Manduca sexta specimens were obtained from the laboratory stock of the Institute of General and Applied Zoology, Justus-Liebig-University, Gießen, Germany. Larvae were kept individually at 24°C under long-day conditions 16:8 h (light:dark), controlled regularly for health status (growth, feces, melanization), and only healthy individuals were used for experiments. Animals were fed with an antibiotic-free artificial diet, modified after Yamamoto (1969): Whole grain flour 5.27% (w/v), soy meal 5.70% (w/v), agar-agar Kobe I 2.55% (w/v), casein 1.40% (w/v), plant oil 0.16% (w/v), ascorbic acid 30.02 mM, methylparabene 11.54 mM, sorbic acid 15.66 mM, formaldehyde 5.33 mM, CaCO³ 5.28 mM, NaCl 9.03 mM, nicotinic acid 97.47 µM, riboflavin 15.94 µM, folic acid 6.34 µM, pyridoxine 16.55 µM, thiamine hydrochloride 9.31 µM and biotin 9.82 µM.

Escherichia coli K12 culture was obtained from the laboratory stock of the Institute of Microbiology, Academy of Science of the Czech Republic, Prague.

Bacillus thuringiensis ssp. kurstaki and E. coli K12 strain D31 (Monner et al., 1971) were obtained from laboratory stocks of the Institute of General Zoology and Developmental Biology, Justus-Liebig-University, Gießen, Germany.

All bacteria were grown in liquid Standard Nutrient Broth I (C. Roth GmbH, Karlsruhe, Germany) prepared according to manufacturer´s protocol.

B. thuringiensis was grown in liquid culture on a rotary shaker at 180 rpm at room temperature for approx. 15 days until spore density reached approx. 50%. Sporulation was controlled by phase contrast light microscopy. After reaching 50% sporulated bacteria, liquid cultures were stored at 4°C and used within one week for infection experiments.

Immune stimulations were conducted 15 hours prior to RNA isolation. Each treatment was performed three times in independent cohorts of n = 4 L5d1 larvae, i.e. between 24 and 48 hours after moulting to L5. Tissues of each cohort were pooled before RNA isolation.

The growth rate of the larvae during the period of immune stimulation was calculated with the following formula: Growth rate = (body weight_t1 – body weight_t0) × body weight_t0 ^-1; with t₀ = time point of immune challenge (injection or feeding) and t₁ = t₀ + 15 hours.

From each animal analyzed in this study, 50 µl hemolymph was collected at the time point of RNA isolation, frozen immediately, and stored at -80°C.

Lytic activity was quantified by radial diffusion agar assay as outlined by Mohrig and Messner (1968). Activity is expressed in µg/ml hen egg white lysozyme equivalents (HEWLE).

Additionally, bactericidal/bacteriostatic activity assay against live E. coli K12 strain D31 was determined by radial diffusion agar assays as described by Hultmark et al. (1982). Serially diluted Gentamicin served as a reference antibiotic. Antimicrobial activity of the hemolymph is expressed in µg/ml Gentamicin equivalents (GE).

All data were tested for normal distribution by Kolmogorov-Smirnov-test. Non-normal distributed data (growth rate, lytic activity, and antimicrobial activity) were tested for significance (p ≤ 0.05) by the asymptotic Mann-Whitney-U-test (R V. 3.0.3, R Development Core Team, 2008; http://www.R-project.org; package coins, Hothorn et al., 2008, function wilcox_test) and data presented as median (Mdn) and interquartile range (IQR). Box-and-whiskers-plots with median (horizontal line), upper and lower quartile (boundaries), and 1.5-fold interquartile range (whiskers) were created with R V. 3.0.3. Normally distributed data (real-time PCR data) were evaluated for significance by two way ANOVA with post hoc Tukey´s multiple comparison (GraphPad Prism V. 5.0, GraphPad Software, San Diego, California, USA). Differences were considered significant with p ≤ 0.05. All values are presented as mean ± standard deviation. Bar plots of real-time PCR data were created with GraphPad Prism V. 5.0.

