Specialized insulin is used for chemical warfare by fish-hunting cone snails

Identification of a Fish-Like Insulin in the Venom Gland of C. geographus.

Analysis of next-generation sequencing of the venom gland transcriptome of C. geographus (12) led to the identification of two transcripts with high sequence similarity to insulin. These transcripts were named C. geographus insulins 1 and 2 (Con-Ins G1 and Con-Ins G2). Like other insulins, Con-Ins G1 and G2 precursors consist of four distinct regions: an N-terminal signal peptide and a B and A chain separated by one or more C peptides (Fig. 1). Based on this precursor organization, it can be predicted that oxidative folding and proteolytic processing of the two C. geographus transcripts yields mature heterodimers that consist of A and B chains connected by disulfide bonds.

The predicted mature Con-Ins G1 shares high similarity with vertebrate insulin, whereas Con-Ins G2 more closely resembles endogenous molluscan insulins. These differences are most obvious in the cysteine pattern and the lengths of the predicted A and B chains. Like vertebrate insulins, Con-Ins G1 has relatively short chains and contains four cysteines in the A and two in the B chain. In contrast, like most other invertebrate insulins (9), Con-Ins G2 contains two additional cysteines, one in each chain, and has longer A and B chains compared with vertebrate insulins.

Con-Ins G1 shows striking sequence similarity to insulins from fish, the prey of C. geographus. This similarity is most apparent within the A chain for which 90% of amino acids are similar between this peptide and insulins from zebrafish (Danio rerio) and the white sucker, Catostomus commersonii (Fig. 2). In particular, the N-terminal half of Con-Ins G1 is almost identical to the fish peptides. Thus, Con-Ins G1 is more similar to fish insulins than to any other vertebrate insulin. Similarities within the B chains are lower, ranging from 68% to 73%.

Although Con-Ins G2 shares a cysteine pattern with molluscan insulin, similarities at the amino acid level are relatively low, ranging from 50% to 71% compared with the A and B chains of the sea hare Aplysia californica and the great pond snail Lymnea stagnalis (Fig. 2). This finding is consistent with the observation that invertebrate insulin sequences are less conserved than their vertebrate counterparts (9).

In molluscs, insulins are typically expressed in neuroendocrine tissues, including ganglia and neurons (9), so the identification of insulin sequences in the venom gland of C. geographus was unexpected. Furthermore, C. geographus insulins represented the most highly expressed non–conotoxin-encoding genes, with expression levels comparable to established venom peptides. (Table 1) (12). The venom gland of cone snails is a long, heterogeneous organ that is highly specialized for the biosynthesis and secretion of conotoxins (11). Even though toxin expression is found across the entire length of the venom gland, expression of Con-Ins G1 was almost exclusively found in the distal region of the venom gland closest to the injection apparatus (99.9% of all RNA-Seq reads that mapped to Con-Ins G1 were found in this region; Table 1). In contrast, read numbers for Con-Ins G2 were very low in this region (0.9%) but high in the proximal section (49%).

Table 1.

Expression levels and distribution of C. geographus insulins across the length of the venom gland

Open in a new tab

The transcriptomes of four regions of the venom gland were sequenced by 454 RNA-Seq and differential expression of insulin transcripts and selected conotoxins was determined by mapping all reads to the transcripts of interest. The (distal) region of the venom gland is highlighted in red in the anatomical schematic. Read numbers for transcripts with highest expression in this region are also shown in red.

The Type of Insulin Expressed in the Venom Gland Correlates with Feeding Preference.

According to their prey preference, cone snails can be grouped into fish-, mollusc-, and worm-hunting snails. Several representatives of each feeding type were selected, and insulin expression was investigated by targeted RT-PCR and mining of cDNA libraries and transcriptome data. The presence of Con-Ins G1 and G2 in the venom gland of C. geographus was confirmed using RT-PCR combined with PCR cloning and Illumina sequencing. An additional sequence, Con-Ins G3, was also identified, albeit at lower expression levels (283 total reads retrieved from transcriptome data). Like Con-Ins G1, Con-Ins G3 contained the vertebrate-like arrangement of cysteines (Fig. S1) and was expressed exclusively in the distal region of the venom gland (100% of all reads). An identical sequence was recently reported in C. geographus as “new superfamily 2 conotoxin G121” but not recognized as an insulin (13). Several variants of Con-Ins G1, G2, and G3 were identified by RT-PCR, a phenomenon that has also been described for other cone snail venom peptides (12) (Fig. S1). All sequences obtained by PCR were also retrieved from the venom gland transcriptome database.

