Metabolic enhancement contributed by horizontal gene transfer is essential for dietary specialization in leaf beetles

Significance

The study uncovers the critical role of horizontal gene transfers (HGTs) in the dietary specialization of a leaf beetle. By identifying two cellulose-degrading genes transferred from bacteria, the research demonstrates how these HGTs enable larvae to thrive on mature leaves with high cellulose content, a dietary preference that reduces competition with adults or other herbivorous insects feeding on new leaves. This metabolic enhancement not only provides a survival advantage but also underscores the broader ecological impact of HGTs in herbivorous insect adaptation and evolution. This finding deepens our understanding of the genetic and ecological mechanisms underpinning dietary specialization and highlights the importance of HGTs in shaping the interactions between herbivorous insects and their host plants.

Abstract

Horizontal gene transfer (HGT) from bacteria to insects is widely reported and often associated with the adaptation and diversification of insects. However, compelling evidence demonstrating how HGT-conferred metabolic adjustments enable species to adapt to surrounding environment remains scarce. Dietary specialization is an important ecological strategy adopted by animals to reduce inter- and intraspecific competition for limited resources. Adults of the leaf beetle Plagiodera versicolora (Coleoptera) preferentially consume new leaves; nevertheless, we found that they selectively oviposit on mature leaves, thereby establishing a distinct dietary niche separation between adults and larvae. Based on the de novo assembled chromosome-level genome, we identified two horizontally transferred genes with cellulose degradation potential, belonging to the glycosyl hydrolase 48 family (GH48-1 and GH48-2). Prokaryotic expression of the HGTs confirmed the cellulose degradation capability of the two genes. Knockdown of GH48 significantly hampered the growth and survival rate of larvae feeding on mature leaves compared to wild-type larvae, with no similar effect observed in adults. Replenishing the GH48-expressing bacteria compensated for the knockdown of these two genes and recurred larval adaptability to mature leaves. Taken together, our results highlight the advantage and metabolic enhancement conferred by the two cellulose-degrading HGTs in P. versicolora larvae, enabling their development on cellulose-enriched mature leaves and underscoring the indispensable role of HGTs in facilitating the adaptation of leaf beetles to plants.

Plants and herbivorous insects collectively constitute the majority of visible biological diversity, and their competition further promotes their joint diversification (1, 2). It is well established that plants have evolved various strategies to counter herbivorous insect invasions, and herbivorous insects have developed multiple means to cope with and adapt plants’ defenses in response (3, 4). Beyond their intrinsic capabilities, insects can also utilize external tools; for instance, insect-associated microbes can assist herbivorous hosts in digesting recalcitrant nutrients, detoxifying toxic chemicals, and down-regulating plant defenses (5). Meanwhile, insects can acquire specific functions through horizontal gene transfer (HGT), which is recognized as an important evolutionary force shaping insect genomes (6, 7). The transferred genes include segments of symbiont genomes to host insects, as well as single or multiple genes from microbes or even plants (6, 8). These transferred genes often have ecologically important functions; for example, carotenoid biosynthesis genes transferred from fungi to aphids partly determine aphid body coloration (9). A gene horizontally transferred from bacteria is key for herbivorous arthropods to detoxify plant-produced defensive chemical cyanide (10).

The order Coleoptera (beetles) is arguably the most speciose group of animal, and previous studies have suggested that microbial-originated HGT may have been crucial to the diversification of beetles (8, 11, 12). However, more cases and actual underlying mechanisms are needed to support the claim. Among beetles, leaf-feeding beetles are one of the most significant groups, and previous research has discovered that they have evolved various behavioral and physiological strategies to better adapt to plants (13–15). Given that the recalcitrance of complex molecules in the plant cell wall (e.g., cellulose) acts as a barrier preventing herbivorous insects from accessing the nutrient-rich cytosol of plant cells (16), the acquisition of plant cell wall degrading enzymes (PCWDEs) by beetles through HGT from microbes would serve as a prominent example of evolutionary adaptation. This process enables these insects to access plant nutrients more effectively. For instance, the genome of the leaf beetle Phaedon cochleariae encodes two active HGT-derived xylanases, originally acquired from bacteria, which potentially allow the insect to process its food more efficiently (17). Additionally, several HGT-derived pectin-digesting enzymes from fungi have been shown to help P. cochleariae access nutrients by degrading previously indigestible pectin, which is vital for the beetle’s survival (18). Unfortunately, concrete evidence of the significance of these HGT-derived PCWDEs for the survival or adaptation of other herbivorous beetles, beyond P. cochleariae, is rarely reported.

In addition to dietary challenges, interspecific and intraspecific competition for food resources in a given ecosystem always occurs and adversely affected phytophagous insects (19). Furthermore, intraspecific competition is generally more intense than interspecific competition (20). Consequently, dietary specialization would theoretically enable herbivorous insects to better access and exploit food resources and avoid competition. While the benefits of dietary specialization are apparent, the mechanisms underlying its formation and maintenance remain largely unclear. Some cases are even counterintuitive. For example, the Salicaceae feeding leaf beetle Plagiodera versicolora employs dietary specialization as a strategy to avoid competition and exploit plants more effectively. This is evidenced by adults’ preference for feeding on new leaves of salicaceous trees while larvae consume mature leaves, even when both leaf types are abundant (21). Although new leaves may contain higher levels of certain defensive compounds, they remain more attractive to herbivorous insects overall (22–24). Adult P. versicolora that feed on new leaves exhibit higher reproductive success compared to those feeding on mature leaves, as new leaves contain more nutrients (21). In contrast, mature leaves, being tougher and higher in cellulose content than new leaves, are more difficult to digest and can cause wear on P. versicolora larval mandible when consumed (22, 25). Thus, based on “mother knows best” principle, larvae should ideally feed on new leaves. Our research indicates that opportunistic pathogens would be enriched in larvae feeding on new leaves, thereby negatively affecting its growth (26). However, this only explains why larvae cannot eat new leaves but does not address why they can consume and adapt well to mature leaves in field.

