Identifying Differential Gene Expression in Wing Polymorphism of Adult Males of the Largest Water Strider: De novo Transcriptome Assembly for Gigantometra gigas (Hemiptera: Gerridae)

Abstract

Wing polymorphism is common in a wide variety of insect species. However, few studies have reported on adaptations in the wing polymorphism of insects at molecular level, in particular for males. Thus, the adaptive mechanisms need to be explored. The remarkable variability in wing morphs of insects is well represented in the water striders (Hemiptera: Gerridae). Within this family, Gigantometra gigas (China, 1925), the largest water strider known worldwide, displays macropterous and apterous males. In the present study, we used de novo transcriptome assembly to obtain gene expression information and compared body and leg-component lengths of adult males in different wing morphs. The analyses in both gene expression and phenotype levels were used for exploring the adaptive mechanism in wing polymorphism of G. gigas. After checking, a series of highly expressed structural genes were found in macropterous morphs, which were related to the maintenance of flight muscles and the enhancement of flight capacity, whereas in the apterous morphs, the imaginal morphogenesis protein-Late 2 (Imp-L2), which might inhibit wing development and increase the body size of insects, was still highly expressed in the adult stage. Moreover, body and leg-component lengths were significantly larger in apterous than in macropterous morphs. The larger size of the apterous morphs and the differences in highly expressed genes between the two wing morphs consistently demonstrate the adaptive significance of wing polymorphism in G. gigas. These results shed light on the future loss-of-function research of wing polymorphism in G. gigas.

Wing polymorphism, which is an evolutionarily successful feature, occurs widely in species of insects, most notably in the orders Coleoptera, Diptera, Hemiptera, Hymenoptera, Lepidoptera, Orthoptera, Psocoptera, and Thysanoptera (Harrison 1980, Roff 1990, Zera and Denno 1997, Grantham and Brisson 2018). This is particularly observed in terms of wing length, including macropterous, brachypterous, and apterous morphs, owing to differences in the completeness of the skeleton, muscles, and nerves of the alary apparatus.

In Hemiptera, the true bugs (Heteroptera) are composed of seven infraorders and all of them contain species with different wing morphs. In particular, species within Gerromorpha exhibit a very distinct polymorphism in the development of their flight apparatus (Andersen 1982). The brachypterous or apterous morphs can be found in most species of all families, among which part of cases are due to wing polymorphism (Andersen 1981, 1982, 2004). However, brachypterous and apterous morphs rarely occur in the same species, at least in the wild (Andersen 1982, Harada 1989, Fairbairn 1992, Hirooka et al. 2016). Because the largest variability in wing development of Gerromorpha is observed among water striders of the family Gerridae, this group is ideal for comparative studies (Brinkhurst 1959; Andersen 1975, 1997). Owing to these traits, the prevalence of wing polymorphism in Gerromorpha, particularly in Gerridae, provides abundant and suitable material for studying the adaptive significance of different wing morphs.

For the water strider species showing wing polymorphism, several studies have provided evidence to support the hypothesis that the corresponding populations exhibit a fitness trade-off between flight capability and reproductive output of females (Andersen 1982). The dispersal ability of macropterous morphs can allow them to fly to suitable habitats, whereas apterous morphs cannot fly, but usually have higher reproductive outputs than macropterous morphs (Harrison 1980; Fairbairn 1986; Zera 2004, 2009; Guerra 2011). However, the adaptive mechanism of different wing morphs of water striders has not been well understood at the level of gene expression, particularly in males. Therefore, to investigate the expression of functional genes, underlying wing polymorphism is crucial for understanding which physiological attributes have major correlation to the adaptive mechanism of wing polymorphism in insects.

In macropterous insects, the rapid movement of the wings is powered by the indirect flight muscles (IFM). These muscles are capable of efficiently affording the rapid oscillatory contraction of thorax because they are activated by stretching in the presence of nearly constant calcium (Ca²⁺) levels (Cammarato et al. 2004, Boussouf et al. 2007, Bullard and Pastore 2011). In this process, the activation by stretching requires particular isoforms of some structural proteins on the thin and thick filaments of IFM (Bullard et al. 2017). In Drosophila, several structural genes have been shown to produce various isoforms of the corresponding structural proteins, which play a key role in muscle contraction and maintenance (Karlik and Fyrberg 1986, Bullard et al. 1988, Vinogradova et al. 2005, Zhu et al. 2009). These structural genes in the IFM are essential for the insects having macropterous morphs, which require good flight capacity to migrate.

