All dissections were performed under an SZ2-ST binocular microscope (Olympus), and images were taken using a mounted HD-Pro VC.3038 camera (Euromex). Animals were weighed using a Genius ME215p analytical balance (Sartorius). Tissues were immediately transferred to screwcapped 2 mL Eppendorf vials containing 0.5 mL RNAlater (Thermo Scientific™) to avoid RNA degradation. Subsequently, tissues were removed from RNAlater, rinsed in S. gregaria Ringer (1 L: 8.766 g NaCl; 0.188 g CaCl₂; 0.746 g KCl; 0.407 g MgCl₂; 0.336 g NaHCO₃; 30.807 g sucrose; 1.892 g trehalose; pH 7.2), and transferred to MagNA Lyser Green Beads vials (Roche), with the exception of suboesophageal ganglia (SOG) and prothoracic glands (PG) which were transferred to screwcapped 1.5 mL Eppendorf vials. All tissues were stored at −80°C until RNA extraction (2.2.2). For ecdysteroid measurements, 10 µL of hemolymph was collected using a glass capillary prior to sacrificing the animals and immediately transferred to 300 µL of phosphate-buffered saline (PBS) (13.7 mM NaCl, 0.27 mM KCl, 1 mM NH₂HPO₄, 0.18 mM KH₂PO₄, pH 7.4) and 300 µL of 1-butanol (Merck). After vortexing and centrifugation at 20,000 × g for 10 min, the upper organic phase was collected, lyophilized using a SpeedVac vacuum concentrator, and stored at −20°C until ecdysteroid extraction (2.6).

Two different RNA extraction methods were used based on tissue type. Total RNA was extracted using the RNeasy Lipid Tissue extraction kit (Qiagen) according to the manufacturer’s protocols. To each MagNA Lyser Green Beads vial (Roche), 1 mL QIAzol lysis reagent buffer (Thermo Scientific™) was added. Next, the tissues were homogenized for 30 s at 6,000 × g using a MagNA Lyser instrument (Roche). Finally, total RNA was eluted in 80 μL of nuclease-free water. By contrast, total RNA was extracted from the SOG and PG using the RNAqueous®-Micro Kit (Ambion) according to the manufacturer’s protocols. First, 100 µL of lysis solution was added to the tissues. Next, tissues were disrupted and homogenized using a Teflon micro pestle attached to a drill. Using either extraction method, a DNase treatment was performed upon elution of the extracted RNA. In both cases, the concentration of the resulting RNA extracts was determined by means of a Nanodrop spectrophotometer (Nanodrop ND-1000, Thermo Fisher Scientific, Inc.). RNA extracts were stored at −80°C.

The S. gregaria whole-body transcriptome of Ragionieri et al. (Ragionieri et al., 2022) (GenBank accession number GJPN000000000.1) contains different gene isoforms and was therefore first processed to reduce the redundancy of transcript sequences. The tool Supertranscripts combines all isoforms of a transcript into one assembled contiguous sequence referred to as a SuperTranscript (Davidson et al., 2017). A SuperTranscript is a single linear sequence constructed from collapsing unique and common sequence regions of all available splicing isoforms of a transcript. This way, all isoforms of a transcript are combined into one assembled contiguous sequence without losing transcript data (Davidson et al., 2017). Furthermore, the original transcripts that make up the SuperTranscripts can be retraced in the original transcriptome (GJPN000000000.1). Therefore, the SuperTranscripts transcriptome version provides a very potent alternative reference for mapping RNA-Seq reads. The SuperTranscripts transcriptome retained 140,093 distinct transcripts. Next, the CD-HIT-EST tool clusters highly similar sequences (>95%) and thereby filters remaining redundant sequences, resulting in the removal of 16,379 redundant contigs (Li and Godzik, 2006; Fu et al., 2012). The resulting processed transcriptome containing 123,714 contigs was submitted to Gene Expression Omnibus (GEO) (accession number GSE226871). Next, the processed transcriptome was annotated in different steps. First, these nucleotide sequences were translated to their predicted best coding sequence using Transdecoder ¹ and annotated using Trinotate according to the procedure described by Bryant et al. (Bryant et al., 2017). Next, all transcript sequences and their predicted coding sequences were used as queries for Basic Local Alignment Search Tool searches (BLASTx and BLASTp, E ≤ 1e⁻⁵) (Boratyn et al., 2013) in the annotated UniProtKB/Swiss-Prot databases (Boeckmann et al., 2005; Bateman et al., 2015). The hits from these homology searches were used to provide the functional annotations of Gene Ontology (GO) and Clusters of Orthologous Groups (COG) terms (Figures 1, 2). The predicted coding regions were further used to annotate domain content in the Pfam database currently hosted by InterPro ² (Mistry et al., 2021) using the HMMER software ³ (version 3.1.2), signal peptides ⁴ (SignalP, version 4.1), and transmembrane domains ⁵ (TMHMM, version 2.0). This processed and annotated transcriptome served as a reference for mapping the sequenced reads and is referred to as the reference transcriptome from this point forward. Finally, the annotated mapped midgut transcripts were organized in an SQLite database using Trinotate (Supplementary Table S1).

