Introduction

Color polymorphism is an interesting phenomenon which could be found commonly throughout the animal kingdom, with functions in camouflage (Barbato et al. 2007), thermoregulation (Heath 1975), mating selection (Houde and Endler 1990), social interactions (Rodgers et al. 2010), desiccation resistance (Parkash et al. 2009), salinity adaptation (Sokolova and Berger 2000), immunity (Scheil et al. 2013), and so on. As one of the most diverse phenotypical traits in nature and one of the driving forces of speciation, coloration, and color pattern formation attract many biologists involved in the study (Bagnara and Matsumoto 2007). And visible coloration polymorphisms provide a tractable system within which to examine the molecular basis of adaptation and evolution because of their often-simple patterns of inheritance and the general ease with which morph/allele frequencies can be estimated (Protas and Patel 2008).

In mollusks, shell color variations are known to exist in many species (Adamkewicz and Castagna 1988; Cain 1988; Kobayashi et al. 2004; Liu et al. 2009; Mitton 1977). These amazing shell color patterns aroused the questions on the mechanism underlying this intricate phenotype, e.g., what is the molecular basis of coloration patterns? How many genes are involved in orchestrating differences in shell coloration between individuals within a population? As one of the marine bivalve molluscan species, clam Meretrix meretrix exhibits highly variable shell color morphs based on the distribution of overlying pigments, which is a suitable species for investigating shell color variation. The methods used to study coloration include genetic crosses, gene expression analyses, complementation studies, and linkage mapping (Sivka et al. 2013; Protas and Patel 2008). To the bivalve mollusks, the existing studies for the shell coloration were few and mainly concerned the genetic basis analysis through selected crosses (Adamkewicz and Castagna 1988; Innes 1977; Winkler et al. 2001). And the genes and pathways underlying shell color formation of the marine bivalve are poorly understood. Considering shell color variation is likely to stem from differences in gene expression, gene expression analysis is a priority for the nonmodel species with rare genetic resources. Next-generation sequencing technologies permit rapid profiling the genes globally and functionally expressed, which has been applied to the coloration studies in the mammal, fish, and insect (Croucher et al. 2013; Fan et al. 2013; Henning et al. 2013; Nie et al. 2014; Wang et al. 2014).

In this study, we employed the powerful tool of high-throughput RNA sequencing and digital gene expression (DGE) analysis in identifying the genes potentially involved in shell color determination of the clam M. meretrix, by comparing the gene expression profiles in the mantle of the clams with different shell color morphs because the pigment is deposited by cells of the mantle. In the wild, the coloration patterns of M. meretrix consist of discrete (e.g., blotching), alternating (e.g., zigzag), radial (e.g., V-shaped), and so on. The shell color morph of one randomly picked clam individual may exhibit elaborate shell color-combined different patterns. In order to eliminate the influence of mixed shell color patterns, we spent 10 years developing three strains of M. meretrix with different and monotonous shell color morphs by artificial selection, and the expression analyses were mainly based on these strains. This study represents the first analysis of the relationship between whole-scale gene expression and shell color variations in the marine bivalve mollusks and would provide primary information for the future study about the molecular basis of the shell color pattern formation.

Materials and Methods

Clams and Sample Collection

Three clam strains separately exhibiting steady and hereditary shell color morph of black blotching (strain SB), purple zigzag (strain SP), or whole maroon (strain WM) were developed by 10-year successive generations selection. Briefly, clams with specific shell color morph (SB, SP, or WM) were collected from the progeny of the wild population. The morph was steady after more than four generations intra-morph artificial breeding during which the unrepresentative individuals were picked out. In this research, clams separately with four kinds of different and monotonous shell color morphs (M. meretrix) (illustrated in Fig. 1) were applied. Three kinds were collected from the three clam strains (strain SB, SP, and WM), respectively. And the other kind, which was nearly white shell color morph (W), was found and collected in the population who were cultured in the same estuarine area (Lingkun County, Wenzhou, Zhejiang Province). All collected clams (adult, 2-year-old) were acclimated in the water (18–20 °C, 25 ‰ salinity) under continuous aeration and fed with Isochrysis galbana daily for 2 weeks. Then for each of the four kinds, mantle from five clams were dissected and immediately ground into a powder mixture in liquid nitrogen and then reserved at −80 °C before processing for RNA extraction.

