Hemipteran insects experience recurrent and synergistic molecular evolution adapting to diet shift

Abstract

Background

Hemiptera is an insect order with extremely high physiological and behavioral diversity. Feeding traits have switched and reversed multiple times, but the molecular basis governing this phenotypical change remains unclear.

Results

We obtained the high-quality genomes and salivary gland transcriptomes of two essential biocontrol agents, Eocanthecona furcellata and Arma custos (Hemiptera: Heteroptera: Pentatomidae: Asopinae). This subfamily represents a typical insect clade with the reversal of feeding traits from phytophagy to zoophagy. Combined with public data of an additional 38 phylogenetically related insects and salivary gland transcriptomes of representative species, we performed a comprehensive analysis on the molecular evolution of feeding traits. We defined a set of diet-related gene groups and found that these genes were repetitively expanded in zoophagous species and experienced fast evolution and positive selection during diet reversal in the E. furcellata–A. custos clade. Transcriptomic analysis revealed dynamic upregulation of diet-related gene expression in zoophagous species, and further endeavor narrowed down the candidates to several genes, like trypsin and carboxypeptidase, which might be involved in diet reversal.

Conclusions

In conclusion, the evolution of zoophagy in Hemiptera and the reversal of feeding traits might require the synergistic regulations at both genomic and transcriptomic levels. Our study provides a potential connection between genotype and phenotype and advances the understanding of the adaptive evolution of zoophagous insects.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12915-025-02429-y.

Background

Biodiversity and the ecological value of zoophagous insects

The origin of biodiversity and phenotypic innovations is a captivating question in evolutionary biology. Insects, as one of the most diversified clades within the animal kingdom, exhibit a variety of highly divergent traits spanning diet, habitat, and morphology, including zoophagous to phytophagous feeding, aquatic to terrestrial habitats, and non-winged to winged forms [1–3]. Understanding the genetic basis and molecular mechanisms governing such phenotypic diversity is crucial for elucidating the genotype–phenotype relationship, advancing our understanding of diverse layers of biodiversity, and potentially contributing to the conservation of biological resources. Insects exhibit a diverse array of feeding strategies, including zoophagous, phytophagous, hematophagous, and omnivorous behaviors. These diverse traits may have independently evolved and frequently reversed within specific lineages in response to selection pressures and environmental adaptations.

Complex evolutionary trajectory of feeding traits in Hemiptera

Hemiptera is the fifth largest insect order with a wide variety of feeding traits [4]. Its suborder Heteroptera (true bugs) is the most phenotypically diversified clade, characterized by specialized feeding adaptations, including phytophagy (feeding on seeds, leaves, roots, and flowers), zoophagy, hematophagy, and omnivory. The feeding trait has reversed multiple times in Hemiptera and particularly in Heteroptera, facilitating the rapid exploitation of different niches (Fig. 1). The common ancestor of Hemiptera was likely to be phytophagous because the two early-diverging suborders, Auchenorrhyncha and Sternorrhyncha, were phytophagous [5] (Fig. 1). Next, the common ancestor of suborder Heteroptera experienced the transition from phytophagous to zoophagous [6] (Fig. 1). An established phylogeny supports that Heteroptera was further divided into seven infraorders [7, 8], among which six infraorders are predicted to be zoophagous at their ancestral nodes [5, 9–11], and only the ancestor of Pentatomomorpha was considered to be phytophagous [7, 9–11] (Fig. 1). This suggests that different infraorders experienced different feeding switches at different times (e.g., Nepomorpha, Cimicomorpha), or not at all (e.g., Gerromorpha, Enicocephalomorpha) [11] (Fig. 1). However, in later splits in Pentatomomorpha, a reversal from phytophagous to zoophagous occurred in a few clades like the subfamily Asopinae (Pentatomidae) [12], whereas in many other families or superfamilies of Pentatomomorpha the phytophagous trait was maintained [6, 10, 13] (Fig. 1). Apart from the frequent switch between zoophagy and phytophagy, other feeding traits like hematophagy and omnivory also emerged during evolution [8, 14, 15]. Unraveling the molecular mechanisms underlying the transition and reversal of feeding traits helps us fill the gap between genotype and phenotype. Particularly, the specimen from the Pentatomidae family in infraorder Pentatomomorpha will be suitable for studying the independent gain or reversal to zoophagy in a smaller evolutionary scale, representing more recent adaptation events (Fig. 1).

Fig. 1.

Transitions of feeding habits across the Hemiptera phylogeny. The branch lengths are unscaled, and only the topology of the tree is informative. Different feeding traits were labeled with different colors of the branches. The four suborders of Hemiptera as well as the seven infraorders of Heteroptera were labeled in the tree. Shaded boxes indicate the infraorders consisting of more than one branch in this phylogenetic tree. The phylogeny and feeding trait of each clade were retrieved from well-acknowledged literatures [9, 10, 14, 16]

Aims and scopes

In Pentatomomorpha, the three currently available genomes (with detailed annotation) belong to phytophagous species, representing the ancestral state of this infraorder. Zoophagy was then regained in this clade (Fig. 1). The genomes of zoophagous Pentatomomorpha species will help us decipher the genetic basis underlying the reversal of feeding traits. Together with the available genomes of species with various feeding traits in other clades, we are able to answer whether the emergence of the zoophagous phenotype shares a similar molecular basis across diverse clades and how the reversal of the feeding trait is subjected to natural selection. In addition to the genomic sequence alone, the transcriptome data also provide another level of valuable information for deciphering the potential molecular evolution underlying the phenotypic adaptation.

This study investigates the molecular mechanisms underlying dietary adaptation in Hemiptera by combining high-quality genomic and transcriptomic data from two zoophagous species, Eocanthecona furcellata and Arma custos. Using a broader comparative framework that includes genomic data from 34 Hemiptera species and an additional four outgroups, we explore how molecular evolution contributes to feeding traits in these insects. We focus on diet-related gene groups and observe their recurrent expansion across different zoophagous lineages. Moreover, our analysis emphasizes the role of salivary gland transcriptomes in capturing dynamic gene expression changes, especially in the context of dietary reversals. These results highlight the complex interplay between genomic evolution and gene regulation at the transcriptomic level, providing a comprehensive understanding of the rapid evolution of feeding traits in response to dietary shifts. This work bridges the gap between genotype and phenotype, enhancing our knowledge of the adaptive evolutionary processes in zoophagous insects.