In this study, we first unveiled the phenotype of adult oviposition and larval feeding preference for mature leaves, while also demonstrating the larvae’s high adaptation to cellulose-rich mature leaves. Subsequently, we sequenced and assembled the chromosome-level genome of P. versicolora and identified two potential cellulase-encoding genes (GH48-1 and GH48-2), which are likely of bacterial origin and were acquired through HGT. Prokaryotic expression of the HGTs was conducted to confirm the cellulose degradation capability of the two genes. We then knocked down these genes, individually and together, to evaluate their influence on the larval adaptation to mature leaves. Finally, we reinoculated GH48-expressing Escherichia coli to verify the ecological function of the two HGTs. Our results provide evidence of how HGTs are involved in dietary specialization, deepening the understanding of fundamental mechanisms by which herbivorous insects successfully adapted and persisted on plants.

Results

Adult Oviposition and Larval Feeding Preferences Converge on Mature Leaves despite Their High Cellulose Content.

The oviposition preference assay indicated that mated females significantly prefer mature leaves over new leaves for egg laying (Fig. 1 A and B). The average number of eggs laid on mature leaves is substantially higher compared to new leaves (Fig. 1C), although the hatching rates are similar (Fig. 1D). The average Preference Index (PI) of newly emerged larvae for mature leaves is 0.32, showing a significant preference for mature leaves (Fig. 1E), despite the mature leaves having significantly higher cellulose content than the new leaves (Fig. 1A). Larvae chosen from either new leaves or mature leaves were subsequently fed with the corresponding leaves. Those fed with mature leaves demonstrated higher fitness, evidenced by increased body weight in the 3rd instar larvae and pupae, as well as a significantly shorter development time from emergence to pupation (Fig. 1 F–I).

Fig. 1.

Adult oviposition and larval feeding preferences of P. versicolora on mature leaves result in enhanced fitness. (A) Schematic representation of adult oviposition preference, adult and larval feeding preferences, and cellulose content in new and mature leaves of Salix babylonica (n = 6). (B) Oviposition PI between new and mature leaves (n = 30). The size of the circles and the numbers within indicate the number of 24-h oviposition choices resulting in a given PI. Solid black vertical lines designate mean PI, with 95% CI ranging from 0.192 to 0.519. (C and D) Number of eggs laid (n = 30) and hatch rate (n = 30 for mature leaves, n = 22 for new leaves) on new and mature leaves. (E) Larval PI between new and mature leaves (n = 19), with 95% CI ranging from 0.156 to 0.477. (F–I) Comparison of larvae feeding on new versus mature leaves: body mass of 3rd instar larvae (n = 57), pupal weight (n = 55), pupation rate (n = 6), and larval development time (n = 30). **P < 0.01 or ***P < 0.001 using independent samples t tests for panels A, C, D, and F–I, and one-sample t test for panels B and E. NS indicates not significant. Error bars represent SD.

Two HGT-Derived Cellulases of Bacterial Origin Are Encoded in the P. versicolora Genome.

To elucidate the mechanisms underlying larval adaptation to diet specialization, we sequenced and assembled the genome of P. versicolora. The assembly spans 233.27 Mb and comprises 17 chromosomes, with a GC content of 35.60% and a scaffold N50 length of 13.93 Mb (Fig. 2 A and B and SI Appendix, Table S1). Comparative genome analysis revealed a low level of genome collinearity even among beetles with close evolutionary relationships (SI Appendix, Fig. S1 and Fig. 2C). Given the abundance of cellulose in mature leaves and the larvae’s well-adapted consumption of these leaves, we conducted genome-wide searches with a particular focus on identifying HGT-derived cellulases. This analysis identified two bacteria-originated genes with cellulose degradation potential (GH48-1 and GH48-2) on the first chromosome of P. versicolora, which are widely recognized as horizontally transferred genes (Fig. 2 D and E) (8). The expression patterns of GH48-1 and GH48-2, along with the seven P. versicolora GH45 genes (GH45-1 to GH45-7, acquired through HGT from fungi), were profiled across different developmental stages and various tissues. Results showed that all genes had significantly higher expression in the gut compared to the carcass (SI Appendix, Fig. S2). Additionally, GH48-1, GH48-2, and GH45-6 displayed similar expression patterns, with significantly higher expression levels in larvae compared to adults, and expression levels increased with larval age (SI Appendix, Fig. S2). Their spatiotemporal expression profile suggests the relevance of these GH genes in digestive processes, particularly in 2nd and 3rd instar larvae.

Fig. 2.

Genome assembly and HGTs identification in the leaf beetle, P. versicolora. (A) genome-wide Hi-C heatmap of chromatin interaction counts. (B) Circos plot of the 17 chromosomes of P. versicolora. From outermost to innermost layer: (a) chromosome length, (b) gene density, (c) repeat density, and (d) GC density. (C) Chromosomal synteny analysis comparing P. versicolora with three other leaf beetle species from the same phylogenetic clade. (D) Maximum likelihood phylogenetic tree constructed using the two HGT-acquired genes and their homologous proteins. The tree was built using the LG+F+R7 model with 1,000 bootstrap replicates. (E) Schematic representation of the two HGT-acquired genes in P. versicolora and their homologous proteins from Streptomyces sp.

The GH48s Can Efficiently Degrade Cellulose In Vitro.