Fig. 1
figure 1

Clams (M. meretrix) separately with four kinds of shell color morphs. SP, clams with purple zigzag shell color; SB, clams with black blotching shell color; WM, clams with whole maroon shell color; W, clams with nearly whole white shell color

Full size image

RNA Extraction and Quality Control

Total RNA was extracted separately from the five mantle powder mixtures of WM, SB, SP, or W using the SV Total RNA Isolation System (Promega, USA) according to the manufacturer’s instructions. RNA degradation and contamination was monitored on 1 % agarose gels. RNA purity was checked using the NanoPhotometer spectrophotometer (Implen, GER). RNA concentration was measured using Qubit RNA Assay Kit in Qubit 2.0 Fluorometer (Life Technologies, CA, USA). RNA integrity was assessed using the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA, USA) and expressed as RNA Integrity Number (RIN). A total amount of 3 μg qualified RNA (RIN > 8) was used. RNA of four shell color morphs (WM, SB, SP, and W) was applied to library preparation for DGE sequencing, respectively, and a pooled RNA including SP, SB, WM, and W were applied to library preparation for de novo transcriptome sequencing.

Library Preparation, Sequencing, and Quality Control

Library was generated using Illumina TruSeq™ RNA Sample Preparation Kit (Illumina, USA) following manufacturer’s recommendations. Briefly, mRNA was purified from total RNA using poly-T oligo-attached magnetic beads. Fragmentation was carried out using divalent cations under elevated temperature in Illumina proprietary fragmentation buffer. First-strand cDNA was synthesized using random oligonucleotides and SuperScript II. Second-strand cDNA synthesis was subsequently performed using DNA Polymerase I and RNase H. Remaining overhangs were converted into blunt ends via exonuclease/polymerase activities and enzymes were removed. After adenylation of 3′ ends of DNA fragments, Illumina PE adapter oligonucleotides were ligated to prepare for hybridization. The library fragments were purified with AMPure XP system (Beckman Coulter, USA) to select cDNA fragments of preferentially 200 bp in length. DNA fragments with ligated adaptor molecules on both ends were selectively enriched using Illumina PCR Primer Cocktail in a 10-cycle PCR reaction. Products were purified using AMPure XP system (Agencourt, USA) and quantified using the Agilent high sensitivity DNA assay on the Agilent Bioanalyzer 2100 system.

Clustering was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS (Illumina, USA) according to the manufacturer’s instructions. After cluster generation, the libraries were sequenced on an Illumina Hiseq 2000 platform. And 100-bp single-end reads were generated for libraries prepared for DGE, 100-bp paired-end reads were generated for libraries prepared for de novo transcriptome. The left-end reads of de novo transcriptome generated were pooled into one big left.fastq file named Mantle_1, while the right-end reads were pooled into one big right.fastq file named Mantle_2.

Raw data were obtained by base calling (all raw tag data have been deposited in Short Read Archive (SRA) of the National Center for Biotechnology Information (NCBI)) and then were filtered through in-house perl scripts, by which clean data were obtained by removing reads containing adapter, reads containing poly-N, and low-quality reads from raw data. At the same time, Q20, Q30, GC-content, and sequence duplication level of the clean data were calculated. All the downstream analyses were based on clean data with high quality.

De novo Transcriptome Assembly and Gene Function Annotation

Transcriptome assembly was accomplished using Trinity (Grabherr et al. 2011), by which transcripts and unigenes (the longest transcript of a set of transcripts that appear to stem from the same transcription locus) were obtained. Gene function was annotated based on the following databases with a cutoff E value of 1.0 × 10−5: Nr (NCBI non-redundant protein sequences); Nt (NCBI non-redundant nucleotide sequences); Pfam (Protein family); COG (Clusters of Orthologous Groups of proteins); Swiss-Prot (A manually annotated and reviewed protein sequence database); KO (KEGG Orthology database); GO (Gene Ontology).

Sequence Mapping and Gene Expression Quantification

The assembled de novo transcriptome was used as the reference database, and gene expression levels were estimated for each sample. Briefly, clean data of DGE were mapped back to the reference transcriptome by Bowtie v0.12.9, and read count for each gene was obtained from the mapping results by RSEM (Li and Dewey 2011). And then RPKM (Reads Per Kilobase of exon model per Million mapped reads) of each gene, which is currently the most commonly used method for estimating gene expression levels, was calculated based on the length of the gene and reads count mapped to this gene (Mortazavi et al. 2008).

