Characterization of D-amino-acid-containing excitatory conotoxins and redefinition of the I-conotoxin superfamily

Authors


B. M. Olivera, Department of Biology, University of Utah, 257 South 1400 East, Salt Lake City, UT 84112, USA
Tel: +1 801 581 8370
Fax: +1 801 585 5010
E-mail: olivera@biology.utah.edu

Abstract

Post-translational isomerization of l-amino acids to d-amino acids is a subtle modification, not detectable by standard techniques such as Edman sequencing or MS. Accurate predictions require more sequences of modified polypeptides. A 46-amino-acid-long conotoxin, r11a, belonging to the I-superfamily was previously shown to have a d-Phe residue at position 44. In this report, we characterize two related peptides, r11b and r11c, with d-Phe and d-Leu, respectively, at the homologous position. Electrophysiological tests show that all three peptides induce repetitive activity in frog motor nerve, and epimerization of the single amino acid at the third position from the C-terminus attenuates the potency of r11a and r11b, but not that of r11c. Furthermore, r11c (but neither r11a nor r11b) also acts on skeletal muscle. We identified more cDNA clones encoding conopeptide precursors with Cys patterns similar to r11a/b/c. Although the predicted mature toxins have the same cysteine patterns, they belong to two different gene superfamilies. A potential correlation between the identity of the gene superfamily to which the I-conotoxin belongs and the presence or absence of a d-amino acid in the primary sequence is discussed. The great diversity of I-conopeptide sequences provides a rare opportunity for defining parameters that may be important for this most stealthy of all post-translational modifications. Our results indicate that neither the chemical nature of the side chain nor the precise vicinal sequence around the modified residue seem to be critical, but there may be favored loci for isomerization to a d-amino acid.

The traditional approach in biochemistry to studying proteins is purification from the natural source; however, the access to genomic information has provided a different, increasingly dominant methodology. The technology for manipulating nucleic acids and expressing polypeptides has advanced so rapidly that fewer and fewer native proteins are being purified; instead, the encoding genes are expressed in recombinant systems. This has become the standard approach used for functionally characterizing gene products.

A potential liability in adopting the modern paradigm is that functionally important post-translational modifications may be overlooked. If the natural gene product is post-translationally modified, but the polypeptide expressed from a cloned sequence is not, then the latter could be functionally deficient to varying extents. For this reason, in the genomic era, it is increasingly important to reliably predict when and where a particular post-translational modification may occur. Some post-translational modifications are easily predicted as they occur on highly specific amino-acid sequences (for example, N-glycosylation or phosphorylation by various kinases). Even when no consensus sequence to predict a post-translational modification can be written down, if enough examples are characterized, an accurate prediction of a post-translational event can often be made; this is the situation with regard to predicting signal sequences and where a signal peptidase will cleave; despite the lack of a rigidly specified consensus sequence, enough data have been collected to make the predictive programs based on the existing data quite reliable.

One of the most subtle and difficult-to-detect post-translational modifications is the epimerization of a standard l-amino acid to its d-isomer. This is a particularly stealthy modification because it remains undetectable by standard proteomic methods such as Edman sequencing or MS. Although post-translational epimerization is a modification with which the larger molecular biological community has not been particularly concerned, there is increasing evidence that its occurrence is surprisingly widespread: it has been found in gene products from very divergent phyla, including chordates, arthropods and molluscs. In vertebrates, this post-translational modification has been discovered in amphibian as well as mammalian systems [1–6]. Additional data on post-translational epimerization are therefore highly desirable; the only reliable predictions for when d-amino acid epimerization may occur are in two small families of highly specialized peptides, notably an opiate peptide family from frog skin (which are all seven amino acids long and highly homologous to each other), and the contryphan family of peptides from Conus (which are eight amino acids long with highly conserved sequence motifs).

Apart from the frog skin opiate peptides and the contryphans, other gene products that have been characterized with d-amino acids are quite a heterogeneous assemblage. Despite this, it has been noted that there are preferential loci for modification, i.e. either on the second amino acid from the N-terminus, or on the third amino acid from the C-terminus. However, there has not been an opportunity to further define in what sequence contexts epimerization is likely to occur.