Animals were cleaned with water, surface disinfected with 70% (v/v) ethanol, and chilled on ice for sedation. Midgut and fat body tissues were dissected with fine forceps and scissors (contralaterally to injection sites for E. coli K12-injected and saline-injected control) at three regions of each animal: anterior ventrolateral (abdomen segment 1), middle part ventrolateral (abdomen segment 4/5), posterior ventrolateral (abdomen segment 6/7) and washed twice with calcium and magnesium-free Manduca sexta saline (MS-, Willot et al., 1994). Hematopoietic organs (HO) were removed completely (four HOs per animal) and washed twice with MS-. Hemolymph from each group was collected in 6 ml ice cold anticoagulant saline (AC saline, Willot et al., 1994), sedimented at 500 x g for 10 min at 4°C, and the resulting hemocyte pellet was used for RNA isolation.

Whole RNA from fat body, hematopoietic organ, midgut, and hemocytes was isolated with an RNeasy Mini RNA purification kit (Qiagen, Hilden, Germany). Pooled tissue from four animals was used for each isolation. The RNA quality, purity, and concentration were evaluated spectrophotometrically with a NanoDrop 1000 (Thermo Fisher, Waltham, USA) and denaturing RNA electrophoresis. Prior to reverse transcription, the contaminating gDNA was removed by DNAse I digestion (DNAse I RNAse free, Thermo Fisher, Waltham, USA). 550 ng purified RNA of each sample was transcribed to cDNA with SuperScript III Reverse Transcriptase (Life Technologies, Carlsbad, USA) using oligo(dT)20 primers.

Quantitative PCR analysis was performed to determine changes in the mRNA levels coding for the immune molecules immulectin-3, scolexin A, PGRP 1A, and lysozyme. Each reaction was prepared as follows: 12.5 µl iQ^TM SYBR^® Green Supermix (Bio-Rad Laboratories Inc., Hercules, USA), 1 µl of each forward and reverse primers ( Table 1 ), and 4 µl of cDNA diluted in RNase-free water (1:20). PCR was done using a CFX96Touch Real-time PCR Detection System (Bio-Rad Laboratories Inc., Hercules, USA). The amplification protocol was as follows: 3 min at 95°C followed by 40 cycles of 94°C for 30 s, 57°C for 40 s, and 72°C for 70 s. The temperature was then gradually increased to 95°C to obtain melting curves of the amplified fragments, which confirmed the specificity and uniformity of the PCR products. Evaluation of amplification efficiency was performed by dilution series of cDNA forming the standard curve. The expression levels were calculated with 2^-ΔΔCq method (Livak and Schmittgen, 2001). To select suitable housekeeping genes, the gene expression stability was assessed using RefFinder (https://www.heartcure.com.au/reffinder/). Two reference genes (ribosomal protein S3, RPS3; translation elongation factor 1 alpha, EF1A) were selected as the most stable internal controls for the normalization of the expression of other genes. Non-template controls were included in each experiment. All parameters were measured in duplicates. The fold change in the mRNA level was related to the change in the settled controls. For the treatments “B. thuringiensis-fed” untreated L5d2 larvae served as a control group. For the treatment “E. coli K12-injected” the “saline-injected” group was used as a control, which resembles the injection procedure without bacteria. The results were expressed as the mean ± SD of the values obtained in three independent experiments.

Digoxigenin-labeled RNA probes against lysozyme were synthesized from a 1301 bp long DNA template, amplified by PCR (initial denaturation 94°C 3 min; 40 repeats of 94°C 25 s, 60°C 30 s, 72°C 45 s; and final extension of 72°C 7 min) with gene-specific forward and reverse primers and added sequences for the SP6 or T7 promoters, respectively ( Table 1 ). Digoxigenin-labeled RNA probes were synthesized with the DIG RNA labeling kit SP6/T7 (Hoffmann-La Roche AG, Basel, Switzerland) according to the manufacturer’s protocol.

Standard in situ hybridization (ISH) of whole hematopoietic organs and hemocyte monolayers was performed as described elsewhere (hematopoietic organs: von Bredow et al., 2020; hemocytes: von Bredow et al., 2021) with 2 ng/µl riboprobe. Bound riboprobes were visualized using alkaline phosphatase-conjugated sheep anti-DIG Fab (Hoffmann-La Roche AG, Basel, Switzerland) and chromogenic substrate (4 mM nitroblue tetrazolium chloride, 5.5 mM 5-bromo-4-chloro-3-indolyl phosphate).