Differential Expression Analysis

For each sequenced library, the read counts were adjusted by edgeR program package through one scaling normalized factor. Differential expression analysis of two samples was performed using the DEGSeq R package (1.12.0). P values were adjusted using the Benjamini-Hochberg procedure (Mortazavi et al. 2008). Corrected P value of 0.005 and log2 (fold_change) of 1 were set as the threshold for significantly differential expression. Volcano plots were applied to intuitively show the differentially expressed genes. Hierarchical cluster analysis of differentially expressed genes union was performed to assess the transcriptional pattern variations among SP, SB, WM, and W using Cluster 3.0 (de Hoon et al. 2004). Venn charts were drawn using VennDiagram R package to exhibit shared or specific differentially expressed genes between different pairwise comparisons.

Gene ontology (GO) enrichment analysis of differentially expressed genes was implemented by the GOseq R package, in which gene length bias was corrected. GO terms with corrected P value less than 0.05 were considered significantly enriched by differential expressed genes (Young et al. 2010).

Quantitative Real-Time Validation

To validate our transcriptome sequencing results, six differentially expressed genes, mainly the ones in the enriched GO terms, were selected for quantitative real-time PCR (qRT-PCR) analysis. RNA was extracted separately from 30 samples (10 samples from each of WM, SB, and SP) and then cDNA was synthesized from RNA, which was applied as the template. Primers for qRT-PCR were designed using Premier Primer 5 and were listed in Supplemental Table S1. Each pair of primers could generate a single fragment which was proved to be identical to the sequence of corresponding gene by sequencing. β-Actin was employed as a reference standard to normalize the expression levels between samples (Mori et al. 2008; Wang et al. 2009). The amplification was carried out on Mastercycler ep realplex 4S (Eppendorf, GER) machine using QuantiFast SYBR Green PCR Kit (Qiagen, GER), with the following profile: 95 °C for 5 min, then 40 cycles of 95 °C for 10 s, 60 °C for 30 s. The measurements were run in four replicates. Relative gene expression data was analyzed using the 2-△△CT method, where CT is the threshold cycle (Livak and Schmittgen 2001). Data were examined for homogeneity of variances (F text), then analyzed by t test using software SPSS13.0. Difference was considered statistically significant if P < 0.05.

Results

De novo Transcriptome Sequencing, Assembly, and Gene Function Annotation

A total of 109,108,128 clean reads filtered from 113,282,886 raw reads (SRR1533149) (Table 1, Mantle_1 and Mantle_2) were generated through Illumina sequencing and assembled into 177,867 transcripts and 85,141 unigenes with a mean length of 1108 and 826 bp, respectively. These unigenes were annotated, among which 21,600 genes were annotated in NR, 1694 genes were annotated in NT, 7832 genes were annotated in KO, 16,007 genes were annotated in Swiss-Prot, 20,928 genes were annotated in PFAM, 20,928 genes were annotated in GO, and 11,102 genes were annotated in KOG. The functions of predicted unigenes were classified in GO, KOG, and KEGG. According to the GO annotation, mantle unigenes of the clam M. meretrix participated in 11 molecular functions (MF) (Fig. 2a). And these unigenes were involved in 26 COG classifications (Fig. 2b) and 31 KEGG pathways (Fig. 2c).

Table 1 Information and quality of RNA-seq
Full size table
Fig. 2
figure 2

Gene function annotation based on different databases. a Molecular function (MF) categories of the unigenes annotated in GO. b Classification of the unigenes annotated in COG. c Classification of the unigenes annotated in KEGG (A, cellular processes; B, environmental information processing; C, genetic information processing; D, metabolism; E, organismal systems)

Full size image

DGE Library Sequencing, Mapping, and Gene Expression Quantification

Four DGE libraries of the clam M. meretrix with different shell color morph were sequenced, and the raw data were deposited in SRA of the NCBI with accession numbers of SRR1533133 (WM), SRR1533145 (SB), SRR1533146 (SP), and SRR153347 (W). The total clean reads in each DGE library ranged from 1.21 to 1.48 G (Table 1, WM, SB, SP, W), and the percentage of the clean reads which could be mapped to the unigenes from the reference de novo transcriptome was 87.24 (SP), 86.66 (SB), 87.31 (WM), and 88.71 % (W), respectively. Read counts and RPKM of each mapped clean read was calculated, which was considered as the expression level of the gene. Density distribution of expression level in each DGE library based on log10 (RPKM) was exhibited in Fig. 3a, which showed that the distribution patterns in WM, SB, and SP were similar and were obviously different from that in W.