While GH45s have been well characterized in previous studies (27), the function of GH48s remains to be confirmed. Therefore, we expressed the proteins in E. coli using a prokaryotic expression system to verify the activity of GH48s. The vector map is shown in Fig. 3A. Both SDS-PAGE and western blot results confirmed the protein expression, with two distinct bands at 110.36 kDa clearly visible between the 100 to 130 kDa markers (Fig. 3 B and C). Additionally, the SDS-PAGE results indicated that the proteins used for enzyme activity assays exhibited high purity. In vitro incubation with microcrystalline cellulose demonstrated average degradation efficiencies of 33.24 U/mg and 35.32 U/mg for the purified GH48-1 and GH48-2 proteins, respectively (Fig. 3D). GH48-2 exhibited significantly higher activity compared to GH48-1, with an average increase of 1.06-fold (Fig. 3D). Furthermore, IPTG-induced E. coli exhibited significant degradation zones on cellulose agar plates, whereas no degradation zones were observed for noninduced E. coli (Fig. 3E).

Fig. 3.

Enzymatic activity of prokaryotically expressed GH48-1 and GH48-2 demonstrates high cellulose degradation. (A) Schematic diagram of the expression vector used for prokaryotic expression, featuring kanamycin resistance, a SUMO solubility-enhancing tag, super-fold GFP, and His tags. (B and C) SDS-PAGE and western blot analysis of GH48-1 or GH48-2 protein expression and purification. SDS-PAGE: lane 1, cell lysate supernatant after sonication; lane 2, purified GH48-1 or GH48-2 protein. M, molecular weight markers (KDa) ranging from 10 to 180 kDa. (D) Enzymatic activity assays of purified GH48-1 and GH48-2 (n = 6). (E) Diameter of degradation zones on cellulose plates by E. coli expressing GH48-1 and GH48-2 with or without IPTG induction (n = 6). Statistical significance was determined using independent samples t tests for panels D and E. *P < 0.05, ***P < 0.001. NS indicates not significant. Error bars represent SD.

The GH48s Are Essential for Larval Adaptation to Mature Leaves.

The knockdown of either one or both GH48s caused significantly higher mortality in larvae that fed on mature leaves (Fig. 4 A and B). The weight of third instar larvae and pupae, as well as the pupation rates, were significantly reduced by the RNAi of the two genes (Fig. 4 C–E). We also observed a significantly prolonged development time in larvae in the dsGH48 groups (Fig. 4F). The cellulose content in larvae was significantly higher when GH48s were knocked down (Fig. 4G). These phenotypes were not observed in larvae that fed on new leaves (Fig. 5). qPCR results confirmed that the two genes were significantly down-regulated by the dsRNAs (Figs. 4 H and I and 5 H and I). Additionally, we compared the performance of larvae with GH45-6 knockdown to that of control larvae when feeding on mature leaves. Our results indicated no statistically significant differences in mortality rate, body mass of third instar larvae and pupae, development time, pupation rate, or cellulose content in larvae (SI Appendix, Fig. S3).

Fig. 4.

Knockdown of GH48 genes, either individually or in combination, significantly reduces P. versicolora performance on mature leaves. (A) Schematic representation of larvae feeding on mature leaves coated with dsRNAs. (B–F) Comparison of larvae feeding on mature leaves coated with dsGFP, either dsGH48-1 or dsGH-48-2, or in combination of the two dsRNAs: survival curves (n = 60), body mass of 3rd instar larvae (n = 43 to 58, excluding dead larvae), pupal weight (n = 40 to 56, excluding dead larvae), pupation rate (n = 6), and development time (n = 30). (G) Cellulose content in larval feces (n = 3). (H and I) Gene expression level in different treatments (n =3). *P < 0.05, **P < 0.01, or ***P < 0.001 using the log-rank (Mantel–Cox) test for panel B and independent samples t tests for panels H and I. Different letters in panels C–G indicate statistically significant differences at P < 0.05 by one-way ANOVA with post hoc Bonferroni multiple comparisons. Error bars represent SD.

Fig. 5.

Individual and combined knockdown of GH48 genes have limited effects on P. versicolora performance on new leaves. (A) Schematic representation of larvae feeding on new leaves coated with dsRNAs. (B–F) Comparison of larvae feeding on new leaves coated with dsGFP, either dsGH48-1 or dsGH-48-2, or in combination of the two dsRNAs: survival curves (n = 60), body mass of 3rd instar larvae (n = 55 to 58, excluding dead larvae), pupal weight (n = 52 to 55, excluding dead larvae), pupation rate (n = 6), and development time (n = 30). (G) Cellulose content in larval feces (n = 3). (H and I) Gene expression level in different treatments (n = 3). Statistical significance was determined using the log-rank (Mantel–Cox) test for panel B, one-way ANOVA with post hoc Bonferroni multiple comparisons for panels C–G, and independent samples t tests for panels H and I. *P < 0.05, **P < 0.01, ***P < 0.001. Different letters indicate statistically significant differences at P < 0.05. NS indicates not significant. Error bars represent SD.

After ingestion of dsGH48-1, dsGH48-2, or a mixture of the two dsRNAs coated on new leaves, there was no observed difference in mortality, egg-laying, or egg hatch rate of adults compared to those fed with dsGFP-coated new leaves (SI Appendix, Fig. S4 A–D). Additionally, no differences in cellulose content were found in their feces (SI Appendix, Fig. S4E), despite achieving significant gene knockdown efficiency (SI Appendix, Fig. S4 F and G). No mortality differences were detected when adults fed on mature leaves (SI Appendix, Fig. S5B). Nevertheless, a significantly lower number of eggs was observed in GH48-silenced females compared to nonsilenced females feeding on mature leaves (SI Appendix, Fig. S5C). Taken together, these results suggest that the presence of GH48s is essential for larvae to adapt to mature leaves.

