands) and hemolymph by estimating the means of the normalized expression (GeTMM) (Figure 1).  Table 1 for product names. Y-axis is in log scale. The lower ends of the error bars are not shown below 10 0 . Genes are arranged in alphabetical order from left to right and top to bottom.
Most of the genes were expressed in the seven sample types. For example, the rel gene, encoding Relish, a key downstream transcription factor involved in the expression of AMP genes [26], is transcribed at a comparable level in the six tissues and slightly higher in hemolymph. Other genes, such as dsp1, eff and SkpA, are highly transcribed in all of the analyzed samples over 10 2 GeTMM.  Table 1 for product names. Y-axis is in log scale. The lower ends of the error bars are not shown below 10 0 . Genes are arranged in alphabetical order from left to right and top to bottom.
Most of the genes were expressed in the seven sample types. For example, the rel gene, encoding Relish, a key downstream transcription factor involved in the expression of AMP genes [26], is transcribed at a comparable level in the six tissues and slightly higher in hemolymph. Other genes, such as dsp1, eff and SkpA, are highly transcribed in all of the analyzed samples over 10 2 GeTMM.
Furthermore, some genes were differentially expressed in one or more sample types (cad, CASP, Duox, Npc2, RYBP and Skp2). For example, the cad gene (encoding Caudal), which in Drosophila functions as a gut-specific transcriptional repressor of NF-κB-dependent induction of AMP genes [27], was almost undetectable in most tissues, except in Malpighian tubules. The CASP gene was only substantially expressed in midgut. It encodes Caspase-3, a member of the caspase family of cysteine proteases, previously described as inducing apoptosis in the midgut of P. americana under starvation conditions [28]. The most highly expressed IMD pathway gene was Npc2 (Niemann-Pick type C-2), but its expression took place mainly in midgut. The most probable reason for such a high expression is its role in the availability of sterol substrate, as is the case for some Drosophila Npc2 genes [29]. In fact, Npc2d was one of the differentially metabolic overexpressed genes detected in Drosophila when comparing normal (with microbiota) to germ-free adult midguts [30].

Expression of AMP Genes in B. germanica Tissues and Hemolymph
The whole expression of AMP genes in six tissues and hemolymph of adult females ( Figure 2) was determined by estimating the means of the total normalized expression (GeTMM). Comparatively, AMP genes were highly expressed in hemolymph and salivary glands, and around two/three orders of magnitude lower in the digestive tract, Malpighian tubules and fat body.
pression took place mainly in midgut. The most probable reason for such a high expression is its role in the availability of sterol substrate, as is the case for some Drosophila Npc2 genes [29]. In fact, Npc2d was one of the differentially metabolic overexpressed genes detected in Drosophila when comparing normal (with microbiota) to germ-free adult midguts [30].