Results

Genome assembly and annotation of E. furcellata and A. custos

Given that the zoophagous species in the Pentatomidae family are favorable for studying the independent gain of zoophagy during Hemiptera evolution, we selected two representative zoophagous species, E. furcellata and A. custos (Hemiptera: Heteroptera: Pentatomidae: Asopinae). We combined long-read (HiFi read), short-read, and Hi-C technologies together with advanced assembly algorithms to construct the genome assemblies (“Methods”). In total, 55.56 Gb of Illumina short reads for E. furcellata and 59.59 Gb for A. custos were obtained after a preliminary filtering step (Additional file 1: Table S1). Based on 17-mer depth, the genome size was estimated to be 1019.36 Mb and 1309.65 Mb, respectively (Additional file 1: Fig. S1). Genome assemblies were performed using 59.28-Gb PacBio HiFi reads for E. furcellata and 64.94 Gb for A. custos. This contig-level genome size of E. furcellata was 1067.43 Mb with N50 of 10.58 Mb and 1023.01 Mb for A. custos with N50 of 4.38 Mb (Additional file 1: Table S2). After physical mapping with 124.60 Gb and 143.10 Gb Hi-C clean data (Additional file 1: Table S1), 92.67% and 94.66% of the total sequence were anchored onto seven chromosome-level scaffolds in E. furcellata (Fig. 2A) and A. custos (Fig. 2B). Finally, two chromosome-level genomes were obtained with a size of 1067.62 Mb for E. furcellata and 1023.33 Mb for A. custos, with scaffold N50 of 138.24 Mb and 136.20 Mb (Table 1). Subsequently, the integrity of the assembly was demonstrated by 99.34% ~ 99.55% mapping rates for Illumina short reads (Table 1). The two genomes had better completeness than other closely related species in Pentatomomorpha (Additional file 1: Fig. S2), with higher BUSCO values (98.50% of E. furcellata and 95.60% of A. custos) (Table 1). These BUSCO results were 0.30% ~ 1.60% of BUSCO fragmented, 1.20% ~ 2.80% being missing, and 1.20% ~ 1.80% being duplicated (Additional file 1: Table S3). We further assessed the base quality of the genome assembly, which showed high QVs of 37.94 and 44.05 in E. furcellata and A. custos (Table 1). Overall, we obtained high-quality genomes of two zoophagous species with highly continuous and complete assemblies.

Fig. 2.

Phylogenetic and genomic comparisons among 40 insects. A, B Circos plots of two newly sequenced species E. furcellata and A. custos. The layer (a) depicts the chromosome ideogram, highlighting individual chromosomes with unique colors and markers, along with chromosome names and lengths, (b) illustrates the distribution of gene density, (c) displays the total content of transposable elements, (d) visualizes the distribution of DNA transposons, and (e) represents GC density across the genome. C Phylogenetic and genomic comparison among 40 insects. The left is a dated species phylogeny tree. Thirty-six hemipterans and 4 outgroup species were included. The diet of each species was indicated beside the species names. The right panel is the composition of TE major families

Table 1.

Statistics for genome assembly and annotation of two zoophagous true bugs E. furcellata and A. custos

Types		Values
		E. furcellata	A. custos
Genome assembly	Estimated genome size (Mb)	1,019.36	1,309.65
	Assembly size (Mb)	1,067.62	1,023.33
	Contig N50 (Mb)	8.61	4.32
	Scaffold N50 (Mb)	138.24	136.20
	Chromosome-anchoring (%)	92.67	94.66
	GC rate (%)	32.40	34.00
Assembly estimation	BUSCO completeness (%)	98.50	95.60
	Quality values (QVs)	37.94	44.05
	Mapping rate (%)	99.55	99.34
Genome annotation	Repeat content (%)	43.50	38.73
	No. of protein-coding genes	20,785	26,680
	No. of mRNA	24,432	28,250
	Average gene length (bp)	8,304	6,757
	Average exon number per gene	5.81	4.45
	Function annotation rate (%)	97.10	94.66
Sex determination	Sex determination system	XY	XY
	Sex chromosome	Chr2	Chr4

Genome annotation was generated by a customized integrated pipeline (“Methods”). Ultimately, 20,785 and 26,680 protein-coding genes, corresponding to 24,432 and 28,250 transcripts, were identified in E. furcellata and A. custos (Table 1). A total of 19,963 (96.05%) and 26,308 (93.13%) genes were located on assigned chromosomes of E. furcellata and A. custos. Most genes (97.10% and 94.66% for the two species) successfully matched at least one public database (Table 1). To rule out a possibility that the large number of identified PCGs (e.g., over 26 K in A. custos) were due to insufficient annotation of genomic repeats, we calculated the “function annotation rate” of PCGs for all the 40 insect species which we will use in the following section (Additional file 1: Table S4). If the PCGs in A. custos suffer from a high false-positive rate, then they should have a remarkably lower annotation rate compared to other species. However, we found that the average annotation rate for all species is 87.9% ± 2.3% (mean ± SE and will be 87.7% ± 2.3% if A. custos is excluded), which is much lower than 97.10% in E. furcellata and 94.66% in A. custos. Together with our delicate annotation on genomic repeats (“Methods”), it suggests that the PCGs identified in our species are reliable.

Phylogenetic tree of 36 hemipterans and 4 outgroups

To understand to what extent the genomic differences could account for the phenotypic divergence between zoophagous and non-zoophagous insects, we searched for high-quality and well-annotated hemipteran genomes from public databases and found an additional 34 hemipteran species (“Methods”). Together with four outgroup species from the orders Thysanoptera and Psocodea, we made up a list of a total of 40 species including 6 zoophagous insects, 29 phytophagous insects, and 5 other insects (Fig. 2C). The “other” category has one omnivorous insect, three hematophagous insects, and one species feeding on feathers. Detailed dietary categories of each species are listed in Additional file 1: Table S5.

We performed a phylogenomic analysis with 40 species based on 221 BUSCO proteins (Fig. 2C and “Methods”). The typology of the phylogeny conforms to previous studies [1, 3]. Most bootstrap values were 100%, except for four nodes showing 81% ~ 92% (Fig. 2C). Molecular clock dating analysis (“Methods” and Additional file 1: Table S6) indicated an early Carboniferous origin of Hemiptera (~ 349 Mya; 95% HPD 301.62–392.45 Mya). The divergence of Sternorrhyncha from the remaining Hemiptera occurred ~ 322 Mya, consistent with previous estimates [9, 16, 17]. Heteroptera diverged from Auchenorrhyncha ~ 259.89 Mya (95% HPD 222.61–295.36 Mya). The divergence time between Halyomorpha halys (phytophagous) and the ancestor of two zoophagous species E. furcellata and A. custos was ~ 107.44 million years ago (95% HPD 68.71–143.06 Mya; Cretaceous period) (Fig. 2C). Notably, the six zoophagous species in our analysis did not share a common zoophagous ancestry since the diet trait has reversed several times within this lineage. This is illustrated more clearly in Fig. 1 where the entire Hemiptera phylogeny was displayed.

Association between feeding traits and genomic content of repetitive elements

Across the phylogeny of 40 insects, we first found that the contents of transposable elements (TEs) varied widely in different species (Fig. 2C). The genome sizes of our two zoophagous true bugs E. furcellata and A. custos were larger than most of the other hemipteran insects, and, accordingly, they contained a high level of TEs that accounted for 43.50% and 38.73% of the genome size (Table 1 and Additional file 1: Table S7). Particularly, class II retrotransposon DNA transposons represented the most dominant TEs in the two species (31.4% of E. furcellata’s genome and 24.3% of A. custos’ genome, Additional file 1: Table S7). To investigate whether genome size is related to the TE content, we performed a phylogenetic generalized least squares (PGLS) regression analysis. Among the 40 species, correlation analysis between TE content and genome size confirmed that TE bursts might have driven the expansion of hemipteran genomes (Pagel’s λ = 0.844, R² = 0.266, P = 0.001). Our results are consistent with the findings of a previous study that proposed the important role of retrotransposons in the genome expansion of hemipterans [18]. Next, to test whether TE contents are related to feeding traits, we performed PGLS analysis and obtained an insignificant result (Pagel’s λ = 0.829, P = 0.256) across 40 species. Although a previous literature reported an increased accumulation of DNA transposons in carnivore lineages of bats compared to their herbivorous relatives [19], our negative result in Hemiptera suggests that the reported connection between dietary traits and genomic TE contents might be a lineage-specific trend in bats.