Three different post-feeding timepoints were selected based on the progression of the digestive process in S. gregaria. Based on descriptions in literature (Baines et al., 1973; Marana et al., 1997) and our own observations, we decided to analyze the transcript profile of the midgut in S. gregaria during the initial phase of digestion (10 min after food uptake) when the ingested food was present in the foregut, and the core phase of digestion when the majority of the food bolus was observed in the midgut lumen (2 h after food uptake). We compared both these timepoints to a reference condition with locusts that had not been fed for 24 h (24 h after food uptake). This condition can therefore also be considered as a group of animals that were food-deprived (or starved) for 24 h. To obtain samples from these timepoints, animals with a balanced sex ratio were bred under gregarine-free conditions (2.1). From the day they molted to the 5th nymphal stage (N5D0), feeding was synchronized by allowing animals to feed on freshly washed cabbage leaves twice daily, in the morning and the afternoon, for 1 h periods (Figure 3A). After feeding on the morning of the 4^th day of the N5 stage (N5D4), three groups were established: one group was deprived of food for the next 24 h and sacrificed the next day (24 h condition). The other animals were allowed to eat for 1 h in the afternoon of N5D4 and received a final meal in the morning of N5D5. They were allowed to eat for 10 min from this meal, which was equal to their typically observed initial feeding stage. A group of animals was sacrificed after an additional 10 min (10 min condition), and a second group of animals was sacrificed after 2 h (2 h condition). In both cases, the time was measured from the end of the initial 10 minutes feeding period. Animals were dissected as described in 2.2.1. Additionally, the dissections of all three groups were executed on the same day around the same time to exclude any potential circadian effects. Midgut tissues of six animals were pooled per sample. For each condition, six biological replicate samples were collected.

From each sample, 1 μg of total RNA was used to perform an Illumina^® sequencing library preparation using the TruSeq^® Stranded mRNA Library prep kit (Illumina^®) according to the manufacturer’s protocol. First, poly-A-containing mRNA molecules are enriched from the total RNA mixture using poly-dT oligonucleotides attached to magnetic beads. Next, these purified mRNA molecules were fragmented into smaller fragments (100–1,000 nt) using divalent metal cations under elevated temperature in a proprietary Illumina^® fragmentation buffer. The smaller RNA fragments resulting from the fragmentation step were used for the first-strand cDNA synthesis using reverse transcriptase and random primers. Next, a single adenosine was added to these cDNA fragments, followed by the ligation of the Illumina^® adapters. The resulting cDNA products were then purified and enriched with 13 PCR cycles to create the final cDNA library. Libraries were quantified by RT-qPCR, according to Illumina’s “Sequencing Library qPCR Quantification protocol guide,” version February 2011. A High Sensitivity DNA chip (Agilent Technologies^®) was used to control the library’s size distribution and quality. A total of six midgut samples per condition (eighteen in total) were sequenced on a high throughput Illumina^® NextSeq 500 flow cell generating 75 bp single reads. The samples were multiplexed, allowing for the sequencing of all samples on one Illumina^® flow cell. Moreover, sequencing was performed in duplicate on two different flow cells to increase sequencing depth (Supplementary Figure S1).

Prior to mapping, Illumina^® adapter sequences were trimmed using Cutadapt (version 1.11) (Martin, 2011). Base calling quality was assessed using FastQC ⁶ . Trimmed sequencing reads were mapped to the reference transcriptome using Bowtie2 (version 2.2.5). Transcript abundance was quantified using the RSEM software package (RNA-seq by Expectation Maximization, version 1.2.31) (Li and Dewey, 2011). Transcripts that did not have an abundance of at least 1 count-per-million in at least six out of eighteen samples were considered to be both biologically and statistically irrelevant and were removed from the dataset prior to downstream analyses. The other transcripts were considered quantifiable. Read counts were normalized using Bioconductor software edgeR’s standard TMM normalization method (Robinson, McCarthy, and Smyth, 2009) in R, version 3.4.2 ⁷ . These TMM-normalized transcript counts were used to calculate differential expression among the three conditions (Supplementary Table S2).

A principal component analysis (PCA) was performed on regularized log-transformed counts using the DESeq2 Bioconductor software in R (Supplementary Figure S2) (Love et al., 2014). Differential expression analysis between groups of samples was performed using the Bioconductor package edgeR in R (Robinson et al., 2009). Three separate differential expression analyses were performed: 10 min vs. 24 h, 2 h vs. 24 h, and 10 min vs. 2 h. A generalized linear model was built for each separate analysis, and statistical testing was done using the empirical Bayes quasi-likelihood F-test. Transcripts having a false discovery rate (FDR) adjusted p-value ≤ 0.05 and a log₂ fold change (FC) ≥ 1 or ≤ −1 were considered to be significantly differentially expressed.

Numerous transcripts predicted to encode proteases were identified amongst putative digestive enzymes, of which the majority code for serine proteases, such as trypsin and chymotrypsin (Supplementary Table S5). Large numbers of the most commonly found types of carbohydrases were identified, namely, α-Amylase, α-Glucosidase, and β-Glucosidase, as well as three transcripts encoding putative cellulases belonging to the GH9 family of endoglucanases (Pfam00759). A relatively large variety of lipid-degrading enzymes was identified, namely, lipases, phospholipases, and sterol dehydrogenases/reductases. Finally, putative DNase, RNase, nucleotidase and nucleosidase encoding transcripts were present in the S. gregaria midgut.