For decades, hormones have been a major focus of studies investigating endocrine regulation of wing polymorphism, such as insulin-like peptides (ILPs), ecdysteroids, juvenile hormone, and biogenic amines (Tawfik et al. 1999, Yan et al. 2004, Jindra et al. 2013, Vellichirammal et al. 2017). Endocrine mechanisms control the coordinate expression of these hormones, which results in differentiation among wing morphs. Recently, a series of studies have reported that some components of the insulin/insulin-like signaling (IIS) pathway, mainly ILPs or insulin-like receptors, are known to regulate wing polymorphism in insects (Teleman 2010, Xu et al. 2015, Xu and Zhang 2017). In the brachypterous and apterous insects, the components of the IIS pathway generally regulate their circulating sugar levels and store excess energy in the form of glycogen and lipids, resulting in larger body size (Sopko and Perrimon 2013).

In water strider species, the enigmatic Gigantometra gigas (Hemiptera: Gerridae) is the largest one with narrow distribution in northern Vietnam and Hainan Island of China (Damgaard et al. 2014), and this species has even been considered extinct by some entomologists (JTP's personal communication). The species exhibits a clear wing polymorphism with both macropterous and apterous adults (Supp Fig. 1 [online only]) and its large individual size is ideal for obtaining high quantity and quality of transcriptome. Few studies had reported on this species (Tseng and Rowe 1999, Damgaard et al. 2014), and its physiology, ecology, and evolution exhibit big blank to fill. Technical advances in the high-throughput sequencing have allowed uncovering the molecular differences underlying wing polymorphism through transcriptomic studies (Yan et al. 2004, Zera 2009, Brisson et al. 2010, Sahraeian et al. 2017). De novo transcriptome analysis of short reads has proven to be a valuable first step for studying genetic traits and allowed researchers to obtain sequence information and expression levels of functional genes involved in metabolic and developmental pathways of nonmodel organisms (Xue et al. 2010, Poelchau et al. 2011, Sloan et al. 2012, Van Belleghem et al. 2012, Li-Byarlay et al. 2014).

In the present study, a de novo transcriptome assembly was used to detect differentially expressed genes (DEGs) in macropterous and apterous morphs of adult G. gigas males, aiming to determine which DEGs still contribute to the adaptive differences between these wing morphs in the adult stage. Furthermore, the size dimorphism of G. gigas males is associated with its adaptive ability. Therefore, the present study examined whether the different phenotypes reflect the same adaptive direction as DEGs do. Uncovering the adaptive mechanism of this rare species would also contribute to the corresponding conservational efforts.

Materials and Methods

Sample Collecting

Male G. gigas used for transcriptome analysis were sampled twice from Yinggeling National Nature Reserve, Hainan Province, China, on 20 July 2013 and 8 July 2015 by Xiao-ya Sun and Yan-hui Wang. The first batch included one macropterous and one apterous morph (Wing_1 and Wgl_1), and the second batch included two macropterous and two apterous morphs (Wing_2/3 and Wgl_2/3).

For measuring body length and leg-component lengths of the two wing morphs in G. gigas males, all the collected individuals were used. The measurements were set as two groups: single site and multiple sites. Within the single-site group (Jianfengling National Nature Reserve), measurements of body length were obtained from 11 macropterous and 16 apterous individuals, and leg lengths were obtained from 10 macropterous and 10 apterous ones. Whereas in the multiple-site group (Jianfengling National Nature Reserve, Yinggeling National Nature Reserve, and Diaoluoshan Nature Reserve), 18 macropterous and 18 apterous individuals were examined for body length, and 16 macropterous and 22 apterous ones were examined for leg lengths.

Sequencing and Transcriptome Assembly

Six individual samples (three macropterous and three apterous morphs) collected from Yinggeling National Nature Reserve were used for sequencing. Total RNA of each sample (Supp Table 1 [online only]) was extracted from the thorax muscles of each individual using TRIzol reagent according to the manufacturer’s instructions (Invitrogen, Carlsbad, NY). RNA integrity number (RIN) was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA), and all the samples had RIN >6.5. Individual libraries were prepared using the Dynabeads mRNA Purification Kit (Thermo Fisher Scientific, Waltham, MA) followed by mRNA reverse transcription using various enzymes (Invitrogen), according to the manufacturer’s instructions. The cDNA libraries were subsequently sequenced on an Illumina HiSeq2000/4000 device (Illumina, San Diego, CA), according to the manufacturer’s protocols.