Defining diet-related OGs in 40 studied insects

To investigate the evolutionary dynamics of protein coding genes (PCGs), we systematically identified the ortho-groups (OGs) of coding genes across the 40 insects. All PCGs in these 40 species were grouped into 121,504 OGs (“Methods”). By annotating these OGs based on protein functions, 1752 OGs (1.44%) were characterized as diet-related OGs, including functions related to chemosensory, detoxification, and digestion (Additional file 1: Fig. S3) (“Methods”). Next, to understand the evolution of diet-related genes in the context of the phylogenetic tree and provide evidence for molecular adaptation to feeding traits, we will first explore the dynamic expansion of the diet OGs in Hemiptera. Moreover, considering that the digestion-related genes might be more closely associated with the zoophagous phenotype, our analyses on diet-related genes will be accompanied by parallel results from using the digestion-related genes alone. If consistent patterns emerge from the analysis of these two diet gene groups, it would suggest that our results and conclusions are generally reliable, which are not affected by different levels of stringency.

Diet-related genes are recurrently expanded in phylogenetic nodes directing to zoophagous insects

Among the phylogeny of the 40 insects used in this study, we determined the expanded and contracted OGs in each evolutionary node (Additional file 1: Fig. S4). We paid special attention to the nodes containing at least one extant zoophagous species, such as nodes no. 61 and no. 63, and the ancestral node of Hemiptera (Fig. 3A for all diet-related genes and Additional file 1: Fig. S5A for digestion-related genes only). The many other nodes containing only the phytophagous species, such as the ancestor of Sternorrhyncha and any inner nodes, were used as controls. The branches leading to hematophagous and omnivorous species were also investigated. We calculated the fraction of diet OGs among the expanded OGs in each node. We found that zoophagous species had the highest fraction of diet OGs. Phytophagous species had significantly lower fractions than zoophagous species. In contrast, hematophagous species, despite the closeness in traits to zoophagy, also exhibited a significantly lower proportion of diet OGs (Fig. 3B and Additional file 1: Fig. S5B). This pattern holds true when we counted the number of total genes instead of OG numbers (Additional file 1: Fig. S6 and Additional file 1: Fig. S7). In light of the phylogeny, these results elaborate the evolutionary dynamics of diet-related OGs and establish a correlation between these genes and the zoophagous phenotype.

Fig. 3.

Gene expansion analysis under phylogenetic context. All diet-related genes were used here. Analogous results were observed when using the digestion-related genes only (Additional file 1: Fig. S5). A Gene expansion in each node of the phylogenetic tree of 40 species. Major taxa were labeled by stars. Diets of each species were indicated by circles beside species names. Pie charts at each node displayed the fraction of diet (pink) and non-diet (blue) OGs among the expanded OGs. B The boxplot shows the fraction of diet OGs for each node, categorized by zoophagous (red), phytophagous (blue), hematophagous (gray), and omnivorous (gray) species. The hematophagous group includes R. prolixus, T. rubrofasciata, and their common ancestor. The omnivorous group contains only one species (A. lucorum), and, therefore, statistical comparisons are not applicable. The P-value for pairwise comparisons was calculated using a two-tailed Wilcoxon rank-sum test. C Fraction of recurrent expansion. For the expanded OGs at node no. 53, the number of OGs expanded at earlier nodes no. 57, no. 61, or no. 63 was calculated. These represent the recurrent expansion in node no. 53. P-value was calculated by two-tailed Fisher’s exact test. D Fraction of OGs with recurrent expansion in node no. 53. X-axis is the number of nodes among no. 57, no. 61, and no. 63 that bear the same expanded OGs as in node no. 53. For the comparison between diet and non-diet OGs, their frequency spectrums of 1, 2, 3 were compared by chi-square test to obtain a P-value

Then, we wonder whether there are particular OGs that are recurrently expanded from the ancestral to recent nodes. We quantitatively counted these cases by focusing on the expanded OGs at node no. 53 (the common ancestor of E. furcellata and A. custos). If the same OG was expanded in either of the three earlier nodes (no. 57, no. 61, or no. 63), then this OG was defined as a recurrently expanded OG in node no. 53. Interestingly, we found that diet OGs were more likely to be recurrently expanded in node no. 53 compared to non-diet OGs, and this difference is significant (Fig. 3C and Additional file 1: Fig. S5C). Moreover, for each expanded OG in node no. 53, we interrogated how many of the three ancestral nodes (no. 57, no. 61, no. 63) showed earlier expansion events. For example, N = 3 means that expansion occurred in all the nodes of no. 53, no. 57, no. 61, no. 63, and thus, a higher frequency represents stronger signals of recurrent expansion. Interestingly, we found that the frequency of recurrent expansion was significantly higher for diet OGs than non-diet OGs (Fig. 3D and Additional file 1: Fig. S5D). Taken together, our findings suggest that diet OGs (or the digestion-related genes in particular) in zoophagous species are under the selection pressure to recurrently increase their gene numbers, presumably in order to adapt to their specialized environment or feeding traits. Nevertheless, the causality beyond the correlation between genotype and phenotype remains to be further investigated.

Diet-related genes are fastevolving during the reversal from phytophagy to zoophagy

Reversal of feeding trait is a frequently seen and worth investigating phenomenon in Heteroptera. While the ancestor of infraorder Pentatomomorpha became phytophagous, a reversal to zoophagy occurred in Asopinae, a subfamily containing two of our sequenced species E. furcellata and A. custos (Fig. 1). However, the genetic bases underlying such reversal remain poorly understood. Using the 12 Heteroptera species studied in this work (6 zoophagous, 3 phytophagous, and 3 others), coupled with a phytophagous outgroup Homalodisca vitripennis, we set out to test whether there are differential evolutionary rates among different species or among genes of different categories that are associated with the reversal of feeding trait.

We calculated the evolutionary rates (dN/dS) for each lineage using the branch model in PAML (“Methods”). Overall, six zoophagous species did not exhibit higher evolutionary rates than seven non-zoophagous species (Fig. 4A). But when we focused on the feeding trait–reversal node within Pentatomomorpha, we found that E. furcellata and A. custos showed significantly higher evolutionary rates than the other three phytophagous species (P = 0.047, Fig. 4A). These observations suggest that although zoophagy itself does not directly associate with the evolutionary rate of a species, the switch of the feeding trait might be driven by selective pressure and lead to fast molecular evolution.

Fig. 4.