Fig. 3
figure 3

a Density distribution of expression level based on log10 (RPKM), exhibiting the overall difference in expression profile among four DGE libraries. b Volcano plots summarizing DEGs in each pairwise comparison according to the gene-specific t test. Each point represented a gene which is ranked by fold change and by t test. The red point was DEG, while the blue one was not)

Full size image

Gene Expression Variations Among Clams with Different Shell Color Morphs

The expression level of genes based on the read counts are compared among four DGE libraries in pairs and the differentially expressed genes (DEGs) were detected (corrected P value <0.005 & |log2 (fold_change)| > 1). The global expression profiles of DEGs union in each DGE library were estimated by hierarchical clustering (Fig. 4), which exhibited that the global expression pattern of DEGs in W was distinguishable from that in other three samples (SP, SB, and WM), and WM clustered more closely with SB rather than with SP. Consistent results were observed in volcano plots (Fig. 3b), which exhibited that there were more DEGs (red dots) between W and any of other three morphs, compared to those between SP, SB, and WM. The detailed numbers of DEGs unique or shared among multiple pairwise comparisons was summarized by Venn diagram (Fig. 5). There were 78, 54, 62, and 790 shared DEGs for the library of SP (Fig. 5a), SB (Fig. 5b), WM (Fig. 5c), and W (Fig. 5d), respectively, when comparing to other libraries. The shared DGEs were listed separately in Supplemental Table S2–S5 and were considered to be specifically expressed in SP, SB, WM, and W, respectively. To the DGE library of W, there were 1351, 1578, and 1408 DEGs detected in the comparison with SP, SB, and WM, respectively, while only 316, 228 and 254 DEGs were detected in the comparisons of WM vs SP, WM vs SB, and SB vs SP, respectively. The results of Venn diagram indicated that there was relatively less DEGs between WM, SB, and SP than those between W and any of other three strains (WM, SB, and SP), which was consistent with the results of clustering and volcano plots.

Fig. 4
figure 4

Hierarchical clustering of DEGs union among four DGEs libraries (SP, SB, WM, and W)

Full size image
Fig. 5
figure 5

Venn diagrams for number comparisons of DEGs among four DGE libraries (SP, SB, WM, and W). Shared or specific DEGs were detected in multiple pairwise comparisons, including SP compared with other three samples in pairs (a), SB compared with other three samples in pairs (b), WM compared with other three samples in pairs (c), and W compared with other three samples in pairs (d). SPvsW, the number of DEGs between SP and W, the same to SBvsW, WMvsW, SBvsSP, SBvsWM, and SPvsWM. The number of the shared DEGs are in the cross area, while the number of the specific DEGs are in the single area

Full size image

For DEGs in each pairwise comparison among four samples, GO enrichment analysis was performed to identify the main molecular function these DEGs exercise. Significantly enriched GO terms of MF (corrected P value less than 0.05) based on upregulated or downregulated DEGs were exhibited in Supplemental Table S6. With respect to the significantly enriched GO terms based on upregulated DEGs, “calcium ion binding” was shared across the comparisons of WM vs W, WM vs SP, WM vs SB, and SB vs W. Go term of “calcium ion binding” was also enriched from the upregulated DEGs in comparison of SP vs W, although it was not significant. DEGs involved in the GO term of “calcium ion binding” for each pairwise comparison were listed in Table 2.

Table 2 Upregulated DEGs involved in the shared enriched GO term (calcium ion binding) for each pairwise comparison
Full size table

qRT-PCR Validation

Illumina sequence data were validated by comparing the gene’s transcript level estimated from the RNA-Seq data with qRT-PCR results of six selected genes (Supplemental Table S7). The results showed that the expression patterns of Mitf, Notch, Myl, and Ryr agreed well between RNA-Seq and qRT-PCR. The expression patterns of other two genes Scp and Spink showed a similar trend in up- or downregulation exhibited by RNA-seq.