The recent discovery of a functionally important d-phenylalanine residue in a peptide of the I-gene superfamily of conotoxins provides a potentially much more diverse sequence database in which to explore the factors required for epimerization of an l-amino acid to a d-amino acid to occur. The I-superfamily peptides have eight cysteine residues in the primary sequence, with the characteristic pattern –C–C–CC–CC–C–C–, and conotoxins belonging to this family are larger than those belonging to other Conus peptide families. I-superfamily conotoxins have been found broadly over many Conus species; although all have the characteristic Cys pattern, these toxins have extremely diverse amino-acid sequences [8–11]. Many appear to be excitatory (and in certain cases, have been shown to target specific potassium channels); their diversity provides a framework for predicting which peptides have a d-amino acid, and which do not.

The I-superfamily peptide previously shown to have a d-phenylalanine residue [7], r11a, is 46 amino acids in length with the d-Phe residue at position 44. In this report, we characterize two peptides, r11b and r11c, purified from venom, that show various extents of sequence homology to r11a; a comparison of the three sequences is shown in Table 1 (the position of the confirmed d-phenylalanine residue in r11a is underlined). There is a notable difference in the sequence divergence from r11a at the C-terminal 12 amino acids (boxed residues in Table 1): the sequences of r11a and r11b are almost identical (with a single conservative substitution), whereas the sequences of r11a and r11c differ in seven out of 12 amino-acid positions, including the amino acid Phe44 in r11a which is racemized. Thus, if there were a stringent sequence determinant for post-translational epimerization in the vicinity of the residue to be modified, or if the enzyme were highly specific for the phenyl side chain, it might be predicted that the native r11b would have a d-phenylalanine at the same locus as r11a, and r11c might have an l-leucine analog at the homologous position.

Table 1.  Amino-acid sequences of I-conotoxins with a d-amino acid near the C-terminus. O, 4-trans-hydroxyproline; F, d-phenylalanine; ^, C-terminal free acid. Peptide r11c is presumably synthesized with a C-terminal arginine as determined from a cDNA clone [8], but the arginine is cleaved by a carboxypeptidase (also, Table 3).
PeptideSequence
inline image

To test these predictions experimentally, we carried out the chemical synthesis of both l and d forms of r11b and r11c to compare these with the native material. All forms have also been functionally characterized. In addition, we carried out a further molecular definition of the I-superfamily. In contrast with all other conotoxin gene superfamilies, we found that the same Cys pattern is apparently shared by two entirely distinct gene superfamilies. Thus, what was previously referred to as the I-conotoxin superfamily should now be divided; a potential connection between the identity of the gene superfamily to which the conotoxin belongs, and the presence or absence of a d-amino acid in the peptide will be discussed.

Results

Purification and synthesis of I-superfamily conotoxins

The excitatory peptides r11b and r11c were purified from the venom of the fish-hunting species Conus radiatus as described previously [8], but with some modification for the purification of r11b. As shown in Fig. 1, r11b was purified to homogeneity after four RP-HPLC runs.

Figure 1.

Purification of conotoxin r11b. (A) Fractionation of crude venom extract using a Vydac C18 semipreparative column eluted with a gradient of 0–60% solvent B90 over 120 min at a flow rate of 5 mL·min−1. (B) Elution of the bioactive fraction marked by an arrow in (A) using a C18 analytical column and a gradient of 30–40% solvent B90 over 40 min. (C) Further purification of the bioactive fraction marked by an arrow in (B) with a gradient of 25–55% solvent B90 over 30 min. (D) Pure r11b was obtained after elution of the fraction marked by an arrow in (C) with a gradient of 25–52% solvent B90 over 27 min. For chromatograms B–D, the flow rate was 1 mL·min−1.

Based on the sequences of the native peptides (Table 1), conotoxins r11b and r11c were chemically synthesized on a solid support using a standard Fmoc protocol. Two versions of both peptides were produced, one with all l-amino acids and the second with a single d-amino acid at the locus homologous to d-Phe44 of r11a, i.e. d-Phe at position 44 of r11b and d-Leu at position 42 of r11c. Both variants of the toxins were folded in the presence of reduced and oxidized glutathione with a yield of 70%, and purified to homogeneity by RP-HPLC. The identities of the peptides were confirmed by MALDI MS, and the molecular masses of the synthetic toxins ([l-Phe44]r11b, 4857.8 Da; [d-Phe44]r11b, 4858.7 Da; [l-Leu42]r11c, 4629.0 Da; [d-Leu42]r11c, 4629.1 Da) were within an atomic mass unit of those calculated from amino-acid sequences.

