A crayfish molar tooth protein with putative mineralized exoskeletal chitinous matrix properties

Biomineralization, the biogenic formation of mineral deposits, plays a central role in the formation and hardening of exoskeletons and occurs in almost all eukaryal phyla, although the morphology and extent of biomineralization vary widely across species (Simkiss, 1975). Invertebrates deposit various mineral compositions, such as iron oxide (Lowenstam, 1967; Weaver et al., 2010), carbonate apatite (Currey et al., 1982; Lowenstam and Weiner, 1985, 1992), silica (van der Wal, 1989; Miller et al., 1990; Michels et al., 2012), protodolomite (Ma et al., 2008), amorphous calcium phosphate (ACP; Greenaway, 1985; Vijayan and Diwan, 1996; Levi-Kalisman et al., 2002; Luquet and Marin, 2004; Neues et al., 2007; Kunkel and Jercinovic, 2013; Huber et al., 2014) and crystalline hydroxyapatite (Weaver et al., 2012), to increase the hardness of exoskeletal working surfaces. In vertebrates, enamel serves this purpose (Lucas, 2004). The mandibles of the arthropod Cherax quadricarinatus (Bentov et al., 2012) contain an unusual crystalline enamel-like apatite layer which resembles vertebrate enamel in both composition and function (Bentov et al., 2012). The appearance of an enamel-like apatite layer in a crayfish molar and the similarities found in its mechanical properties to mammalian teeth suggest a possible case of convergent evolution as previously described in other species (Lowenstam and Weiner, 1992; Bentov et al., 2012; Weaver et al., 2012). The apatite layer is presented in the form of fluorapatite (FAP) and is mounted on chitin reinforced with amorphous calcium carbonate (ACC) (Bentov et al., 2012). The mineral composition of the molar continuously transforms from the FAP to ACP and from ACP to ACC (Bentov et al., 2012).

Unlike vertebrate teeth, the complex functional tooth structure of the crayfish develops quickly, and is shed and regenerated with each molt cycle (Shechter et al., 2008a; Tynyakov et al., 2015). The molt cycle in crustaceans can be divided into four major stages, whereas a new exoskeleton is assembled during pre-molt and post-molt and sclerotized and calcified during post-molt (Shechter et al., 2008b). During pre-molt, the new epicuticle and exocuticle are built beneath the old cuticle (Roer and Dillaman, 1984), but these layers do not calcify until post-molt (Travis, 1963; Travis and Friberg, 1963). Thus, the mineral deposition of epicuticle and exocuticle begins at post-molt together with endocuticle synthesis (Drach, 1939; Travis, 1963,, 1965; Travis and Friberg, 1963). Unlike the exoskeleton, the crayfish molar is assembled and mineralized during pre-molt (Tynyakov et al., 2015). Similarly to crustacean cuticle, the process of crayfish molar formation is entirely extracellular and presumably controlled by proteins, lipids, polysaccharides, low-molecular weight molecules and calcium salts (Willis, 1987; Lowenstam and Weiner, 1989; Horst and Freeman, 1993; Luquet, 2012).

As in vertebrates (Fujisawa and Tamura, 2012), skeletal proteins also play vital roles in mineral formation in invertebrates (Glazer and Sagi, 2012; Luquet, 2012). Previous in vitro experiments identified crustacean proteins thought to be associated with calcium carbonate calcification (Nagasawa, 2011). In contrast to vertebrates, the mineral in crustaceans is precipitated on chitin, which forms their exoskeletal structural components. As such, many exoskeletal proteins contain chitin-binding domains (Tabor, 1986), such as chitin-binding domain 2 (Tellam et al., 1999) and the Rebers–Riddiford (R–R) consensus sequence (Karouzou et al., 2007), and/or interact with calcium carbonate. Three known forms of the R–R consensus sequence, RR-1, RR-2 and RR-3, have all been recognized in proteins (Andersen, 2000) and shown experimentally to bind chitin (Rebers and Riddiford, 1988). However, there remains a lack of knowledge about the interactions of R–R sequence-containing proteins with chitin and minerals and their localizations in the exoskeleton.

In vertebrates, different proteins have been suggested to play important roles in the mineralization of bones and teeth (George and Veis, 2008). In vivo studies suggested that some of these proteins may influence enamel plasticity (deformation behavior) and morphology, as well as the mineral content of teeth (Robinson et al., 2004; Gruenbaum-Cohen et al., 2009). Accordingly, mineralization experiments revealed that recombinant versions of these proteins contribute to the regulation of the structure and configuration of mineral particles in vitro (Deshpande et al., 2010; Fang et al., 2011).

The aim of this study was to identify and characterize a novel protein from the crustacean molar by means of the transcript tissue expression pattern, in vivo silencing of the transcript, mineral precipitation and chitin-binding assays with a recombinant protein. While in vivo silencing of the transcript was lethal, a recombinant version of the protein was found to bind chitin and interact in calcium carbonate and hydroxyapatite precipitation kinetics in a concentration-dependent manner in vitro.