Expression of AMP Genes in B. germanica Tissues and Hemolymph
The whole expression of AMP genes in six tissues and hemolymph of adult females ( Figure 2) was determined by estimating the means of the total normalized expression (GeTMM). Comparatively, AMP genes were highly expressed in hemolymph and salivary glands, and around two/three orders of magnitude lower in the digestive tract, Malpighian tubules and fat body. The average levels of expression of individual AMP genes in each sample type revealed large differences among samples and among genes ( Figure 3). In general, the expression of most AMP genes was very heterogeneous among sample types, although two genes, defensin_g10 and drosomycin_g5, were expressed in the seven samples with GeTMM values > 10 1 . Other genes with expression in most sample types were defensin_g9, defen-sin_g11-g12, drosomycin_g2 and drosomycin_g6. Some genes displayed very high expression (>10 3 GeTMM) in only one sample (hemolymph or salivary glands). However, it was relevant that genes that are relatively highly expressed in whole bodies, such as blattel-licin_g4 and attacin-like genes ( Figure 3), were expressed hardly at all in any of the analyzed tissues. The expression pattern of each AMP gene in the seven analyzed samples is shown in Figure S1. The average levels of expression of individual AMP genes in each sample type revealed large differences among samples and among genes ( Figure 3). In general, the expression of most AMP genes was very heterogeneous among sample types, although two genes, defensin_g10 and drosomycin_g5, were expressed in the seven samples with GeTMM values > 10 1 . Other genes with expression in most sample types were defensin_g9, defensin_g11-g12, drosomycin_g2 and drosomycin_g6. Some genes displayed very high expression (>10 3 GeTMM) in only one sample (hemolymph or salivary glands). However, it was relevant that genes that are relatively highly expressed in whole bodies, such as blattellicin_g4 and attacin-like genes ( Figure 3), were expressed hardly at all in any of the analyzed tissues. The expression pattern of each AMP gene in the seven analyzed samples is shown in Figure S1.
Even though the expression of AMP genes in fat body was relatively low ( Figure 5), they were analyzed because it is an important organ that harbors bacteriocytes, specialized cells containing the ancient bacterial endosymbiont Blattabacterium, and in some insects the fat body is an important organ involved in the production of AMPs for the hemolymph [31]. The average expression of attacin-like_g2 and attacin-like_g3 was due to one out of the four replicas. In this, the GeTMM value was higher than 10 2 , while in the three others, the value was smaller than 10 1 . Although low, fat body was the sample type with the highest expression level of some genes such as drosomycin_g9, termicin_g2 and termicin_g3 ( Figure S1). All expressed genes, except defensin_g1 and defensin_g2, show a differential expression between the two sample types (|log2FoldChange| > 2). Differential expression was significant (adjusted p-value < 0.001) except for defensin_g7, defensin_g8 and drosomycin_g9.
Even though the expression of AMP genes in fat body was relatively low ( Figure 5), they were analyzed because it is an important organ that harbors bacteriocytes, specialized cells containing the ancient bacterial endosymbiont Blattabacterium, and in some insects the fat body is an important organ involved in the production of AMPs for the hemolymph [31]. The average expression of attacin-like_g2 and attacin-like_g3 was due to one out of the four replicas. In this, the GeTMM value was higher than 10 2 , while in the three others, the value was smaller than 10 1 . Although low, fat body was the sample type with the highest expression level of some genes such as drosomycin_g9, termicin_g2 and termicin_g3 ( Figure S1).
Most AMP genes were not expressed in foregut, midgut, hindgut and Malpighian tubules ( Figure S2). Hindgut and midgut were the tissues with the lowest total AMP gene expression (Figure 2). In both cases, only the widely expressed drosomycin_g5 and defensin_g10 displayed an expression level higher than 10 1 ( Figure S2). Most AMP genes were not expressed in foregut, midgut, hindgut and Malpighian tubules ( Figure S2). Hindgut and midgut were the tissues with the lowest total AMP gene expression ( Figure 2). In both cases, only the widely expressed drosomycin_g5 and defen-sin_g10 displayed an expression level higher than 10 1 ( Figure S2).

Discussion
AMPs are the main effector molecules of the innate immune system in insects and enable them to fight microbial infections (bacteria, fungi and viruses). The expression of the different AMP genes is triggered as a consequence of the detection of the invasive microorganisms (infecting microbes) by two major NF-kB-mediated signaling pathways: IMD (which provides protection mainly against Gram-negative bacteria and some Grampositive bacteria) and Toll (which provides protection mainly against fungi and other Gram-positive bacteria) [32]. Two main modes of producing AMPs appear to exist in insects. While in holometabolous insects (with complete metamorphosis), AMPs are mainly synthetized in the fat body and secreted to the hemolymph, in hemimetabolous insects (with incomplete metamorphosis), AMP genes expression takes place mainly in the hemocytes upon infection [33,34].
B. germanica lives in a hostile environment in the presence of (potentially) pathogenic bacteria and in consequence displays a large arsenal of AMPs. The AMP gene repertoire of B. germanica (39 genes) has derived from the expansion of the five types of AMP gene families [4,35]. Comparatively, this species is among those with the largest number of AMP genes, such as the 44 of Nasonia vitripennis (Walker) [36] or the 31 of Musca domestica (Linnaeus) [37] (summarized by [38]), and much more that those reported in P. americana [2]. The phylogenies in B. germanica of the most expanded defensin and drosomycin gene families [4] revealed the presence of both identical (or almost identical) and very divergent