Reintroduction of GH48-Expressing E. coli Restores Larval Adaptation to Mature Leaves.

To verify whether the HGT-mediated cellulose degradation capability is essential for larval adaptation to mature leaves, we reintroduced GH48-expressing E. coli into GH48s knockdown larvae and further assessed their performance. For larvae with GH48-1 or GH48-2 knockdown, reintroducing E. coli reduced the mortality rate when feeding on mature leaves, although the difference was not statistically significant (Fig. 6 B and C). For larvae with both GH48-1 and GH48-2 knocked down, reintroducing E. coli significantly improved their survival rate (Fig. 6D). Similar trends were observed in larval weight, pupal weight, and developmental period (Fig. 6 E–J). Additionally, we found a notable reduction in cellulose content in larval feces after the reintroduction of GH48-expressing E. coli in the gene knockdown larvae (Fig. 6K). The cellulase activity assays revealed that while cellulase activity remained markedly suppressed in GH48-knockdown larvae, reintroducing GH48-expressing E. coli significantly restored intestinal enzyme activity (Fig. 6L). Fluorescent imaging of the gut confirmed the successful ingestion of E. coli by the treated larvae (SI Appendix, Fig. S6).

Fig. 6.

Reintroduction of GH48-expressing E. coli partially rescues the adverse phenotype of larvae feeding on mature leaves after GH48 knockdown. (A) Schematic representation of larvae feeding on mature leaves coated with dsRNAs and GH48-expressing E. coli. (B) Survival curves of larvae fed with dsGH48-1 and either IPTG-induced GH48-expressing E. coli (denoted as dsGH48-1+Re) or E. coli containing empty vector (denoted as dsGH48-1) (n = 60). Control groups were fed with dsGFP and E. coli containing empty vector (denoted as dsGFP, n = 60). (C) shows results for dsGH48-2 (n = 60); (D) shows simultaneous knockdown of GH48-1 and GH48-2 (n = 60). (E–G) Body mass of 3rd instar larvae (n = 43 to 58, excluding dead larvae), pupal weight (n = 38 to 57, excluding dead larvae), and pupation rate of larvae (n = 6) under the aforementioned treatments. (H–J) Larval development time under the aforementioned treatments (n = 30). (K) Cellulose content in larval feces (n = 3). (L) Cellulase activity in larval gut under different treatments (n = 6). Statistical significance was determined using the log-rank (Mantel–Cox) test for panels B–D, one-way ANOVA with post hoc Bonferroni multiple comparisons for panels E–L. *P < 0.05, **P < 0.01. Different letters indicate statistically significant differences at P < 0.05. NS indicates not significant. Error bars represent SD.

Discussion

Considering limited food availability and both interspecific and intraspecific competition of animals, food choice plays a vital role at individual, population, and species levels, significantly defining an organism’s fitness (28). In this context, individual dietary specialization is a result of species adapting to their environment, and this ecological strategy is not limited to higher animals with advanced cognitive abilities but is also present in herbivorous insects (29, 30). In this study, behavioral choice experiments with P. versicolora revealed that while adults feed on young leaves, they prefer to oviposit on mature leaves. Interestingly, a similar preference for feeding on young leaves of Salicaceous plants has been observed in five other leaf beetle species, such as Agelastica alni (23, 25). Additionally, when given an artificial choice, the larvae also showed a preference for mature leaves. Thus, we can conclude that dietary specialization exists in P. versicolora. We then propose a mechanism for maintaining the dietary specialization in the leaf beetle; the acquisition of bacterial-origin genes through HGT has likely been a key factor in enabling P. versicolora larvae to exploit a niche that would otherwise be nutritionally and physically challenging (31). This adaptation not only allows the larvae to access a less competitive food source but also potentially reduces their exposure to plant defenses that are typically more concentrated in new leaves (21), highlighting the ecological significance of HGT in shaping insect–plant interactions and dietary specialization. Furthermore, our study confirms that, at least from the perspective of cellulose degradation, the dietary specialization of P. versicolora adheres to the mother knows best or “preference-performance” principle, as female adults prefer mature foliage for oviposition (32). Similar observations have been reported in the leaf beetle Phratora vitellinae, where adults tend to oviposit on plants that are more favorable for larval growth (33). Adults feed on new leaves to obtain more nutrients to ensure higher oviposition rates, while larvae, endowed with strong cellulose-degrading capabilities through HGTs, can efficiently adapt to and consume mature leaves, thereby avoiding competition with adults or other herbivorous insects for the more nutritious new leaves. However, there may be other reasons for larvae feeding on mature leaves, such as better protection from ultraviolet radiation, reduced exposure to predators, or lower concentrations of plant secondary metabolites that act as feeding deterrents (34, 35). Further research is needed to determine which factor is the primary reason.