Using the same methodology, three insulin sequences were identified from the venom gland of Conus tulipa, another member of the Gastridium clade and the closest known relative of C. geographus (14). These sequences were named Con-Ins T1, T2, and T3. All three exhibit the cysteine arrangement of vertebrate insulin and share high overall sequence similarity with Con-Ins G1 and therefore fish insulins (Fig. S2). Sequence identities to Con-Ins G1 across the entire length of the prepropeptide range from 83% to 87%. The predicted mature peptide of the most similar transcript, Con-Ins T1, differs by only one conservative substitution in the A chain (Tyr-19 to Phe) and five substitutions in the B chain, of which three are conservative (Fig. S2). No sequences resembling endogenous molluscan insulins were found in the venom gland of C. tulipa, despite deep sequencing of PCR products (a total of 274,520, 187,930, and 132,975 reads were obtained for Con-Ins T1, T2, and T3, respectively).

Five additional fish-hunting cone snails were examined by RT-PCR using the same oligonucleotides and PCR conditions used for C. geographus and C. tulipa (Table S1). Unlike C. geographus and C. tulipa, which use a “net-hunting” strategy for prey capture, these species belong to the “hook and line hunters” (1). RT-PCR did not identify insulin transcripts in any of the five tested species. For two of these snails, C. bullatus and C. consors, the absence of insulin transcripts was also confirmed by mining next-generation transcriptome data (15, 16) [C. consors transcriptome data have not been made publically available, but Violette et al. (16) do not report any insulin expression in the venom gland of this species].

The venom glands of two mollusc and four worm-hunting cone snails were also investigated; all expressed molluscan-like insulins, but not the fish-like peptides (Table S1 and Fig. S1). As observed for Con-Ins G2, these transcripts were characterized by an additional cysteine residue in each chain and longer predicted A and B chains. Sequence diversity ranged from one transcript for Conus marmoreus, Conus textile, and Conus memiae to four in Conus floridulus. These findings suggest that insulin expression in the venom gland of a particular species can be correlated with its feeding strategy. Molluscan-like insulins are broadly distributed among snail and worm hunters, whereas the expression of fish-like insulins appears to be limited to a subset of fish hunters.

Con-Ins G1: High Abundance and Novel Posttranslational Modifications.

To determine whether C. geographus insulin transcripts were translated and posttranslationally processed into mature insulins, the venom of this species was interrogated by MS. Mature peptides derived from the Con-Ins G1 precursor but not G2 were identified by MS sequencing performed on reduced and alkylated (RA) venom using a high-resolution Orbitrap Elite instrument. Tandem MS identified the individual A and B chains of Con-Ins G1, demonstrating that cleavage of the precursor peptide at basic or dibasic sites liberates an A and B chain, as observed for other insulins (Fig. 3; MS/MS spectra in Fig. S3). Con-Ins G1 carries several posttranslational modifications that are unprecedented in insulins but commonly observed in other cone snail venom peptides (17). Glu4 of the A chain, one of the most conserved amino acids in vertebrate insulin (18), is modified to γ-carboxylated glutamic acid (Gla or γ). Another Gla modification (Gla12) was also identified in the B chain. Moreover, Pro5 in the A chain is modified to hydroxyproline (Hyp5), and the C terminus of this chain is amidated (Fig. 3).

MS sequencing further revealed the presence of several variants of Con-Ins G1 in fractionated venom (Fig. S4). Modifications in these variants include Gla15 and Hyp10 in the A chain, Hyp5 in the B chain, and alternative proteolytic cleavage at the C terminus of the B chain. The presence of multiple variants deriving from the same peptide precursor has recently been described for several conotoxins (16, 19) but is exceptional for insulins. The other vertebrate-like insulin, Con-Ins G3, was also identified by MS sequencing (Fig. S5). Similar to Con-Ins G1, this peptide is modified, and several variants could be identified in the venom (Fig. S4).