D-Amino acids are present in native r11b and r11c

HPLC co-injections of each of the two synthetic variants of r11c conotoxin, [l-Leu42]r11c and [d-Leu42]r11c with the natural form were performed. As illustrated in Fig. 2, only the d-Leu42-containing variant coeluted with the natural peptide. As previously shown, r11a, another I-conotoxin [7,8], has a d-Phe at position 44 of the 46-residue peptide; conotoxins r11a and r11b have high sequence homology (Table 1), thereby leading to the strong prediction that d-Phe is located at position 44 in r11b. Indeed, both natural and synthetic r11b with d-Phe44 have an identical elution time, which differs from that of the all-l isomer by about 1.0 min under the same HPLC conditions (Fig. 3). Thus, the correct amino-acid sequences of conotoxins r11b and r11c have d-Phe and d-Leu at positions 44 and 42, respectively. Henceforth, we will refer to the d version of the peptides simply as r11b and r11c, and the all-l versions as [l-Phe44]r11b and [l-Leu42]r11c.

Figure 2.

HPLC analysis. Top panel, natural r11c (r11c N); middle panel, a mixture of two synthetic variants, r11c (r11c S) and [l-Leu42]r11c; bottom panel, a mixture of natural and synthetic r11c (r11c N and r11c S). Elutions were carried out using a C18 analytical column at 45 °C with a gradient of 30–50% B over 20 min at a flow rate of 1 mL·min−1.

Figure 3.

HPLC analysis. Top panel, natural r11b (r11b N); bottom panel, a mixture of two synthetic variants, r11b (r11b S) and [l-Phe44]r11b. Elutions were carried out using a C18 analytical column at 45 °C with gradient of 30–50% B over 20 min at flow rate of 1 mL·min−1.

Increased biological potency of D-amino-acid-containing peptides

A whole animal behavioral assay with 15-day-old mice was used to assay r11b, r11c and their l isomers. In the case of r11b, a fivefold difference was observed in potencies of the d-Phe44 and l-Phe44 forms. The estimate is based on comparison of the threshold dose that induced hyperactivity symptoms (tail raising, rapid walking, circular motion, sensitivity to touch), which was 20 pmol for r11b and 100 pmol for [l-Phe44]r11b. An approximately fivefold difference in potency was likewise observed between the d-Leu42 and l-Leu42 forms of r11c. Similarly, this estimate is based on a comparison of the dose at which hyperactivity symptoms appeared (tail raising, convulsions), which was 20 pmol for the r11c and 100 pmol for the [l-Leu42]r11c.

The synthetic forms of peptides r11b and r11c were also tested on a frog nerve–muscle preparation as previously described [7,8]. The results are shown in Figs 4 and 5. Both r11c and [l-Leu42]r11c were active in inducing repetitive activity in this preparation. As a measure of potency, the on rates of the peptides were compared by observing the time it took for repetitive nerve activity to be manifested (the ‘induction time’). When tested at a concentration of 100 nm, both l-Leu42 and d-Leu42 forms of r11c had induction times of ≈ 2 min (Fig. 4). On the other hand, the induction time for 100 nm r11b was longer, ≈ 4 min (not illustrated, see legend of Fig. 4 for statistics). This is similar to the induction time previously reported for 100 nm r11a [7]. In contrast with its d-counterpart, [l-Phe44]r11b did not induce repetitive activity even at a concentration of 10 µm (not illustrated). This is consistent with the observation that [l-Phe44]r11a is also inactive on the frog preparation [7].

Figure 4.

The synthetic d-Leu42 and l-Leu42 forms of r11c have comparable activities on the frog nerve–muscle preparation. Responses to nerve stimulation were obtained simultaneously from muscle (lower panels) and nerve (upper panels) every 30 s, except during solution exchange (Experimental procedures). Successive responses are shown in each panel from bottom to top, before (lowest trace in each panel) and during (remaining traces) exposure to 100 nm of [l-Leu42]r11c (panels A1 and A2) or r11c (B1 and B2). In each panel, the gap between the lowest trace and the one immediately above it reflects the time during which peptide was being applied, and no recordings were made. Vertical and horizontal units are mV and ms, respectively (1 mV vertical displacement of traces corresponds to an elapsed time of 30 min). To show the repetitive action potentials clearly, a y-axis range was used that truncated the evoked compound action potential. The latter is seen in each panel almost immediately after the nerve stimulus, which was applied 20 ms after the start of each trace. The first nerve trace to show repetitive activity occurred after about 1.5 min for both peptides. This ‘induction time’ for repetitive activity in the nerve with [l-Leu42]r11c was 1.9 ± 0.2 min (mean ± SD, N = 6), and with r11c it was 1.6 ± 0.2 (N = 5). The induction time with r11b was significantly longer (not illustrated), 3.8 ± 2 min (N = 3). Note that, with r11c, repetitive activity in the muscle can be seen after peptide application (panel B2), without any repetitive activity in the nerve (panel B1); this is consistent with r11c having a target in muscle, as well as in nerve (Fig. 6). The effects of these peptides were reversible after washout (not illustrated).