Raw data were preprocessed by removing reads of poor quality (adaptor reads, repeated reads, and reads with low-quality sequencing), and the resulting 5 Gb (approximately) of clean data obtained for each sample were further screened for poor quality bases and adaptor sequences using Trimmomatic v0.32 (http://www.usadellab.org/cms/index.php?page=trimmomatic;Bolger et al. 2014). Sequence quality, including per base sequence quality and per sequence GC content, was assessed using FastQC v0.11.3 (Andrews 2010). Clean data were deposited at the National Centre for Biotechnology Information (NCBI) Sequence Read Archive (SRA) (SRP: 142557; BioProject ID: PRJNA453444).

Clean data were used for de novo transcriptome assembly using Trinity v3.0 (Haas et al. 2013). To obtain nonredundant transcript sequences, contigs of each sample were further clustered to unigenes using the TGI Clustering Tool (Pertea et al. 2003). All unigenes were used for further analysis of differential gene expression.

Functional Annotation of Assembled Transcripts

The protein coding regions in the unigenes were predicted by using TransDecoder v3 (Haas et al. 2013) based on the most likely longest Open Reading Frame. The predicted nucleotides and proteins were searched in the SwissProt database using basic local alignment search tool (BLAST) homologies in Trinotate v3 (https://trinotate.github.io/) considering an E-value ≤10^–5. Both BLASTX (search protein database using a nucleotide query) and BLASTP (search protein database using a protein query) were used, and their top-hit matches (E-value ≤ 10^–5) were kept for the downstream analysis. Trinotate was also used in searches considering specific releases of NR (RefSeq nonredundant proteins) and Pfam databases (available at https://data.broadinstitute.org/Trinity/Trinotate_v3_RESOURCES/). Homologous proteins found in the SwissProt database were used to retrieve functional annotations from the Evolutionary Genealogy of Genes: Non-supervised Orthologous Groups (EggNOG), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Gene Ontology (GO) databases via Trinotate. The EggNOG (Powell et al. 2012) is a database of orthologous groups with general functional annotation codes, largely derived from the Clusters of Orthologous Groups (COG) database (Tatusov et al. 1997). The GO terms were searched in both the SwissProt and Pfam results, and annotated GO terms were categorized in Blast2GO v5.0 (Conesa et al. 2005). The EggNOG annotations were filtered to keep COGs and those were categorized using the current version of the COG database (ftp://ftp.ncbi.nih.gov/pub/COG/COG2014/data). KEGG annotations were filtered to keep KEGG orthologs (KOs) and those were categorized using the tool ‘Reconstruct Pathway’ (http://www.kegg.jp/kegg/tool/map_pathway.html). Annotation results for the NR, EggNOG, KEGG, and GO databases were displayed using VENNY v2.1 (Oliveros 2007).

Identification and Profiling DEGs for Wing Polymorphism

To align short reads directly to all unigenes, we generated gene indices and mapped reads separately for the six individual samples collected from Yinggeling National Nature Reserve using STAR v2.5.3a (Dobin et al. 2013). In the mapping step of STAR, the maximum number of multiple alignments allowed for a read was set at 20; if this value was exceeded, the read would be considered unmapped.

Gene reads count files of the six samples obtained by STAR were merged into a total file for the downstream analysis. Identification of DEGs between the different wing morphs of G. gigas was performed using DESeq2 (Love et al. 2014). This count-based tool accurately detects DEGs based on negative binomial distribution with log2 transformed fold change (log2Fold Change). Due to the different batches of the six samples, batch information was included in the analysis. We explored DEGs between the different wing morphs using all unigenes. The six samples were divided into two sample groups, each containing apterous or macropterous morphs only. Samples were visualized using a principal component analysis (PCA) and clustering analysis for data quality assessment and visualization of replicate concordance between samples. We extracted different genes in two parts, consisting of upregulated genes and downregulated genes considering log2Fold Change ≥ 1 or ≤ −1, respectively, with an adjusted P-value threshold (FDR threshold) of 0.05. The annotated transcripts of DEGs were aligned with genes available in FlyBase database (http://flybase.org/). The GO information for the transcripts described different biological process, molecular function, and cellular component categories. The GO enrichment analysis of DEGs was performed using clusterProfiler (Yu et al. 2012), with a cutoff q-value of 0.05 to determine significant differences.