Fractionated C. geographus venom was analyzed by high-resolution LC-MS and searched for masses corresponding to the intact insulin dimers. A heterodimer that consisted of the originally sequenced A and B chains (Fig. 3) was not identified, but MS sequencing successfully identified one of the variants that contained an additional γ-carboxylated glutamic acid in the A chain (Gla15) and did not have the N-terminal arginine in the B chain (Fig. S6A). The molecular mass of this variant was 5,030.06 Da, consistent with the presence of three disulfide bonds. A second variant containing an additional hydroxyproline (Hyp10) in the A chain (5,046.05 Da; Fig. S6B) and a third variant that had no hydroxylated prolines in either chain were also identified (5,014.05 Da; Fig. S6C). Fractions were reduced and alkylated, and sequences were confirmed by tandem MS sequencing (Fig. S5). These data clearly show that Con-Ins G1 forms a heterodimer consisting of A and B chains that are linked by disulfide bonds. The intact heterodimer of the other vertebrate-like peptide, Con-Ins G3, was also identified by MS sequencing. Intact Con-Ins G3 has a molecular mass of 4,817.10 Da, consistent with the presence of three disulfide bonds (Fig. S6D). With a total length of 43 amino acids, to our knowledge, native Con-Ins G1 and G3 are the shortest native insulins reported to date (human, zebrafish, and Aplysia insulins are 51, 50, and 80 residues long, respectively).

Venom fractionation and sequencing further confirmed that Con-Ins G1 and its variants are abundant peptides in the venom of C. geographus. The venom fraction containing Con-Ins G1 was 1 of 25 collected fractions and predominantly consisted of venom insulins.

Con-Ins G1 Was Successfully Synthesized Using a Selenocysteine-Based Strategy.

To obtain sufficient material to investigate its bioactivity in fish, Con-Ins G1 was chemically synthesized. Since the first chemical synthesis in 1963 (20), insulin has remained challenging to synthesize with the correct intra- and intermolecular disulfide bonds. A strategy incorporating a pair of selenocysteines in the A chain was used to synthesize Con-Ins G1. Replacement of one pair of cysteines with selenocysteines (Sec) has been used for the synthesis of various bioactive peptides, including conotoxins (21). Diselenide bond-containing peptide analogs have similar biological activities to their native peptides and, in some cases, even improved potency or selectivity (21). Due to the lower redox potential, diselenide bond formation is favored over the disulfide bond formation under acidic conditions (21). This Cys-to-Sec replacement strategy was combined with orthogonal protection of the remaining two pairs of cysteines to sequentially form intra- and intermolecular disulfide bridges, as shown in Fig. 4. Cys6 and Cys11 of chain A were replaced with Sec residues, which formed an intradiselenide bridge after peptide cleavage and reduction (21, 22) (Fig. 4A). Cys20A and Cys21B of the respective chains [synthesized with a trityl (Trt) protecting group] were used to form the first intermolecular disulfide bridge on DMSO treatment (Fig. 4C). Cys7A and Cys9B, bearing acetaminomethyl (Acm) groups, were subjected to iodine oxidation to remove Acm protection and simultaneously form the second disulfide bridge (Fig. 4E). The two-step ligation protocol led to the final purified product, synthetic selenocysteine-Con-Ins G1 (sCon-Ins G1), in a total yield of 10.5%, based on the starting amount of purified chain A. RP-HPLC confirmed the purity (95%), and electrospray ionization MS sequencing confirmed the correct identity of the product (Fig. S7). To our knowledge, this is the first time a selenocysteine-based strategy has been applied to the synthesis of insulin.

sCon-Ins G1 Lowers Blood Glucose and Alters Swimming Activity in Fish.

The streptozotocin (STZ)-induced model of hyperglycemia was used to assess whether sCon-Ins G1 could effectively lower blood glucose levels in a model prey: adult zebrafish. Animals were first rendered hyperglycemic through i.p. injection of the β-cell poison STZ (1.5 g/kg) (23), and the effect of subsequent injection of sCon-Ins G1 was examined. Following STZ treatment, blood glucose levels were significantly elevated from 65.9 ± 4.75 (n = 11) to 273.0 ± 34.08 mg/dL (n = 5, P < 0.01; Fig. 5A). Administration of sCon-Ins G1 at 65 ng peptide/g body weight significantly lowered blood glucose to 74.0 ± 8.051 mg/dL (n = 8, P < 0.01). The ability to effectively lower blood glucose was similar to human insulin (92.0 ± 17.35 mg/d, n = 6, P < 0.01).

To determine whether sCon-Ins G1 elicits an effect when applied to the water column, the spontaneous swimming behavior of fish larvae (7 d after fertilization, n = 3) was monitored before (baseline) and following application of 25 nmol/mL peptide. Administration of sCon-Ins G1 reduced overall locomotor activity, observed as a significant decrease in the percentage of time spent swimming (P = 0.038) and movement frequency (P = 0.003) compared with baseline values for the same animals (Fig. 5B and Movie S1). It has previously been reported that externally applied human insulin can induce insulin shock in fish via direct absorption through the gills into the bloodstream (24). Our findings suggest that sCon-Ins G1 also efficiently crosses the endothelium of the gill plexus.