Cherax quadricarinatus (Von Martens 1868) males were grown in artificial ponds at Ben-Gurion University of the Negev, Beer-Sheva, Israel, under conditions described previously (Shechter et al., 2008a). Intermolt crayfish were held in individual cages and endocrinologically induced to enter pre-molt through removal of the X-organ–sinus gland (XO–SG) complex in the eyestalk or injection of ecdysone into the hemolymph (Shechter et al., 2005,, 2007). Growth of the gastrolith (the calcium storage organ) during the molt cycle was non-invasively evaluated using X-ray in terms of the molt mineralization index (MMI: gastrolith width/crayfish carapace length), which is used as an accurate molt stage marker (Shechter et al., 2007). For all dissection procedures, crayfish were placed on ice for 5–10 min until they were anesthetized.

Inter-molt stage animals were endocrinologically induced as above and, 14 days later, their molar teeth were dissected at late pre-molt. Molars were pulverized in liquid nitrogen using a mortar and pestle and proteins extracted as previously described (O'Brien et al., 1991; Tynyakov et al., 2015). In brief, protein extract was obtained using 5 mol l⁻¹ guanidine thiocyanate, 0.5 mol l⁻¹ EGTA, 5 mmol l⁻¹ Pipes, pH 7 (1:10, w:v), at room temperature for 5 h. The extract was centrifuged at 16,000 g for 5 min in an Eppendorf microfuge, and the supernatant was dialyzed against 10 mmol l⁻¹ ammonium bicarbonate, pH 7.8, overnight at 4°C and concentrated using a Vivaspin 20 (MWCO 10,000; Vivascience).

The molar protein extract was separated in two dimensions as described previously (Arnott et al., 1998) with a few modifications. Briefly, for isoelectric focusing, an immobiline dry strip (11 cm, pH 3–10) was rehydrated and aligned on an isoelectric focusing tray. The molar protein extract was loaded adjacent to the anode, and voltage was applied for a total of 11.9 kV h. Following isoelectric focusing, the gel strip was equilibrated in a buffer containing 6 mol l⁻¹ urea, 0.375 mol l⁻¹ Tris-HCl, pH 8.8, 2% SDS and 20% glycerol, and incubated in 2% dithiothreitol for reduction following 8% iodoacetamide for alkylation. Then the strip was mounted on a 15% SDS-PAGE gel with the Tris/glycine running buffer system as described previously (Lucas, 2004). Spots were visualized with Coomassie Blue staining and the corresponding gel pieces were excised for mass spectrometry.

After undergoing two-dimensional gel separation and visualization, proteins from molar extracts were trypsin (Gold Trypsin, Promega) digested, and the peptides were purified on a C18 column and dissolved in 0.1% formic acid. Nanoliquid chromatography and mass-spectrometry analysis were performed as described previously (Glazer et al., 2010) using a 75 μm internal diameter fused silica column, packed with C18 (NewObjective, Woburn, MA, USA) connected to an Eksigent nano-LC system (Eksigent, Dublin, CA, USA). Mass spectra were acquired using an LTQ-Orbitrap XL (Thermo Fisher Scientific, San Jose, CA, USA) and analyzed using the PEAKS Mass Spectrometry Software, including a de novo peptide sequencing approach. Further Cq-M15 protein validation was performed by the Sequest and Mascot algorithms operated under Proteome Discoverer 2.0 (Thermo Fisher Scientific) using an NCBI database containing the newly identified M15 sequence.

Degenerative primers Cq-M15 dF (5′-CAY TTY ACN GCN YTI WSI GC-3′) and Cq-M15 dR (5′-CGN CGN ATR GGY GGYGGY GG-3′) were generated based on the sequence of a partial fragment of Cq-M15 obtained by tandem mass spectrometry. A fragment of the cDNA was amplified using Cq-M15 dF as a forward primer and Cq-M15 dR as a reverse primer. The PCR conditions using the REDTaq ReadyMix™ PCR Reaction Mix (Sigma) were: 94°C for 3 min, followed by 30 cycles of 94°C for 30 s, 55°C for 30 s and 72°C for 3 min, followed by a final elongation step of 72°C for 10 min. The fragment was cloned into the pGEM-T Easy vector (Promega, Madison, WI, USA) and sequenced. 5′- and 3′-RACE (rapid amplification of cDNA ends) was carried out with the SMART rapid amplification of cDNA ends kit (Clontech), according to the manufacturer's protocol to obtain the entire sequence of Cq-M15.

Cq-M15 sequence data have been deposited in the GenBank™ Data Bank with accession number KF739426.

The deduced amino acid sequence of the gene was obtained bioinformatically by the ExPASy Translate tool (Artimo et al., 2012). Signal peptide and phosphorylation sites were predicted using SignalP 4.0 (Petersen et al., 2011) and NetPhos 2.0 (He et al., 2005), respectively. Identification of putative domains in the deduced protein sequence was predicted by SMART (Schultz et al., 1998; Suderman et al., 2006) and Motif Scan (Suzuki et al., 2013) and its type by a CuticleDB server (Karouzou et al., 2007). Physico-chemical parameters of the Cq-M15 protein sequence, including molecular weight and amino acid composition, were predicted using the ProtParam tool (Fujisawa and Tamura, 2012). Secondary structure predictions were done by PSIPRED (Futahashi and Fujiwara, 2008).