Statistical Analysis of Body and Leg-Component Lengths

Software SPSS v22.0 (SPSS, Inc., Chicago, IL) was used for statistical analysis. The Student’s t-test analysis was used for determining significant differences in body length and leg-component lengths between the macropterous and apterous morphs of G. gigas. The threshold P-value for statistical significance was 0.001.

Results

RNA-seq and Transcriptome Assembly

After quality filtering, the first batch of samples contained 55.5 million paired-end reads, and the second batch contained 30.5 million paired-end reads (Table 1). A summary of the results of the transcriptome assembly is presented in Table 2. The FastQC reports showed that the Illumina reads for cDNA libraries were generally of an acceptable quality in terms of GC distributions and per base sequence quality (Supp Fig. 2 [online only]). In addition, no poor quality bases and adaptor sequences were filtered by Trimmomatic. The assembly resulted in a transcriptome consisting of 55,784 unigenes, and the size of the unigenes ranged from 200 to 25,874 bp with a mean length of 625 bp and a minimum unigene length needed to cover 50% of the transcriptome (N50) of 955 bp. The length distribution of unigenes from G. gigas is shown in Fig. 1. The unigenes in the transcriptome had a GC content of 35.87%. Moreover, unique reads accounted for an average of 63% when mapping all unigenes (Table 1).

Table 1.

Read counts (number of read pairs) per library and percentage of unique reads mapped on unigenes

Libraries		No. of read pairs	Mapped reads (%)
Macropterous morph, replicate 1	Wing_1	55,457,714	63.18
Macropterous morph, replicate 2	Wing_2	30,617,432	62.42
Macropterous morph, replicate 3	Wing_3	30,591,242	65.41
Apterous morph, replicate 1	Wgl_1	55,457,714	64.89
Apterous morph, replicate 2	Wgl_2	30,617,432	63.82
Apterous morph, replicate 3	Wgl_3	30,591,242	63.61

Table 2.

Statistics from the de novo transcriptome assembly of G. gigas

	Assembly statistics
Number of unigenes	55,784
Total length of unigenes	34,898,737
Shortest unigene	297
Longest unigene	24,945
Mean unigene length	625
N50 unigene length	955
GC-content of all unigenes (%)	35.87

Except for the number of unigenes and GC-content of all unigenes, all statistics were given in numbers of base pairs.

Fig. 1.

Length distribution of unigenes from G. gigas.

Functional Annotation

Among assembled unigenes, 18,061 out of 55,784 (32%) were annotated by Trinotate with statistically significant (E-value < 10^–5) hits. The annotation results in NR, EggNOG, KEGG, and GO databases were shown in Fig. 2. The BLASTX and BLASTP searches against the SwissProt database resulted in total reliable annotations for 13,033 unigenes. Annotations in NR and Pfam databases were obtained for 17,304 and 7,665 genes, respectively. The BLAST searches against the EggNOG database resulted in annotations for 8,261 genes. The GO terms assigned to the unigenes were distributed across the biological process, cellular component, and molecular function categories, with a mean level of two (Supp Fig. 3 [online only]). Among these, 10,351 unigenes were assigned to at least one GO term. In the KEGG database, 12,764 unigenes showed orthologs in the KO database.

Fig. 2.

The Venny diagram profiles for all transcripts annotated in four databases. The gene annotations obtained by NR, EggNOG, KEGG, and GO databases are shown in all transcripts of the two wing morphs of G. gigas.

