Snake venom NAD glycohydrolases: primary structures, genomic location, and gene structure

Historical overview of studies on snake venom NADase

Seven years after the Bhattacharya study, Suzuki, Iizuka & Murata (1960) examined 9 Asian snake venoms for NADase activity, using the UV detection method (340 nm) of Zatman, Kaplan & Colowick (1953) and discovered this enzyme in the venoms of Bungarus multicinctus and Trimeresurus gramineus. Venoms reported as negative for NADase included Gloydius blomhoffii, Deinagkistrodon acutus, Ovophis okinavensis, Protobothrops mucrosquamatus, Naja atra, Naja naja, and Ophiophagus hannah.

In a study of 37 elapid, viperid and crotalid venoms, also using UV detection, Tatsuki et al. (1975) confirmed the earlier findings of NADase activity in venoms of the two Bungarus species and further identified it in venoms of Agkistrodon c. contortrix, A. c. mokasen, A. c. laticinctus, A. p. piscivorus, Gloydius blomhoffii, Deinagkistrodon acutus, and Causus rhombeatus, the first viperid examined. All other taxa were reportedly negative for NADase activity (Table 1). Using a succession of five liquid chromatographic procedures they isolated the enzyme from G. blomhoffii venom and characterized it biochemically. The Gloydius enzyme readily hydrolyzed β-NAD and NADP+, and cleaved 3-acetylpyridine adenine dinucleotide, but it did not hydrolyze NADH, NADPH, α-NAD+, or β-nicotinamide mononucleotide (β-NMN) (Tatsuki et al., 1975). They did not estimate the enzyme’s molecular weight.

Yost & Anderson (1981) characterized the NADase from Bungarus fasciatus venom, and reported that it was a homodimeric glycoprotein of 120–130,000 Da, having a monomeric molecular weight of 62,000 (denaturing SDS PAGE). The enzyme comprised approximately 0.1% of Bungarus fasciatus venom by mass (Yost & Anderson, 1981; Anderson, Yost & Anderson, 1986) while a value of 0.5% was reported from Deinagkistrodon acutus venom (Wu et al., 2002) using a simpler chromatographic procedure with more sophisticated resins.

Huang et al. (1988) investigated the NADase from Deinagkistrodon acutus venom. That enzyme is a homodimeric glycoprotein of about 98,000 Da, having a minimum monomeric molecular mass of 33,500 Da, allowing for a carbohydrate content estimated at 33%. The authors reported that the N-terminal amino acid was proline. As with the Gloydius enzyme (Tatsuki et al., 1975), NADP was the optimal substrate (Huang et al., 1988). The Deinagkistrodon NADase is a metalloenzyme, containing a single, essential copper ion.

Despite these studies, no structural information has been reported for any snake venom NADase. Because our group has completed a series of elapid and crotalid venom gland transcriptomic studies employing high-throughput techniques (Aird et al., 2013; Aird et al., 2015; Aird et al., 2017b), we searched these transcriptomes for the presence of NAD glycohydrolase. It was found in all of them, and we here report their primary structures, possible 3D structures, genomic arrangement, and gene structure.

Snake venom NADase amino acid sequences

The NCBI Protein site was searched for vertebrate NAD glycohydrolase sequences and the sequence of chicken ADP-ribosyl cyclase (ADQ89191.1), also known as CD38, was downloaded for use as a query sequence. TBLASTN searches of venom gland transcriptomes of 10 elapid and crotalid species were performed using Geneious 8.1.9. A highly similar sequence was identified in each transcriptome, eight of which were complete (Fig. 1), and NADase transcripts were present at low levels in venom gland transcriptomes of all 30 Protobothrops mucrosquamatus examined by (Aird et al., 2017a). A partial (30-residue), unidentified sequence also occurs in the Ophiophagus hannah genome (L345_15802). The former sequences were aligned with CD38 sequences from Gallus gallus, Xenopus laevis, Anolis carolinensis, and Homo sapiens, using Geneious (Fig. 1).

None of the venom NADases appears to have a signal peptide, based upon sequence analyses using SignalP 4.1 (Petersen et al., 2011). All are readily distinguished from the former four vertebrate CD38 sequences, as all venom sequences commence with the N-terminal sequence, MPFQNS, rather than with proline, as reported by Huang et al. (1988). Like other vertebrate CD38 sequences, snake venom sequences possess a hydrophilic 14-residue N-terminus (Fig. 1). In all NADases, immediately C-terminal to this hydrophilic block, there is a hydrophobic, 25-amino acid segment containing 17 aliphatic residues (L, V, I, and G), 3 threonine residues, 1 lysine, and 2-3 phenylalanines. These are followed by another 20 residues that are nearly all hydrophilic. The aliphatic segment almost has the appearance of a signal peptide (Fig. 1).