Multiple transcripts encoding putative amino acid, carbohydrate, and lipid transporters were identified (Supplementary Table S6). Additionally, the midgut transcriptome was scanned for other transcripts known to contribute to the intestinal physiology of herbivorous insects, including those involved in peritrophic membrane formation, transmembrane trafficking, and xenobiotic metabolism (Supplementary Table S7). Transcripts predicted to encode proteins active in all three phases of detoxification of xenobiotics were identified, namely, P450s and CEs (phase I), UGTs and GSTs (phase II) as well as putative transmembrane transporters such as members of the ATP-binding cassette (ABC) family of transporters (phase III).

SgVAHa1 is predominantly expressed in the gastric caeca and in the midgut and only low transcript levels were detected in other parts of the digestive tract, namely, foregut, Malpighian tubules, and hindgut (Figure 4A). Transcript levels of SgVAHa1 were significantly higher in caeca and midgut compared to all other tissues (n = 4; one-way ANOVA with Holm-Šídák’s post hoc test; midgut and caeca compared to all other tissues: p < 0.0001; hindgut, foregut and Malpighian tubules compared to all other tissues: p < 0.005). Upon injection of dsVAHa1, the midgut transcript levels of SgVAHa1 on N5D6 were significantly reduced by 88.5% compared to animals injected with dsGFP (Figure 4B; dsGFP: n = 5, dsVAHa1: n = 5; unpaired Student’s t-test, p = 0.0003).

An RNAi-induced knockdown of SgVAHa1 (dsGFP: n = 30, dsVAHa1: n = 31) significantly affected survival throughout the N5 stage (Figure 4C, logrank Mantel-Cox test, p < 0.0001), as well as molting from the nymphal to the adult stage (Figure 4D, logrank Mantel-Cox test, p = 0.0003). The first animals died on N5D6 (dsGFP) and N5D7 (dsVAHa1). The highest mortality occurred in the dsVAHa1 condition on N5D9 (37.93%) and N5D10 (24.14%). On N5D10, only 24.14% of animals were still alive (dsVAHa1) compared to 90% in the control condition (dsGFP). The first animals molted to the adult stage at N5D6 (dsGFP) and N5D7 (dsVAHa1). On N5D9, only 28.85% of animals had molted (dsVAHa1) compared to all of the surviving dsGFP injected animals (90%). At the end of the experiment, a total of eighteen out of thirty-one animals (58.06%) from dsVAHa1 injected animals had reached the adult stage. However, these animals were visibly unhealthy (our observations) and fifteen of these eighteen adults died within the duration of this experiment (15 days after the molt to the N5 stage). By contrast, all of the twenty-seven dsGFP injected animals that successfully molted to the adult stage remained healthy throughout this experiment.

The majority of SgNPC1b expression was measured in the gastric caeca and the midgut. By contrast, SgNPC1b transcript levels were low in other parts of the digestive tract, namely, foregut, Malpighian tubules, and hindgut (Figure 5A). Transcript levels of SgNPC1b were significantly higher in caeca and midgut compared to all other tissues (n = 4, one-way ANOVA with Holm-Šídák’s post hoc test, midgut vs. foregut: p = 0.0002, all other comparisons: p < 0.0001). Transcript levels of SgNPC1b were measured in the midgut on N5D6, 24 h after a single injection with one of the (non-overlapping) dsRNAs targeting SgNPC1b (dsNPC1b1 or dsNPC1b2) in 24 h starved animals, and were significantly reduced by 69.5% (dsNPC1b1) and 53.1% (dsNPC1b2) compared to dsGFP injected animals (Figure 5B; dsGFP: n = 9, dsNPC1b1: n = 9, dsNPC1b2: n = 9; one-way ANOVA with Holm-Šídák’s post hoc test, dsNPC1b1: p = 0.0004, dsNPC1b2: p = 0.0303).

An RNAi-induced knockdown of SgNPC1b (dsGFP: n = 10, dsNPC1b1: n = 11, dsNPC1b2: n = 9) significantly affected survival throughout the N5 stage (Figure 5C, logrank Mantel-Cox test, dsNPC1b1: p < 0.0001, dsNPC1b2: p = 0.0003), as well as molting from the nymphal to the adult stage (Figure 5D, logrank Mantel-Cox test, dsNPC1b1: p < 0.0001, dsNPC1b2: p = 0.0001). The first animals died at N5D3 (dsNPC1b1) and N5D8 (dsNPC1b2). All but one dsNPC1b injected animal died during the N5 stage. The only surviving knockdown animal, injected with dsNPC1b2, molted to the adult stage at N5D15. By contrast, no control animals (dsGFP) died during the experiment. All control animals molted to the adult stage between N5D8 and N5D12. The total body weight was monitored from N5D0 onwards (Figure 5E). Weight gain was determined as the difference in total body weight between N5D0 and N5D6. Control animals gained significantly more weight, approximately doubling in weight, compared to both knockdown conditions (one-way ANOVA with Holm-Šídák’s post hoc test, dsNPC1b1: p < 0.0001, dsNPC1b2: p < 0.0001). Finally, ecdysteroid levels were measured in the hemolymph at N5D6 and expressed as 20E equivalents (Figure 5F, dsGFP: n = 9, dsNPC1b1: n = 8, dsNPC1b2: n = 10). Upon knockdown of SgNPC1b, ecdysteroids levels in the hemolymph were significantly reduced (one-way ANOVA with Holm-Šídák’s post hoc test, dsNPC1b1: p = 0.0184, dsNPC1b2: p = 0.0184). No significant differences in survival rate, molting success, weight gain or hemolymph ecdysteroid levels were observed between the treatments with non-overlapping dsNPC1b constructs.