Identification of DEGs Based on FlyBase

Based on the DESeq2 results, the PCA and clustering analysis evidenced the close relationship of samples (Supp Fig. 4 [online only]). Both up- and downregulated DEGs were found for apterous and macropterous morphs, as measured by log2Fold Change (P-value ≤ 0.05, Supp Fig. 5 [online only]). In macropterous morphs, 142 of the 423 DEGs were annotated, whereas in apterous morphs, 152 of the 385 DEGs were annotated. For both wing morphs, the corresponding 42 genes available in FlyBase database were found from the 294 annotated DEGs. Among the DEGs matching FlyBase, 10 were correlated with wing development (Table 3). The top 40 DEGs identified for each wing morph were shown in Fig. 3. Imaginal morphogenesis protein-Late 2 (Imp-L2; log2Fold Change = 2.4) and lipid storage droplet-2 (Lsd-2; log2Fold Change = 1.6) were found in differentially expressed transcripts within the apterous morphs. For the macropterous morphs, Z band alternatively spliced PDZ-motif protein 52 (Zasp52; log2Fold Change = 13.6), flightin (fln; log2Fold Change = 11.6), troponin C at 41C (TpnC41C; log2Fold Change = 8.5), tropomyosin 2 (Tm2; log2Fold Change = 5.7), myosin heavy chain (Mhc; log2Fold Change = 5.1), and wings up A (wupA; log2Fold Change = 1.3), which coregulate flight muscle development, were differentially expressed. The hormone receptor-like in 38 (Hr38; log2Fold Change = 3.8) and trehalose transporter 1–2 (Tret1-2; log2Fold Change = 2) were also differentially expressed in the macropterous morphs of G. gigas.

Table 3.

List of wing development genes blasted against FlyBase database found in DEGs of G. gigas

FlyBase ID	Gene symbol	Gene name	Function	log2Fold Change	P-value
Apterous morph
FBgn0001257	Imp-L2	Imaginal morphogenesis protein-Late 2	Inhibiting ILP2	2.4	9.64E-05
FBgn0030608	Lsd-2	Lipid storage droplet-2	A barrier for lipases and preventing the mobilization of lipid store	1.6	0.003
Macropterous morph
FBgn0265991	Zasp52	Z band alternatively spliced PDZ-motif protein 52	Myofibril assembly	13.6	1.24E-22
FBgn0005633	fln	flightin	Correct assembly of the myosin thick filament	11.6	4.55E-63
FBgn0013348	TpnC41C	Troponin C at 41C	Calcium ion binding	8.5	3.17E-55
FBgn0004117	Tm2	Tropomyosin 2	Actin filament binding	5.7	1.85E-12
FBgn0264695	Mhc	Myosin heavy chain	Muscle contraction	5.1	4.23E-19
FBgn0014859	Hr38	Hormone receptor-like in 38	Development of the adult cuticle, uptake and storage of glycogen in larva	3.8	9.64E-05
FBgn0033644	Tret1-2	Trehalose transporter 1-2	Mediating the bidirectional transfer of trehalose	2	0.019
FBgn0283471	wupA	wings up A	Calcium-dependent regulation of muscle contraction	1.3	0.048

ILP2, insulin-like peptide 2.

Due to one or more transcripts corresponding to a DEG, the log2fold change of each DEG shown in this table was the mean value of those of all transcripts, which were annotated as the same DEG (P-value ≤ 0.05).

Fig. 3.

Heatmap in macropterous and apterous morphs of G. gigas. Red cells indicate high levels of expression and blue cells indicate low levels of expression. The pink bar stands for the first batch, and the blue bar stands for the second batch. The purple and the green bars correspond to the apterous morph condition, and the macropterous morph condition, respectively. All genes are shown in gene symbols, and NA stands for the transcript, which is not annotated.

Enrichment Analysis

The function enrichment analysis was used for finding over-represented GO terms after controlling for False Discovery Rate (FDR; q-value ≤ 0.05), and predicting differences in functional trends between the two wing morphs (Fig. 4). In the apterous morphs, GO terms focused on phosphotransferase activity and inhibited performance, whereas in the macropterous morphs, GO terms focused on the flight muscle and cuticle development.

Fig. 4.

GO enrichment analysis of the two wing morphs in G. gigas. (a) Result of GO enrichment in macropterous morph and (b) result of GO enrichment in apterous morph. The color bar stands for the adjust q-value of each GO term. The size of circle stands for the number of genes in each GO term.

Body and Leg-Component Lengths

The body and leg-component lengths of macropterous and apterous morphs showed significant differences both in single- and multiple-site groups (P < 0.001). The mean lengths of body and leg components and their SDs are shown in Fig. 5. The apterous males of G. gigas had significantly longer leg-component and body than macropterous males. In the single-site group, the body length of apterous males was 4.4% greater than that of macropterous males, and leg-component lengths of apterous males were 11.9–27.8% longer than those of macropterous males (Supp Table 2 [online only]). In the multiple-site group, the body length of apterous males was 8.1% greater than that of macropterous males, and leg-component lengths of apterous males were 14.9–28.4% longer than those of macropterous males (Supp Table 3 [online only]).