Discussion

Pigmentation has received much attention due to its strong implications for speciation and adaptation. Apart from a few model species, the molecular mechanism of this phenotypic diversification remains largely unknown, especially for the pigmentation in the shell of mollusks, which forms a variety of shell color patterns. In this study, we expect to investigate the genes involved in the shell color polymorphism and get information about the molecular mechanism of shell color determination by comparative transcriptome analyses among clams (M. meretrix) with different shell color morph.

Mollusk shell is a natural biomaterial secreted by the mantle (Addadi and Weiner 1985). Like another species of hard clam Mercenaria mercenaria, the shell color of M. meretrix roots in periostracum pigmentation, which originates in the mantle (Hillman 1961). To obtain an integrated view of the transcriptional events involved in the periostracum pigmentation, we used high-throughput sequencing technologies to analyze the transcripts in the mantle of M. meretrix. About 109,108,128 clean reads were obtained and assembled into 177,867 transcripts and 85,141 unigenes. This represents the most extensive transcriptional data set for the mantle of hard clams so far available. The functional classification of these transcripts according to the GO database showed that “binding” and “catalytic activity” were the dominant molecular functions involved (Fig. 2a), which was consistent with previous studies in other bivalves (Freer et al. 2014; Joubert et al. 2010; Shi et al. 2013). It is worth noting that a significant proportion of sequences in our mantle transcriptome were implicated in the function classification of “signal transduction” based on both COG and KEGG databases (Fig. 2b, c). Signal transduction occurs when an extracellular signaling molecule activates a specific receptor located on the cell surface or inside the cell and triggers a biochemical chain of events inside the cell in turn (Hlavacek et al. 2003). It is therefore hypothesized that there are complex interactions between cells in mantle, which is potentially involved in the formation of variant shell color patterns of M. meretrix.

In this study, digital gene expression (DGE) analysis was applied to identifying the genes potentially involved in the shell color determination of the clam M. meretrix, by comparing the gene expression profiles in the mantle among clams with four kinds of shell color morphs. All results including density distribution of expression level (Fig. 3a), volcano plots (Fig. 3b), hierarchical clustering (Fig. 4), and Venn charts (Fig. 5) indicated that W was distinguishable from the other three morphs (SP, SB, and WM). There were more DEGs detected in the comparison between W and each of other three morphs than those detected in the comparison between SP, SB, and WM. SP, SB, and WM were three clam strains obtained by our artificial selection in successive generations, which made them with lower genetic diversity than W. The increased number of DEGs would be derived from the shell color difference between W and any of the other three morphs, and from the other difference owning to genetic background difference, the latter potentially brought in false-positives. Similar phenomenon happened in other reports. A total of 2235 DEGs (479 genes upregulated and 1756 genes downregulated) were detected in black versus white sheep skin, and the sheep were randomly sampled (Fan et al. 2013). A total of 1161 DEGs were detected in q mutants versus wild-type silkworm integument (Nie et al. 2014). While much fewer DEGs (200–300) were detected in full-sib Midas cichlids with different body color (Henning et al. 2013). Combined with these reports, our data proposed that using the samples with lower genetic diversity was a solution to avoid false-positives for DEGs detecting. The application of the three clam strains we developed to the DGE analyses makes it possible for mining the real and effective DEGs responsible for the different shell color formation.

The shell color morph of a clam was influenced by the amount of overlying pigments, and it is visible that the amount of pigment in the four kinds of shell color morphs was very different from each other. It appears that WM has the largest amount of pigment, SB comes the second, SP has the third, and W has the minimal (nearly none). The results of hierarchical clustering (Fig. 4) conformed to this trend, which exhibited that the global expression profile of DEGs union in WM clustered more closely with SB rather than with SP, and was almost completely opposite to that in W. Hence, it supported that there was a relationship between gene expression and shell pigmentation in the clam M. meretrix, and the DEGs detected would give clues on the molecular mechanism of shell color determination.