HGTs are widely present across various organisms and often transfer from microorganisms to insects (6). In contrast to symbiotic relationships, these HGTs may provide increased adaptability to the recipient host, potentially reducing its reliance on certain microorganisms. As a result, the absence of the donor microorganisms might have less impact on the host. In our study, sequence similarity comparisons and phylogenetic analysis showed that the two HGTs are most similar to those found in Actinomycetes. However, related bacteria were not found in the culture-dependent and culture-independent microbiome analyses of P. versicolora (36), suggesting that the donor microorganisms of these HGTs were lost by the leaf beetle long ago. On the other hand, these two HGTs belong to plant cell wall-degrading enzymes, which are significant for other beetles as well (8). They greatly enhance the ability of beetles to digest and utilize plants, potentially contributing to the diversity found within the order Coleoptera (8). Besides providing direct capabilities for digesting recalcitrant materials, some HGTs can enhance the beetles’ ability to acquire other plant nutrients as evidenced by the aforementioned GH28 in P. cochleariae (18). For the highly diversified leaf beetles, there may be many other HGTs performing various roles. In this study, we also identified seven fungal-originated cellulase genes acquired through HGT. Among these, GH45-6 was selected as a target for further RNAi studies due to its similar expression patterns with GH48s. Phylogenetic analysis revealed that this gene clusters together with five other genes (LDE6, PCO1, DVI1, DVI2, and DVI4, SI Appendix, Fig. S2A), of which four exhibit endo-β-1,4-glucanase activity (SI Appendix, Table S3), suggesting that GH45-6 may share the same enzymatic function. Although knockdown of GH45-6 did not affect larval performance on mature leaves, the ecological functions of these genes merit further investigation. Besides, the presence of these genes in microbes and other beetles suggests that HGT events may be more common and widespread than previously thought. Understanding the mechanisms and evolutionary pressures that facilitate HGT could provide deeper insights into the adaptive strategies of herbivorous insects and their interactions with plants.

The significant cellulose-degrading activity of HGT GH48 enzymes was confirmed through both in vitro enzymatic activity assays (using prokaryotically expressed GH48 proteins) and in vivo gene knockdown experiments. Furthermore, we attempted to rescue the knockdown effects by reintroducing bacteria expressing GH48 genes. This approach partially alleviated the maladaptation phenotype observed in GH48 knockdown larvae when fed on mature leaves, and the compensatory effect was most pronounced in larvae with double knockdown of GH48-1 and GH48-2. This indirectly suggests that cellulose-degrading activity mediated by GH48 is crucial for larval adaptation to mature leaves. Although adults rarely feed on mature leaves in the wild, it is noteworthy that while GH48 knockdown in adults did not affect their survival rate or egg hatching rate, it significantly reduced their oviposition output when fed mature leaves. This reduction may be attributed to the limited cellulose-degrading capacity, which restricts the adults’ ability to obtain sufficient nutrients to ensure high oviposition rates (31). This hypothesis is further supported by the observation that similar oviposition outputs of P. versicolora females were recorded between RNAi-treated and control groups when fed new leaves, which contain lower cellulose levels and more nitrogen compared to mature leaves (23). This finding also partly explains the adults’ preference for new leaves in natural conditions. Collectively, our results show that while GH48 genes are crucial for larval adaptation to mature leaves, they appear to be less critical for adult beetles. This differential importance suggests that the HGT-acquired genes have been integrated into the developmental program of P. versicolora in a stage-specific manner, with both GH48s exhibiting higher expression in larvae compared to adults. A very similar phenomenon was observed for HGT-derived genes (cdtB and aip56) in various Drosophila species, where the cdtB single copy is exclusively expressed in embryos, while cdtB::aip56 fusion A and B genes are expressed only in larvae to enhance resistance against parasitoid wasps (37). This case also suggests that HGT events in insects appear to be incorporated into developmentally regulated programs that are crucial not only for adaptation to the first trophic level (plant hosts) but potentially also to the third trophic level (natural enemies). However, further research is needed to confirm this speculation. Both studies emphasize the need to consider life stage when studying the impacts of HGT on insect ecology and evolution. In conclusion, our findings highlight the critical role of HGT in the dietary specialization and ecological adaptation of P. versicolora. This study not only advances our understanding of insect–plant interactions but also opens broad avenues for exploring the evolutionary significance of HGT in shaping the genomes and ecological strategies of herbivorous insects.

Materials and Methods

Insect and Plant Sources.

A wild population of P. versicolora was collected from Sha Lake Park, Wuhan (Hubei Province, China) and maintained on fresh willow leaves at 27 ± 1 °C, 70 ± 5% relative humidity, and a 16:8 h light:dark photoperiod. Willow (S. babylonica) leaves, cultivated in a greenhouse at a constant 25 °C, were used as the plant material. New leaves were defined as the first and second leaves from the apex with the top buds removed, while mature leaves were the fourth, fifth, and sixth leaves.

Total 300 mg new leaves or mature leaves was collected (n = 6) and analyzed for cellulose content using the Cellulose Content Assay Kit (Solarbio, Beijing, China). Briefly, leaves ground in liquid nitrogen were dispersed in extraction solution I. The mixture was incubated at 90 °C for 20 min and centrifuged (6,000 g, 10 min, room temperature). Pellets were washed twice with extraction solution I and twice with acetone. The crude cell wall fraction was treated with extraction solution II for 15 h to remove starch, followed by centrifugation (6,000 g, 20 min). Then, rinsed twice with distilled water. Dried pellets (60 °C) were dissolved in distilled water and concentrated sulfuric acid, then centrifuged (8,000 g, 10 min). Supernatants were diluted 20-fold with distilled water. Cellulose content was determined by measuring absorbance at 620 nm, using glucose as the standard.

Preference Assays and Impact of Dietary Specialization on Larval and Adult Fitness.

Oviposition preference was evaluated using 90 mm diameter polystyrene Petri dishes. Each dish contained two rectangular mature leaves and four rectangular S. babylonica new leaves (with the area of two new leaves approximately equivalent to one mature leaf, See Fig. 1B), along with moist filter paper to maintain humidity. A pair of sexually mature adults (5 d postemergence) was introduced, and the oviposition choices of females were recorded over a 24-h period, starting from the onset of egg-laying. Thirty replicates, each containing six females, were analyzed (n = 30). A PI was calculated using the formula: PI = (A − B)/(A + B), where A represents the total number of females ovipositing on mature leaves, and B represents the total number ovipositing on new leaves. Females that did not oviposit or laid eggs on surfaces other than leaves were excluded from the analysis. A positive PI indicates a preference for mature leaves, while a negative PI suggests a preference for new leaves (38). The total number of eggs on mature or new leaves was counted for each replicate, and the average number of eggs per leaf was calculated (n = 30). Hatching rates were then recorded for each replicate after 3 d (n = 30 for mature leaves, n = 22 for new leaves).