A search for homologous sequences to Cq-M15 (www.ncbi.nlm.nih.gov) was performed using BLAST, and phylogenetic analysis among arthropod sequences was performed by the neighbor-joining method (Saitou and Nei, 1987) using MEGA5 software (Andersen, 2000). Evolutionary distances were computed using the Poisson correction method (Lu et al., 2007) and are expressed as amino acid substitutions per site. All positions containing gaps and missing data were eliminated from the analysis.

To identify tissue-specific temporal and spatial expression of Cq-M15, RNA was extracted from the forming epithelia of cuticle, mandible, molar and basal segment tissues, as well as from the hepatopancreas, muscle and testis, using an EZ-RNA Total RNA Isolation Kit (Biological Industries, Beit Haemek, Israel), according to the manufacturer's protocol. First strand cDNA was generated with oligo (dT) 22 VN using expand RT (Roche Diagnostics). Cq-M15 was amplified by means of PCR with REDTaq ReadyMix™ PCR Reaction Mix (Sigma; one cycle at 94°C for 1 min; 30 cycles at 94°C for 1 min, 55°C for 2 min, 72°C for 3 min; one cycle at 72°C for 10 min) with the Cq-M15F (5′-CAACATGAAGTCTATTGTAATGTTGGC-3′) and Cq-M15R (5′-TGTTTCAAATAAAGTATATTGAACC-3′) primers. 18S rRNA was evaluated as the control as described previously (Glazer et al., 2010).

Reference C. quadricarinatus transcriptome was constructed from sequencing of four sets of samples as described previously (Tynyakov et al., 2015): (1) RNA extracted from hypodermis-, gastrolith- and molar-forming epithelium sequenced by 100 bp paired-end sequencing (Illumina HiSeq2000, USA); (2) RNA extracted from molar- and gastrolith-forming epithelium from different molt stages (intermolt, early pre-molt, late pre-molt and post-molt), barcoded and sequenced by 50 bp single-end sequencing (Illumina HiSeq2000, USA); (3) RNA sequenced by 454 technology, as described by Glazer et al. (2013); and (4) RNA sequenced ‘single pass’ by Sanger technology (expressed sequence tags, ESTs) as described by Yudkovski et al. (2010). All sequences (also called ‘reads’) were de novo assembled with CLC Genomics Workbench 6.51 (CLC Bio, Aarhus, Denmark) using default parameters. The resulting contigs were submitted to a second assembly by CAP3, yielding a collection of putative transcripts regarded as the ‘reference C. quadricarinatus transcriptome’. The 50 bp sequences from molar and gastrolith samples in different molting stages described above were aligned to the reference C. quadricarinatus transcriptome using STAR 2.3. For estimation of transcript expression levels, the number of sequences aligned to each transcript in each stage was counted, and further normalized to the total number of sequences in that stage. A statistical test for assessing the differences in expression among stages was carried out using DESeq (Anders and Huber, 2010). The Cq-M15 sequence was searched against the reference transcriptome using BLAST, and the transcript with the highest score and lowest e-value was regarded as the Cq-M15 transcript in different molt stages: inter-molt (pool of animals, N=1), early pre-molt (pool of animals, N=1), late pre-molt (two single animals and one pool, N=3) and post-molt (two single animals, N=2).

Cq-M15 and CqVg (C. quadricarinatus vitellogenin) dsRNAs were synthesized as described previously (Rosen et al., 2010). For a preparatory short-term silencing experiment, inter-molt males (MMI=0; 5–10 g) were injected with 5 µg g⁻¹ of Cq-M15 dsRNA (N=5) or CqVg dsRNA (N=5). Four consecutive injections were given once a day. On the fifth day, the animals were dissected to isolate the cuticle- and molar-forming epithelia and testis. For long-term silencing, the experimental group (N=10) was injected daily with Cq-M15 dsRNA, while the control group (N=10) was injected with CqVg dsRNA. In both experiments, the daily injections included 1 ng µl⁻¹ ecdysone (Sigma) (Glazer et al., 2010) to induce molting, applied into the sinus of the first abdominal segment.

RNA was extracted as described above from the testis and the forming epithelium of the molar and cuticle at the following molt stages: mid- and late pre-molt (N₁, N₂=9), early and late post-molt (N₃, N₄=4) and inter-molt (N₅=4). The same procedure was performed with experimental and control males (N₆, N₇=5) of the short-term preliminary in vivo Cq-M15-silencing experiment. First strand cDNA was generated by reverse transcription using random hexamer primers (VersoTMcDNA kit, Thermo Fisher Scientific). Relative quantification was performed with the following primers: Cq-M15 F (5′ATGTTGGCTCTGCTTGCTG-3′), Cq-M15 R (5′-TGGTGCTCCATAGGTTGTGA-3′) and Universal Probe Library Probe no. 83 (Roche). Relative quantification of M15 transcript levels was obtained using the following primers: Cq-M15qPCR_F (5′ATGAATTACATCATCTGTGAGGAACC-3′), Cq-M15qPCR_R (5′CAATGAGCACATAACAGCCCA-3′) and Probe TM FAM-CTTGATTTTTACTGCCGTTGACACACTGA-BBQ, especially designed by TIB MOLBIOL Synthese labor (Roche). Cq-18S (accession no. AF235966), used as the normalizing agent, was evaluated in all the experiments by real-time RT-PCR in similar conditions (Glazer et al., 2010).

