Introduction

Crabs are one of the most species-rich groups among extant crustaceans, with over 7250 described species, and many crabs play an important role as a food source for humans with commercial value in fish markets worldwide (Wang et al. 2018c; De Grave et al. 2009). In the broadest sense, crabs could be classified into three functional feeding groups: herbivores (HV), omnivores (OV), and carnivores (CV). Many species of mangrove crabs, e.g., sesarmid crabs, are known to be herbivorous, and they can feed on mangrove pneumatophores, bark, and macroalgae (Dahdouh-Guebas et al. 1999; Wang et al. 2018a, 2019b). Portunid crabs are reported as being mainly carnivorous, preying on gastropods, crustaceans, bivalves, polychaetes, and sometimes fish (Choy 1986; Cannicci et al. 1996; Figueiredo and Anderson 2009). Yet most crabs are, to some degree, omnivorous, feeding on resources including aquatic plants, algae, plankton, molluscs, fish, worms, and organic matter (Figueiredo and Anderson 2009; Jin et al. 2003). As such, different crabs may have different feeding adaptations, yet little is known about the genetic basis of physiological adaptations.

Digestive enzyme activity is the most common indicator for evaluating the capacity for digestion (Dai et al. 2009). The digestion of food to obtain nutrients is a core physiological function (Wei et al. 2014), and digestive enzymes of animals with different feeding habits (herbivores, omnivores and carnivores) have different characteristics (Wang et al. 2016a, b). In crabs, the hepatopancreas is an important organ for the absorption and storage of nutrients, and can synthesize digestive enzymes for food digestion. Studies on hepatopancreatic secretions have identified trypsin, chymotrypsin, amylase, and lipase as the most important components for digestion in crustacean species (Fernández et al. 1997; Wei et al. 2014; Hammer et al. 2000). Despite hepatopancreas have important role in digestion (Figueiredo and Anderson 2009; Dammannagoda et al. 2015), research on digestive enzyme gene diversity and gene regulation in crabs’ hepatopancreas is rudimentary.

Next-generation sequencing technologies make it possible to generate large amounts of transcript sequences and gene expression data, including for non-model species without a sequenced genome (van Dijk et al. 2014; Zhu et al. 2017). In addition, de novo transcriptome provides great breadth and depth of information that can be used to allow cataloging of all genes expressed in a tissue, and facilitate detailed functional research regarding various proteins (Bain et al. 2016; Wang et al. 2018b). There are many transcriptomic analyses of crustaceans that help build a more complete understanding of regulatory mechanisms, such as heavy metal detoxication (Wang et al. 2019c), osmoregulation (Azam et al. 2016; Wang et al. 2018a), and oxidation resistance (Hui et al. 2017). However, transcriptomic profiling is thus far limited regarding molecular characteristics of crabs’ digestive enzyme genes.

In the present study, we performed adaptive evolution analysis on digestive enzyme protein-coding genes of 12 crab species based on hepatopancreas transcriptome data. We collected all the positively selected genes among CV and HV crabs, and compared them with a representative OV crab (Eriocheir japonica sinensis) to examine whether adaptive evolution is apparent. These results will reveal the genetic basis for some feeding adaptations of crabs, and improve our understanding of their genetic and evolutionary architecture.

Materials and methods

Sample collection, RNA extraction, and Illumina deep sequencing

The Illumina paired-end transcriptome data were generated from 12 crab species, including 3 CV, 4 HV, and 5 OV crabs (Table 1). All species were collected from a coastal mudflat wetland or surrounding waters in Shanghai, China. Biological information of the species is presented in Table S1. Crabs were placed in an ice bath for 1–2 min until anesthetized. The hepatopancreas was removed through surgery, immediately frozen in liquid nitrogen, and stored at − 80 °C until RNA extraction. All procedures followed guidelines for the care and use of laboratory animals from the Institutional Animal Care and Use Committee of Yancheng Teachers University, Yancheng, China. Total RNA was isolated from the hepatopancreas samples using Trizol reagent (Invitrogen, San Diego, CA, USA) following the manufacturer's protocol. The RNA quality and concentration were determined with a Nanodrop-2000 spectrophotometer (NanoDrop products, Thermo Fisher Scientific, Inc., Wilmington, DE, USA). Samples for whole transcriptome analysis were prepared using an Illumina kit (Illumina Inc., San Diego, CA, USA) following the manufacturer's recommendations. First, mRNA was purified using oligo (dT) magnetic beads, and the mRNA was split into short fragments (about 200 bp) using a fragmentation buffer. The first strand of cDNA was synthesized with random hexamer primers using the mRNA fragments as templates. Then, buffer, dNTPs, RNase H, and DNA polymerase I were added to the mixture to synthesize the second strand. Subsequently, sequencing adapters were ligated to the 5′ and 3′ ends of the fragments. The fragments were purified by agarose gel electrophoresis and enriched by PCR amplification to create a cDNA library. Eventually, the cDNA library was sequenced on an Illumina HiSeq X Ten platform (Illumina Inc.) and 100 bp paired-end reads were generated.