Figure 5.

r11b Acts on nerve, but [l-Leu42]r11c acts on both nerve and muscle. Stimulation and recording were as in Fig. 4. (A) In 100 nm r11b, each action potential in the muscle recording (lower trace) has a corresponding action potential that just precedes it in the nerve recording (upper trace), indicating that this peptide's target is in nerve, but not muscle. (B) In contrast, in 100 nm[l-Leu42]r11c, there are many events in the muscle recording without a corresponding event in the nerve recording, indicating that this peptide has targets in both nerve and muscle. Note the spontaneous activity in muscle (and single event in nerve) preceding the stimulus, which was applied at 20 ms from the start of each trace. Repetitive events in muscle, without corresponding events in nerve, were also induced by r11c (Fig. 4B1,2). The effects of these peptides were reversible after the washout (not illustrated). Repetitive activity could not be elicited in either nerve or muscle with [l-Phe44]r11b, even at a concentration of 10 µm (not illustrated).

Figure 4 also shows that r11c can induce repetitive activity not only in nerve, but also in muscle. Figure 5 shows that [l-Leu42]r11c, like its d-Leu counterpart, can induce repetitive activity in muscle. On the other hand, r11b was able to induce repetitive activity in nerve, but not muscle per se (Fig. 5). To further examine the peptides' activity, the muscle was directly stimulated, and the participation of nerve activity was avoided by using d-tubocurare to block any synaptic activity as described in Experimental Procedures. Furthermore, the application of peptide was confined to one end of the muscle, where there are few, if any, neuromuscular synapses. Figure 6 shows that both l-Leu42 and d-Leu42 forms of r11c were about equipotent in inducing repetitive action potentials under these conditions. In contrast, neither r11a nor r11b were active in this assay when tested with concentrations up to 1 µm (not illustrated). The differences in the effects of r11a, r11b, r11c and their l isoforms are summarized in Table 2.

Figure 6.

[l-Leu42]r11c and r11c have comparable effects on directly stimulated muscle. Action potentials were evoked by directly stimulating one end of the muscle every 30 s while recording from the other end, as described in Experimental procedures. Recordings are presented as in Fig. 4. Thus, the lowest trace in each panel represents the control response, and all others are those during progressive exposure to 100 nm l-Leu42 peptide (A) or d-Leu42 r11c (B). d-Tubocurare (10 µm), a postsynaptic acetylcholine receptor blocker, was present in both control and peptide solutions to block any possible, although unlikely, contributions of synaptic activity. Exposure to peptide was confined to one end of the muscle, opposite the end that was electrically stimulated. A 1-mV vertical displacement of the traces represents an elapsed time of 30 s. The muscle was stimulated 20 ms after the start of each trace, and after a latency of ≈ 5 ms, a compound action potential was observed. After exposure to the peptide, events are seen with a polarity opposite that of the evoked-compound action potential, indicating that these events were action potentials propagating in the opposite direction, as would be expected if they were generated at the end of the muscle to which toxin had been applied. It is evident that both peptides induced repetitive firing within ≈ 2 min of peptide exposure. These results demonstrate that both peptides have muscle targets and are about equally potent. The actions of both peptides were reversed after washout (not illustrated).