Because of the negligible amounts of Cq-M15 and difficulty of extraction from tissue samples, in order to elucidate Cq-M15 functions in vitro, recombinant Cq-M15 (rCq-M15) was produced and purified, and verified through western blot by GenScript (GenScript Inc., Piscataway, NJ, USA). In brief, rCq-M15 was chemically synthesized, cloned into pUC57-Kan vector and expressed in E. coli BL21 (DE3) cells at 37°C with 200 rpm agitation. SDS-PAGE and western blot were used to detect protein expression.

A chitin-binding assay was performed as described previously (Inoue et al., 2001). Briefly, a solution of rCq-M15 was incubated with chitin powder and further successively washed starting with DDW and then with 0.2 mol l⁻¹ NaCl and finally boiled for 10 min in 2% SDS, 20% (v/v) β-mercaptoethanol. The supernatant of each wash was electrophoresed on 15% SDS-PAGE. The bands were visualized by silver staining. Bovine serum albumin (BSA) served as a negative control.

Circular dichroism (CD) was conducted with a J750 spectropolarimeter (Jasco, Easton, MD, USA) equipped with a Peltier device. Spectra in the absence and presence of calcium were measured at pH 7.4 using a 0.1 cm optical path Suprasil quartz cuvette (Hellma, Müllheim, Germany) in 400 µl of 0.028 mmol l⁻¹ rCq-M15 prepared in 0.1 mmol l⁻¹ Hepes. Then, 12 µl calcium chloride solution (1 mol l⁻¹) was added such that a molar ratio of rCq-M15 to Ca²⁺ ions of 1:1071 was obtained. Spectra were collected at a wavelength range of 190–260 nm at ambient temperature with a bandwidth of 1 nm, scan speed of 10 nm⁻¹ min⁻¹ and time constant of 4 s. Each spectrum represents an average of five successive scans over the 190–260 nm wavelength range.

Different concentrations of rCq-M15 were tested in a calcium carbonate precipitation kinetics assay as described previously (Inoue et al., 2001), with slight modifications. In brief, 100 µl of 22 mmol l⁻¹ NaHCO₃, pH 8.7, was added to 20 µl rCq-M15 in DDW at different concentrations (57, 230 and 459 µmol l⁻¹), to which 100 µl of 22 mmol l⁻¹ CaCl₂ was added. Addition of DDW instead of rCq-M15 served as a negative control (blank). Turbidity was measured at 570 nm with a spectrophotometer.

Apatite precipitation experiments were performed as described previously (Beniash et al., 2011), with slight modifications. First, rCq-M15 was dialyzed against DDW and concentrated to 3 mg ml⁻¹ using a SpeedVac apparatus. Then, 100 mmol l⁻¹ stock solution of CaCl₂ was added to 20 µl of different concentrations (308, 615, 923 and 1850 µmol l⁻¹) of the dialyzed rCq-M15 in DDW, and 100 mmol l⁻¹ (NH₄)₂HPO₄ was added to yield final concentrations of 2.5 mmol l⁻¹ Ca²⁺ and 1.5 mmol l⁻¹ phosphate. Droplets (20 µl) of the solution were placed on quartz slides in a humidity chamber for incubation at 37°C overnight at 100% humidity. In addition, 20 µl droplets of rCq-M15 at 923 µmol l⁻¹ were placed on quartz slides in a humidity chamber for incubation at ambient temperature for 24 h and 2 weeks. After incubation, the quartz slides were dipped in DDW, air-dried and subjected to Raman spectroscopy. Raman spectra were acquired on a Jobin–Yvon LabRam HR 800 micro-Raman system equipped with a liquid nitrogen-cooled detector as described previously (Bentov et al., 2012).

Qualitative evaluation of the abundance of apatite and its precursor octacalcium phosphate (OCP) was performed as described previously (Crane et al., 2006). Calcium phosphate precipitation was performed in solution by mixing 1 mol l⁻¹ calcium chloride (CaCl₂) and 0.5 mol l⁻¹ sodium phosphate (Na₂HPO₄, NaH₂PO) (the two salts were used in different ratios to adjust the pH to 7.4) in 1.5 ml reaction volume, to a final concentration of 20 and 10 mmol l⁻¹, respectively. The precipitation assay was performed in the presence of 150 nmol l⁻¹ or 4.1 µmol l⁻¹ BSA, and 620 nmol l⁻¹ or 3.1 µmol l⁻¹ rCq-M15. All the reactions were performed in duplicate and were compared with the control (without the addition of protein). Each tube was loaded first with CaCl₂, H₂O (adjust to the volume of 1.5 ml) and the tested protein in the final desired concentration. Then, the Na₂HPO₄/NaH₂PO₄ solution was added to start the precipitation. After 24 h of incubation, the samples were centrifuged at 10,000 rpm and the pellet was separated and washed with DDW and centrifuged again. The pellet was than smeared on Opal slides (for low background) and allowed to dry completely. Samples were analyzed using Raman spectroscopy and qualitative evaluation of the abundance of apatite and its precursor (OCP) was performed by subtracting the intensity of the 1010 cm⁻¹ peak from the main PO₄ ʋ1 vibration of 960 cm⁻¹. Micro-Raman analyses of specimens were performed on a multichannel bench Renishaw InVia Reflex spectrometer (Renishaw InVia Raman, UK) coupled with a Peltier-cooled CCD detector. Excitation was provided by the 785 nm line of a diode laser. The samples were scanned from 200 to 1200 cm⁻¹ wavenumber shift at a spectral resolution of 2 cm⁻¹. The scattered light was analyzed by a spectrograph with holographic grating (600 g mm⁻¹), slit width 150 mm and opened confocal hole (1000 mm). The time of acquisition of a particular spectral window was 10 s.