Venom NADases are also clearly resolvable into elapid and crotalid sequences. The Micrurus NADases have 303 amino acids, while crotalid enzymes have 304 and may be readily distinguished based upon sequence differences at various positions. Coralsnake and pitviper sequences display the following respective differences: Q/E52, R or Q/W91, S/R105, H/N145, N/D146, K/N167, D/N170, E/M218, I/T217, N/S219, Q/K224, K/E240, and probably also G/D253, S/N276, I/S278, T/A301, and T/K303, although the Ovophis okinavensis and Protobothrops flavoviridis sequences are incomplete at this point (numbered as in Fig. 1). At several positions, some Micrurus sequences show the crotalid residue while others have a different amino acid.

The M. surinamensis apoprotein has a monomeric molecular weight of 34,206 Da and a predicted pI of 7.93, using the Expasy Compute pI/MW tool. The coralsnake enzymes have masses only slightly higher than the 33,500 predicted for the Deinagkistrodon enzyme (Huang et al., 1988). The Micrurus sequences contain three potential N-glycosylation sites (NKSL, position 125; NGSI, position 214; NRSI, position 274), based upon results obtained using the Expasy NetNGlyc 1.0 Server (http://www.cbs.dtu.dk/services/NetNGlyc/). The first, which is common to elapid and crotalid sequences alike (Fig. 1), has the highest likelihood of being glycosylated, although it may not be since venom NADases lack signal peptides. Nonetheless, Huang et al. (1988) reported a carbohydrate content of about 33% for the Deinagkistrodon NADase, so it seems likely that at least the NKSL at position 125 is N-glycosylated. Based upon analysis with the Expasy NetOGlyc server, there are no probable O-glycosylation sites (Steentoft et al., 2013).

Higher-level structural attributes of vertebrate NADases

Human CD38 is an ecto-enzyme with a long, helical membrane anchor and an intracellular N-terminal segment believed to be a random coil (Malavasi et al., 1992; Prasad et al., 1996; Lee, 2006) (Fig. 2). In addition to its enzymatic activity, CD38 also transduces signals to the cytoplasm. It is thought to regulate metabolism and participates in the pathogenesis of diverse maladies such as inflammation, obesity, diabetes, heart disease, asthma, and aging (Chini et al., 2018). Its enzymatic activity is involved in many of these functions. Moreover, CD38 has been identified as a cell–surface marker in multiple myeloma and other blood-related cancers (Chini et al., 2018).

In contrast, snake venom enzymes are soluble rather than membrane-bound (Tatsuki et al., 1975; Yost & Anderson, 1981), like the Aplysia ADP-ribosyl cyclase (Lee & Aarhus, 1991). While venom NADases could conceivably be membrane-bound in exosomes, they have not been reported as exosomal enzymes (Ogawa et al., 2008), and their elution on Sephadex G-100 is appropriate for soluble enzymes of ∼100 kDa rather than for exosomes (Tatsuki et al., 1975). Exosomal embedding seems further unlikely in that all of the venom NADases reported here possess a very short N-terminal α-helix and a random coil, instead of the long α-helical membrane anchor of human CD38 (Lee, 2006) (Fig. 2A). Snake venom NADase residues 45–302 superimpose almost perfectly upon the crystal structure of the soluble extracellular domain of human CD38, except for the divergent C-termini and the truncated N-termini (3F6Y_A) (Fig. 2B).

Yost & Anderson (1981) found that the Bungarus NADase is a homodimer with a dimeric mass of 125–130 kDa, and a subunit mass of 62 kDa on reducing SDS-PAGE. Huang et al. (1988) reported slightly lower values of 98 kDa and 50 kDa for the Deinagkistrodon enzyme, but apparently employed non-reducing SDS-PAGE, suggesting that the homodimer in that taxon is non-covalent. Interestingly, our Micrurus sequences have 12 cysteine residues, arranged in the 3D structure in a manner consistent with 6 disulfide bonds (C68–C84; C101–C181; C121–C202; C162–C175; C256–C277; C289–C298) (Fig. 2) (Egea et al., 2012), coinciding exactly with the disulfide bonds in CD38 (Fig. 2C); however, the two Protobothrops sequences both have a 13th cysteine in the penultimate position (C-terminus), such that the Protobothrops enzymes could be covalent homodimers (Fig. 1).