Discussion
AMPs are the main effector molecules of the innate immune system in insects and enable them to fight microbial infections (bacteria, fungi and viruses). The expression of the different AMP genes is triggered as a consequence of the detection of the invasive microorganisms (infecting microbes) by two major NF-kB-mediated signaling pathways: IMD (which provides protection mainly against Gram-negative bacteria and some Grampositive bacteria) and Toll (which provides protection mainly against fungi and other Gram-positive bacteria) [32]. Two main modes of producing AMPs appear to exist in insects. While in holometabolous insects (with complete metamorphosis), AMPs are mainly synthetized in the fat body and secreted to the hemolymph, in hemimetabolous insects (with incomplete metamorphosis), AMP genes expression takes place mainly in the hemocytes upon infection [33,34].
B. germanica lives in a hostile environment in the presence of (potentially) pathogenic bacteria and in consequence displays a large arsenal of AMPs. The AMP gene repertoire of B. germanica (39 genes) has derived from the expansion of the five types of AMP gene families [4,35]. Comparatively, this species is among those with the largest number of AMP genes, such as the 44 of Nasonia vitripennis (Walker) [36] or the 31 of Musca domestica (Linnaeus) [37] (summarized by [38]), and much more that those reported in P. americana [2]. The phylogenies in B. germanica of the most expanded defensin and drosomycin gene families [4] revealed the presence of both identical (or almost identical) and very divergent genes (at CDS level) that may have arisen by old or recent duplications or by events of gene conversion homogenizing the CDS sequences.
The expression of all the genes involved in the IMD signaling pathway has been detected in adult B. germanica transcriptomic projects, with the exception of the IKKbeta, IKKg and key genes (Table 1). While IKKg and key genes are not found in several insect genomes, homologues of IKKbeta are detected in all insect genomes described in the 4IN database, except Acyrthosiphon pisum (Harris).
We observed that the highest levels of total AMP gene expression, among the analyzed tissues, were detected in salivary glands and hemolymph. Most genes display tissue specificity. Those with the highest expression in hemolymph were defensin_g9, defensin_g10, and drosomycin_g5, and those in salivary glands were defensin_g3-g5 and defensin_g15-16.
Although we did not detect the expression of all AMP genes in B. germanica hemolymph, it must be taken into consideration that the samples come from individuals grown under controlled conditions, and that in a challenge situation with diverse pathogenic bacteria, both the levels and the diversity of AMPs produced could be increased.
The large amounts of several AMPs produced by the salivary glands imply that the control of ingested pathogens in the digestive system starts after mixing the saliva with the ingested food. This expression of AMP genes agrees with the idea that anterior parts of the digestive system encounter ingested bacteria first and might need to invoke a stronger response than distal parts [39]. In fact, our results show that in B. germanica, this response is not observed in the different parts of the gut (foregut, midgut or hindgut), where total AMP gene expression is very low. Moreover, some AMPs (symbiotic AMPs) are also produced by eukaryotic hosts during their symbiotic interaction with bacteria, counterintuitively not to kill but to control the symbiotic bacterial population [18,19]. Furthermore, it has been shown that AMPs could act by controlling the microbiota present in the intestine, made up mainly of mutualistic bacteria, determining its composition. In fact, it has been proposed that the immune system may have evolved initially to control interaction with this type of mutualistic bacteria in contact with epithelia, such as the intestine, and was later specialized in the fight against pathogenic bacteria [40].
In Hydra, the composition of the microbiota is determined by AMPs produced by the epithelial cells in contact with the microbiota [41]. In some insects, there is evidence that AMPs determine the composition and structure of the gut microbiota. This is true of Drosophila, where if the expression of some AMP genes is canceled, a change in the microbiota is triggered [20,42]. As compared to Drosophila, where some production of AMPs takes place in the midgut [17], our results demonstrate that in B. germanica, the majority of AMP production in the digestive system is detected in the salivary glands. Similarly, high expression of an AMP gene has also been observed in the salivary glands and midgut of Bemisia tabaci (Gennadius) [43], and some defensin genes are expressed in both salivary glands and midgut, and others are specific to one of these tissues in the tick Ornithodoros turicata (Dugès) [44]. In insects, as a general rule, AMP gene expression is silent in the absence of an immune challenge [45], although RNAseq or qRT-PCR methodologies can detect miniscule levels of gene expression. If the AMPs released by salivary glands at the top of the digestive tract were not able to control most of the ingested microbes, they would induce an immune response in other parts of the system, such as foregut, midgut or hindgut. For that reason, we cannot rule out the possibility that many genes with null or basal expression in our samples could experience a strong increase in expression.
In most insects, the fat body is an organ that can produce large amounts of AMPs during infection that are released to hemolymph [45,46]. Alternatively, in other insects, the largest production of AMPs occurs in the hemocytes, which are secreted to the hemolymph [34]. The low AMP expression levels in B. germanica fat bodies may increase if different types of microbes reach their cells, since in insects it is an immune-responsive tissue [45]. However, we cannot rule out the existence of mechanisms that protect the population of the endosymbiont Blattabacterium in bacteriocytes.
It has been proposed that AMPs are also involved in endosymbiotic associations, controlling the bacterial population and allowing metabolic complementation between the eukaryotic host and the endosymbiont. Symbiotic AMPs can affect endosymbionts through various mechanisms, including interaction with the bacterial membrane (causing disruption or permeation and allowing metabolite leakage) or targeting intracellular machinery and interfering with bacterial metabolism [19]. In the symbiosis between legume plants and nitrogen-fixing Rhizobium bacteria, several hundred AMPs (nodule-specific cysteine-rich peptides) allow the plant to manipulate the physiology of the bacterial symbiont [24]. In insects, the relationship between coleoptericin A (an attacin-related AMP) and the control of a bacterial endosymbiont has been reported in weevils of the genus Sitophilus [21], while in the aphid A. pisum, some cysteine-rich AMP-like polypeptides are specifically expressed in bacteriocytes, cells containing the bacterial endosymbiont Buchnera aphidicola [47], and some of them display antimicrobial activities [48]. In B. germanica, the massive production of AMPs that could carry out this function has not been detected, although the existence of unidentified AMP-like genes, previously undetected [4], cannot be ruled out. The use of various machine-learning algorithms to predict genes with antimicrobial activity in fat body transcriptomes may serve to identify non-canonical AMP-like gene candidates involved in this process.
Finally, in adult whole bodies, attacin-like and blattellicin genes are among the most highly expressed (Figure 3). However, almost no expression of these genes was detected in our study, except a moderate expression of attacin-like_g2 and attacin-like_g3 in one out of the four samples of fat body, perhaps due to a contamination of surrounding tissues. The most probable place for their expression in adult bodies is the abdomen close to the heart, where the periostial hemocytes are attached and induce, after infection, the aggregation of additional hemocytes. This integrated immune and circulatory system was recently described as a general trend in insects [49]. One of the analyzed insects was B. germanica, in which it was detected that hemocytes were located in the dorsal tergum and almost absent in the ventral sternum [49]. However, it has recently been reported that the expression of attacin-like_g1, attacin_like_g2 and blattellicin_g1 genes in B. germanica gut is significantly increased after feeding with live Salmonella enterica serovar Typhimurium [50]. In this work, the expression of these three genes as well as defensin_g1-g2 was also detected at very low levels in control individuals by qRT-PCR (approximate range 10 −4 to 10 −5 ) [50]. The levels of detection by this method compared to our RNAseq analysis were similar (in both cases relative to the expression of EF1A), except for attacin-like_g1, which was not detected at all in our samples (attacin_like_g2 in the three gut sections: range 2 × 10 −3 to 1 × 10 −4 ; blattellicin_g1 in hindgut: 3 × 10 −4 ; defensin_g1-g2 in midgut: 8 × 10 −5 ).
In summary, as we have described, the expression of AMP genes in adult B. germanica is mainly located in hemolymph and salivary glands. We have also shown that although some genes are expressed in most tissues, others are tissue-specific. This differential expression in the digestive system and fat body may be related to avoiding negative effects on the beneficial microbiota and the endosymbiont, while maintaining the removal of pathogens.