Data are expressed as means±s.e. Non-parametric tests were used as follows: statistical analysis for relative transcript levels among different tissues and molt stages, differences in survival rates after silencing with Cq-M15 dsRNA and changes in turbidity of calcium carbonate precipitation with rCq-M15 final concentrations, as well as for qualitative evaluation of the apatite abundance, was performed using the Kruskal–Wallis rank sum test, followed by multiple pair-wise comparisons using the Wilcoxon rank sum test (with Bonferroni correction) for a two-sided test of significance. P<0.05 was considered statistically significant.

Initially, efforts were directed at defining the protein content of C. quadricarinatus molars. Separation of EGTA-extracted molar proteins by two-dimensional PAGE revealed the presence of at least 12 distinct spots. These spots were excised from the gel and subjected to mass spectrometry for peptide de novo sequencing. Degenerate primers corresponding to the obtained peptide sequences were constructed. Then, cDNA fragments were amplified and eventually full transcripts were obtained. While screening the protein profile of the molar tooth, a protein termed Cq-M15 was identified as a spot of ∼17 kDa and pI of ∼5.4 (Fig. S1, circled spot, modified) (Tynyakov et al., 2015). Its deduced 740 bp transcript includes a 5′ UTR of 18 bp, an ORF of 465 bp encoding a deduced 155 amino acid protein, and a 3′ UTR of 257 bp (Fig. 1). The first 15 amino acids are predicted to be a signal peptide. A chitin-binding R–R domain is suggested for residues 53–113. This domain is assigned to the RR-1 family, members of which have been found in soft and flexible cuticles (Kim et al., 2005). The predicted pI of Cq-M15 is 5.2, and approximately 5.8% of the total amino acids of the protein correspond to possible phosphorylation sites. Amino acid composition analysis of Cq-M15, moreover, revealed an abundance of non-polar, aliphatic amino acids, as well as polar but uncharged amino acids (Table 1).

Following the assembly of its sequence, the similarity of the putative Cq-M15 protein to other proteins in protein databases was assessed, revealing Cq-M15 to be similar to cuticular proteins found in crustaceans and insects (Table 2). Evolutionarily, the above proteins were divided into two clades (Fig. 2). Clade i includes Cq-M15 and cuticular proteins from the mineralized parts of crustacean cuticles and clade ii encompasses cuticular proteins expressed in the crustacean arthrodial cuticles (non-calcified and flexible cuticles) and cuticular proteins expressed in insects. From an evolutionary perspective, Cq-M15 is thus much closer to crustacean proteins from mineralized cuticles than to arthrodial and insect cuticles.

As suggested by its similarity to exoskeletal proteins from mineralized crustacean cuticles, and as the molar tooth is being built during pre-molt (Tynyakov et al., 2015), the specific expression of Cq-M15 was tested in a number of target tissues at different molting stages. RT-PCR revealed that Cq-M15 is transcribed at the mid- and late pre-molt stages in exoskeletal elements forming tissues, such as the cuticle, molar, basal segment, maxillae and gastrolith, as well as in non-exoskeletal tissue, such as the muscle, as compared with the transcription seen during inter-molt and post-molt (Fig. 3A). In most non-exoskeletal tissues, including testes, no transcription was detected. RNA sequencing of hypodermis-, gastrolith- and molar-forming tissue by Illumina (560 million reads), and additional 454 and Sanger sequencing (287,900 reads) was followed by de novo assembly, yielding 56,490 contigs as described previously (Tynyakov et al., 2015). Based on the above database, a differential expression analysis of the Cq-M15 transcript showed a significant increase (P<0.05) in normalized read count of early and late pre-molt compared with the post- and inter-molt stages at the molar-forming epithelium samples as described in Fig. 3B. This was confirmed by real-time RT-PCR, which was performed next to evaluate differences in the relative transcription levels of Cq-M15 between cuticle- and molar-forming tissues at different molt stages, with the level of transcript in testicular tissue serving as a control (Fig. 3C). The results show how Cq-M15 transcription significantly differs as a function of tissue and molt stage (Kruskal–Wallis χ²=79.3018, d.f.=14, P<0.001). Multiple pair-wise comparisons using the Wilcoxon rank sum test revealed that at mid-pre-molt, Cq-M15 transcription in the molar-forming tissue was significantly higher than in cuticle-forming tissue (P<0.05), while such transcription in testis was significantly lower than in the latter (P<0.001). In late pre-molt, testicular tissue transcription of Cq-M15 was significantly lower than in cuticle- and molar-forming tissues (P<0.001), while there was no significant difference between the latter two (W=60, P=0.09). Overall, Cq-M15 transcription in cuticle- and molar-forming tissues at the late pre-molt stage was highest, relative to other molt stages (P<0.05). In contrast, Cq-M15 was almost undetectable in the late post-molt and inter-molt stages (P<0.05).