Table 1 Characteristics and assembly statistics for 12 crabs’ transcriptomics
Full size table

Transcriptome de novo assembly

Clean reads were obtained by removing reads containing an adaptor, reads containing poly-N (the ratio of ‘N’ to be more than 10%), and low quality reads (quality score < 20). Transcriptome assembly was accomplished based on clean reads using Trinity (Grabherr et al. 2011), with min_kmer_cov set to 2 and all other default parameters. The longest copy of redundant transcripts was regarded as a unigene. Unigenes were aligned to databases, including NR, String, Pfam, the Swiss-Prot, and the COG for eukaryotic complete genomes database, separately, using BLASTX with E values < 1E−5 (Altschul et al. 1997).

Identification of digestive enzyme genes and orthology inference

According to the annotation result, all unigenes related to most digestive enzymes were identified with the following criteria: the annotations of all-unigenes were obtained by a BLASTX (Camacho et al. 2009) search against the NR or Swiss-Prot database to match the corresponding digestive enzymes with E value < 1E−10. All candidate genes were selected manually and reconfirmed using the BLASTX network server in the National Center for Biotechnology Information (NCBI). The open reading frames (ORFs) of the putative digestive enzyme genes were predicted using an ORF finder (https://www.ncbi.nlm.nih.gov/orffinder/). In order to identify groups of putative orthologs, we adopted an approach based on sequence similarity and a tree-based approach (Yang and Smith 2014). Considered short sequences were not useful for the following analyses; only high quality and high integrity sequences were collected. Ultimately, 10 genes were chosen: Alpha-amylase (AMYA), N-aminopeptidase (APN), Carboxypeptidase B (CPB), Chymotrypsin-like proteinase (CTRL), Maltase-Glucoamylase (MGAM), Pancreatic lipase (PNLIP), Retinoid-inducible serine carboxypeptidase (RISC), Trypsin (TRY), Triacylglycerol lipase (TL), and xaa-Pro dipeptidase (XPD). All have a well-studied structure and function and are known enzymes in digesting lipid, protein, and carbohydrates. The detail functional information of the 10 genes is presented in Table S2, and the length of each gene is in Table S3. The sequences of the 10 digestive enzyme genes were verified by the genome data of Eriocheir japonica sinensis (PRJNA555707) and Portunus trituberculatus (PRJNA555262) (Tang et al, 2020a, b). The two species’ cDNA sequences are identical to the genomic data. Nucleotide sequences of each gene examined, and their deduced amino acid sequences, were aligned using MEGA 7.0 (Kumar et al. 2016) and manually adjusted with GeneDoc. The all sequences were deposited in GenBank under accession numbers MN964137–MN964248.

Adaptive evolution analysis

Estimating the nonsynonymous (dN)/synonymous substitution (dS) rate (ω = dN/dS) is considered a useful way to quantify the impact of natural selection on adaptive evolution (Ohta 1992; Wang et al. 2017, 2018a). The change of selective pressures can be indicated by the value of ω, where ω < 1, ω = 1, and ω > 1 correspond to purifying selection, neutral evolution, and positive selection, respectively (Wang et al. 2018a). Here, the selective pressure was tested based on phylogeny by using the codon-based maximum likelihood (CodeML) program in the PAML 4.7 package (Yang 2007). The well-supported phylogeny of Brachyura (Wang et al. 2019a; Shen et al. 2013) was used as the input tree in all analyses (Fig. 1).