Selection analysis of different species and genes. A A phylogenetic tree of the species chosen to conduct the selection analyses. Evolutionary rates (dN/dS) of each species were shown as a barplot. T-test was used to determine the significant difference shown in the plot. *P < 0.05. B Barplots showing the -log₂FDR values. Significance was determined by Wilcoxon rank-sum tests between dN/dS of diet genes versus dN/dS of non-diet genes. Species with diet genes > non-diet genes have bars to the right (orange). Other species have bars to the left (dark blue). Nonsignificant bars are in gray. C Ratio of median dN/dS values of positively selected genes/other genes. This ratio at the common ancestor node of E. furcellata and A. custos was also displayed by the triangle in the plot. Significance was determined by Wilcoxon rank-sum tests between selected genes versus other genes. D Comparison of expression level of positively selected genes in E. furcellata, A. custos, and two other species. RPKM was log2-transformed. P-value was calculated using T-test. E An example of a selected gene of digestion enzymes in insects. Trypsin plays a key role in breaking down large protein molecules into smaller peptides and amino acids, facilitating their absorption. It is central to protein digestion and also contributes to the insect’s immune response [20, 21]

Next, we investigated the evolutionary rates of different groups of genes. To measure whether diet-related genes had differential evolutionary rates with non-diet genes in a particular species, we calculated the ratio of mean dN/dS of diet genes divided by mean dN/dS of non-diet genes and determined the significance between the two groups. A ratio significantly higher than one suggests an overall faster evolutionary rate for diet genes. Among 13 species, only 3 zoophagous species, E. furcellata, A. custos, and Orius laevigatus, had significantly higher evolutionary rates for diet genes than non-diet genes (Fig. 4B). For E. furcellata and A. custos with a reversal of the feeding trait, the fast evolution of diet-related genes again indicates a potential link between genomic changes, selective pressure, and such phenotypic transition.

To further understand whether positive selection played a role in the genome evolution during diet reversal, we defined 181 positively selected genes at the common ancestor node of E. furcellata and A. custos using the aBSREL model of software Hyphy (“Methods”). Not surprisingly, this set of genes showed significantly higher evolutionary rates than the 2198 non-positively selected genes in the common ancestor node of E. furcellata and A. custos (Fig. 4C). Moreover, in the branch leading to A. custos, this trend is also observed. In contrast, in other species or evolutionary nodes, the 181 positively selected genes did not show higher evolutionary rates than the remaining genes (Fig. 4C). These patterns suggest an overall selection force acting on the diet-related genes in E. furcellata and A. custos.

Then, by incorporating the salivary gland transcriptomes which will be used in the following section, we found that positively selected genes in E. furcellata and A. custos, including nine diet-related genes, had significantly higher expression levels compared to other non-zoophagous species in Pentatomomorpha (Fig. 4D). This indicates that fast evolution of genome sequence and high expression levels of particular genes might both be favorable. There might be synergistic effects between genome evolution and transcriptome regulation to facilitate adaptation of feeding traits. Despite that, a causality beyond correlation needs to be further validated by functional experiments. An example of a fast-evolving, positively selected, and highly expressed gene related to feeding traits is trypsin (Fig. 4E). Trypsin plays a key role in protein digestion and contributes to the insect’s immune response.

Dynamic regulation of salivary gland gene expression in zoophagous insects with reversed feeding trait

Since we observed a higher expression of positively selected genes in E. furcellata and A. custos compared to phytophagous Pentatomomorpha, this leads to the hypothesis that diet-related genes are more highly expressed in zoophagous species than in phytophagous ones, which can be tested by a direct comparison of gene expression profiles. We sequenced the transcriptomes of salivary glands for two zoophagous species, E. furcellata and A. custos, and obtained the salivary gland transcriptomes of an additional 11 hemipteran insects from a public database (“Methods,” Additional file 1: Table S8). Only the two species generated in this study (E. furcellata and A. custos) were zoophagous, while the 11 publicly available species included 9 phytophagous insects, 1 omnivorous insect, and 1 hematophagous insect. We found that the diet-related genes were moderately expressed in all species, with median RPKM values ranging from 0.352 to 7.38 (Fig. 5A for all diet-related genes and Additional file 1: Fig. S8A for all digestion-related genes only, and Additional file 1: Fig. S9 and Additional file 1: Fig. S10 for different strategies in calculating gene expression, see “Methods”). Then, to provide a quantitative evolutionary dynamics of gene expression, we employed a phylogeny-based software, CAGEE, to determine the up- and downregulation of genes (Fig. 5B and Additional file 1: Fig. S11). In E. furcellata, 10.5% of the diet-related genes were significantly up-regulated, and only 3.9% of the non-diet genes were significantly up-regulated; the difference between these two percentages was significant (P = 7.44e-20; Fig. 5C and Additional file 1: Fig. S8B; Additional file 1: Fig. S12A and Additional file 1: Fig. S13A). Moreover, three up-regulated diet genes in E. furcellata (including chemosensory receptor, CUB domain protein, and ABC transporter) are positively selected, exhibiting synergistic effects at the genome and transcriptome level. The same goes for the zoophagous species A. custos, where 10.4% and 3.0% significantly up-regulated genes were obtained for diet genes and non-diet genes (P = 3.30e-27; Fig. 5C and Additional file 1: Fig. S8B; Additional file 1: Fig. S12B and Additional file 1: Fig. S13B). These results suggest that under a phylogenetic context, the two zoophagous species tend to upregulate diet-related genes in salivary glands to align with their reversed feeding trait. Given the trait similarities with zoophagy, we also analyzed the fractions of significantly up-regulated diet-related and non-diet OGs in Apolygus lucorum (omnivorous) and Triatoma rubrofasciata (hematophagous). In both species, the proportion of significantly up-regulated diet-related OGs was notably higher than that of non-diet OGs (Additional file 1: Fig. S14). This pattern mirrors what has been observed in zoophagous species.

Fig. 5.

Evolutionary dynamics of gene expression in salivary glands of representative hemipterans. All diet-related genes were used here. Analogous results were observed when using the digestion-related genes only (Additional file 1: Fig. S8). A Boxplot showing the expression of diet-related (red) and non-diet (blue) genes in all species. B Numbers of significantly up- (red) and down-regulated (blue) genes under phylogenetic context. C Barplots showing the fractions of significantly up-regulated diet or non-diet genes in E. furcellata and A. custos. P-values were calculated by Fisher’s exact tests. D Six-hundred thirteen genes showing the top two highest expressions in E. furcellata and A. custos. E Fraction of the above category of genes among all genes. The fractions in diet and non-diet genes were compared with Fisher’s exact test. F Heatmap of seven diet-related genes with top expression in E. furcellata and A. custos. G Functional schematic of the top-expressed diet-related gene. Carboxypeptidase enzymes, including carboxypeptidase A (CPA) and B (CPB), play an essential role in protein digestion by sequentially cleaving C-terminal amino acids from endopeptidase-generated peptides, such as trypsin. This exopeptidase activity completes proteolysis and produces absorbable amino acids via specific transporters [22]

To narrow down potential functional genes involved in the adaptation to diet reversal in E. furcellata and A. custos, we carried out a more stringent filter in the expression profile. We ranked the expression of each gene across 13 species and looked for genes where E. furcellata and A. custos had the top two highest expressions (Fig. 5D and Additional file 1: Fig. S8C). We obtained 613 such genes, including 35 diet genes and 578 non-diet genes. Interestingly, the fraction of genes belonging to this group is significantly higher for diet genes (3.3%) than for non-diet genes (1.1%) (P = 2.716e-8; Fig. 5E and Additional file 1: Fig. S8D), which again indicated the association between diet genes and the feeding trait. Moreover, 7 of the 35 diet genes were also identified as significantly up-regulated genes in E. furcellata or A. custos under phylogenetic context (Fig. 5F and Additional file 1: Fig. S8E). One of the seven genes, OG0010419 (carboxypeptidase A, CPA), was also significantly up-regulated in the common ancestor node of E. furcellata and A. custos (Fig. 5G). Carboxypeptidase enzymes are primarily involved in protein digestion, peptide hormone processing, and chitin synthesis in insects [22].