There are so many DEGs detected, from which we need to pick out the genes of interest. We presented GO enrichment analysis to identify the main molecular functions the DEGs exercise and the genes involved, which would help us understand the potential cause of different shell color morph formation. It is of great interest that there was a shared GO term (calcium ion binding) enriched in the upregulated DEGs across all pairwise comparisons except SB vs SP (Supplemental Table S6). The upregulated DEGs for these comparisons represented the genes with higher expression level in the samples with a deeper shell color (colored degree: WM > SB > SP > W). The genes involved in the shared GO term were listed in Table 2, among which Notch-related genes were presented in all comparisons, and thus arguing for the vital role of Notch signaling in the pigmentation of different shell color.

The Notch signaling pathway is an essential cell-cell interaction mechanism, which regulates processes such as cell proliferation, cell fate decisions, differentiation, or stem cell maintenance (Artavanis-Tsakonas et al. 1999). In this pathway, receptors and ligands are single-pass transmembrane proteins with large extracellular domains that consist primarily of epidermal growth factor (EGF)-like repeats (Bray 2006). Although the function of Notch signaling pathway in development has been well studied and widely accepted (Artavanis-Tsakonas et al. 1999), nothing was known on this pathway in the pigmentary system until a few years ago, and the related reports were mainly focus on its role in the hair pigmentation of mammals. In the pigmentary system, members of the Notch signaling pathway are expressed in melanocytes (Moriyama et al. 2006) and seem to be upregulated in melanoma cell lines (Nickoloff et al. 2005). Transgenic and knockout mice were instrumental in analyzing the roles of Notch signaling pathway in the pigmentary system. Moriyama et al. (2006) reported that conditional deletion of RBP-Jκ encoding recombination signal binding protein Jκ (an important transcriptional regulatory factor in Notch signaling pathway) in melanocytes of mice would cause elimination of melanocyte stem cells and thereby resulted in hair graying, which demonstrated the crucial role of Notch signaling in the maintenance of melanocyte stem cells, by preventing apoptosis. Conditional deletion of Notch1 and Notch2 in mice could also result in obvious coat color dilution, and this dilution was dose-dependent since it was influenced by the number of intact Notch1 and Notch2 alleles. No hair graying was observed when only one Notch allele was absent in melanocytes. In contrast, dispersed gray hairs were discernible when two Notch alleles are floxed (Notch1 flox/+, Notch2 flox/+; Notch1 +/+, Notch2 flox/flox; Notch1 flox/flox, Notch2 +/+). The coat was nearly white in the absence of both Notch1 and Notch2 (Notch1 flox/flox and Notch2 flox/flox) (Schouwey et al. 2007). Data from these reports demonstrated that Notch pathway was essential for the pigment cell homeostasis and had a gene-dosage effect to induce a coat color dilution. Similar function of Notch pathway may be present in the shell pigmentation. Our data revealed that the expression of Notch was upregulated in the mantle of clam with the darker shell color. In other words, the lighter the shell color was, the less the Notch was expressed. It is therefore proposed that Notch pathway was involved in the shell pigmentation in a gene-dosage dependent pattern.

It was also different in shell color pattern among the four kinds of shell color morphs except for the colored degree. Spatial patterns are specified by cell interactions, so signaling pathways must be involved (French and Brakefield 2004). It has been reported that Notch pathway was involved in the color patterning. For mice, conditional deletion of Notch1 in the embryonic ectoderm resulted in a mosaic pattern of hair growth (Pan et al. 2004). More direct evidence for the involvement of Notch pathway in body color patterning comes from the studies in the butterfly. Notch-mediated lateral inhibition was implicated in the spatial organization of butterfly wing scales which served the function of color pattern formation (Reed 2004). Moreover, the concentric eyespot is an intriguing color pattern on the wing of butterfly (Beldade and Brakefield 2002), and eyespot pigment patterns are induced by a signal that originates from a group of focal cells at the center of the eyespot (Nijhout 1991). Reed and Serfas (2004) reported that Notch upregulation was an early event in the development of the eyespot pattern across multiple species of butterflies, and in a loss-of-eyespot mutant Notch expression was reduced at missing eyespots site. These reports support that Notch was probably involved in the shell color patterning, and it may be an upstream component of the shell color pattern determining process by defining the boundary of pigment occurs.

