Topology of the disulfide bonds in the antiviral lectin scytovirin

Abstract

The antiviral lectin scytovirin (SVN) contains a total of five disulfide bonds in two structurally similar domains. Previous reports provided contradictory results on the disulfide pairing in each individual domain, and we have now re-examined the disulfide topology. N-terminal sequencing and mass spectrometry were used to analyze proteolytic fragments of native SVN obtained at acidic pH, yielding the assignment as Cys7–Cys55, Cys20–Cys32, Cys26–Cys38, Cys68–Cys80, and Cys74–Cys86. We also analyzed the N-terminal domain of SVN (SD1, residues 1–48) prepared by expression/oxidative folding of the recombinant protein and by chemical synthesis. The disulfide pairing in the chemically synthesized SD1 was forced into predetermined topologies: SD1A (Cys20–Cys26, Cys32–Cys38) or SD1B (Cys20–Cys32, Cys26–Cys38). The topology of native SVN was found to be in agreement with the SD1B and the one determined for the recombinant SD1 domain. Although the two synthetic forms of SD1 were distinct when subjected to chromatography, their antiviral properties were indistinguishable, having low nM activity against HIV. Tryptic fragments, the “cystine clusters” [Cys20–Cys32/Cys26–Cys38; SD1] and [Cys68–Cys80/Cys74–C-86; SD2], were found to undergo rapid disulfide interchange at pH 8. This interchange resulted in accumulation of artifactual fragments in alkaline pH digests that are structurally unrelated to the original topology, providing a rational explanation for the differences between the topology reported herein and the one reported earlier (Bokesh et al., Biochemistry 2003;42:2578–2584). Our observations emphasize the fact that proteins such as SVN, with disulfide bonds in close proximity, require considerable precautions when being fragmented for the purpose of disulfide assignment.

Introduction

Scytovirin (SVN) is a potent antiviral lectin isolated from the cyanobacterium Scytonema varium,1 reported to exhibit significant activity against human immunodeficiency virus (HIV).1, 2 Such potency makes SVN a promising candidate for development as a component of a female–Controlled, anti-HIV virucide. The structure of SVN has been studied by both NMR and X-ray crystallography.3, 4 For these investigations, SVN was either directly isolated from the cyanobacterium or produced as a recombinant protein in Escherichia coli.5 However, the initial mass spectrometry analysis and NMR structure of the recombinant SVN and crystallographic structures of both the natural and recombinant SVN exhibited significant differences, raising a possibility that this lectin might adopt different structures in solution and in the solid state.3, 4 In particular, the topology of the disulfide bonds differed between the crystal structure and the initial mass spectrometry data on the recombinant protein, although it was identical in the crystal structures of the natural and recombinant SVN.4

The primary structure of SVN consists of a single polypeptide chain containing 95 amino acids, with a highly conserved internal repeat. Both the NMR and crystallographic structures identified two clearly delineated domains, named SD1 (residues 1–48) and SD2 (residues 49–95). Each domain contains five cysteines, with Cys7 of SD1 and Cys55 of SD2 making an inter-domain disulfide bond. Both the NMR and crystal structures were in agreement about the identity of the pairing of these two cysteines. However, the two structures diverged in their assignments of the other disulfide bonds. The disulfides reported on the basis of mass spectrometry1, 5 and NMR3 indicated the pairing Cys20–Cys26 and Cys32–Cys38 in SD1, and Cys68–Cys74 and Cys80–Cys86 in SD2 [Fig. 1(A), Mode A], whereas in the crystal structures they were reported to be Cys20–Cys32 and Cys26–Cys38 in SD1 and Cys68–Cys80 and Cys74–Cys86 in SD2 [Fig. 1(A), Mode B]. In principle, a third mode of disulfide pairing is also possible [Fig. 1(A), Mode C], but was never reported.

Figure 1. Amino acid sequence, putative location of disulfides, and theoretical tryptic digests of scytovirin. (A) Amino acid sequence, with the two domains shown in separate lines. The three possible topologies of disulfide pairing are marked by thin lines. The positively charged residues that determine the location of tryptic cleavage of SVN are highlighted. Mode A: The topology corresponding to previous disulfide assignment by mass spectrometry and NMR.1, 3 Mode B: The topology consistent with the crystallographic results.4 Mode C: The remaining possible topology, not reported to date. (B) Peptides resulting from tryptic digestion of SVN. Peptide numbers (in parentheses) correspond to Tables I and II. Cystines are colored red and disulfide bonds are shown with lines. Fragments that are not involved in reshuffling at pH 8.0 are highlighted in gray. All other fragments are highlighted in green, blue, and yellow. The reshuffled pattern at pH 8.0 is shown in the bottom two panels. The theoretical monoisotopic masses are listed for each identified peptide.

Despite such differences in the results of crystallographic and NMR/mass spectrometry investigations, the biological properties of the proteins used in these experiments were the same. For example, simultaneous whole–Cell anti-HIV assays performed with the sample of SVN used for crystallography and with a recombinant sample prepared previously for the NMR experiments indicated that the anti-HIV activity of both samples was indistinguishable from each other, with both test samples providing EC₅₀ values of ∼10 nM (data not shown).