We identified various transcripts predicted to code for putative proteases, carbohydrases, lipases, and nucleases (Supplementary Table S5). Such broad repertoires of digestive enzymes are frequently observed in insects (Campbell et al., 2005; Pauchet et al., 2009; Ma et al., 2012; Spit et al., 2016a; Zhang et al., 2016). Nevertheless, important species-specific differences in digestive enzyme repertoires are observed amongst insects, partially due to different dietary preferences and midgut luminal pH (Holtof et al., 2019). The fact that the majority of putative proteolytic enzyme encoding transcripts identified here code for serine proteases is in line with previous research (Campbell et al., 2005; Spit et al., 2012; Bonelli et al., 2020; Coutinho-Abreu et al., 2020). The proteolytic activity in the midgut of S. gregaria has indeed been mainly attributed to serine proteases, such as trypsins and chymotrypsins, but was also revealed to contain minor cysteine and carboxypeptidase activity (Spit et al., 2012). This high serine protease activity has also been found in other orthopterans (Spit et al., 2014; Huang et al., 2017). However, this is not the case for the more acidic midgut of Coleoptera, and Hemiptera, where the most abundant proteolytic activity is due to cysteine peptidases (Pauchet et al., 2009; Valencia et al., 2016; Oppert et al., 2018; Holtof et al., 2019; Pimentel et al., 2020).

Interestingly, three putative cellulases were identified in the locust midgut transcriptome. This is in accordance with the more recent findings that an endogenous cellulase activity is present in some insect species, such as many cockroaches and termites (Blattaria), the cricket Teleogryllus emma, the beetle Tribolium castaneum and the fly Musca domestica, amongst others (Watanabe and Tokuda, 2010; Holtof et al., 2019). However, genome encoded cellulase enzymes appear to be absent in Anopheles gambiae, Drosophila melanogaster and B. mori (Watanabe and Tokuda, 2010). In conclusion, our data align with the general observation that insects deploy divergent sets of enzymes to mediate the enzymatic digestion of food.

In addition to putative digestive enzymes, we observed the presence and upregulation upon feeding of putative transporters for all of the most abundant nutrient classes, namely, amino acids, carbohydrates, and lipids (Table 2; Supplementary Table S6). We identified several transcripts encoding proteins putatively involved in dietary lipid and sterol uptake. Fatty acid transport proteins, as well as scavenger receptors class B type I, have been suggested to interact in the active transport of dietary fatty acids across the midgut membrane, which is supported by their high abundance in the midgut transcriptome (Holtof et al., 2019). Additionally, various putative sterol transporters were identified, namely, sterol carrier protein x (SCPx), SCP2, and NPC1 proteins.

Insects often encounter various toxic or xenobiotic compounds that can be harmful if they are not properly dealt with. Herbivorous polyphagous species, like S. gregaria, have to typically rely on a wide range of detoxification mechanisms. This is especially true for insects feeding on complex diets with a high diversity of allelochemicals, such as green leaves, in contrast to diets less diverse in allelochemicals, such as plant sap, nectar and pollen (Rane et al., 2019). Our transcriptome data indeed predicted the presence of various counter-defensive mechanisms in the S. gregaria midgut. This is based on the abundant presence of transcripts predicted to encode NPSs, UGTs, GSTs, CEs, P450s as well as ABC transporters (Supplementary Table S7), all of which have been shown previously to play important roles in xenobiotic metabolism and detoxification in a wide range of species (Rane et al., 2019; Denecke et al., 2021). Furthermore, multiple transcripts for each of these protein classes, with the exception of P450, were upregulated upon feeding (Table 2). Such differential gene expression patterns upon exposure to xenobiotics have been observed in several other species of insects (Herde and Howe, 2014; Spit et al., 2016b; Dhania et al., 2019; Shu et al., 2021; Jin et al., 2022). Our findings suggest that S. gregaria actively detoxifies its ingested food in the midgut by swiftly increasing the transcript levels of several detoxification systems. Additionally, we hypothesize that the broad repertoire of putative detoxification mechanisms encountered here is a consequence of the polyphagous herbivorous lifestyle of this locust species, implying that it can cope with a large diversity of plant toxins in its diet.