IMD Signaling Pathway Characterization
In order to identify the genes that constitute the IMD immune signaling pathway in B. germanica, we performed a search for homologous sequences, using as starting data the Innate Immunity Genes in Insects (4IN) database of the University of Lyon, which is a comprehensive database of innate immunity genes in a diverse range of available insect genomes. The multifasta files obtained after the 4IN search were subjected to multiple alignment with MAFFT v7.475 [51] using the L-INS-i alignment method. Subsequently, three HMMER package 3.3.2 [52] programs were used to perform the homologous sequence search: hmmbuild to create HMM alignment profiles from the multiple alignments, hmmpress to compress the HMM profiles into binary data, and hmmsearch to contrast these profiles with B. germanica gene databases. Sequence searches were performed on the B. germanica genome CDS set, which was obtained from GCA_003018175.1 (PYGN01 project [3]), and on several transcriptomes: SRR5458588.1, SRR5458589.1, SRR5458590.1, SRR5458591.1 (PRJNA382128 project) and SRR6784710 (PRJNA389591 project). All samples corresponded to the Orlando Normal from American Cyanamid strain and to whole-body sequences of females at different developmental stages (nymphs or adults), to cover various gene expression patterns and to identify as many genes as possible. This process was automated with a Python 3.9.5 pipeline (https://docs.python.org/3/reference/, accessed on 13 June 2022) to recursively perform the processing for each gene and for each database used. The final results were confirmed with blastx searches and modified when required.