Larval feeding preference was tested in 60 mm diameter Petri dishes. Circular leaf discs (6 mm diameter) were punched from mature and new leaves, and four of each type were placed equidistantly around the dish, kept moist with filter paper (Fig. 1E). A starved first instar larva was placed at the center, and feeding choices were recorded after 12 h. Nineteen replicates, each with eight larvae, were analyzed, and PI was calculated similarly, with larvae that did not feed excluded from the analysis (n = 19). Subsequently, a total of 120 newly hatched larvae were divided into two groups: One fed mature leaves and the other fed new leaves of S. babylonica to measure the body mass of 3rd instar larvae (n = 57) and pupal weight (n = 55). Additionally, 60 larvae, in groups of 10, were treated the same way to calculate the pupation rate (n = 6), and 30 larvae were used to determine the development time for each instar (n = 30).

Genome Analysis.

The genomic DNA of P. versicolora was isolated using fresh pupa from a female adult employing the CTAB method. The High-fidelity (HiFi) libraries were constructed with SMRTbell Prep Kit 3.0 and sequenced on Pacbio Revio system (Pacifc Biosciences, Menlo Park). The HiFi reads were then de novo assembled by using Hifiasm software v0.19.5 (39, 40) with default parameters. The haplotypic duplication in the initial assembly was removed using purge_dups v1.2.5 with default parameters (41). We then used High-throughput chromosome conformation capture (Hi-C) to assemble the P. versicolora genome at the chromosomal level. In brief, the high-quality sequencing raw reads were filtered using fastp v0.23.4 (42). Then, the cleaned Hi-C reads were mapped to the HIFI assembly mentioned above using Juicer v1.6 (43). The unique high-quality paired-end reads were taken as input for 3D-DNA v190716 pipeline (44) with parameters “-r 0.” Chromosome interaction matrix was corrected manually by using JuicerBox v1.11.08 (43). The Hi-C heatmap was generated using HiCExplorer v3.7.2 (45).

Larvae and adult female were collected for RNA extraction. Total RNA was extracted using Vezol reagent (Vazyme, China), and sequencing libraries were generated using the VAHTS Universal V6 RNA-seq Library Prep Kit (Vazyme, China). The library was sequenced on the Illumina X Ten platform with PE150 strategy, and the short-read RNA-seq raw data were used to annotate the P. versicolora genome. For genome annotation of P. versicolora, we primarily relied on a genome annotation pipeline developed by the Institute of Insect Sciences at Zhejiang University (46). In detail, a repeat library, focusing on insect specialization, was initially created using RepeatModeler v2.0.4 (47) and RepeatMasker v4.1.5 (48) for the annotation of repeat sequences. Subsequently, the genome was masked using RepeatMasker v4.1.5 with the “-xsmall -gff -html -lib” parameters. For the prediction of protein-coding genes, homology proteins from various insect species were obtained from InsectBase 2.0 (http://v2.insect-genome.com/). Transcriptome data were aligned to the genome using HISAT v2.2.1 (49), and the successfully aligned reads were converted into a BAM file using Samtools v1.17 (50). Both homology proteins and transcriptome-based evidence served as inputs for BRAKER v3.0.3 (51). The functional annotation of protein-coding genes was assessed using eggNOG-mapper version 2 (http://eggnog-mapper.embl.de/). All raw sequencing data used for genome assembly and analyses have been deposited in the Sequence Read Archive database of NCBI and can be accessed through BioProject: PRJNA1045849.

Comparative Genomics Analysis.

For phylogenetic reconstruction, we sourced chromosome-level genomes of 14 Chrysomelidae insects (GCA_950111635.1, GCA_949316355.1, GCA_958502065.1, GCA_944452925.2, GCA_027562985.1, GCA_949320105.2, GCA_958507055.1, GCA_946251905.1, GCA_029229535.1, GCA_026250575.1, GCA_951812935.1, GCA_949126875.1, GCA_947563755.1, and GCA_917563875.2) from NCBI genome database (https://www.ncbi.nlm.nih.gov/genome/?term=). We also downloaded the Tribolium castaneum genome (GCA_031307605.1) as a reference outgroup. OrthoFinder v2.5.4 was employed to examine orthologs and homologs, using “-M msa -T fasttree” parameters (52). A total of 784 single-copy genes were concatenated into a supergene for multiple sequence alignment with ClustalW with default parameters (53). The aligned sequences were then trimmed using TrimAl v1.4 with the following parameters: -gt 0.6 and -cons 60 (54). The phylogenetic tree was then constructed using IQTREE v2.2.2.6, incorporating 1,000 boot-strap replicates and the Q.insect model (55). We also incorporated fossil data from The Paleobiology Database (https://paleobiodb.org/#/) to refine our divergence estimates, using the MCMC tree program in PAML v4.10.7 (56). Besides, we performed a synteny analysis to evaluate the evolutionary relationships of closely related insects and validate the reliability of this assembly.

Identification of HGT-Derived Cellulases and Their Expression Pattern in P. versicolora.