One of the plant toxins most often encountered by the locusts in our colony is glucosinolate from cabbage (Barba et al., 2016). We showed the upregulation of several putative NSPs, UGTs and GSTs, which have all been shown to be specifically involved in defense against glucosinolate-derived toxins (Burow et al., 2006; Li et al., 2018), suggesting that these proteins may also have a detoxifying function in the S. gregaria midgut. High levels of NSPs have been identified in the midguts of Lepidoptera as well as the locust L. migratoria (Fischer et al., 2008; Spit et al., 2016a). Interestingly NSPs isolated from other insect species, including A. gambiae and Blattella germanica, showed binding capacity to human immunoglobulin E, provoking an allergic reaction in atopic persons (Wang, Lee, and Wu, 1999). Locust NSPs might therefore provoke allergies in a similar manner. Six upregulated transcripts are predicted to belong to the multidrug resistance protein 1 (MRP1) family, including the ABC transporter subfamilies ABCB, ABCC and ABCG, which are described to have important functions in the xenobiotic metabolism of a wide range of insect species (Labbé et al., 2011; Sodani et al., 2012; Dermauw and Van Leeuwen, 2014; Qi et al., 2016; Denecke et al., 2021). Among these MRPs, one was further categorized as a P-glycoprotein 1 (ABCB1), which has previously been shown to be present in the brain barrier of S. gregaria (Andersson et al., 2014). P-glycoprotein 1, and other MRPs, such as MRP49 (ABCB1) and MRP1 (ABCC1), have been shown to play a role in pesticide resistance (Tian et al., 2013; Denecke et al., 2021). In summary, we showed the presence of various transcripts encoding proteins putatively involved in xenobiotic metabolism, as well as their upregulation upon feeding. Our findings therefore illustrate that polyphagous herbivorous insects are able to rely on various defense mechanisms against a large array of toxins. Interestingly, insects have been shown to rely on similar mechanisms for their protection against natural plant toxins and synthetic pesticides (Hawkins et al., 2019). How they succeed in doing so is still an important subject of study. Identifying and characterizing the underlying patterns orchestrating metabolic resistance against pesticides will be crucial in combatting insect pesticide resistance going forward.

In summary, based on our differential gene expression analysis in S. gregaria as well as those described in literature, it is apparent that gene expression in the insect midgut can respond rapidly to environmental changes, such as the presence of food, infection or exposure to xenobiotics. Based on our data, as well as reports in literature, multiple transcripts encoding putative digestive enzymes, nutrient transporters and other digestion-related genes are abundantly upregulated upon feeding (Sanders et al., 2003; Xu et al., 2012; Cui and Franz, 2020). By contrast, immunity-related genes and detoxification enzymes are strongly upregulated upon infection or exposure to xenobiotics, respectively, while digestion-related genes are downregulated, although these reactions are often complex (Noland et al., 2013; Herde and Howe, 2014; Dhania et al., 2019; Coutinho-Abreu et al., 2020; Kuang et al., 2021; Shu et al., 2021; Jin et al., 2022).

Based on our multiple sequence alignment (Supplementary Figure S6) and phylogenetic analysis (Supplementary Figure S7), we can conclude that SgVAHa1 is a true ortholog of insect H⁺ V-ATPase subunit-a. We observed a significant upregulation of a H⁺ V-ATPase subunit-a 2 h after feeding, indicating that this subunit is under transcriptional regulation. The H⁺ V-ATPases are highly conserved transmembrane proton pumps which in insects occur mostly in various epithelial tissues, such as the salivary glands, the gut, Malpighian tubules, and the nervous system (Toei et al., 2010). These protein complexes use the energy released from hydrolyzing ATP to translocate protons across the apical membrane towards the extracellular fluid or across lysosomal or endosomal membranes towards the lumen of these intracellular compartments. This proton transport generates an electrochemical gradient enabling numerous biological processes, such as regulation of the internal pH of intracellular compartments, lysosomal degradation, loading of cargo into secretory vesicles, and maintaining the polarity of midgut epithelial cells (Banerjee and Kane, 2020). Furthermore, it is well established that in insects this proton motive force drives cation/H⁺ exchangers (i.e., K⁺/H⁺ and Na⁺/H⁺ antiporters) and therefore the transepithelial secretion of ions, which in turn stimulates Na⁺ or K⁺ coupled nutrient uptake (e.g., Na⁺ coupled glucose transporters) (Torrie et al., 2004; Wieczorek et al., 2009; Halberg and Møbjerg, 2012). However, the interaction between the electrochemical gradient generated by H⁺ V-ATPase and the secondary transport of inorganic ions by Na⁺/K⁺-ATPase at the basal membrane of epithelial cells is still unclear (Halberg and Møbjerg, 2012). The H⁺ V-ATPase protein complex consists of various subunits, further subdivided into a peripheral V₁ domain consisting of eight different subunits (A-H) and an integral, membrane-bound, V₀ domain consisting of six different subunits (a-e) (Toei et al., 2010). The V₁ region contains the catalytic site converting ATP into energy, which is used for proton transport, in which subunit-a plays a crucial role. In addition, subunit-a was shown to be important for coupling specific V₁ and V₀ domains to different cellular sites (Kawasaki-Nishi et al., 2001; Sun-Wada and Wada, 2022) through interaction with specific phospholipids (Banerjee and Kane, 2020). Furthermore, the N-terminal domain of subunit-a has been described as a regulatory hub, incorporating environmental input, such as the glycolytic enzyme aldolase, and thereby regulating both assembly/disassembly and localization of the H⁺ V-ATPase complex to specific cellular membranes (Banerjee and Kane, 2020).

The SgVAHa1 tissue profile highlights large differences in relative transcript levels (Figure 4A) and suggests a specialized function of SgVAHa1 in the midgut and associated caeca. In addition to SgVAHa1, two other unique H⁺ V-ATPase subunit-a isoform sequences were also identified in our study. However, their tissue-dependent expression profiles were not evaluated since they were not differentially enriched upon feeding. It is currently unclear whether these isoforms perform different (tissue-specific) functions compared to SgVAHa1. More research is therefore needed to functionally characterize the different S. gregaria H⁺ V-ATPase subunit-a isoforms.