As the expression of Cq-M15 was significantly increased in relevant molt stages and locations related to exoskeleton and teeth assembly, a study of Cq-M15 function in vivo through an RNAi loss of function approach was attempted by Cq-M15 dsRNA administration. First, the Cq-M15 dsRNA sequence was compared with all known transcripts listed in C. quadricarinatus databases, in order to rule out non-specific silencing of non-Cq-M15 transcripts in the crayfish (Yudkovski et al., 2007,, 2010); no similar sequences were detected. The effectiveness of the RNAi-mediated gene knock-down was evaluated by qPCR (Fig. 4A). Examination of the relative quantification values obtained revealed significant differences between the experimental and the control groups (Kruskal–Wallis χ²=24.6676, d.f.=5, P<0.001). When multiple pair-wise comparisons were conducted using the Wilcoxon rank sum test, a significantly lower level of Cq-M15 transcription of the experimental group, compared with the control, was noted in the cuticle- and molar-forming tissues, as well as in testis (P<0.05).

Once the evaluation of the silencing effect of Cq-M15 dsRNA injection had been established, a long-term silencing experiment addressing Cq-M15 functionality was performed. This approach revealed a high mortality rate in the Cq-M15 dsRNA-injected group (Fig. 4B). As such, the experiment was terminated after the death of the first seven experimental subjects, with no molt cycle differences noted between the groups. As Cq-M15 was found to be transcribed at the pre-molt stage, the molt cycle was induced to initiate Cq-M15 transcription. As the old exoskeleton does not undergo any changes during the newly induced molt cycle, the effects of silencing Cq-M15 would be visible only at the newly formed exoskeleton. None of the experimental group crayfish built a new exoskeleton, including the molar. Thus, we could not perform any microscopic or spectral analysis of newly constructed molar morphology and composition. The rate of survival of the experimental group was 30% versus 100% for the control group (Fig. 4B), reflecting a significant difference starting from as early as the third day of the experiment (Wilcoxon rank sum test, W=2.5, P<0.05).

rCq-M15 and BSA (which served as a control protein, lacking a chitin-binding domain; Shimizu et al., 1989) were separately incubated with chitin powder in in vitro studies to assess whether rCq-M15 binds to chitin. Unbound rCq-M15 and BSA were recovered in the supernatant after consecutive washes with DDW (Fig. 5, lanes 4 and 1 , respectively) and were also detected in the supernatant following further washing with weak salt solution (Fig. 5, lanes 5 and 2, respectively). After boiling the washed chitin in 2% SDS with 20% 2-mercaptoethanol, only rCq-M15 was recovered (Fig. 5, lane 6), indicating the strong binding of rCq-M15 to chitin, in accordance with the prediction of the presence of a chitin-binding domain.

Evolutionary and sequence evidence point to the potential involvement of Cq-M15 in biomineralization, specifically in the precipitation of Ca²⁺ ions and its carbonate and phosphate salts. Thus, the interaction of rCq-M15 with Ca²⁺ ions and its carbonate and phosphate salts found in the mandible and molar tooth were next tested in vitro. Initially, rCq-M15 was assessed by CD. According to bioinformatics predictions, the first 96 residues of Cq-M15 contain a mixture of well-spaced loops and short β-sheets, whereas the region spanning residues 120–140 is predicted to fold as an α-helix. The CD curve obtained in calcium-free solution is characterized by minima at 213 and 216 nm (Fig. 6). In the presence of calcium, rCq-M15 undergoes a conformational transition, reflected by the appearance of prominent absorbance shoulders at 209, 215 and 218 nm.

Finally, to investigate the role of Cq-M15 in the biomineralization processes, in vitro calcium carbonate and phosphate precipitation assays were conducted with rCq-M15. No significant differences between calcium carbonate precipitation rates were noted in the absence of rCq-M15 (blank, control) and in its presence at concentrations of 57 and 230 µmol l⁻¹. However, rCq-M15 at a concentration of 459 µmol l⁻¹ significantly promoted calcium carbonate precipitation. At the end of the experimental period, the amount of deposited mineral (indicated by the absorbance) was shown to be concentration dependent with the blank showing the lowest value and 459 µmol l⁻¹ rCq-M15 showing the highest (Wilcoxon rank sum test, P<0.05) (Fig. 7).