In this part, we systematically analyzed the salivary gland gene expression profile in various Heteroptera species and identified a set of up-regulated genes in E. furcellata and A. custos under phylogenetic context. By further linking the expressional change with positive selection, we suggested the synergistic effect of genome evolution and transcriptome regulation and provided candidates of crucial genes involved in the diet reversal of zoophagous insects. Again, we acknowledge that a causality beyond correlative observations remains to be further investigated by functional validation.

A nuanced classification of phytophagy reveals dynamic evolution of feeding traits in Sternorrhyncha

In the above analysis on zoophagous species, in order to use phytophagous species as a control, we treated phytophagy as a single uniform trait. However, phytophagy can be subdivided into more refined categories [10]. To investigate the genomic changes associated with phytophagous species during evolution, we performed ancestral state reconstruction based on a nuanced classification of phytophagous types (Additional file 1: Fig. S15). Six subtypes of phytophagy were classified. Our results demonstrate that the ancestral feeding trait of Hemiptera was vegetative feeding, a trait retained in the basal clade Sternorrhyncha (node no. 35). While most Sternorrhyncha species are strict vegetative feeders, we observed that in Aphididae (node no. 27), different feeding traits have evolved. Three species, Aphis craccivora, Rhopalosiphum maidis, and Sitobion miscanthi, have independently acquired the ability to feed on both seeds and vegetative tissues. Additionally, Aphididae species Daktulosphaira vitifoliae evolved root-feeding behavior (Additional file 1: Fig. S15). For other suborders, the ancestral node of Auchenorrhyncha also exhibited vegetative feeding, and this trait has been maintained in all four Auchenorrhyncha species analyzed in this study. In contrast, the ancestral node of Heteroptera was zoophagous, and a transition to seed feeding occurred at the ancestor of the Pentatomomorpha clade (Additional file 1: Fig. S15).

To further explore the molecular mechanisms underlying these feeding transitions in phytophagous species, we focused on genomic changes at key evolutionary nodes in Sternorrhyncha. In A. craccivora, R. maidis, and S. miscanthi, where seed feeding was independently acquired, 2429, 441, and 924 OGs were expanded, among which 106, 21, and 13 OGs showed significant results (P < 0.05). Thirteen OGs were significantly expanded in at least 2 of the 3 species. Two OGs, OG0001624 (sensory perception of taste) and OG0000256 (ABC transporter), belong to the diet-related OGs as we defined (Additional file 2: Table S9). In root feeder D. vitifoliae, 904 OGs were expanded, and 14 of them were significant. Six of the significant OGs were diet related, including trypsin genes (Additional file 2: Table S9).

These results suggest that, although Sternorrhyncha species are mainly phytophagous, a more nuanced classification of phenotype might reveal the involvement of diet-related OGs during the evolution of key nodes (see “Discussion”). Our analysis offers a deeper understanding of the molecular basis underlying the transition and adaptation of feeding traits.

Discussion

Summary of main findings

Understanding the molecular basis underlying the biodiversity of the tree of life is one of the ultimate goals of life science. Insects, particularly those in the order Hemiptera, represent a highly diverse group with complex morphologies, behaviors, and physiologies. The evolution of diet traits, particularly the shift to zoophagy, remains a fascinating question in this order. Answers to these questions can help clarify basic principles in evolutionary biology.

In this study, we have systematically investigated the multi-level molecular basis underlying the evolution of diet traits in Hemiptera. With our high-quality genome assemblies of E. furcellata and A. custos together with the publicly available data, we first found a significant and recurrent expansion of diet-related genes (particularly digestion-related genes) in zoophagous species, providing a correlation between genotype and phenotype. Then, by focusing on E. furcellata and A. custos, the evolutionary node in Pentatomomorpha that reversed from ancestral phytophagy to current zoophagy, we found faster evolutionary rates and dynamically up-regulated gene expression for part of the diet-related genes. These results suggest a potential synergistic regulation at genomic and transcriptomic levels to govern the evolution of diet traits.

Notably, in several previous literatures, zoophagous Heteroptera species were described as “predaceous” or “predatory” [9, 10]. In this study, considering that predation is a behavioral suite of traits that includes a complex set of physiological and neurological interactions that allow the sensing, overpowering, and consumption of prey, we used “zoophagy” and “zoophagous” to specifically refer to the consumption of animal material.

Repetitive expansion, selection, and adaptation of diet-related genes in zoophagous species

We proposed the collective effects of genomic and transcriptomic variations to shape the biodiversity of feeding traits. The particular set of diet-related genes mainly includes detoxifying, metabolic, chemical sensory, and digestive genes. For example, trypsin is positively selected in the nodes directing to zoophagous species, and its expression level in salivary glands is also significantly elevated in two zoophagous species E. furcellata and A. custos. Trypsin plays a primary role in digestion across insects. Besides digestion, trypsin-like enzymes also participate in other physiological processes, such as molting [23], tissue remodeling [24], innate immunity, and activation of enzyme precursors of chitinase. Previous studies have also shown positive selection on trypsin in zoophagous insects or spiders [25–27]. Our study makes several steps forward by discovering the transcriptomic evolution as well as the recurrence of selection on these genes. The selection of trypsin may not be exclusively due to dietary shifts but could also reflect broader evolutionary pressures.

Diet shifts have occurred multiple times in Hemiptera, as zoophagous species do not form a monophyletic group. The fact that the ancestral node of Pentatomomorpha was actually phytophagous further supports the multiple independent origins of zoophagy. The question is whether different clades of zoophagous insects share similar convergent molecular signals. Convergent evolution occurs when distantly related species are subjected to similar selection pressures [28]. At the molecular level, convergent evolution could be driven by factors functioning in development or physiology [29], genetic modifications [30, 31], and similar genes or pathways [32–34]. Our results do not directly test convergent evolution because the six zoophagous species do not have a parallel relationship. Our sequenced species, E. furcellata and A. custos in Pentatomomorpha, experienced a zoophagous–phytophagous–zoophagous reversal. Nevertheless, we found that the different zoophagous species do share a similar molecular basis, such as the expansion of a particular set of diet-related genes and regulation in salivary glands. Future studies with more genomes from zoophagous species may reveal convergent molecular mechanisms underlying diet-related traits.