Fig. 5.

The statistical results of body length and leg-component lengths of G. gigas in the different wing morphs. Body length was defined as the distance from the anterior of the labrum to the posterior of the last abdominal segment, and the femur and tibia of fore, middle, and hind leg lengths were separately measured. The body and leg-component lengths of the two wing morphs are separately measured in the single- and multiple-site groups. The mean lengths of body and leg-component of the individuals in single-site group (a) and in multiple-site group (b) are shown. The blue rectangles of both panels stand for the macropterous morph of G. gigas, and the red rectangles stand for the apterous morph. The error bar stands for the SD in individuals of each wing morph. The stars (***) stand for the significant differences of body length and leg-component lengths between two wing morphs (P < 0.001).

Discussion

Most species of water striders exhibit various wing morphs, and some earlier researches have revealed many ecological factors of wing polymorphism of water striders in various aspects, including the effects of population density (Harada and Spence 2000), nutrient conditions (Harada and Nishimoto 2007, Hirooka et al. 2016), photoperiod (Guthrie 1959; Harada and Numata 1993; Harada et al. 2005, 2011), temperature (Harada et al. 2003), or geography (Ahlroth et al. 1999). In contrast, previously, molecular studies on the mechanism of wing morph adaptation have been scarce in water striders. With the increasing prevalence of transcriptomic data, such mechanism can be slightly elucidated at the gene expression level.

To our knowledge, this is the first de novo transcriptome assembly for the different wing morphs of water striders. In the present study, a few DEGs were found between macropterous and apterous morphs of G. gigas in the adult males. Several structural genes, such as Zasp52, fln, TpnC41C, Mhc, Tm2, and wupA genes, which coregulate the development of flight muscles, were still highly expressed in the adult macropterous males. Genes Imp-L2 and Lsd-2, which are involved in the endocrine regulation of wing polymorphism, were still highly expressed in the adult apterous males. The DEGs found between the macropterous and apterous morphs provide gene expression information on the adaptive mechanism of wing polymorphism of adult G. gigas.

In the present study, wing polymorphism in G. gigas has been investigated considering two opposite directions of adaptation. The wing development genes aligned against FlyBase revealed that the regulation of flight muscle contraction still plays a key role in the macropterous morphs by mediating the expression of certain structural genes in the adult stage (Fig. 6a). In this morph, the highly expressed genes Zasp52 (log2Fold Change = 13.6), fln (log2Fold Change = 11.6), and TpnC41C (log2Fold Change = 8.5) are IFM genes for production of flight muscles. Zasp52 is the major Zasp PDZ domain protein that mediates actin cross-linking in α-actin, and it is essential for myofibril assembly and maintenance at Z-discs (Luther 2009, Liao et al. 2016). This helps other structural genes to regulate muscle contraction and relaxation in the myofibrils. The IFM-specific protein fln, which is not expressed in other muscle types, is involved in the correct assembly of the thick filament of myosin during flight muscle development (Guruharsha et al. 2011). Protein TpnC41C is predominantly expressed in the pupal and adult tubular muscles, with minor amounts in the flight muscles, and it binds Ca²⁺ to regulate muscle contraction (Herranz et al. 2004). Muscle becomes fibrillar owing to the expression of the structural genes TpnC41C and fln, which enable the muscles to function as highly oxidative muscles for high endurance and low force (Bryantsev et al. 2012). Other structural genes also play a role in the regulation of flight muscle contraction. Gene wupA is responsible for the ATPase inhibitory subunit of troponin in the thin filament regulatory complex, and therefore involved in the regulation of muscle and nervous systems. Gene Mhc regulates muscle contraction, and gene Tm2 plays a role in the calcium-dependent regulation of muscle contraction. In short, these structural genes may coregulate the development and maintenance of flight muscles (Tapanes-Castillo and Baylies 2004, Guruharsha et al. 2011, Sopko and Perrimon 2013). Moreover, biological processes in the development of flight muscles accounted for a large proportion of terms in the GO enrichment analysis (Fig. 4a), whereas the structural constituent of cuticle is most enriched in the GO enrichment analysis, which might affect or develop the dorsal and ventral layers of epidermal cells in the alary apparatus (Brower and Jaffe 1989, Moreno et al. 2002). Thus, in the giant water strider, these highly expressed structural genes likely contribute to macropterous males’ fully developed flight capacity allowing their migration to more suitable habitats.