With respect to calcium phosphate precipitation, rCq-M15 showed a concentration-dependent effect on hydroxyapatite precipitation at 37°C, compared with the blank (in the absence of rCq-M15), in which brushite (a hydroxyapatite precursor) (Park et al., 2002) was precipitated exclusively (Fig. 8A). At the initial concentration of 308 µmol l⁻¹, rCq-M15 precipitated brushite with characteristic Raman peaks at 904 and 990 cm⁻¹ (Fig. 8A), while at a concentration of 615 µmol l⁻¹, rCq-M15 precipitated hydroxyapatite with a characteristic Raman peak at 960 cm⁻¹ (Fig. 8A), indicative of the presence of both compounds in the precipitate. In the presence of 923 and 1085 µmol l⁻¹ rCq-M15 (Fig. 8A), only the hydroxyapatite peak at 960 cm⁻¹ was observed. To follow more realistic conditions of biomineralization in the crayfish, the above experiment was repeated for longer intervals and under ambient temperature. This experiment demonstrated that at a concentration of 923 µmol l⁻¹, rCq-M15 seems to interact in the kinetics of crystallization by enhancing the level of OCP (an additional precursor in the crystallization pathway of hydroxyapatite; Grynpas and Omelon, 2007), with a characteristic peak at 985 cm⁻¹ at ambient temperature, compared with the blank in which brushite, with characteristic peaks at 904 and 990 cm⁻¹, was precipitated exclusively (Fig. 8B). In the presence of rCq-M15, OCP remained stable during the 2 week duration of the experiment (Fig. 8B), whereas in the blank, a mixture of brushite and OCP gradually appeared (Fig. 8B). Fig. 9 presents the numerical values of Raman intensity at 960 cm⁻¹ (ʋ1 vibration of PO₄) after subtraction of the intensity of the 1010 cm⁻¹ (typical OCP) peak. A significantly higher apatite/OCP ratio is shown in the presence of 3.1 µmol l⁻¹ rCq-M15 compared with the control. No statistical difference in the apatite/OCP ratio was observed in the presence of 150 nmol l⁻¹ or 4.1 µmol l⁻¹ BSA.

In the present study, a novel protein termed Cq-M15 was identified from the crayfish molar tooth and found to be expressed in several exoskeleton-forming epithelia.

Characterization of the protein suggests its possible involvement in biomineralization. Indeed, this is the first report of an invertebrate protein suggested to interact in the precipitation of hydroxyapatite. The deduced protein sequence contains a predicted signal peptide, in accordance with the detection of the protein in the extracellular matrix. Cq-M15 has low percentages of acidic and putatively phosphorylated amino acids and possesses an RR1-type chitin-binding motif, which are more compatible with the properties of structural cuticular proteins (Willis, 2010) and, as such, do not favor an interaction with ions. However, Cq-M15 could be acidic enough to interact in mineral precipitation as has been discovered with other hydrophilic and acidic molecules involved in the regulation of crystal nucleation and growth during biological mineralization in vertebrates (He et al., 2005; Gayathri et al., 2007), as well as in invertebrates (Glazer and Sagi, 2012; Luquet, 2012). Moreover, several crustacean proteins associated with calcification processes present amino acid composition profiles similar to that of Cq-M15 (Inoue et al., 2004; Glazer et al., 2010). In fact, such shared patterns of amino acid composition were also found in proteins from organic matrices involved in invertebrate calcium carbonate precipitation (Faircloth and Shafer, 2007; Gayathri et al., 2007; Marie et al., 2007; Inoue et al., 2008; Suzuki et al., 2013), as well as in protein extracted from vertebrate enamel (Robinson et al., 2005).

The expression of Cq-M15 during the pre-molt stage coincides with molar tooth mineralization and the synthesis of the new cuticle during this particular molt stage (Tynyakov et al., 2015) and presumably indicates the involvement of Cq-M15 in exoskeleton assembly and, specifically, in the early calcification of the molar tooth and other exoskeletal elements. Elevation of Cq-M15 transcript expression in molar-forming epithelium at mid-pre-molt coincides with the formation of the apatite part (Tynyakov et al., 2015), which could implicate Cq-M15 interaction in calcium phosphate precipitation. Folowing the molt cycle progression, the highest Cq-M15 expression was detected at late pre-molt both in molar- and cuticle-forming epithelium, which coincides with exocuticle synthesis (Drach, 1939; Roer and Dillaman, 1984). The latter could indicate the presence Cq-M15 in exocuticle and its interactions in calcium carbonate and ACC precipitation. We assume that in the molar, Cq-M15 is secreted by epithelial cells, the extensions of which presumably form an array of channels within the apatite layer (Bentov et al., 2012).

Cq-M15 is transcribed in muscle as well. Previously it was published that crustacean muscle growth is associated with the molt cycle. Previous studies have demonstrated that muscle protein synthesis rate is elevated during pre-molt (Mykles and Skinner, 1982; El Haj and Houlihan, 1987). Together with the protein synthesis elevation, claw muscles have been shown to undergo atrophy (Skinner, 1965; Mykles and Skinner, 1982), which enables the claw to be drawn through the narrow basi-ischium joint (Mykles and Skinner, 1990). It was beyond the scope of this study to investigate the relationship between Cq-M15 expression, muscle type and changes in muscle through the molt cycle. However, based on the published literature, we would suggest the possible involvement of Cq-M15 in atrophy and restoration of muscle proteins to facilitate withdrawal from the carapace during molting.