Table 2.  Activities of I-conopeptides in inducing repetitive action potentials in two target tissues from frog. A plus sign indicates that the peptide induces repetitive action potentials within 5 min of its application at a concentration of 100 nm (e.g. Fig. 4).
PeptideMotor nerveSkeletal muscleReference
r11a
 d-Phe44+[7]
 l-Phe44[7]
r11b
 d-Phe44+This work
 l-Phe44This work
r11c
 d-Leu42++This work
 l-Leu42++This work

cDNA cloning and identification of I-superfamily conotoxins

We previously identified cDNA clones, R11.6, R11.14 and R11.4, encoding the I-superfamily conotoxins r11a, r11b and r11c, respectively [8]; this report described the initial characterization of the I-conotoxin superfamily. Other cDNA clones belonging to the I-superfamily were identified from both C. radiatus and other species. Homologous sequences including those from Conus figulinus and Conus betulinus, which are worm-hunting species, and Conus episcopatus, a mollusc-hunting species, are shown in Table 3. In Group A, all peptides have the consensus pattern –CX6CX5CCXCCX4CX8−10C–. The high sequence homology and the confirmed presence of a d-amino acid in peptides r11a, r11b and r11c strongly suggest that other peptides belonging to the first group in the table undergo post-translational isomerization at the third amino acid from the C-terminus; these include clones Fi11.1 from a worm-hunting Conus species and M11.1 and S11.2 from fish-hunting Conus species.

Table 3.  Amino-acid sequences of I-superfamily peptides determined from cDNA clones. Peptides with the following precursor consensus sequence: MKLCVTFLLVLMILPSVTG/EKSSERTLSGALLRGVKRR; defined as the I1 superfamily. Peptides with the following precursor consensus sequence: MMFRVTSVGCFLLVIVFLNLVVLTDA; defined as the I2 superfamily. O, 4-trans-hydroxyproline; F, d-phenylalanine; L, d-leucine; ^, C-terminal free acid. γ, γ-carboxyglutamate; *, C-terminal amidation. The underlined parts of sequences are presumed to be post-translationally cleaved by carboxypeptidase.
Conus speciesClone/peptide SequenceReference
Group I
 A (presence of d-amino acid either demonstrated or strongly predicted)
  C. radiatus
 
R11.6/r11aGPSFCKADEKPCEYHADCCNCCLSGICAPSTNWILPGCSTSSFFKI
GOSFCKADEKOCEYHADCCNCCLSGICAOSTNWILPGCSTSSFFKI^
[8]
  C. radiatusR11.14/r11bGPSFCKANGKPCSYHADCCNCCLSGICKPSTNVILPGCSTSSFFRI
GOSFCKANGKOCSYHADCCNCCLSGICKOSTNVILPGCSTSSFFRI^
[8]
  C. radiatusR11.4/r11cGPSFCKADEKPCKYHADCCNCCLGGICKPSTSWI–GCSTNVFLTR
GOSFCKADEKOCKYHADCCNCCLGGICKOSTSWI–GCSTNVFLT^
[8]
  C. figulinusFi11.1aGHVSCGKDGRACDYHADCCNCCLGGICKPSTSWI–GCSTNVFLTRThis work
  C. magusM11.1aGAVPCGKDGRQCRNHADCCNCCPIGTCAPSTNWILPGCSTGQFMTRThis work
  C. striatusS11.2aGCKKDRKPCSYQADCCNCCPIGTCAPSTNWILPGCSTGPFMARThis work
 B (presence of d-amino acid unlikely)
  C. radiatusR11.3GPRCWVGRVHCTYHKDCCPSVCCFKGRCKPQSWGCWSGPT[8]
  C. betulinusBt11.1MCLSLGQRCERHSNCCGYLCCFYDKCVVTAIGCGHYThis work
  C. episcopatusEp11.1GDWGMCSGIGQGCGQDSNCCGDMCCYGQICAMTFAACGPThis work
Group II
 C
  C. betulinusBtX/BtXCRAEGTYCENDSQCCLNECCWGGCGHPCRHPGKRSKLQEFFRQR
CRAgGTYCgNDSQCCLNgCCWGGCGHOCRHP*
[10]
  C. virgoViTx/ViTxSRCFPPGIYCTSYLPCCWGICCSTCRNVCHLRIGKRATFQE
SRCFPPGIYCTSYLPCCWGICCSTCRNVCHLRIGK^
[9]
  C. emaciatusEm11.10CFPPGIYCTPYLPCCWGICCGTCRNVCHLRIGKRATFQEThis work
  C. figulinusFi11.11CHHEGLPCTSGDGCCGMECCGGVCSSHCGNGRRRQVPLKSFGQRThis work
  C. episcopatusEp11.12CLSEGSPCSMSGSCCHKSCCRSTCTFPCLIPGKRAKLREFFRQRThis work
  C. striolatusSx11.2CRAEGTYCENDSQCCLNECCWGGCGHPCRHPGKRSKLQEFFRQRThis work

A second group of I-conotoxin clones (Group B in Table 3) was highly homologous to Group A in their precursor sequences. However, all of these peptides have a slightly different consensus pattern –CX6CX5CCX3CCX4CX6C–, and the presence of a d-amino acid at the third position from the C-terminus is impossible or improbable. Note that either a Gly residue or a Cys involved in a disulfide bond is present at that locus.