Fig. 1
figure 1

A well-supported phylogeny of crabs used for selective pressure analysis in PAML. Positively selected genes identified in branch-site model are indicated

Full size image

We first used the free-ratio and branch-site models (Zhang et al. 2005) implemented in CodeML to evaluate whether positive selection was restricted to specific HV or CV lineages. The free-ratio model (M1) assumes an independent ω ratio for each branch, and the one-ratio model (M0) which set up the same ω for all branches was used as the null hypothesis (Yang 1998). The improved branch-site model A (test 2) was applied to investigate the evolutionary rates of lineages of herbivores (HV), E. j. sinensis (OV), and carnivores (CV). Crabs of the three groups were respectively assigned as foreground, and compared with the null model, which assumes that all branches have the same evolutionary rate. The likelihood ratio test (LRT) statistic (2ΔL) approximates a chi-square distribution, and was used to test if there were significant evolutionary rate differences between foreground and background lineages. The Bayes empirical Bayes (BEB) approach (Yang et al. 2005) was used to identify amino acids under selection for CODEML.

Furthermore, to test for evidence of positive-selected sites in the digestive enzyme genes, we used the Datamonkey online server (https://www.datamonkey.org), that has the advantage of improving the dN/dS ratio estimate by incorporating variation in the rate of synonymous substitution (Yang et al. 2005). The internal fixed-effect likelihood (IFEL) and random-effect likelihood (REL), two different codon-based maximum likelihood methods, were used estimate the dN/dS ratio at every codon in the alignments (Pond and Frost 2005). The IFEL model estimates the ratio of dN/dS on a site-by-site basis, without assuming a priori distribution across sites. The REL model first fits a distribution of rates across sites and then infers the substitution rate for individual sites (Poon et al. 2009). Sites with p values < 0.1 for IFEL, and Bayes factor > 50 for REL, were considered as candidates under positive selection.

Finally, to further validate the result of PAML and Datamonkey, a complementary protein level approach was used in the TreeSAAP program, that compares non-synonymous residue property changes and identifies positive selection amino acid properties (Woolley et al. 2003). The TreeSAAP program detected sites based on 31 physicochemical amino acid properties, which were all magnitude category 6–8 changes; p values ≤ 0.01 were used as an index for the degree of radical amino acid substitution and positive selection (Wang et al. 2018a).

Mapping of positively selected sites onto protein structures

To gain insights into the functional significance of the putatively positive-selected sites, we mapped them onto the protein secondary and three-dimensional structures. We used the RaptorX (https://raptorx.uchicago.edu/) to predict the secondary structures of the implicated proteins, such as helix, beta-sheet, and coil (Källberg et al. 2014; Wang et al. 2016a, b). The protein structure domains were predicted by the PredictProtein web server (https://ppopen.rostlab.org/) (Yachdav et al. 2014). The 3D structures of genes under positive selection were predicted using the homology modeling software provided by the I-TASSER server (Zhang 2008). The radical amino sites under positive selection identified by more than one method were mapped onto the 3D structure using PYMOL.

Results

Transcriptome sequencing and de novo assembly

Obtaining cDNA sequences by transcriptome sequencing for evolutionary analysis has become an efficient option (Yuan et al. 2017; Ma et al. 2019). Here, the 12 newly generated crabs’ hepatopancreas transcriptomes contained a total of ~ 587.92 million clean reads after filtering, ranging from 38.38 to 57.10 million clean reads (Table 1). The assembled transcriptomes had an average of 228,925 transcripts with a N50 length of 307 to 1077 bp (Table 1). Number of predicted unigenes ranged from 113,860 to 399,653 per species, with an average length of 326.43 to 565.04 bp.

Molecular evolution of digestive enzyme genes in crabs

To test the selection constraints of different crab lineages for the 10 digestive enzyme genes, we used likelihood models of coding sequence evolution implemented in Codeml of the PAML package. With the one-ratio model (M0), that only allows a single ω ratio for all crab branches, the ω value of 10 genes ranged from 0.0695 to 0.270 (Table 2). The free-ratio (M1) model was significantly better than the M0 model (p < 0.05, Table 2) for eight genes (AMYA, APN, CPB, MGAM, PNLIP, RISC, TRY, XPD), suggesting heterogeneous selective pressures on the different lineages.