For fluorescence in situ hybridization (FISH), the riboprobe concentration was 5 ng/µl. Instead of a chromogenic reaction, the sheep anti-DIG Fab AP was labeled with biotin-labeled goat anti-sheep immune serum (Vector Laboratories, Burlingame, USA) followed by signal enhancement with the Vectastain ABC Kit Standard (Vector Laboratories, Burlingame, USA) and the TSA Amplification Renaissance Kit (PerkinElmer Inc., Waltham, USA). Nuclei were stained with 360 nM 4′,6-diamidino-2-phenylindole (DAPI). The specimens were mounted with Fluoromount G (Southern Biotech, Birmingham, USA) and analyzed with a Leica TCS SP2 confocal laser scanning microscope (Leica Microsystems GmbH, Wetzlar, Germany).

For both lysozyme ISH and FISH, a digoxigenin-labeled sense riboprobe specific for a 319 base sequence of PGRP-1A mRNA was used as negative control since we were not able to synthesize a sufficient amount of digoxigenin-labeled lysozyme sense riboprobe. Control reactions were performed for each hybridization i) by incubation with digoxigenin-labeled PGRP-1A sense riboprobes to rule out non-specific binding of riboprobes, and ii) by omitting any riboprobes to rule out non-specific antibody binding, endogenous alkaline phosphatase activity (ISH), and non-specific fluorochrome adhesion or autofluorescence (FISH).

Hematopoietic organs and hemocyte monolayers were immunolabeled with rabbit-anti-Hyalophora cecropia lysozyme immune serum (Trenczek and Faye, 1988), diluted 1:1000 in 3% BSA in TBS. Hematopoietic organs were additionally labeled with mouse monoclonal antibody MS75 (Beetz et al., 2004). Nuclei were stained with 360 nM DAPI. For a detailed immunolabeling protocol, please see von Bredow et al., 2020. The slides were mounted with Fluoromount G^® and analyzed by confocal laser scanning microscopy.

The increase in body weight of larvae served as an indicator of the overall health status. Injection of E. coli K12 did not alter the growth rate significantly when compared to saline-injected control group (growth rate E. coli-injected Mdn 0.22, IQR 0.06; growth rate saline-injected Mdn 0.29, IQR 0.11; p = 0.273, Mann-Whitney-U-test; Figure 1 ). Infection with B. thuringiensis caused a loss of body weight within 15 hours after the first pathogen uptake, therefore resulting in negative growth rates differing significantly from untreated controls (growth rate B. thuringiensis-fed Mdn = -0.04, IQR 0.03; growth rate untreated Mdn 0.37, IQR 0.13; p = 0.000, Mann-Whitney-U-test; Figure 1 ).

Lytic activity of the hemolymph was observed for all treatment and control groups ( Figure 2A ). Both injection of E. coli K12 and feeding B. thuringiensis resulted in increased lysozyme activity in the hemolymph compared to the respective control. The lytic activity (µg/ml HEWLE) of E. coli K12-injected animals was significantly higher compared to saline-injected controls (E. coli-injected, Mdn 2090 µg/ml HEWLE, IQR 706; saline-injected Mdn 825 µg/ml HEWLE, IQR 582; with p = 0.000, Mann-Whitney-U-test; Figure 2A ). In B. thuringiensis-fed animals, the lytic activity was also significantly higher than in untreated controls (B. thuringiensis-fed, Mdn 1189 µg/ml HEWLE, IQR 789; untreated, 458 µg/ml HEWLE, IQR 929; with p = 0.016, Mann-Whitney-U-test; Figure 2A ).

The hemolymph antibacterial activity against live E. coli K12 strain D31 was measured for each treatment. Most hemolymph samples of untreated animals and saline-injected animals did not inhibit bacterial growth (n = 7 of 12 for each control cohort). Conversely, inhibition was observed for all samples from both E. coli K12-injected and B. thuringiensis-fed animals (n = 12 of 12 for each treatment cohort) ( Figure 2B ). The antibacterial hemolymph activity was significantly increased in both bacterial treatments compared to the respective control (E. coli K12-injected Mdn. 13.36 µg/ml GE, IQR 7.06; saline-injected Mdn. 0.00 µg/ml GE, IQR 7.76; with p = 0.000, Mann-Whitney-U-test; B. thuringiensis-fed Mdn. 10.17 µg/ml GE, IQR 5.45; untreated Mdn. 0.00 µg/ml GE, IQR 9.37; with p = 0.013, Mann-Whitney-U-test; Figure 2B ).