Transcriptome Assembly
Transcriptomes used as databases in which to search for homologous sequences were assembled from Sequence Read Archive (SRA) projects. The assembling was performed de novo rather than with mapping due to the fact that the genome of B. germanica contains many assembly gaps. Data extraction from SRA projects was performed with fasterqdump (NCBI SRA Toolkit v2.11.0 package). The quality of reads was checked with FastQt v0.2.3 (https://github.com/labsquare/fastQt, accessed on 13 June 2022). Correction of poor quality reads was performed with Rcorrector v1.0.4 [53]. Trimmomatic v0.39 [54] was used to remove residual adapters that may not have been removed in the standard sequencing protocol. Finally, Trinity v2.12.0 [55] was used to perform de novo assemblies with paired-end reads. Data processing was performed on a CentOS v7.9.2009 server.

Review of Results and CDS Gathering
While HMMER was able to detect which sequences had a higher degree of homology in B. germanica, it was not able to identify the proteins unequivocally. To avoid false positives and hits with incomplete sequences, the results were thoroughly checked. The CDS harbored by the individual transcripts were obtained with getorf, from the EMBOSS v6.5.7 package [56], with a minimum sequence size of 300 nucleotides. Incomplete genome sequences were filled in with data from SRA projects. The resulting sequences were aligned to the genome with Splign v2.0.1 [57] and Spidey in Unipro UGENE v37 [58]. The resulting sequences were also aligned to reviewed 4IN sequences using MAFFT. The online version of Pfam v34.0 [59] was used to contrast the Pfam domains of the resulting sequences with those of the reviewed 4IN sequences.