Table 2 Free-ratio (M1 vs M0) analyses of selective pattern on the digestive enzyme genes in crabs
Full size table

Then, we used the branch-site model to test for positive selection in individual codons for the three functional feeding groups (i.e., HV, OV, and CV). Considering that most crabs are omnivores, E. j. sinensis was set to represent OV group and the control for the other two groups. Results show that two (AMYA and MGAM) genes were identified under positive selection for HV, whereas no significant sign of positive selection was detected for CV (Table 3 and Fig. 1). Conversely, LRT tests showed that four genes (CPB, PNLIP, RISC, and XPD) were indicated as positive-selected in the ancestral branch of CV, but these genes were not detected as positive-selected in the HV group (Table 3 and Fig. 1). The ANY gene was determined to be positively selected in both HV and CV groups. 4 (HV), 2 (HV), 7(CV), 1(CV), 6 (CV), 2 (CV), and 38 (16 in HV, 22 in CV) sites were separately identified in AMYA, MGAM, CPB, PNLIP, RISC, XPD, and ANY, respectively, to be under selection using the BEB approach (posterior probabilities 0.60). Notably, no significant signs of positive selection for all 10 digestive enzyme genes were detected in the E. j. sinensis lineage (Table 3 and Fig. 1). These results are consistent with results drawn from free-ratio models.

Table 3 Selective pressure analyses (branch-site model) of 10 digestive enzyme genes in crabs
Full size table

Significant evidence of positive selection was further identified by the other two ML methods (IFEL and REL) implemented in Datamonkey, in which ω values were calculated based on the dS. The IFEL model results showed 33 positively selected sites in 7 genes (5 in AMYA, 9 in ANY, 4 in CPB, 8 in MGAM, 1 in PNLIP, 3 in RISC, and 3 in XPD) at a significance level < 0.1 (Table 4). In addition, REL also identified 11 codons in 3 genes (2 in AMYA, 6 in APN, and 3 in PNLIP) under positive selection at a level of Bayes factor > 50 (Table 4).

Table 4 Datamonky estimates for positive selection sites of digestive enzyme genes
Full size table

To support the ML method results, a complementary protein-level approach was implemented in TreeSAAP (Woolley et al. 2003). In TreeSAAP, the number of radical changes in the amino acid properties was used as a proxy for determining the strength of positive selection at a particular amino acid position (Sunagar et al. 2012). A total of 94 positively selected codons identified at three genes (17 in AMYA, 28 in MGAM, and 49 in APN) in the HV group (Table S4). Moreover, a series of putative positively selected sites from 5 osmoregulatory genes were identified in the CV group, i.e., 12 in APN, 19 in CPB, 14 in PNLIP, 14 in RISC, and 13 in XPD (Table S4).

Structural analyses of positively selected sites

To obtain insight into the functional significance of the putatively selected sites, we mapped all positively selected sites onto secondary structures of the corresponding digestive enzymes. It was found that the most of the positively selected sites were detected to fall in the regions of the functional regions within structures of the digestive enzymes (Table S5). One positively selected site in PNLIP (codon 267) was in the active site (Fig. 2).

Fig. 2
figure 2

Distribution of positively selected sites in the three-dimensional (3D) structure of PNLIP

Full size image

Discussion

Animals are thought to be adaptive in their digestive enzyme production in response to differences in diet (German et al. 2004). Herbivorous animals normally exhibit higher carbohydrase activities (Horn 1989), and carnivorous animals frequently show higher lipid and protein enzyme activities (German et al. 2004; Hidalgo et al. 1999). Yet there is little data on the genetic basis underlying different feeding habits. To this end, we investigated selection pressure on the digestive enzyme genes of 12 crab species based on transcriptome data.

In the one-ratio model analysis, the ω values of all digestive enzyme genes were significantly less than 1, suggesting the general evolutionary pattern for crab digestive enzymes is conservative (Table 2). Even so, the free-ratio and branch-site model analyses still provide strong evidence that several digestive enzyme genes have been subjected to positive selection in HV and CV crabs (Tables 2, 3, and Fig. 1). According to previous studies, the positive selection signs are usually swamped by continuous negative selection that occurs on most sites in a gene sequence because positive selection mainly acts on only a few sites and for a short period of evolutionary time (Shen et al. 2010; Zhang et al. 2005). These reasons may partly explain why we can detect positive selection in some genes using the free-ratio and branch-site models, but not with the one-ratio model.

Selection analysis results showed significant evidence for positive selection at 7 of 10 digestive enzyme genes, i.e., AMYA (in HV), MGAM (in HV), CPB (in CV), PNLIP (in CV), RISC (in CV), XPD (in CV), and ANY (in HV and CV). Moreover, a series of sites were detected to be under positive selection in these genes using the codon-based maximum likelihood methods in Datamonkey (Table 4), which provides accessional evidence for positive selection. Adaptive evolution was also supported by evidence that the positively selected sites were identified by the protein-level approach in TreeSAAP (Table S4). Particularly, many of positively selected sites were localized on or near important structural regions (i.e., protein-binding region, alpha-helix, and beta-sheet) in predicted secondary and tertiary protein structures of the digestive enzyme, suggesting that these positively selected sites might have influenced protein properties and functions. Therefore, these nucleotide and protein-level analyses indicated positive selection may be a major driving force for evolution of digestive enzyme genes with respect to different functional feeding strategies.