Animals treated the same way as for the qPCR experiments were tested for bacterial hemocoel invasion 15 hours after feeding B. thuringiensis, and survival of injected E. coli K12 in the hemocoel, respectively. Hemolymph samples were spread on nutrient agar plates and the number of colony-forming units (CFUs) was determined. Colonies occurred rarely in untreated samples or saline-injected animals. They most likely originated from contamination during the sampling procedure ( Supplementary Figure 1 ).

Bacteria colonies occurred only seldom in samples of B. thuringiensis-infected animals. The morphology of the colonies resembled that observed for B. thuringiensis colonies in one case (data not shown) with 40.4 CFU/µl recovered from the hemolymph. This was the only case (of four in total) that indicates an invasion of microorganisms from the gut after experimental gut infection ( Supplementary Figure 1 ). A second sample exhibited a very low CFU count (0.2 CFU/µl), which is within the range of CFUs recovered in untreated controls ( Supplementary Figure 1 ).

Since the extent of gut damage is likely proportional to the concentration of B. thuringiensis toxins ingested, we administered two additional cohorts (n = 4 animals each) with higher B. thuringiensis concentrations (5×10⁶ and 1×10⁷ spores per 1cm³ diet cube). No CFUs with the typical morphological appearance of B. thuringiensis occurred in any sample, indicating that only rarely (i.e. in one of twelve cases across all concentrations) an invasion of gut bacteria into the hemocoel appears within 15 hours after B. thuringiensis infection.

Fifteen hours after E. coli K12-injection CFUs resembling the morphology and size of E. coli K12 colonies were isolated from the hemolymph in various densities (0.8, 1.2, 3, and 55.2 CFU/µl) from n = 4 animals tested ( Supplementary Figure 1 ). These data suggest that injected E. coli K12 were not completely cleared from the hemocoel within 15 hours, indicating a persisting experimental infection in the hemocoel.

During preparation, animals were examined for the following parameters: Body rigidity, melanotic capsules in the body cavity, and integrity of the peritrophic membrane of the gut.

Untreated controls, saline-injected controls, and E. coli K12-injected animals were turgescent whereas B. thuringiensis-fed animals were flabby and responded with convulsive muscular contractions to tactile stimuli. Melanotic capsules were present in the hemocoel of each E. coli K12-injected larva, but absent in untreated, saline-injected, and B. thuringiensis-fed animals. The peritrophic membrane of the gut was stable in untreated controls, saline-injected controls, and E. coli K12-injected animals, allowing the removal of the gut content by taking out the peritrophic membrane tube. However, in B. thuringiensis-fed animals the peritrophic membrane was fragile and disintegrated into small fragments during preparation.

In response to B. thuringiensis infection each examined tissue increased expression of one or more immunity-related genes ( Figure 3A ). Hematopoietic organs exhibited highly significantly increased PGRP-1A mRNA levels (10.96 ± 3.34, p < 0.0001). Hemocytes responded by slightly but significantly increased lysozyme mRNA level (1.89 ± 0.29, p = 0.047) ( Figure 3A ). The levels of PGRP-1A (17.37 ± 8.45, p = 0.0317) and scolexin A mRNA (19.47 ± 12.65, p = 0.014) were significantly increased in the fat body ( Figure 3A ). In the midgut tissue, only lysozyme mRNA level was significantly increased (9.52 ± 2.14, p = 0.034) ( Figure 3A ). IML-3 mRNA level was below the detection threshold in midgut tissue (Ct > 35 for both untreated control and B. thuringiensis-fed animals).

Fifteen hours after E. coli K12-injection significant differences in mRNA levels occurred in hematopoietic organs and the fat body. The hematopoietic organs increased the mRNA level of lysozyme (3.03 ± 0.44, p < 0,0001), PGRP-1A (2.63 ± 0.63, p = 0.0002), and scolexin A (7.18 ± 0.71, p < 0,0001) significantly after this systemic immune challenge ( Figure 3B ). The fat body responded with a remarkably increased scolexin A mRNA level (88.29 ± 47.65, p = 0.004) ( Figure 3B ). In the midgut, again IML-3 mRNA was not detectable (Ct > 35 for both saline-injected and E. coli K12-injected animals).