Insect Rearing
A population of B. germanica originating from a laboratory population housed by Dr. X. Bellés's group at the Institute of Evolutionary Biology (CSIC-UPF, Barcelona, Spain) was raised in plastic containers at a temperature of 25 • C and a relative humidity of 60%, in 12/12 h light/dark cycles in climatic chambers at the Institute for Integrative Systems Biology (University of Valencia-CSIC). The diet was based on dog food (Teklad Global 21% protein dog diet, 2021C, Envigo) and water provided ad libitum. Adult individuals were collected from different containers to avoid population bias, between 0 and 48 h after the transition from nymph to adult, which is considered day 0. Only females were selected to avoid bias due to sexual dimorphism, and all the sampling was carried out at day 5 of adult life.

Insect Dissection and Collection of Samples
The cockroaches were subjected to CO 2 sedation and cleaned with 10% bleach, 70% ethanol and double washed with Type II water. They were fixed with entomological needles in supine position on a silicone plate, and the following tissues were extracted: salivary glands (pool of ten individuals, four replicates), foregut (pool of five individuals, four replicates), midgut (pool of five individuals, four replicates), hindgut (pool of five individuals, four replicates), Malpighian tubules (pool of six individuals, four replicates), and fat body (pool of three individuals, four replicates). Samples were collected in previously cooled Eppendorf tubes, treated overnight with RNAlater, frozen the next morning with liquid nitrogen and stored at −80 • C until RNA extraction.
For the collection of hemolymph samples (pool of forty individuals, three replicates), individuals were sedated with CO 2 and fixed with entomological needles in supine position on a silicone plate. Subsequently, one of the legs of the third pair was cut at the beginning of the second segment. Hemolymph was collected by capillarity by massaging the abdominal area. Hemolymph was collected directly in TRI Reagent (hemolymph to reagent ratio 1:3), and samples were frozen in liquid nitrogen and stored at −80 • C until RNA extraction.

RNA Extraction and Sequencing
RNA was extracted from the tissues with the NucleoSpin RNA kit (Macherey-Nagel), while hemolymph RNA was extracted with Direct-zol RNA miniprep kit (Zymo Research), both according to the manufacturer's recommended instructions. Samples were quantified with Nanodrop and Qubit. Sequencing was carried out at the Servei Central de Suport a la Investigació Experimental (SCSIE, University of Valencia). RNA integrity and quality, as well as mRNA enrichment, were analyzed before proceeding with library preparation. Sequencing was performed with an Illumina NextSeq 550, resulting in single-end reads of 150 nucleotides each.

Generation of Normalized Expression Levels
To make a side-by-side comparison of the expression levels of the different AMP genes (genes with identical CDS sequences were grouped) in each of the tissues, the reads were quantified and normalized. Residual adapters were cleaned with Trimmomatic, and poor-quality bases were corrected with Rcorrector, both using standard parameters. Quantification was performed by pseudoalignment with Kallisto v0.46.2 [60], with a mean sequence size of 136 nucleotides and a standard deviation of 10 nucleotides. The index used by Kallisto was generated by adding the mRNA sequences of B. germanica AMPs to the database.
To be able to make both an intersample comparison and an intrasample comparison, gene length corrected trimmed mean of M-values (GeTMM) normalization was used [61]. This normalization accounts for gene length, sequencing depth and RNA composition. An estimation of GeTMM values in all CDS of the B. germanica genome (those annotated in the genome [3] plus the sets of AMPs and IMD pathway genes from this study) showed that, on average, only 0.44% of the genes displayed an expression equal to or higher than 10 3 , 3.42% higher than 10 2 and 19.15% higher than 10 1 . We have used these threshold values to define very high, high and moderate expression. Data processing was performed with R v4.1 and the edgeR v3.34.0 package [62]. Figures were created with ggplot2 v3.3.4 [63] and lattice v0.20-44 [64] in R. In order to compare expression levels in salivary glands and hemolymph, a differential expression analysis was performed using DESeq2 1.36.0 [65] with standard parameters, using the quantification values from Kallisto. AMP genes were considered differentially expressed if |log2FoldChange| > 2, and adjusted p-value cutoff was set to 0.001.