Diet-related genes were also possible to contribute to non-zoophagous diet traits

Although diet OGs tend to expand in gene number and exhibit higher expression levels in zoophagous species, some exceptions exist. Certain diet OGs are also expanded or up-regulated in phytophagous species. This suggests that the recruitment of these genes during evolution is not exclusive to zoophagy. Indeed, some diet-related genes are involved in both phytophagy and zoophagy, contributing to the evolution of these traits. For example, digestion-related genes in Sternorrhyncha show how genes associated with one feeding strategy can be co-opted for another. Moreover, even for the expanded diet OGs in zoophagous species, particularly members of these OGs, might also exist in phytophagous species, indicating that their potential role in phytophagous diet traits should not be overlooked. This phenomenon, termed exaptation, typically occurs when a species with some preexisting functional genes has adapted to novel niches. Taken together, these findings suggest that while distinct gene sets may underlie different feeding strategies, some genes possess functional versatility that enables their contribution to multiple dietary adaptations.

Future perspectives

Future studies could benefit from expanding the sample size of zoophagous species, particularly by incorporating species that have independently evolved zoophagous traits across different evolutionary lineages. This would enable a more precise evaluation of the convergence or divergence in the evolution of zoophagous traits. Moreover, Pentatomomorpha includes fungivorous lineages as depicted in Fig. 1, which are not included in our analysis but might potentially complicate the ancestral state reconstructions and comparisons between the phytophagous and zoophagous lineages. Nevertheless, we maintain that the current phylogenetic framework, along with the ancestral trait reconstruction at key nodes, is robust and supported by previous literature [9, 10, 14, 16], and the definition of diet transition in this study should be generally sound. It is compelling to incorporate fungivorous species in future analyses to gain a more comprehensive understanding of diet evolution in Heteroptera.

Additionally, ecological factors, such as food availability, temperature, and humidity, should be integrated into future studies, as these environmental factors can influence feeding behaviors and preferences in insects [35–38]. Integrating ecological and genomic approaches would provide a more comprehensive understanding of the evolution of feeding traits in Hemiptera.

Apart from bioinformatic analysis, functional validation of diet-related genes in particular species is highly appreciated. With the well-established CRISPR system in model insects or the RNAi knockdown system in a few non-model hemipteran insects [39], one may silence a particular diet gene to test whether there are phenotypic changes in their feeding traits like diet spectrum, hunting behavior, and chemical perception. The validation of single or several genes would largely support the bioinformatic pipeline and evolutionary theory proposed in this study.

In addition, our analyses mainly focus on the evolution and regulation of coding genes, but the noncoding genes or intergenic regulatory regions are not investigated. Promisingly, when the genomes of more species are available (especially when the numbers of zoophagous versus phytophagous species are balanced), a whole genome alignment analysis is favorable to identify conserved and non-conserved genomic elements that correlate with diet traits. Then, similar experimental validation could be carried out to examine the effect of regulatory elements on insect behavior, physiology, fitness, and other phenotypes.

Conclusions

In conclusion, our work shows potential genetic and molecular bases underlying the evolution and reversal of feeding traits in Heteroptera and proposes a synergistic regulation at both genomic and transcriptomic levels. Our study provides a potential connection between genotype and phenotype and advances the understanding of the adaptive evolution of zoophagous insects.

Methods

Sample rearing and collection

E. furcellata and A. custos were collected in Kunming, Yunnan Province, China (102.80°E, 24.96°N). All samples were reared from the third to fifth instars of Spodoptera litura under controlled laboratory conditions. The environment was maintained at a temperature of 26 ± 1 °C, a relative humidity of 70 ± 5%, and a photoperiod of 14–h light/10–h dark. To ensure a sufficient yield of nucleic acids for sequencing, E. furcellata and A. custos adults were used for extractions. An individual was promptly transferred to collection tubes, flash-frozen on liquid nitrogen, and subsequently stored at − 80 °C until further processing.

DNA and RNA extraction

Given the greater body size of the two species and the need for nucleic acids by sequencing technologies, the genomic DNA was extracted from a female adult using a Blood and Cell Culture DNA Midi kit (QIAGEN, Germany). The quantity and integrity of DNA were determined by a Qubit 2.0 (Thermo Fisher Scientific) and 0.75% agarose gel electrophoresis, respectively.

For E. furcellata, pooled adults comprising one female and one male were prepared for RNA extraction. Total RNA was extracted with TRIzol reagent (Thermo Fisher Scientific, USA). For A. custos, RNA-seq data were downloaded from NCBI SRA under accession numbers SRR10098905, generated from a pooled library containing adults and nymphs. These two sets of RNA-seq data were used for genome annotation of the two species. The RNA-seq used for expressional analysis will be described in later parts.

Genome and transcriptome sequencing of E. furcellata and A. custos

The genomes of E. furcellata and A. custos were sequenced by a combination of Illumina sequencing, HiFi sequencing (PacBio), and Hi-C sequencing.

A DNA library with 150-bp paired-end (PE) reads was constructed and sequenced using the Illumina NovoSeq 6000 platform. According to PacBio’s standard protocol of SMRTbell Template Prep Kit 2.0 (Pacific Biosciences, USA), a ~ 15-kb genomic library was also performed and sequenced on the Pacific Biosciences Sequel II platform. After being filtered by ccs v5.0.0 (https://github.com/PacificBiosciences/ccs), highly accurate single-molecule consensus (HiFi) reads were obtained for E. furcellata and A. custos.

To further improve the continuity of the assembled genome and anchor the assemblies into chromosomes, an E. furcellata female and an A. custos female were used for the library of chromosome conformation capture (Hi-C). In brief, fresh tissues from living insects were cross-linked using a 2% formaldehyde isolation buffer. The purified nuclei were subsequently digested with DpnII (NEB), and biotinylated nucleotides were used to repair tails. After ligated DNA was interrupted into 350-bp fragments using a focused ultrasonicator, the Hi-C library was sequenced on the Illumina NovaSeq 6000 platform.

To obtain transcriptomic evidence for genome annotation, we constructed an RNA-seq library and full-length transcript isomers (Iso-Seq) library. The paired-end library was constructed via the TruSeq RNA Library Preparation kit (Illumina, USA), followed by sequencing with the Illumina NovoSeq 6000 platform. The Iso-Seq library with an insert of 1–10 kb was performed using the SMRTbell Express Template Prep Kit 2.0 (Pacific Biosciences, USA), and the target size was sequenced on the PacBio Sequel II platform.

Genome assembly of E. furcellata and A. custos

Illumina short reads were used for genome characteristics estimation (genome size, genome heterozygosity, and repeat content) by a kmer-based statistical analysis using JELLYFISH v2.1.3 [40] and GenomeScope v2.0 [41].