Fig. 6.

The gene networks for DEGs in different wing morphs. (a) These structural genes shown regulate the flight muscle development and maintenance. The intensity of color stands for the value of log2Fold Change. The much deeper color stands for the much larger value of log2Fold Change. (b) The Imp-L2 gene is involved in the IIS pathway. The square nodes stand for the gene inhibiting the pathway, and the circle nodes stand for the gene activating the pathway. The width of connected lines in both panels stand for the confidence of the interacted genes. The thick lines indicate the strong interaction of the two connected genes, and the thin lines indicate the weak interaction.

In the apterous morphs of G. gigas, the Imp-L2 gene, which is a component of the IIS pathway (Fig. 6b), was highly expressed in the adult stage (log2Fold Change = 2.4). Protein Imp-L2, produced by corpus cardiacum cells, binds insulin-like peptide 2 and inhibits its function, thereby reducing nonautonomous tissue growth in the larval stage (Okamoto et al. 2013). By suppressing the insulin signaling pathway, Imp-L2 increases lipid storage resulting in large body size. Overexpression of Imp-L2 is known to reduce phosphatidylinositol (3,4,5)-trisphosphate levels (Honegger et al. 2008). Much evidence implies that physiological effects of manipulating insulin signalling differ largely depending on the tissue in which the process occurs (Kitamura et al. 2003, Teleman 2010). Considering insulin signaling in a specific tissue, the GO enrichment analysis (Fig. 4b) demonstrated that both phosphorylation and inhibition regulation are enriched in apterous males of G. gigas, probably resulting from the activation of the IIS pathway in some organs, regardless of Imp-L2 overexpression. Thus, it might demonstrate that Imp-L2 overexpression inhibits specifically downstream signaling of the IIS pathway in the wing disc, producing the apterous morph. Moreover, the Lsd-2 gene (log2Fold Change = 1.6), also overexpressed in the apterous morph and expressed in fat bodies, is involved in lipid storage in these tissues (Fauny et al. 2005). This gene is expressed during all stages of Drosophila development, and it is associated with the accumulation of lipid droplets (Grönke et al. 2003). The overexpression of Imp-L2 and Lsd-2 genes in the adult stage has similar effects on the maintenance of large body size in apterous G. gigas.

Concerning the morphological data, the body and leg-component lengths of the apterous morphs were significantly larger than those of the macropterous morphs (Fig. 5). Previous studies have reported that sexual selection favors larger males, owing to the improvement of mating success in water striders (Sih and Krupa 1992, Fairbairn and Preziosi 1996, Danielsson 2001). Besides, in Aquarius remigis, another species of water strider, apterous males still have additional mating advantage over macropterous males, likely because the macropterous males migrate after diapause, immediately before reproduction, thereby providing the apterous morphs more chances for mating (Kaitala and Dingle 1993, Fairbairn and Preziosi 1996). Evidence of mating advantage for apterous males with their larger body size compared with macropterous males indicates a trade-off between the dispersal and reproduction in males. The larger body size of apterous morphs has also been reported as more adaptive to cold environments (Ahlroth 1999). Moreover, leg-component lengths of apterous G. gigas were larger than those of macropterous morphs. Previous studies have reported that the evolution of new gene interactions contribute to the shape of the legs that enables water striders to dodge predator strikes. The predation escape strategy relies primarily on a jump reflex, enabled by the long slender legs (Armisén et al. 2015). It would be inferred that apterous male possessing longer legs is more suitable to live on water surface than macropterous male. Thus, highly expressed genes in the apterous morphs of G. gigas and the phenotype displayed by these males are consistent in terms of adaptation.

De novo transcriptome assembly of the present study uncovered a few DEGs in the adult males, which still play a key role in wing polymorphism of G. gigas. The gigantic species is hard to find and collect in the field and the larvae of this species have never been reported by far, and the present research tries to integrate the understanding on the adaptations of water striders with different wing morphs revealed by molecular and morphological evidence. In this study, a series of structural genes found in macropterous males contribute to the maintenance of flight capacity to migrate and find a suitable habitat, while the gene Imp-L2 found in apterous males might keep the larger body size to increase mating success or surviving through adverse environment, which is consistent with the adaptation displayed by the phenotype of these males. These DEGs would elucidate the adaptive mechanism in wing polymorphism of G. gigas, and perhaps could be one of the factors for wing polymorphism of other water striders.