Differential distribution is typical for H⁺ V-ATPase subunits isoforms, including subunit-a (Kawasaki-Nishi et al., 2001; Toei et al., 2010; Chintapalli et al., 2013; Sun-Wada and Wada, 2022). All four isoforms of subunit-a in mice show different localization in embryos (Wada and Wada, 2022). In yeast two subunit-a isoforms occur, Stv1p which is localized to the vacuole and Vph1p which is localized to the Golgi apparatus (Kawasaki-Nishi et al., 2001). In insects, such tissue-specific expression of H⁺ V-ATPase subunit-a isoforms has been analyzed in D. Melanogaster, T. castaneum, and L. migratoria (Chintapalli et al., 2013; Mo et al., 2020; Liu et al., 2022; Naseem et al., 2023). In D. melanogaster, a midgut-specific H⁺ V-ATPase subunit-a isoform, called vha100-4, was identified. Other D. melanogaster H⁺ V-ATPase subunit-a isoforms are more uniformly expressed across tissues, with the exception of vha100-3 which is almost exclusively expressed in the testes (Chintapalli et al., 2013). Moreover, differential expression of subunit-a isoforms within the midgut has been observed in D. melanogaster (Overend et al., 2016). The midgut-specific vha100-4 is expressed exclusively in the acidic region of the midgut, while other isoforms are expressed over the entire length of the midgut (Overend et al., 2016). Multiple D. melanogaster subunit-a isoforms were shown to perform different functions in wing development within the same tissue (Mo et al., 2020). Similarly to what is observed in D. melanogaster, a H⁺ V-ATPase subunit-a isoform from T. castaneum shows differential expression across different epithelia with the highest expression being measured in the posterior midgut, while three out of five isoforms are not expressed in the midgut (Naseem et al., 2023). Two isoforms were identified in L. migratoria, showing distinct tissue-dependent and developmental expression profiles, although they were not highly expressed in the midgut or the gastric caeca (Liu et al., 2022). One isoform is most highly expressed in the hindgut and to a lesser extent in the fat body and trachea, while the second isoform is most highly expressed in the fat body. Tissue-specific expression patterns of different subunit isoforms was suggested to be most likely linked to local, tissue-specific V-ATPase functions (Sun-Wada and Wada, 2022). In many cases however, distinct H⁺ V-ATPase subunit isoforms have not been fully functionally characterized, especially in insect species. Furthermore, it is still unclear whether different subunit isoforms perform different functions within the same tissue, as so far this has only been described in the wing of D. melanogaster (Mo et al., 2020). Therefore, it would be of great interest to study the tissue-dependent expression profiles of three SgVAHa isoforms identified in the midgut in this study as well as possible other isoforms and their subsequent functional characterization. This would add an interesting perspective from a hemimetabolan insect species to the current knowledge available from the holometabolan model organisms, D. melanogaster and T. castaneum (Chintapalli et al., 2013; Naseem et al., 2023).

RNAi-induced gene silencing of SgVAHa1 resulted in high mortality rates and impaired molting (Figure 4), illustrating the crucial importance of SgVAHa1 in the midgut. The majority of the experimental animals died shortly after molting to the adult stage, while some died during the molting process. However, based on our results, it is currently unclear what exactly is causing lethality. We propose two hypotheses, which may not be mutually exclusive. First, lethality might have resulted from a defective uptake of essential nutrients due to the silencing of the H⁺ V-ATPase complex, since its formation and the subsequently induced favorable membrane potential, may have been inhibited upon silencing SgVAHa1. Secondly, SgVAHa1 is important for the integrity and function of the midgut epithelium, including the polarity of the epithelial cells and optimal functioning of various cell organelles. The first hypothesis is supported by the upregulation of SgVAHa1 as soon as 10 min after feeding (Supplementary Figure S5C) in conjunction to the upregulation of various transporters for carbohydrates, amino acids, and lipids 2 h after feeding (Table 2). The swift upregulation of a H⁺ V-ATPase V₀ domain subunit in the midgut in response to feeding has only been previously described in the mosquito A. aegypti, namely, the upregulation of subunits a and c (as well as various subunits of the V₁ domain) 12 h after a bloodmeal (Sanders et al., 2003). In addition, to our knowledge only one other study in insects reported the upregulation of subunits of the V₁ domain in response to feeding, namely, the upregulation of subunit C after a bloodmeal in the mosquito Lutzomyia longipalpis (Pitaluga et al., 2009). Furthermore, starvation in Manduca sexta caused disassembly of the H⁺ V-ATPase complex, possibly to reduce energy expenditure (Sumner et al., 1995). It would therefore be valuable to study the nutrient uptake at the midgut epithelium or nutrient levels in the hemolymph upon silencing SgVAHa1. In the locust, L. migratoria, a combined knockdown of two H⁺ V-ATPase subunit-a isoforms was obtained by injecting dsRNA directed against a sequence common to both isoforms. This treatment resulted in mortality and molting defects, similarly to what we observed here (Liu et al., 2022). Targeting these isoforms led to a decrease in food intake and body weight. Interestingly, researchers found no food in the midgut lumen of treated animals, fewer columnar epithelial cells and the absence of the brush border in some regions of the midgut epithelium (Liu et al., 2022). Similarly, knockdown of H⁺ V-ATPase subunit-a genes (vha100) or a subunit-B gene (vha55) in D. melanogaster led to increased pH in the acidic region of the midgut (Overend et al., 2016; Koyama et al., 2021), and vha55 knockdown resulted in impaired nutrient absorption and mortality (Koyama et al., 2021). It is remarkable that subunit-a is the only H⁺ V-ATPase subunit that was significantly upregulated in the S. gregaria midgut 2 h post-feeding. It is possible that subunit-a is under strict transcriptional control because of its critical role in H⁺ V-ATPase assembly as well as in targeting specific V₁ domains to specific membranes.