The N-terminal 50 residues of the M. surinamensis NADase are slightly hydrophobic, with a Gravy score of 0.142, while the N-terminal 50 residues of human CD38 (BAA18966.1) have a significantly more hydrophobic Gravy score of 0.710. As a result, the soluble venom enzymes appear to have a slightly more compact N-terminal domain than human CD38 (Fig. 2D).

Many invertebrate and vertebrate enzymes exhibiting NADase activity are multifunctional, not only hydrolyzing β-NAD(P)⁺ to nicotinamide and ADP-ribose, but also exhibiting ADP- and GDP-ribosyl cyclase, and cADPR/cGDPR hydrolase activities (Howard et al., 1993; Lee, Graeff & Walseth, 1997; Ziegler et al., 1997; Ziegler, Jorcke & Schweiger, 1997; Augustin, Muller-Steffner & Schuber, 2000; Ferrero et al., 2014). However, unlike ADP-ribosyl cyclase, human CD38 converts very little β-NAD to cADPR (Lee, 2006).

In contrast to many invertebrate and vertebrate CD38 homologs, Yost & Anderson (1981) found that when β-NAD was hydrolyzed by Bungarus fasciatus NADase, nicotinamide and ADP-ribose were the sole products. The B. fasciatus enzyme does not catalyze the conversion of β-NAD to cyclic ADP-ribose (cADPR). This lack of cyclase activity results in part from the presence of a disulfide bond (C124–C206 in Fig. 1; C121–C202, actual) which is absent in Aplysia ADP-ribosyl cyclase (Tohgo et al., 1994). Moreover, this disulfide bond is present in all snake venom NADases for which we have sequences. Graeff et al. (2009) reported that mutation of Phe221 in Aplysia ADP-ribosyl cyclase (Phe227 in Fig. 1) reduced cADPR production and increased ADPR liberation. Consistent with this conclusion, all snake NADases and at least some vertebrate CD38s have isoleucine in this position (Fig. 1), effectively preventing cADPR formation. Snake venom NADases also all have the conserved TLEDTL (144–149) sequence (Fig. 2A) (residues 149–154 in Fig. 1) that forms the bottom of the active site pocket (Lee, 2006), and the catalytic residue, Glu226 (Glu232 in Fig. 1), which are present in all CD38 molecules (Graeff et al., 2001). Substitution of Glu-146 with Phe, Asn, Gly, Asp, Leu, or Ala resulted in cyclase activity up to 9x higher than of wild-type CD38 (Graeff et al., 2001).

Genome location of vertebrate NAD glycohydrolases/ADP-ribosyl cyclases

We performed genome-wide BLAST searches to locate the NAD glycohydrolase gene in the genomes of Homo sapiens, Gallus gallus, Alligator mississippiensis, Anolis carolinensis, Protobothrops mucrosquamatus, Python bivitattus, Thamnophis sirtalis, and Xenopus laevis. After locating the genes, we manually checked their sequences and compared them with existing annotations of transcriptomic and proteomic data.

In the genomes surveyed, NAD glycohydrolase, CD38, is located in the vicinity of the CC2D2A and PROM1 genes (Ch1L in Xenopus laevis, Ch4 in Homo sapiens and Gallus gallus), usually directly downstream from the FGFBP1 gene. Non-squamate vertebrates have a duplicate gene, called BST1, located upstream from CD38. Squamates apparently lack BST1 in this region, probably due to clade-specific gene loss.

Gene structure of vertebrate NAD glycohydrolases/ADP-ribosyl cyclases

Human CD38 displays similar intron-exon architecture to that seen in the invertebrate, Aplysia kurodai ADP-ribosyl cyclase (Nata et al., 1995), suggesting that this architecture is highly conserved. It comprises eight exons, extending more than 77 kb in the human genome (Nata et al., 1997). However, in Protobothrops mucrosquamatus (and presumably in other venomous snakes) the 8-exon gene is transcribed in two variants, a long NADase similar to CD38 of other vertebrates (Fig. 3), and a shorter form that lacks exon 6 (Fig. 4). Exon 6 contains no cysteines, and its deletion results in the loss of 31 amino acids from Ser223-Gln253 (Fig. 4A). Owing to the presence of split arginine codons at both the exon 6–7 and exon 7–8 boundaries, no amino acid substitution occurs in the shortened (272-residue) structure. Instead, the effect of removing these 31 residues is to delete the loose helical region (Ser223-Gln253) from the center of the molecule, including the essential catalytic residue, Glu229 (Fig. 4A; Glu232 in Fig. 1), while leaving the remainder of the structure essentially intact (Fig. 4B). If this protein is expressed, it is difficult to imagine what its function might be, but with its precise alternate splicing, this does not seem like a pseudogene.