Surprisingly, we have also identified I-conotoxin family clone sequences (Group C in Table 3) that have signal sequences completely unrelated to those of Groups A and B, and which lack the canonical propeptide region that has been a hallmark of all conotoxin precursors. This group includes cDNA clones of Em11.10, Fi11.11, Ep11.12 and Sx11.2 identified from Conus emaciatus, Conus figulinus, Conus episcopatus and Conus striolatus, respectively. All the mature peptides belonging to this group have the following consensus cysteine pattern: –CX6CX5CCX3CCX2−3CX3C–. Examples of complete precursor sequences from all three groups are shown in Table 4.

Table 4.  Complete precursor sequences of I-superfamily peptides presented in Table 3. Conus species: R, C. radiatus; Ep, C. episcopatus; Bt, C. betulinus; Sx, C. striolatus.
  1. a  For reference see [10]; all other complete precursor sequences came from the present report.

inline image

Discussion

The presence of a functionally important d-Phe residue in position 44 of the I-conotoxin superfamily peptide r11a was established previously [7]. In this work, we determined whether other I-conotoxin superfamily peptides have a d-amino acid at the homologous locus, specifically, r11b and r11c (which differ from r11a at 6 and 11 positions, respectively). Because of the considerable sequence similarity in the C-termini of r11a and r11b, we had predicted that r11b would have a d-Phe residue at the position homologous to d-Phe44 in r11a. In the case of r11c, however, seven out of 12 amino acids at the C-terminus of r11a are not conserved, including the amino acid to be modified (Phe in the case of r11a compared with Leu in the case of r11c). Furthermore, in the mature toxin purified from venom, the homologous Leu residue is the penultimate residue, and not three amino acids from the C-terminus, the previously postulated favored locus for modification. Thus, we could not confidently predict whether or not the homologous Leu residue would be post-translationally modified. The experimental data presented above unequivocally establish that both native r11b and r11c peptides purified from venom do indeed have a d-amino acid at the position homologous to d-Phe44 in r11a.

A functional comparison of these I-superfamily conotoxins was also carried out. It is clear that in the amphibian nerve–muscle preparation used, r11c has different molecular targeting specificity from r11a and r11b, as it elicits repetitive action potentials in directly stimulated muscle, whereas the other two peptides do not (Table 2). For r11a and r11b, converting the d-amino acid into the corresponding l isomer resulted in a complete loss of potency in eliciting repetitive action potentials in motor nerve. An additional noteworthy feature of r11c is that, unexpectedly, the [l-Leu42]r11c analog was just as active as the d-Leu-containing peptide.

The fact that the epimerized Leu residue is only two amino acids from the C-terminus of the mature toxin can be rationalized. The cDNA clone that encodes r11c, R11.4 (Table 3), has an additional amino acid at the C-terminus, an Arg residue; presumably, this was cleaved off by a carboxypeptidase. Thus, the discovery of a d-Leu residue in the sequence of r11c is consistent with the postulate that the third amino acid from the C-terminus is preferentially modified if the epimerization took place before the carboxypeptidase digestion.

Chemical synthesis has been reported for only one other I-conotoxin, the K channel inhibitor from Conus virgo, ViTx [9]. BtX, purified from the venom of Conus betulinus, has been characterized as a specific BK channel modulator [10]. These peptides were reported to be biologically active with no indication that a d-amino acid was present (for sequences, see Table 3). The second substantive conclusion from the data presented above is that, surprisingly, the peptides that we characterized in this work belong to a gene superfamily different from the one to which ViTx and BtX belong. As, in all previous studies, the Cys pattern of a conotoxin is strongly predictive of the gene superfamily to which the particular peptide belongs, the discovery that I-conotoxins really comprise two different gene superfamilies was unexpected. The diagnostic difference between the two superfamilies is that they have completely different signal sequences (Table 3); we will refer to these as the I1 superfamily (Group A and B in Table 3, including r11a, r11b and r11c) and the I2 superfamily (Group C in Table 3, including ViTx and BtX). A noteworthy feature of the I2 superfamily is that it lacks a propeptide region between signal sequence and mature toxin, differing from all other conotoxin superfamilies in this respect. As shown in Table 3, and as previously noted by Kauferstein et al. [11], another characteristic of the I2 superfamily is that a small region at the C-terminus is proteolytically excised.