In this study, “signal transduction” was observed as the dominant function classification for the mantle expression data, and Notch signaling pathway may be just one component of a long-range signal for the shell pigmentation. It has been reported that Notch signaling pathway could be activated in a Ca2+-dependent manner. Source of Notch activation was an accumulation of calcium (Raya et al. 2004). Calcium was considered as an important second messenger. Both the receptor (NOTCH) and ligand (DELTA) of Notch pathway contain EGF repeats in tandem, of which some are capable of Ca2+ binding (Fehon et al. 1990). Receptor-ligand interaction can initiate Notch pathway and is extremely sensitive to Ca2+ concentration (Rand et al. 1997). Elevation of Ca2+ concentration can be due to Ca2+ release from internal stores, which are located in the sarcoplasmic reticulum (SR). Internal stores provide Ca2+ mainly by two types of release channels, the inositol 1,4,5-trisphosphate receptor (IP3R) and the ryanodine receptor (RYR). In this study, as listed in Table 2, the expression of Ryr was upregulated across all comparisons except SB vs W. The increase expression of Ryr may induce Ca2+ release, and thereby determine the shell color via activating Notch signaling. RYR-dependent Ca2+ enrichment had been reported to affect laterality through the activation of Notch signaling (Garic-Stankovic et al. 2008). To the comparison of SB vs W, although Ryr was not detected to be upregulated, the gene encoding 1-phosphatidylinositol-4,5-bisphosphate phosphodiesterase gamma-1 (PLCG1) was present in the list (Table 2). PLCG1 catalyzes the formation of inositol 1,4,5-trisphosphate, which stimulates another channel IP3R and therefore mediates calcium release (Dent et al. 1996). Therefore, calcium signaling process might equally be implicated in shell color formation via activation Notch pathway.

Although there is few information on the mechanism of shell color formation, some contributing factors for the body color formation in other animals has been proposed in the past few years, which affect diverse processes, such as development of pigment-producing cells, synthesis of pigment and its transfer from pigment-producing cells to surroundings (Schouwey and Beermann 2008). Notch combined with calcium signaling, as an upstream component of the shell color-determining process, mainly affects the development of pigment-producing cells, such as maintenance of melanocyte stem cells. Our transcriptome data also revealed the DEGs potentially involved in other two processes, the synthesis of pigment and pigment transfer from pigment-producing cells to surroundings. As listed in Table 2, the expression of Myl encoding the protein of myosin regulatory light chain was upregulated across all comparisons except SB vs W. As mentioned above, the increase expression of Ryr may induce Ca2+ enrichment. Elevated Ca2+ encourages the formation of Ca2+-calmodulin complex to activate the myosin light chain kinase to phosphorylate serine 19 of myosin light chain, which in turn removes inhibition of the myosin ATPase. This event is followed by ATP hydrolysis and sliding of myosin on actin filaments to generate cytoskeleton actin-myosin contraction (Somlyo and Somlyo 1994), which is necessary for the trafficking of the pigment organelles where pigment is made (Chang et al. 2012). The distribution of pigment is achieved via pigment organelles transport along cytoskeleton by the mean of contraction from the perinuclear region to dendrites and the subsequent transfer to neighboring environment (e.g., keratinocytes) (Boissy 2003). Myl functions in the pigment organelles (e.g., melanosome) translocation, and the significantly different expression of Myl observed among clams with shell color variations in this research implied that Myl exerted an influence on the pigment distribution which determined the shell color pattern. There were also some DEGs related to the process of the synthesis of pigment detected in our data, such as the gene encodes microphthalmia-associated transcription factor (MITF) (Levy et al. 2006). This transcription factor has been reported to play a pivotal role in melanocyte differentiation through the direct transcriptional control of Tyr, Tyrp1, and Dct genes, encoding three enzymes involved in pigment synthesis. In this study, the expression of Mitf in the clam strain WM was significantly higher than that in SB, SP, and W (Supplemental Table S7), which implied that Mitf was also involved in the shell color variations.

In conclusion, this study represents the first analysis of the relationship between whole-scale gene expression and shell color variations in the marine bivalve mollusks. Clam strains with different and monotonous shell color morph developed by artificial selection in successive generations were applied to comparative transcriptome analyses, which minimized the false-positives. The results of the comparative transcriptome analyses supported the idea that there was a relationship between gene expression and shell coloration in the clam M. meretrix. The possible involvement of Notch combined with calcium signaling pathway in shell color determination is a novel finding in clams. Other DEGs potentially implicated in shell color pigmentation were also discussed. This study sheds light on the color formation mechanism of the shell.