The whole NR and NT databases in the NCBI were downloaded for reference database. We then performed a BLAST analysis using default parameters, employing both CDS and protein sequences of the leaf beetle. From these results, we identified sequences that aligned with bacterial sequences. These selected sequences were subsequently subjected to a second BLASTP search against the NCBI clustered NR database. The results were then filtered by the following parameters: Coverage >= 90% and identify >= 60%. After this, two remained sequences were tentatively designated as horizontal transfer genes of bacterial origin.

To evaluate the reliability of the two HGT-derived cellulases in leaf beetles, a total of 100 top-scoring amino acid sequences obtained from BLASTP results (using default parameters) were subjected to phylogenetic analysis. For the seven fungal-originated GH45s in P. versicolora, 33 additional GH45 sequences from five other leaf beetle species, which have already been functionally characterized, were included in the phylogenetic analysis, with a GH45 sequence from Geotrichum candidum serving as the outgroup (27). Sequence alignment was performed using ClustalW with default parameters, followed by manual adjustments to improve alignment accuracy, resulting in a final character matrix of 102 sequences for GH48 (SI Appendix, Table S2) and 41 sequences for GH45 (SI Appendix, Table S3). We then constructed a phylogenetic tree using maximum likelihood method with LG+F+R7 model (for GH48s) or WAG+F+G4 model (for GH45s), with 1,000 bootstrap replicates. The phylogenetic tree was visualized using iTOL (https://itol.embl.de/).

Newly hatched larvae were fed mature leaves of S. babylonica under controlled conditions. Samples were collected from 1st to 3rd instar larvae, with three larvae per sample in triplicate. Additionally, larvae were dissected in PBS buffer to obtain gut and body tissue samples (n = 3; five guts or carcasses per sample). Adult beetles, newly emerged and fed on new willow leaves, were sampled after 4 d. Five males or females constituted one sample, with gut and body tissues also collected (n = 3; five guts or carcasses per sample). The expression profiles of GH48s and GH45s at different developmental stages and tissues of the leaf beetle were assayed using the above samples by RT-qPCR (SI Appendix, Table S4).

Prokaryotic Expression and Enzyme Activity Assay.

We constructed a 6×His-tagged sfGFP-sumo-GH48 fusion protein expression plasmid using the pET-28a vector. Recombinant plasmids were transformed into E. coli BL21 cells, selected, sequenced, and incubated overnight at 37 °C in 20 mL LB medium with kanamycin (50 mg/L). A 2 mL culture was inoculated into 200 mL LB medium with kanamycin and grown at 37 °C, 200 rpm. At OD₆₀₀ 0.5, protein expression was induced with IPTG (0.5 mM) and incubated at 18 °C, 160 rpm for 16 h. Cells were harvested (6,000 rpm, 10 min, 4 °C), washed, and resuspended in 10 mL buffer A (50 mM HEPES and 150 mM NaCl). After 20 min sonication, lysates were centrifuged (12,000 rpm, 10 min, 4 °C). The supernatant was purified using a Ni-NTA column and eluted with buffer B (50 mM HEPES, 150 mM NaCl, and imidazole gradient 50 to 300 mM). Proteins were concentrated using a 30 kDa ultrafiltration tube (Millipore) at 4,000 rpm, and imidazole was removed by PBS dialysis. Purity was assessed via 10% SDS-PAGE. Proteins were transferred to PVDF membranes, blocked with 5% skimmed milk for 1 h, and incubated overnight at 4 °C with primary antibody. After three Tris HCl Tween (TBST) rinses, membranes were incubated with HRP-labeled secondary antibodies for 1.5 h, rinsed thrice with TBST, and protein signals detected using ECL western blotting substrate.

The enzyme activity of recombinant proteins was evaluated by measuring the glucose content using 3,5-dinitrosalicylic acid (DNS) after incubating the purified proteins with microcrystalline cellulose (57). Briefly, the reaction mixture contained 0.5 mL of 1% microcrystalline cellulose (sourced from wood pulp or cotton linters, Sinopharm Chemical Reagent, China) in 0.05 M sodium citrate buffer (pH 4.8) and 0.5 mL 4 mg/mL of purified protein in a 1.5 mL tube. Heat-inactivated proteins served as blank controls. All samples and blank tubes were incubated at 40 °C for 2 h. Postincubation, samples were centrifuged at 12,000 rpm for 15 min. Then, 0.5 mL of the supernatant was mixed with DNS reagent and heated in a boiling water bath for 10 min. The development of an orange color indicated the generation of free glucose residues from the digestion of the substrate by cellulase. The enzyme activity of cellulase was expressed as μmol of reducing sugar per minute per milligram of protein (U/mg, n = 6).

To further evaluate the cellulose-degrading capability of GH48-expressing E. coli, we inoculated the bacteria onto a medium containing microcrystalline cellulose as the sole carbon source, supplemented with Congo red. A 10 µL of IPTG-induced or not induced E. coli suspension was dropwise added onto a 6 mm diameter filter paper disk, which was then placed at the center of the medium (n = 6). The degradation zone size was measured after incubating the plates at 37 °C in a constant temperature incubator for 48 h. The degradation area was calculated using the formula: Degradation area = (diameter of the transparent zone − 6 mm) ×10.

Impact of GH48s and GH45-6 Silencing on Dietary Specialization.

The DNA templates of three genes (GH48-1, GH48-2, GH45-6, and EGFP) were synthesized from recombinant plasmids using primers with T7 promoter sequences (SI Appendix, Table S4). The PCR products were purified for dsRNA synthesis using the T7 RiboMAX™ Express RNAi System (Promega). Integrity and quantity of dsRNAs were assessed via electrophoresis and spectrophotometry (NanoDrop 2000, Thermo Scientific).