Thus, all of the d-amino-acid-containing peptides characterized above belong to the I1 superfamily. Previous work on a post-translational modification of conotoxins, the γ-carboxylation of glutamate residues, has indicated that there may be a recognition signal sequence that serves as an anchor site for the modification enzyme [12]. Thus, one possible scenario that must be considered in light of the results presented above is that, as in the previous work on Conus peptide γ-carboxylation, the isomerase that acts on newly translated precursors of r11a, r11b and r11c might have a recognition sequence present in I1 family peptides; once anchored to the recognition sequence, the enzyme then acts on the preferred locus for modification, three amino acids from the C-terminus. However, the I2 superfamily may lack a recognition signal, and, consequently, the same pattern of conversion of l-amino acids into d-amino acids might not be operative for peptide precursors in that superfamily.

A key insight from this work, given the considerable divergence in sequence at the C-terminus, is that it appears that the enzymatic machinery that post-translationally modifies the l- amino acid to the d-amino acid does not stringently recognize a long vicinal sequence around the residue to be modified, nor is it specific for a particular amino-acid side chain in the residue epimerized. The data do not preclude the presence of a sequence-recognition motif at the C-terminal region; however, the considerable divergence in sequence clearly eliminates a majority of amino acids in the vicinity of the modified Phe or Leu residue as being essential for recognition.

It is somewhat exceptional to have in hand both the native peptide confirmed to have a d-amino acid, and synthetic or expressed versions with both l-amino acids and d-amino acids at the correct locus in order to predict when and where d-amino acids may be found post-translationally in polypeptide chains. The I-conotoxins are a large and diverse group, and they provide one of the rare opportunities for examining and defining the parameters that may be important for this most stealthy of all post-translational modifications.

Experimental procedures

Purification of conotoxins r11b and r11c from C. radiatus venom

Crude venom extract was prepared as described previously [13]. The extract was loaded on to a Vydac C18 semipreparative HPLC column and eluted at 5 mL·min−1 with a gradient of solvent A (0.1% trifluoroacetic acid) and solvent B90 (0.085% trifluoroacetic acid in 90% acetonitrile). Peptides r11b and r11c were further purified on a Vydac C18 analytical column at a flow rate of 1 mL·min−1, as previously [8], with the modifications for peptide r11b indicated in the legend to Fig. 1.

Mass spectrometry

MALDI mass spectra were obtained using a Bruker REFLEX time-of-flight mass spectrometer (Bruker Daltonics, Billerica, MA, USA) fitted with gridless reflectron, an N2 laser and a 100 MHz digitizer, courtesy of the Salk Institute for Biological Studies (La Jolla, CA, USA). Alternatively, MALDI mass spectra were obtained through the Mass Spectrometry and Proteomic Core Facility of the University of Utah.

Peptide synthesis

Peptides were synthesized on solid support by an automated peptide synthesizer using N-Fmoc-protected amino acids, 2-(2-N-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate and N,N-di-isopropylethylamine, courtesy of Dr Robert Schackmann of the DNA/Peptide Facility, University of Utah. All cysteine residues were trityl-protected. The coupling time was 1 h. Peptide cleavage/deprotection was accomplished with reagent K (82.5% trifluoroacetic acid/5% phenol/5% water/5% thioanisole/2.5% ethane-1,2-dithiol) for 2 h at room temperature. Soluble crude peptides were precipitated with cold methyl-t-butyl ether and centrifuged. The pellet was washed with methyl-t-butyl ether, centrifuged again, then dissolved in 25% acetonitrile in 0.1% trifluoroacetic acid and lyophilized. The linear peptides were purified on a Vydac C18 semipreparative HPLC column.

Oxidative folding

Oxidative folding of peptides was carried out with 0.1 m Tris/HCl, pH 8.7, containing 1 mm EDTA, 1 mm oxidized glutathione (GSSG) and 1 mm reduced glutathione (GSH) at room temperature for 3 h. The reaction was initiated by adding the linear synthetic peptide to a final concentration of 20 µm, then quenched by adding formic acid to a final concentration of 8%. The oxidized form was purified on a Vydac C18 HPLC column with a linear gradient 15–60% solvent B (0.1% trifluoroacetic acid in 90% acetonitrile) for 40 min.