A number of studies have evaluated the insecticidal potential of targeting H⁺ V-ATPase subunits in several insect species, including a variety of insect pests, showing strong negative impacts on growth and survival (Ahmed, 2016; Sato et al., 2019; Rahmani and Bandani, 2021; Guo et al., 2022; Liu et al., 2022; Shi et al., 2022; Zeng et al., 2022). So far, the majority of these studies focused on targeting the different subunits of the V₁ domain of these proton pumps. To the best of our knowledge, only two studies describing the targeting of subunit-a of the H⁺ V-ATPase V₀ domain in insects exist (Ahmed, 2016; Liu et al., 2022). In both L. migratoria and the lepidopteran Pectinophora gossypiella a H⁺ V-ATPase subunit-a knockdown resulted in strong negative impacts on insect growth and survival, clearly illustrating the strong insecticidal potential of targeting insect H⁺ V-ATPases, including SgVAHa1.

Both NPC1 and NPC2 proteins were identified in the S. gregaria midgut. Based on our multiple sequence alignment (Supplementary Figure S8) and phylogenetic analysis (Supplementary Figure S9), we conclude that SgNPC1b is a true ortholog of insect NPC1b. We observed a significant upregulation of SgNPC1b 2 h after feeding. Niemann-pick proteins are a family of transmembrane sterol-binding proteins known to play an important role in sterol absorption in the digestive tract of both mammals and insects (Davies et al., 2005; Voght et al., 2007; Zheng et al., 2018; Jing and Behmer, 2020). Even though the sterol uptake pathways in mammals and insects are similar, the exact role of Niemann-pick proteins in insects has so far only been studied in D. melanogaster and Helicoverpa armigera. Functional research in other insect species is currently lacking. NPC1 proteins typically contain three conserved domains which all show sterol-binding activity: N-terminal domain, patched domain and SSD (Voght et al., 2007; Zheng et al., 2018; Jing and Behmer, 2020). Due to gene duplication, the majority of insects possess two NPC1 homologues, namely, NPC1a and NPC1b (Zheng et al., 2018). NPC1b was shown to be crucial for dietary sterol absorption in the midgut of D. melanogaster and H. armigera (Voght et al., 2007; Zheng et al., 2020). After absorption, sterols are taken up in the lysosomes and are transferred by NPC2 to NPC1a which is integrated in the lysosomal membrane (Jing and Behmer, 2020). In insect herbivores, such as S. gregaria, plant sterols are dealkylated to form cholesterol, likely in the smooth endoplasmic reticulum of the midgut epithelium (Li and Jing, 2020).

SgNPC1b shows a midgut and associated caeca specific expression (Figure 5A), similar to SgVAHa1, implying a specialized function of SgNPC1b in the midgut, which is consistent with findings in the holometabolan species, D. melanogaster, H. armigera, T. castaneum and Tenebrio molitor (Voght et al., 2007; Oppert et al., 2018; Zheng et al., 2020). In line with these reports, we found two NPC1-like sequences in T. castaneum in the BeetleAtlas ⁹ database (Naseem et al., 2023), one putative NPC1a sequence (TC005088) with uniform distribution and one putative NPC1b sequence (TC033918) with midgut-specific expression. Furthermore, NPC1b expression in the digestive system is probably not restricted to insects, since NPC1L1, the NPC1b ortholog in mammals, is highly expressed in the human liver and to a lesser extent the small intestine (Davies et al., 2005).

Injection of dsNPC1b caused impaired weight gain, lowered hemolymph ecdysteroid titers, impaired molting, and high mortality (Figure 5). As NPC1b has been shown to be a midgut sterol transporter (Voght et al., 2007; Zheng et al., 2020), the phenotypes we observed here are expected to be the consequence of impaired sterol uptake. Arthropods do not synthesize sterols de novo, and thus rely entirely on intake of dietary sterols (Behmer, 2017; Li and Jing, 2020). Therefore, we propose three plausible, not mutually exclusive, hypotheses explaining the phenotypes we observed: 1. Sterols are crucial for normal growth and development since they are an integral component of biological membranes; 2. Sterols play a signaling role, being an important cue for nutritional status and thereby regulating growth and development; 3. Plant sterols are converted to cholesterol, which is the main precursor for ecdysteroid biosynthesis and therefore required for ecdysteroid hormone signaling.