Strong evidence for positive selection was noted at AMYA and MGAM in the HV group, in agreement with previous studies that carbohydrate-degrading enzymes (e.g., amylase, Maltase-Glucoamylase) tend to be higher in activity in guts of herbivorous crabs than in those of omnivores and carnivores (Dahdouh-Guebas et al. 1999). The AMYA gene encodes an alpha-amylase produced by the hepatopancreas. Alpha-amylases are common hydrolytic enzymes that break down polysaccharides by hydrolyzing alpha-d-(1,4)-glucan bonds, and thus catalyze the initial step in digestion of dietary starch, glycogen, and other related carbohydrates (Franco et al. 2000; Ramzi and Hosseininaveh 2010). The MGAM gene encodes maltase-glucoamylase, which is a brush-border membrane enzyme that plays a role in the final steps of digestion of starch. In addition, maltase-glucoamylase has an important role in herbivore evolution because of its importance in the breakdown of secondary metabolites of plants (Hemming and Lindroth 2000). Thus, the observed positive selection in AMYA and MGAM genes in HV crabs is suggestive of an enhanced capability for plant tissue digestion.

For CV crabs, lipids and proteins are the major nutritional components of food (Cannicci et al. 1996). Here, we detected positive selection of digestive enzyme genes involved with lipids and proteins might partly explain the molecular basis of CV crabs’ digestive adaptation mechanisms. PNLIP has a very important function in dietary lipids absorption by hydrolyzing triglycerides into diglycerides and subsequently into monoglycerides and free fatty acids (Mun et al. 2007). PNLIP only presented evidence of positive selection in CV crabs (Table 3), and most of positively selected sites in this gene were located primarily in functional domains that facilitated ligand-receptor interactions (Fig. 2). Particularly, codon 267 of PNLIP was in an active site (Fig. 2), which is hydrolyzed as the initial step in the activation process. Therefore, these positively selected amino acids in the key residues of PNLIP may have a positive impact on CV crabs’ lipid absorption ability. The detection of positive selection with PNLIP suggested that CV crabs may have acquired an enhanced capacity for lipid digestion. Three proteases (CPB, RISC, and XPD) were also only found to be under positive selection in CV crabs (Table 3), and they mainly play an important role in the intermediate step of protein digestion (Sakharov et al. 1997; Chen et al. 2001; Kumar et al. 2014). By contrast, proteases (CTRL and TRY) at the initiation stage of digestion (Perera et al. 2015) were not determined to be under positive selection (Table 3). These results suggested that CV crabs might have acquired an enhanced capacity for intermediate steps of protein digestion. In general, lipases and proteases subject to positive selection in CV crabs may be related with their complex diet and capabilities of digesting proteins and lipids.

Surprisingly, APN was determined to have undergone positive selection in both HV and CV crabs. APN (used as an indicator of protein digestive capacity) plays a major role in the final stages of dietary protein digestion in animals’ intestines, yielding various products, such as peptides which are finally digested by intracellular peptidases (Michiels et al. 2017; Tang et al. 2016). Thus, APN was consistent with evidence of positive selection in CV crabs and that these protease genes have important roles in enhancing digestion of protein. In comparison, positive selection identified in HV crabs seems to be more difficult to understand. In fact, herbivores also need a certain amount of plant protein from their food (Simpson et al. 2004). APN was also positively selected in HV crabs, suggesting that APN may play an important role in the digestion of plant protein. Further studies should be conducted regarding this interesting phenomenon to interpret roles of APN in HV crabs.

Conclusions

This study represents a preliminary survey of the molecular genetic basis underlying different feeding strategies in crabs. Significant positive selection genes were examined in HV crabs (carbohydrate digestive enzymes: AMYA and MGAM) and CV crabs (lipid digestive enzymes: PNLIP; protein digestive enzymes: CPB, RISC, and XPD). APN was detected with strong positive selection signals in both HV and CV crabs. Most of the putatively selected sites were localized in the important functional regions of these digestive enzyme genes. These results are consistent with the complex adaptations of crabs to digestion and absorption of diverse food resources.