All values are given as fold expression mean ± SD; and level of significance after 2-way ANOVA with post hoc Tukey´s multiple comparison.

The localization of lysozyme mRNA and lysozyme protein was determined on hemocytes and hematopoietic organs by in situ hybridization and antibody labeling ( Figure 4 ). Based on published data (Mulnix and Dunn, 1994) to enhance the signal strength for in situ hybridization, some animals were wounded or injected with E. coli K12 15 hours before tissue extraction. The antibody labeling was conducted only on untreated animals.

mRNA coding for lysozyme was detected in circulating plasmatocytes and granular cells ( Figure 4A ). The signal appeared to be stronger in plasmatocytes than in granular cells when alkaline phosphatase was used as a reporter ( Supplementary Figure 2A ), indicating plasmatocytes as the main lysozyme synthesizing hemocyte type. The specificity of the riboprobe has been confirmed by a lack of signals when the cells were incubated with a control (i.e. non-mRNA specific) digoxigenin-labeled riboprobe ( Supplementary Figure 2 ). In the hematopoietic organs only a few cells exhibited an extraordinarily strong signal for lysozyme mRNA whereas most cells showed no signal ( Figure 4B ).

The lysozyme protein is localized in circulating hemocytes in granular cytoplasmic inclusions in both plasmatocytes and granular cells ( Figure 4C ). Similarly, the protein was localized in granular cytoplasmic inclusions in most, but not all hematopoietic organ cells ( Figure 4D ).

The B. thuringiensis toxin Cry1A impairs the integrity of the gut epithelium and subsequently allows gut microbiota to enter the hemocoel. The killing mechanism of Bacillus thuringiensis is not finally solved to date, and observations from different species contradict each other. For example in Lepidoptera, injection of B. thuringiensis spores is deadly even in small quantities for M. sexta (Johnston and Crickmore, 2009; own observations), while gut-mediated B. thuringiensis infection fails to kill the host in the absence of enterobacteria in the moth Lymantria dispar (Broderick et al., 2006).

The translocation of gut bacteria to the hemocoel is not immediate. Mason et al. (2011) found that the larval M. sexta gut becomes permeable for gut bacteria within 24 to 48 hours after feeding Cry1A toxin, which fits our observations that 15 hours after feeding B. thuringiensis no colonization of the hemolymph appeared ( Supplementary Figure 1 ). To elicit a strong immune response of the midgut we infected larvae with a sublethal dose of sporulated live B. thuringiensis ssp. kurstaki. The experimental infection was not lethal within 15 hours but resulted in a total stop of food uptake within one hour and a severe and significant loss of body weight (negative growth rate, Figure 1A ). Immune challenges decrease the growth rate of M. sexta (Adamo et al., 2007; Mason et al., 2011). Additionally, lysozyme activity ( Figure 2A ) as well as antimicrobial activity ( Figure 2B ) were significantly increased, and the integrity of the peritrophic membrane was impaired. These data indicate that the B. thuringiensis infection model was sufficient to study the effects of a severe gut inflammation. Furthermore, absence of melanotic capsules in the hemocoel and lack of CFUs from hemolymph samples ( Supplementary Figure 1 ) indicate that the infection can be characterized as a severe gut infection without septic injury of the gut.

On the transcript level, the local effect was restricted to increased lysozyme expression; increased levels of mRNA coding for scolexin A and PGRP-1A were not significant. Even though the gut infection did not provoke sepsis, mRNA levels of hemocoelic tissues were altered in response to B. thuringiensis infection. The pattern recognition receptor PGRP-1A mRNA synthesis was upregulated in the hematopoietic organ and fat body ( Figure 3A ), and lysozyme mRNA synthesis was upregulated in hemocytes ( Figure 3A ).

An increased lysozyme mRNA synthesis in hemocytes after exposure to B. thuringiensis toxins is described for the cabbage looper Trichoplusia ni, too (Tamez-Guerra et al., 2008). In D. melanogaster larvae, oral application of the entomopathogenic bacteria Erwinia carotovora ssp. carotovora strain 15 (Basset et al., 2000) or Pseudomonas entomophila (Vodovar et al., 2005) led to the increased expression of antimicrobial peptides in gut tissue (local contact zone) as well as in the fat body, indicating cross-talk between the gut and the hemocoel. These data confirm the involvement of both local tissue and tissue not in direct contact with the gut pathogens to react to the gut-borne infection before a severe invasion of the hemocoel occurs.