Our data revealed several important phenotypes. We measured a significantly lower ecdysteroid titer in the hemolymph of SgNPC1b knockdown versus control locusts on N5D6, a developmental time point at which increased ecdysteroid levels are expected to occur in hemolymph (Verbakel et al., 2021). A possible explanation for this observation is that silencing SgNPC1b causes impaired sterol uptake in the midgut, resulting in a lack of cholesterol as precursor for ecdysteroid biosynthesis. However, this may only partially explain the observations. Locusts may still have sufficient access to cholesterol as a precursor for ecdysteroid biosynthesis based on the following arguments: 1. We knocked down SgNPC1b, meaning that we did not generate a complete knockout and presumably sterol uptake was not completely inhibited; 2. Locusts might still rely on stored steryl esters, which could potentially suffice to maintain ecdysteroid biosynthesis (Korber et al., 2017); 3. Most cholesterol is incorporated into cellular membranes and only a small fraction is needed for ecdysteroid biosynthesis (Jing and Behmer, 2020); 4. Alternative dietary sterol uptake mechanisms might still be in place, such as SCP (Guo et al., 2009). This suggests that it is unlikely that the SgNPC1b knockdown only affected ecdysteroid biosynthesis directly, due to the lack of cholesterol precursor. It is well described in insects that the attainment of a critical weight is an important trigger to stimulate ecdysteroid biosynthesis for metamorphosis or molting (Pan et al., 2021). Therefore, we suggest that the impaired weight gain during the N5 stage may also have inhibited or postponed ecdysteroid biosynthesis indirectly, resulting in the observed lower ecdysteroid titers in hemolymph. Anyway, it seems unlikely that the observed phenotypic effects, namely, reduced weight gain, impaired molting, and high mortality, were solely the consequence of a lower ecdysteroid titer. The effects on weight gain already occurred at the start of the N5 stage, long before the expected rise in ecdysteroid levels (around N5D6). Interestingly, in D. melanogaster NPC1b mutants, administration of exogenous ecdysone did not affect mortality, indicating that reduced ecdysteroid biosynthesis by itself does not explain all phenotypes observed upon impaired sterol uptake (Voght et al., 2007).

Instead, we suggest that the lethal phenotype we observed may be a consequence of impaired membrane integrity and functionality, especially at the midgut epithelium. Lack of sterols in the cellular membrane is known to cause mortality by compromising membrane integrity, affecting the function of membrane-bound enzymes, endocytosis, receptor binding and signal transduction (Behmer and Elias, 1999; Gasque et al., 2005). Nevertheless, in D. melanogaster sterols appeared to be dispensable for maintaining membrane properties such as impermeability during periods of extreme dietary sterol restriction. During such periods these membrane properties were supplied by non-sterol lipids, such as sphingolipids (Carvalho et al., 2010). In contrast to D. melanogaster, acridid grasshoppers (Orthoptera) do not tolerate high proportions of non-cholesterol sterols as structural components in their membranes, while the tolerance in the lepidopterans Heliothis virescens and Helicoverpa zea is somewhere in between (Behmer and Elias, 2000; Jing et al., 2014; Jing and Behmer, 2020). This suggests that acridid grasshoppers (Orthoptera), such as S. gregaria, are more vulnerable to reduced cholesterol levels as they need mostly cholesterol for maintaining membrane integrity.

Interestingly, cholesterol has been shown to perform an important signaling function in D. melanogaster (Texada et al., 2022). Cholesterol serves as a nutritional signal, mediated by the target of rapamycin pathway in the fat body, which stimulates growth and development by increasing insulin-like peptide transcription and release in the insulin-producing cells of the brain. Furthermore, the growth-promoting effect of cholesterol requires insulin signaling and is not the result of effects on ecdysteroid biosynthesis. It is currently unclear whether cholesterol has a growth-promoting effect in S. gregaria as well, and whether this effect is facilitated by the insulin signaling pathway. A SgNPC1b knockdown did not affect transcript levels of insulin-related peptide in the locust brain (Supplementary Figure S10), although this result does not exclude possible effects on actual insulin-related peptide release from the insulin-producing cells of the brain. Further investigation is therefore required to find out whether SgNPC1b depletion might have affected the insulin signaling pathway in S. gregaria.

Our data show that SgNPC1b performs a crucial function in the midgut. Furthermore, if compensatory mechanisms were available, either alternative sterol transporters or the incorporation of alternative sterols or non-sterol lipids in the cellular membranes, then these mechanisms were insufficient to rescue the knockdown of SgNPC1b in a timely fashion. In addition to SgNPC1b, several SCP-x and SCP2 transcripts were identified in the S. gregaria midgut. However, none of these putative sterol transporters were upregulated upon feeding. Functional characterization of these proteins is therefore necessary to determine their possible roles in S. gregaria. Our data suggest that SCPs are either unable to compensate for reduced SgNPC1b function or are not involved in sterol uptake in S. gregaria. Finally, it cannot be excluded that SgNPC1b performs another crucial function, for example, the uptake of a different essential nutrient. This was suggested earlier by Voght et al., although so far no direct support for this hypothesis was published (Voght et al., 2007). Therefore, investigating the substrate specificity of SgNPC1b as well as the exact sterol depletions caused by the knockdown of SgNPC1b may contribute to a better understanding of its role in midgut physiology. In summary, we hypothesize that injection of dsNPC1b reduced dietary sterol uptake in the midgut and may have affected membrane integrity, resulting in reduced weight gain, impaired molting, and death. Additionally, we suggest that a SgNPC1b knockdown affected ecdysteroid biosynthesis, probably more indirectly (through an effect on weight gain and metabolism) than directly (through a lack of cholesterol as ecdysteroid precursor). Overall, our experiments were the first to demonstrate the crucial role and insecticidal target potential of NPC1b in an hemimetabolan (orthopteran) insect species.