We used a common pipeline to assemble the genomes of E. furcellata and A. custos. Contigs were assembled by Hifiasm v0.13 [42] with the following settings (− 2 -a 70) and removed heterozygous duplication by Purge_dups v1.2.3 [43] tool. Then, Hi-C data were mapped to the contig genome by BWA-MEM v0.7.17 [44] with the following parameters (mem -SP5M). The DpnII site was generated using the script “generate_site_position” in Juicer v1.5 [45]. Finally, the contigs were scaffolded, ordered, and clustered into chromosomes via the 3D-DNA pipeline with default parameters [46]. Assembly errors detected during the Hi-C scaffolding were further corrected visually using Juicebox v1.11.08 (https://github.com/aidenlab/Juicebox).

We evaluated the completeness of the chromosomal-level genome using Benchmarking Universal Single-Copy Orthologs (BUSCO v5.2.2) under the insecta_odb10 [45]. Furthermore, BWA-MEM v0.7.17 [44] and Merqury v1.1 [47] were used to assess the accuracy and base error of E. furcellata and A. custos genomes.

Genome annotation of E. furcellata and A. custos

To systematically characterize repetitive sequences in the E. furcellata and A. custos genomes, we employed the extensive de novo TE annotator (EDTA) pipeline [48]. Briefly, long terminal repeat retrotransposons (LTR-RTs) were identified by an integrated approach combining LTR_FINDER [48], LTRharvest [49], and LTR_retriever [50]. DNA transposons were classified by TIR-Learner [51] and HelitronScanner, respectively [52]. To annotate known repetitive sequences, we performed homology-based searches using RepeatMasker v4.0.7 [53] and RepeatProteinMasker v4.0.7 [53] against the repbase v21.12 [54]. De novo repeat libraries were constructed using RepeatModeler to identify repetitive elements. Finally, tandem repeats were identified using Tandem Repeats Finder v4.07b [55] with the parameters as follows: 2 7 7 80 10 50 500 -f -d -m.

Genes in the assembled genome were predicted by homology-based, transcriptome-based, and de novo-based methods.

1. Homology-based predictions involved downloading homologous proteins and transcripts from Gerris buenoi, A. lucorum, Cimex lectularius, Orius laevigatus, Rhodnius prolixus, T. rubrofasciata, H. halys, Riptortus pedestris, Oncopeltus fasciatus (covering all available public genomes in Heteroptera), and the model insect Drosophila melanogaster (NCBI, https://www.ncbi.nlm.nih.gov/; InsectBase v 2.0). The IsoSeq v 3.4.0 workflow was used to generate high-quality full-length transcripts with quality parameters of 0.99 (https://github.com/PacificBiosciences/IsoSeq).

2. For transcriptome-based methodology, RNA-seq data were mapped to the reference genome using HISAT2 v2.2.1 [56] and assembled into transcripts using StringTie v2.4.0 with default parameters [57]. Homologous proteins and transcripts were then aligned using Exonerate v 2.4.0 for training gene sets [58]. Meanwhile, a sorted and mapped BAM file of RNA-seq data was transferred to a hint file using the bam2hints program in AUGUSTUS v3.2.3 (–intronsonly –in = rnaseq.bam –out = hints.gff) [59].

3. Self-trained sets were combined with hint files as inputs for AUGUSTUS to predict de novo coding genes from the assembled genome. Finally, the homology-based, transcriptome-based, and de novo-based results were merged in MAKER v2.31.10 to generate a high-confidence gene set [60].

Gene structure and annotations were determined using eggnog-mapper v2.0.1 [61], InterProscan v5.0 [62], BLAST v2.2.28 [63], and HMMER v3.3.2 [64] to search against NCBI nonredundant protein (nr), Gene Ontology (GO), Clusters of Orthologous Groups of Proteins (COG), Kyoto Encyclopedia of Genes and Genomes (KEGG), Swiss-Prot, and PFAM.

Phylogeny and divergence time analysis

To investigate to what extent the genomic differences could account for the phenotypic divergence between zoophagous and non-zoophagous insects, we obtained all available public genomes of 34 hemipterans and an additional four outgroup species (two Thysanoptera and two Psocoptera) from NCBI, InsectBase, Figshare, and GigaDB (Additional file 1: Table S4).

A molecular phylogeny tree was constructed with two newly assembled genomes (E. furcellata and A. custos) plus 38 species above using BUSCO proteins. In brief, BUSCO sets are defined as collections of near-universal single-copy genes, which are rarely lost or duplicated. We downloaded 1367 insect coding genes from the database insecta_odb10 in BUSCO v5.2.2 [45]. Orthologous protein sequences of each candidate BUSCO group were aligned using MAFFT v7. 487 with auto strategy [65]. Sequence alignments were trimmed and concatenated by TRIMAL v1.4 [66] and FASCONCAT-G v1.0.4 [67]. To reduce the possible systematic errors in large genomic data sets, we calculated the compositional heterogeneity of loci by BACOCA v1.1 [68]. Next, orthologous groups with single-copy orthologues present in 90% of the species were used for the phylogenetic tree using IQTREE v2.2.0 under the following parameters: -m MFP -B 1000 –alrt 1000 [69].

We selected fossil records as minimum age calibration points for some lineages (listed in Additional file 1: Table S6 [16, 70–73]). Then, the age of each node was estimated using a correlated rates clock in MCMCTREE of PAML version 4.4 [74].

Ortho-group (OG) identification

To cluster families of protein-coding genes, we extracted the longest protein sequences from the 40 species above. OrthoFinder v2.5.4 was used to identify orthologous and paralogous genes across these protein sequences with the parameters “-a blast -M msa” [63]. Functional annotation of proteins was carried out via InterProScan v5.0 [62] against GO, InterPro, and PFAM databases. Orthologous groups (OGs) were analyzed using KinFin v1.0 by providing function annotation. In the next step [75], these OGs were further used for defining the OGs involved in feeding habits (the diet-related genes).

In the identification of OG with significantly different gene numbers in zoophagous and phytophagous species, the gene numbers in each OG were transformed with log₂(OG number + 1), and then T-tests were used to examine the difference between zoophagous and phytophagous species. P-values were adjusted for multiple testing correction [76].

Definition of diet-related OGs

Candidates of diet-related genes were obtained by searching previous literatures with keywords [77–80]. The terms involved detoxification, chemosensory, and digestion adaptation. For these gene families, their domain information was downloaded from the Pfam database: (1) Detoxification-related families include cytochrome P450 (P450s, PF00067), glutathione S-transferase (GSTs, PF00043), carboxylesterases (CCEs, PF00135), and ATP-binding cassette transporters (ABCs, PF00005); (2) chemosensory-related families include ionotropic receptors (IR, PF00060), gustatory receptors (GR, PF06151/PF08395), odorant receptors (OR, PF13853/PF02949), odorant-binding proteins (OBP, PF01395), chemosensory proteins (CSP, PF03392), and sensory neuron membrane proteins (SNMPs, PF01130); and (3) digestion-related families include serine protease (PF00089), serpin (PF00079), carboxypeptidase (PF00246), aspartic protease (PF00026), lipase (PF00151/PF01764/PF06350/PF04083/PF01734/PF00657/PF13472), alpha-amylase (PF00128), thioredoxin (PF00085), CUB (PF00431), and Ptu family (PF08117).