How the infection is signaled from the gut epithelium to the hemocoel is unknown. Two principal ways how a gut-borne infection may be sensed and the signal transmitted to the hemocoel in insects are known. (1) Microorganisms or immunogenic molecules may cross the barrier of the infected gut as proposed for PGN fragments (Zaidman-Rémy et al., 2006), or (2) an active signaling process of the gut epithelium is involved. As an example of the latter, it was shown in D. melanogaster that nitric oxide (NO) triggers AMP synthesis in the fat body via the Imd-pathway after the gut comes in contact with pathogenic bacteria (Foley and O'Farrell, 2003). Nitric oxide plays also a role in the local response of the midgut to enterobacteria Photorhabdus luminescens in M. sexta, leading to a reduced number of pathogens in the hemolymph and the gut when upregulated (Eleftherianos et al., 2009). This dual role as an antimicrobial substance and signaling molecule is a promising model to explain the transmission of the infection signal sent by the inflamed gut to the hemocoelic organs.

However, it is imaginable that sensing an unusually high load of microorganisms in the midgut may trigger a preemptive immune activation. To test this hypothesis, we administered live E. coli K12 per os and determined changes in the immune gene mRNA levels. Only a slight increase of PGRP-1A mRNA level occurred, and scolexin A mRNA expression of the midgut was decreased ( Supplementary Figure 3 ). In another set of experiments, E. coli supplemented artificial diet even increased the growth rate of younger (L2/L3) instars (Trenczek, own observation). This indicates that an unusually high abundance of microbes alone may not be sufficient to trigger a (preventive) immune response, but that increased immune activity in the midgut and hemocoel occurs only in response to a true infection or inflammation.

To simulate a systemic infection, e.g. as it might occur after gut leakage provoked by B. thuringiensis toxins, or when microorganisms enter hemocoel through an epidermal wound, we injected live E. coli K12 into the hemocoel. Injection of B. thuringiensis ssp. kurstaki through a cuticular wound turned out to be fatal for M. sexta larvae (data not shown; cf. Johnston and Crickmore, 2009) and was therefore not suitable for our experiments. E. coli is a typical gut commensal and shares characteristics of its cell wall composition (DAP-type PGN) with B. thuringiensis (Kingan and Ensign, 1968). The injection procedure itself (saline-injection) lowered the larval growth rate while the lytic and antibacterial activity of the hemolymph did not increase. Hence, we conclude that in our infection model, wounding and saline-injection lead to a physiological response but only minor immunological changes (as measured in activity changes).

The injection of E. coli K12 resulted in the formation of numerous melanotic capsules (noduli) found throughout the body cavity. Moreover, bacteria were recovered in high numbers from the hemolymph 15 hours post injection ( Supplementary Figure 1 ). Additionally, the lysozyme activity and the antimicrobial activity of the hemolymph increased significantly ( Figures 2A, B ). On the other hand, neither the body weight growth rate ( Figure 1 ) nor the integrity of the peritrophic membrane was affected, as opposed to B. thuringiensis-fed animals. Taken together, the injection of live E. coli K12 provoked a similar increase in both the lytic activity and the antimicrobial activity of the hemolymph compared to the gut infection, while the peritrophic membrane and growth rate remained largely unaffected.

In the hematopoietic organ, mRNA coding for lysozyme, PGRP-1A, and scolexin A was increased, the hemocytes increased lysozyme mRNA synthesis, and an elevated level of mRNA coding for lysozyme and scolexin A was detected in the fat body, the latter being dramatically increased compared to saline-injected controls ( Figure 3B ). Most importantly, other than in B. thuringiensis-infected animals, the midgut did not alter the mRNA expression of the observed immune-related genes significantly.