The newly identified Cq-M15 was found to share significant similarity with known proteins from arthropod cuticles and its sequence was more closely related to crustacean proteins from mineralized parts of the cuticle (Gosline, 1980; Inoue et al., 2008; Suzuki et al., 2013) and more distantly related to proteins found in crustacean non-mineralized arthrodial membranes (Gosline, 1980) and insect cuticles (Kim et al., 2005; Futahashi and Fujiwara, 2008; Werren et al., 2010). Such close homology further supports the involvement of Cq-M15 in biomineralization in the crayfish. Furthermore, in vivo silencing of Cq-M15 was lethal, suggesting it has a vital function during the molt cycle. While numerous gene-silencing studies assessing the functionality of novel crustacean genes have been published previously (Ventura et al., 2009,, 2012; Rosen et al., 2010), to the best of our knowledge, the lethal effect of gene silencing has been previously reported in crustaceans only once (Christoffersen et al., 1989). We had shown earlier that silencing of particular transcripts in C. quadricarinatus affects the formation of the extracellular matrix and the calcification process (Shechter et al., 2008b; Glazer et al., 2010), as well as the duration of molt cycles (Pamuru et al., 2012; Tynyakov et al., 2015). However, the absence of the C. quadricarinatus sequenced genome does not permit us to completely rule out potential effects of non-specific small interfering RNAs derived from the Cq-M15 sequence acting within target cells, even though a complete sequence search was done prior to the silencing study.

Cq-M15 is predicted to contain the RR1-type chitin-binding motif. However, unlike most such proteins found in soft cuticular elements, i.e. insect cuticles (Rebers and Riddiford, 1988), Cq-M15 was extracted from a hard exoskeletal organ, i.e. the molar tooth. This is thus only the second report of an RR1-type protein being present in a hard cuticle (Suderman et al., 2006). In accordance with the bioinformatics-based prediction, rCq-M15 chitin-binding ability suggests its direct association with the extracellular chitinous cuticular matrix in different parts of the crayfish exoskeleton. In accordance with computer modeling predicting that β-sheet folds of cuticular proteins form chitin-binding sites (Iconomidou et al., 2005), the ellipticity data obtained here confirmed the presence of β-sheets and a helical structure in rCq-M15. In addition, CD analysis revealed that Cq-M15 undergoes a conformational transition in the presence of calcium, possibly to provide the proper environment for mineral nucleation.

In vitro mineralization studies under conditions that support the spontaneous formation of calcium phosphate crystals (Margolis et al., 2006) suggest that Cq-M15 interacts in the processes of calcium carbonate and calcium phosphate crystal nucleation. ACP and/or OCP have been suggested as corresponding to precursor phases of apatite crystals in vertebrates (Grynpas and Omelon, 2007; Beniash et al., 2009) and invertebrates (Lowenstam and Weiner, 1985; Bentov et al., 2012; Weaver et al., 2012). As the ability of rCq-M15 to interact in the initiation of hydroxyapatite nucleation depends on temperature, it seems that additional physiological factors regulating the transition to hydroxyapatite, yet to be discovered in the crayfish molar apatite, might also be involved in this transition.

In conclusion, the data presented here suggest that Cq-M15 possesses physical traits that would allow it to play a role in molar tooth and cuticle biomineralization. Because transcript silencing was lethal, we could not demonstrate the involvement of Cq-M15 in ACP or AFP precipitation in vivo. Cq-M15 spatial and temporal expression and its participation in mineral precipitation in vitro point to a possible role in calcium carbonate precipitation, including ACC precipitation. Thus, the above-discussed characteristics of Cq-M15 point to the involvement of the protein in the formation of the chitin–protein–mineral complex of the molar tooth and cuticle, through interactions with the chitinous matrix and calcium ions, and possibly with other proteins yet to be discovered. In addition, the level of the interaction depends in a concentration-dependent manner on the amount of calcium ions. Recombinant proteins that are produced in E. coli lack post-translational modification (Hannig and Makrides, 1998; Jonasson et al., 2002), and therefore we suppose that the Cq-M15 recombinant protein could possess a different tertiary and/or quaternary conformation compared with native Cq-M15 protein in the in vitro medium. Moreover, the Cq-M15 recombinant protein interacts in hydroxyapatite precipitation, whereas the molar is covered by fluoroapatite (Bentov et al., 2012), both of which are molecular forms of apatite but with a different in vitro conformation. Thus, this protein could be involved in calcium phosphate and/or carbonate precipitation. Further study of Cq-M15 properties and neighboring proteins will increase our knowledge on the involvement of proteins in exoskeletal calcified tissues and may prove instrumental for research in the fields of bone and teeth repair/regeneration and nano-structured materials.

We thank Leila Zeiri for technical assistance with Raman studies and Vered Chalifa-Caspi for her professional support in next generation sequencing analysis. Animals for the study were supplied by the Department of Fisheries and Aquaculture, Dor Agriculture Center.