Expansion and contraction of gene families

Gene-family expansion and contraction were estimated using CAFÉ v4.2 with parameters “lambda -s -t,” based on maximum likelihood and reduction methods [81]. Phylogenetic tree topology and branch lengths were considered when inferring the significance of changes to gene-family size in each branch. KO-Based Annotation System (KOBAS) program was used to analyze GO and KEGG enrichment [82].

Salivary glands collection and transcriptome sequencing

Five adults from both sexes of E. furcellata and five adults from both sexes of A. custos were dissected on dry ice after anesthetization by chilling under a stereomicroscope, separately. Then, salivary glands (SGs) were isolated and snap-frozen in liquid nitrogen. The pooled salivary gland tissues from each species were prepared for RNA extraction. Total RNA was extracted with TRIzol reagent (Thermo Fisher Scientific, USA). A 150-bp paired-end RNA seq library was constructed and sequenced on the Illumina 6000 platform.

Besides the new transcriptomic data from 2 zoophagous species with salivary glands, we obtained available transcriptome data with salivary glands of an additional 11 non-zoophagous insects in Hemiptera from NCBI (9 species with phytophagous and 2 species of other feeding types [1 omnivorous and 1 hematophagous]; Additional File 1: Table S8). The sample information, including the gender and developmental stages of the salivary gland samples, was also provided in the above supplementary table.

Gene expression and differential expression analysis

RNA-Seq reads from salivary glands were mapped to reference genomes using STAR v2.7.6a with default parameters [83]. Reads count for each gene was calculated with FeatureCounts v0.3 [84]. Gene expression was measured by RPKM (reads per kilobase per million mapped bases). Only the exonic reads were used to calculate RPKM.

The CAGEE program was then employed to infer evolutionary patterns of gene expression changes (up- or downregulation) across the phylogeny. The method has been widely applied to diverse studies [85, 86]. To minimize potential bias introduced by gene copy, we further performed an analysis using only the longest transcript (highest RPKM) per OG as a representative expression value [87].

Selection analyses for genes and sites

To understand the natural selection pressures between zoophagous and non-zoophagous, we estimated the rates of nonsynonymous (dN) to synonymous (dS) substitution rates in OGs across 13 species, including 12 heteropteran species and the outgroup H. vitripennis. We first identified single-copy gene families for each clade. Protein sequences of each gene family were aligned by MAFFT v7. 487 with auto strategy [65]. The amino acid alignments were transformed into codon-based nucleotide alignments by Pal2Nal [88]. And then evolutionary rates (dN/dS) for each lineage were inferred using the branch model in PAML v4.4 [74]. We then calculated dN/dS for each OG and compared evolutionary rates between diet OGs and non-diet OGs using Wilcoxon rank-sum tests. Additionally, the adaptive branch-site random effects likelihood (aBSREL) method implemented in Hyphy was employed to detect positive selection along specific branches [89]. The chi-square program was performed to calculate likelihood ratio tests (LRTs) with P < 0.05 defined as a significant difference.

Ancestral state reconstructions

We used ancestral state reconstruction (ASR) to estimate the evolution of feeding strategies across the phylogeny. First, we quantified the phylogenetic signal of the diet trait by calculating Pagel’s lambda (λ) using the phytools package [90]. ASR was then performed using the ace function in the ape package [91]. The best-fitting model of trait evolution was selected by comparing the Akaike information criterion (AIC) scores, with the equal rates (ER) model. Finally, the ancestral state likelihoods were visualized as pie charts on the phylogeny together using the ggtree package [92].

Abbreviations

AA

Amino acid

CDS

Coding sequence

dN

Nonsynonymous substitution rate

dS

Synonymous substitution rate

TE

Transposable element

LTR

Long terminal repeat

BUSCO

Benchmarking Universal Single-Copy Orthologs

OG

Ortho-group

RPKM

Reads per kilobase per million mapped reads

Authors' contributions

Conceptualization & supervision: W.C. and H.L. Data acquisition: F.G. and X.L. (sample collection); T.Z. and S.W. (DNA/RNA extractions and library preparations); L.M. (data organization). Data analysis: Y.D. and L.M. Interpretation of data: Y.D., L.M., F.G., X.L., T.Z., S.W., F.S., L.T., W.C., and H.L. Writing – original draft: Y.D. and L.M. Writing – review & editing: Y.D., L.M., F.G., X.L., T.Z., S.W., F.S., L.T., W.C., and H.L. All authors read and approved the final manuscript.

Funding

This study is financially supported by the National Natural Science Foundation of China (No. 31922012); the Pests and Diseases Green Prevention and Control Major Special Project (No. 110202201018[LS-02]); Sanya Yazhou Bay Science and Technology City (No. SYND-2022–04); Expert Workstation in Zhaotong, Yunnan (No. 2021ZTYX05); the 2115 Talent Development Program of China Agricultural University; and the Postdoctoral Fellowship Program of CPSF under Grant Number (No. GZC20241945).

Data availability

For Eocanthecona furcellata, the PacBio long-read sequencing data were deposited in the NCBI (https://www.ncbi.nlm.nih.gov/) SRA database under accession number SRR22405948 [93]. The Hi-C data are available through the NCBI SRA database under accession number SRR22405944 [94]. The Illumina short-read DNA-sequencing data are available in NCBI SRA under accession number SRR22405945 [95]. The transcriptome data are available in the NCBI SRA under accession numbers SRR22405946 and SRR22405947 [96, 97]. The salivary gland transcriptome data are deposited in the NCBI SRA database under accession number SRR35264038 [98].

For Arma custos, the PacBio long-read sequencing data were deposited in the NCBI SRA database under accession number SRR27748988 [99]. The Hi-C data are available through the NCBI SRA database under accession number SRR27748986 [100]. The short-read DNA-sequencing data are available in NCBI SRA under accession number SRR27748987 [101]. The salivary gland transcriptome data are deposited in the NCBI SRA database under accession number SRR35264468 [102]. The chromosome-level genome assembly sequence is available at NCBI GenBank through accession number JBEJUF000000000 [103].

All other genomic resources analyzed in the Additional File 1: Table S4 were obtained from previous studies [26, 78, 79, 104–123] or download from NCBI (https://www.ncbi.nlm.nih.gov/) [124–133], InsectBase (http://v2.insect-genome.com/), Figshare and GigaDB (http://gigadb.org) [134–137]. The SRA accession numbers for the salivary gland transcriptomes of 11 non-zoophagous hemipteran species, downloaded from NCBI (https://www.ncbi.nlm.nih.gov/sra/), are as follows: Apolygus lucorum: SRR10411604; Triatoma rubrofasciata: DRR205094; Halyomorpha halys: SRR12286913; Riptortus pedestris: SRR15184056; Nilaparvata lugens: SRR13958479; Sogatella furcifera: SRR3211109; Laodelphax striatellus: SRR19351358; Acyrthosiphon pisum: SRR15862168; Hormaphis cornu: SRR11428768; Diaphorina citri: SRR11801820; Bemisia tabaci: SRR10527110. Details listed in Additional File 1: Table S8.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.