The midgut did not react with significant mRNA level alterations on E. coli K12-injection into the hemocoel, while it did on a local infection of the midgut itself. On the other hand, cells located within the hemocoel respond to both midgut and systemic infections. Therefore, we propose the activation of organs in the hemocoel in response to a gut-borne infection before direct contact of the hemocoel with the pathogen occurs, which is mediated through a yet not analyzed pathway. Either immunogenic molecular patterns cross the gut epithelium and elicit the immune response of the fat body, the hemocytes, and the hematopoietic organs or signaling takes place through mediators, probably released by the gut epithelium, activating these organs. Conversely, an infection of the hemocoel mimicked by injection of E. coli K12 did not alter the mRNA level of the observed genes in the midgut. Data from other insects, however, are contrary to our findings. In the common bedbug Cimex lectularius (Heteroptera, Cimicidae) for example, the expression of defensin genes of the gut tissue was increased in response to a hemocoelic injected bacterial mixture (Meraj et al., 2021). Similarly, in the mosquito Anopheles stephensi (Diptera, Culicidae), an injection of E. coli K12 into the hemocoel increased AMP expression in the midgut (Das De et al., 2018). Since both species are hematophagous, with often long starving periods between blood meals, their gut defense may play a more important role in systemic infections than in the herbivorous, constantly feeding M. sexta larva. We do not state that the midgut of M. sexta does not react at all to a systemic immune threat, e.g. by increased AMP expression as was shown for larvae inoculated with a mixture of heat-killed bacteria and fungi mixture (McMillan and Adamo, 2020) but rather suggest to compare the gut response against systemic infections in a broad range of insect species from different ecological niches. Especially detritivores such as many Diptera larvae, feeding on a diet with a high load and diversity of microbes, and parasites that may encounter highly adapted pathogens and parasites through food uptake, e.g. mosquitos, can be expected to exhibit stronger involvement of the gut immune system to both gut and systemic infections as well as a stronger anticipation of systemic immune events after sensing gut mediated infections.

The hemocytes of insects possess lysozyme-containing phagosomes (Zachary and Hoffmann, 1984) and secrete lysozyme (Anderson and Cook, 1979). Lysozyme may therefore be a marker for immune-competent hemocytes. Interestingly, we found lysozyme to be expressed in granular cells and plasmatocytes, the most abundant hemocyte types in M. sexta ( Figures 4A, C ), both cell types involved in phagocytosis and encapsulation (Horohov and Dunn, 1983; Levin et al., 2005; Ling and Yu, 2006). The hematopoietic organ, however, forms and releases plasmatocytes (Nardi et al., 2003; von Bredow et al., 2021). Because some of the cells of the hematopoietic organ contain a higher amount of lysozyme mRNA these cells likely represent mature plasmatocytes ( Figure 4B ). Similarly, not all cells of the organ contain lysozyme protein ( Figure 4C ). Upregulation of lysozyme expression in response to a systemic infection indicates either a putative role of the hematopoietic organs in the humoral defense by secreting lysozyme or an increased hemocyte maturation, forming more immune competent cells (plasmatocytes) in response to an infection.

Hematopoietic organ cell lysates exhibited lytic activity (data not shown), further supporting the idea of the role of the organs in humoral immune activity.

Research on the roles of hematopoietic tissues in insect immunity has been restricted mostly to increased proliferation (e.g. Duressa et al., 2015), differentiation of specific hemocyte types (e.g. Lanot et al., 2001), or its phagocytic properties (e.g. Nutting, 1951; Hoffmann, 1970). However, to our knowledge, no studies concerning changes in the expression of distinct immunity-related molecules, or involvement in humoral immune responses were conducted before for lepidopteran hematopoietic organs. Hence, we present the first description of differential expression of immune molecules in the hematopoietic organ of M. sexta, and show that the mRNA levels of lysozyme, PGRP-1A, and scolexin A alter when systemic infections occur and that this tissue reacts by increased PGRP-1A expression in response to gut infection.

Since the hematopoietic organs form plasmatocytes (Nardi et al., 2003; von Bredow et al., 2021), the expression of lysozyme within a subset of hematopoietic organ cells may be indicative of mature plasmatocytes contained by the organs. The increase in immune-related gene expression could therefore be due to an increased differentiation of the hematopoietic organ cells to mature plasmatocytes. Moreover, we cannot rule out that the organ itself plays a role in the humoral defense, being a minor source of receptors and effectors secreted to the hemolymph. However, a direct comparison of the reaction of hemocytes and the hematopoietic organ is difficult, since the hemocytes in circulation comprise a mixture of granular cells, plasmatocytes, spherule cells, oenocytoids, and prohemocytes whereas the hematopoietic organs do not contain granular cells, spherule cells or oenocytoids (Nardi et al., 2003; von Bredow et al., 2021).