Bioassay

Conotoxins were dissolved in normal saline solution. Swiss Webster mice (15 days old) were injected intracranially with 20 µL conotoxin solution using a 29-gauge insulin syringe. Control animals were similarly injected with normal saline solution only. Proper animal care and use protocols were followed in accordance with the guidelines set by the University of Utah Institutional Animal Care and Use Committee.

Electrophysiology

The cutaneus pectoris muscle preparation of Rana pipiens was used as previously described [8]. Briefly, the recording/stimulating chamber was made of Sylgard (Dow Corning, Midland, MI, USA), and consisted of a rectangular trough (≈ 1 mm × 15 mm × 4 mm deep) and an adjacent set of four abutting cylindrical wells (4 mm diameter × 4 mm deep), with the well furthest from the trough designated as well 1 and that closest as well 4. Each compartment was separated from the adjacent one by 1 mm. A mini-muscle preparation [14] was pinned in the trough, and the attached motor nerve was draped into the adjacent wells, with the proximal end of the nerve in well 1. Portions of the nerve overlying the partitions were covered with vaseline. A pair of electrodes in wells 1 and 2 were used to stimulate the motor nerve. Well 2 also contained a ground electrode. Pulses, 0.1 ms in duration, 6 V in amplitude and applied through a stimulation isolation unit, were used to evoke action potentials in the nerve at a frequency of 2 min−1. Activity of the nerve was recorded with a pair of electrodes located in wells 3 and 4. At the same time, the activity of the muscle was recorded with a pair of electrodes: one electrode was located in the middle, and the other at one end, of the trough. Each pair of recording electrodes was connected to a differential A/C amplifier, and the signal bandpass-filtered (1 Hz to 3 kHz) and digitized at a sampling frequency of 10 kHz. Home-made software programmed in labview (National Instruments, Austin, TX, USA) was used for data acquisition. All electrodes were 0.01 inch diameter stainless-steel wires. Solutions in all compartments were static and refreshed periodically by manual replacement. Peptides were dissolved in frog Ringer's containing 0.1 mg·mL−1 BSA. To apply peptide, the fluid in the muscle-containing trough was removed and replaced with the peptide solution. The process took about 1.5 min, and accounts for the gap in the records shown in Fig. 4. Recordings were performed at room temperature (≈ 21 °C).

In experiments involving direct stimulation of the muscle, instead of placing the muscle in the trough, the muscle was denervated and stretched over all four wells, with one end pinned in well 1 and the other end in well 4. Segments of muscle in wells 2 and 3 were kept submerged by a hold-down pin protruding horizontally from the wall of each well and overlying the muscle. Exposed portions of the muscle atop partitions were covered with vaseline, which minimized tissue desiccation and also served as a barrier that prevented fluid exchange between adjacent wells. Stimulating electrodes were located in wells 1 and 2, and a ground electrode in well 2. Recording electrodes were located in wells 3 and 4. Peptide was added by exchange of the fluid contents of well 4. To eliminate any contribution from synaptic activity, well 4 contained 10 µm d-tubocurare at all times to block postsynaptic acetylcholine receptors. Stimulation and recording parameters were as described in the previous paragraph.

Cloning and sequencing

Cloning and sequencing of cDNAs encoding I-conotoxins cDNAs were prepared by reverse transcription of RNAs isolated from Conus venom ducts as previously described [15]. cDNA from each species was used as a template for PCR with oligonucleotides corresponding to either the conserved 5′ UTR and 3′ UTR sequences of prepropeptides for peptides I1, or the conserved signal sequence and 3′ UTR sequence of prepropeptides for peptides I2. The resulting PCR products were purified using the High Pure PCR Product Purification Kit (Roche Diagnostics, Indianapolis, IN, USA) following the manufacturer's suggested protocol. The eluted DNA fragments were annealed to pAMP1 vector, and the resulting products transformed into competent DH5α cells, using the CloneAmp pAMP System for Rapid Cloning of Amplification Products (Life Technologies/Gibco BRL, Grand Island, NY, USA) following the manufacturer's suggested protocols. The nucleic acid sequences of the resulting toxin-encoding clones were determined according to the standard protocol for automated sequencing.

Acknowledgements

This work was supported by Program Project GM 48677 from the National Institutes of Health.

Ancillary