To investigate the effect of HGTs on dietary specialization, GH48s were silenced by feeding dsRNA to larvae and adults. For GH48s, adults were fed new leaves or mature leaves coated with 40 ng/cm² dsRNA (dsGFP, dsGH48-1, dsGH48-2, or dsGH48-1+dsGH48-2). Briefly, a 10 ng/μL dsRNA solution was dropped onto the surface of a detached and trimmed S. babylonica mature (~4 cm²) or new leaf (~2 cm²) and spread evenly using a cell scraper. The leaves were replaced daily, and survival was recorded (n = 60). Another 10 pairs of adults were treated with the same procedures to measure oviposition (n = 10) and hatching rates (n = 10). At the end of the experiment, three individuals from each group were randomly selected to assess RNAi efficiency (n = 3).

For larval assays, larvae were fed either mature leaves or new leaves coated with 40 ng/cm² dsRNA (dsGFP, dsGH48-1, dsGH48-2, or dsGH48-1+dsGH48-2). Various metrics were recorded, including survival rate (n = 60), 3rd instar body mass (n = 43 to 58, excluding dead larvae), and pupal weight (n = 40 to 56, excluding dead larvae). Deceased larvae were excluded from body mass and pupal weight measurements. The developmental time of 30 larvae in each treatment group was monitored (n = 30). An additional 60 larvae, divided into six groups of 10, underwent the same treatment to determine the pupation rate (n = 6). For RNA extraction, three larvae were sampled from each group six days after dsRNA feeding (n = 3). To investigate the potential effect of GH45-6 on larval adaptation to mature leaves, dsGH45-6 was fed to larvae with dsGFP as a control using the same procedures (n = 60 for survival rate, n = 53 to 57 for 3rd instar larval body mass, n = 50 to 55 for pupal weight, n = 6 for pupation rate, n = 30 for developmental time, and n = 3 for qPCR).

Feces from insects subjected to gene knockdown and their corresponding control groups were collected and analyzed for cellulose content using a Cellulose Content Assay Kit (Solarbio, Beijing, China) according to the manufacturer’s instructions. Due to the minimal fecal output of individual insects, feces from 50 larvae or adults were pooled to form a single sample group (n = 3 for each treatment). Fecal samples were collected after six days of dsRNA feeding for larvae and 3 d for adults.

Rescue Experiment.

To verify HGT-mediated cellulose degradation’s role in larval adaptation to mature leaves, rescue experiments were conducted using GH48-expressing E. coli or E. coli containing an empty vector. The E. coli cultures were induced with IPTG (0.5 mM) at OD₆₀₀ 0.5 before the cells were collected for leaf painting. The larvae were fed S. babylonica mature leaves coated with 4 μL/cm² mixture (10 ng/μL dsRNA and 10⁸ cells/mL E. coli). Various parameters were recorded using the aforementioned methods, including survival (n = 60), 3rd instar body mass (n = 43 to 58, excluding dead larvae), pupal weight (n = 38 to 57, excluding dead larvae), pupation rate (n = 6), and development time (n = 30). Cellulose content in faces was assayed using the above-mentioned method (n = 3). Fecal cellulose content was assayed using the previously described method (n = 3). To assess cellulose enzyme activity, we dissected the guts of larvae treated with dsGFP, dsGH48, and dsGH48+GH48-expression E. coli. After 6 d of feeding, the guts from 10 larvae were pooled as a single sample (n = 6). Enzyme activity was then measured using the methods described above.

To confirm successful ingestion and presence of GH48-expressing E. coli in the larval guts, gut tissues were collected and immediately fixed in a formalin–acetic acid–alcohol (FAA) solution for 48 h. The fixed tissues were dehydrated through a series of ethanol solutions (75%, 85%, 95%, and 100%), followed by clearing in a 1:1 ethanol/xylene mixture and 100% xylene. Then, the tissues were embedded in paraffin wax. A microtome was used to cut the tissues into 3 μm sections. The sections were permeabilized with 0.1% Triton X-100 for 10 min, blocked with 1% BSA for 30 min, and incubated with an anti-eGFP antibody overnight at 4 °C. Alexa Fluor 488 goat anti-rabbit IgG/FITC (1:400 in PBS) was used for 50 min at room temperature to detect GFP fluorescence (green). β-actin (red) was stained with TRITC-Phalloidin. Images were analyzed using a fluorescence microscope with excitation wavelengths of 545 nm and 488 nm, and emission wavelengths of 570 nm and 511 nm, respectively.

RT-qPCR.

Total RNA was extracted with RNAiso Plus (Takara, Japan) and quantified using a Nanodrop ND-2000 spectrophotometer (NanoDrop). cDNA was synthesized from 2 μg of total RNA using the Hifair 1st Strand cDNA Synthesis for qPCR Kit (YEASEN, China) with random primers. RT-qPCR was conducted on a CFX96 Real-Time System (Bio-Rad Laboratories, Hercules, CA) with each 10 μL reaction containing 5 μL Mon-Amp™ SYBR® Green qPCR Mix (Monad Biotech, China), 2 μL cDNA, 0.4 μL of each primer (10 ng/μL), and 2.2 μL ddH₂O. The PCR conditions were 95 °C for 3 min, followed by 40 cycles of 95 °C for 10 s, and 60 °C for 30 s. Primer sequences are listed in SI Appendix, Table S4.

Data Analysis.

Survival curves were analyzed using the Kaplan–Meier method, and log-rank tests evaluated significance. The one-sample t test analyzed adult and larval preference experiments. Independent samples t tests or one-way ANOVA with Bonferroni were used for other data. Statistical significance was set at P < 0.05. Analyses were performed using SPSS 19.0, and figures were created using GraphPad Prism 8 and Adobe Illustrator CS6 or Photoshop CS6.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.