These discrepancies raised a number of important questions. The topology of disulfide pairing in the crystal structures was beyond any doubt, since it was based on highly refined atomic-resolution data and was confirmed by anomalous signal of sulfur atoms.4 This left two possibilities—either the existence of two forms of SVN with different disulfide topology, with different techniques somehow selecting only a single protein form, or the presence of errors in the determination of disulfide topology in both mass spectrometry and NMR. Since mass spectrometry is the usual way of assigning disulfide pairings in the absence of detailed structural data, finding an explanation for the possible failure of this technique might be of general interest. To reconcile these differences and establish whether the disulfide pattern in SVN and its single-domain constructs is unique, we reinvestigated the topology of the disulfides, using not only the recombinant full-length SVN and its SD1 domain, but also two synthetic versions of the latter protein with predetermined disulfide patterns.

Results

Proteins used in the experiments described here were either native, recombinant, or synthetic, but since we have previously shown that the structures of SVN isolated from the parent organism and of its recombinant equivalent are virtually identical,4 we are confident that there is no difference in the disulfide topology between the proteins from different sources, unless such changes were forced during synthesis. The amino acid sequence of SVN with three feasible disulfide pairings is shown in Figure 1(A). The topology of disulfides in SVN was originally assigned by matching the masses of its tryptic fragments obtained by LC-MS,1 as well as of fragments of expressed SD1.2, 5 Starting with assignment of species ionizing as m/z = 1318.6 (residues 30–43; C32–C38; SD1), m/z = 1553.6 (residues 79–95; C80–C86; SD2), and m/z = 3157.5 (residues [1–19] to [51–59]; C7-55), the remaining two species, termed here “cystine clusters” of m/z = 2511.0 and m/z = 2719.1, each containing two disulfides, C20–C26/C32–C38 (SD1) and C68–C74/C80–C86 (SD2), were assigned by process of elimination.

The experiments described below were based on a combination of N-terminal sequencing and mass spectrometric characterization of highly purified SVN fragments and led to unambiguous disulfide assignments. Reflecting upon previous experience with disulfide interchange of cystine-rich fragments at alkaline pH (for review see Ref. 6), all fragmentations and purifications of the peptide fragments leading to the assignments were performed at acidic pH, to minimize any possible artifacts that could result from sample treatment. In these experiments, we utilized several forms of the protein, including recombinant full-length SVN, recombinant SD1 domain, as well as two synthetic versions of the latter, in which the disulfide pairing was forced into either the SD1A or SD1B topology, respectively.

Limited proteolysis of SVN with pepsin

Cleavage of full-length SVN with pepsin split the native polypeptide into two major fragments. The larger fragment was identified by N-terminal sequencing and MS to be comprised of residues 1–58, and it contained the first six half–Cystines (C7, C20, C32, C26, C38, and C55). This fragment was isolated in two forms, a single–Chain polypeptide comprised of residues 1–58 (found m/z = 6133.5, Table I), and its hetero-dimeric product, residues [1–8] to [9–58], resulting from cleavage of the peptide bond Trp8-Asn9 (Table I). The smaller fragment, identified by N-terminal sequencing and MS to be comprised of residues 59–95, contained the remaining four half–Cystines paired in two disulfides (C68, C80, C74, C86; found m/z = 3592.4). This cleavage pattern is consistent with the three-dimensional structure of the protein as determined by both NMR and crystallography.

Table I. Molecular Characteristics of SVN Polypeptides, and of Peptic and Tryptic Peptides that Do Not Contain Cysteines

Sequence	Residues	Cystine bridge	Figure 2 Fraction No.	[M + H] Observed (monoisotopic)	[M + H] Theoretical (monoisotopic)	Confirmation method	Conditions
Scytovirin	1–95		11	9707.050	9707.038	MS	Full length
rSD1	1–48, C7S	C20–C32	N/A	4943.082	4943.076	Edman	Recombinant
		C26–C38
SD1A	1–48, C7S	C20–C26	N/A	4943.143	4943.076	Synthesis	Synthetic
		C32–C38
SD1B	1–48, C7S	C20–C32	N/A	4943.137	4943.076	Synthesis	Synthetic
		C26–C38
AKNPGGPNRCSNSKQ CDGARTCSSSGFCQGTAGHAAA	59–95	C68–C80	N/A	3592.409	3592.528	Edman	Pepsin, C-term, pH 2.5
		C74–C86
GSGPTYCWNEANNPGGP NRCSNNKQCDGARTCSSS GFCQGTSRKPDPGPKGPTYCWDE	1–58	C7–C55	N/A	6133.547	6133.528	Edman	Pepsin, N-term, pH 2.5
		C20–C32
		C26–C38
GSGPTYCW NEANNPGGPNRCS NNKQCDGARTCSSSGFCQG TSRKPDPGPKGPTYCWDE	1–8	C7–C55	N/A	6151.589	6151.539	Edman	Pepsin, N-term, pH 2.5
	9–58	C20–C32
		C26–C38
GSGPTYCWNEANNPGGPNR	1–19	Reduced	N/A	1990.810	1990.846	MS	3157, TCEP
GPTYCWDEAK	51–60	Reduced	N/A	1169.477	1169.493	MS	3157, TCEP

Assignment of the C7–C55 disulfide bond

A tryptic peptide map of SVN obtained at pH 6 is shown in Figure 2(A) and is graphically summarized in Figure 1(B). As can be seen from fragment assignments [Fig. 2(A), Table II], a heterodimeric peptide, residues [1–19]-[51–60], containing C7–C55, has been identified (fragment 10). Its Edman degradation yielded two N-terminal sequences, including di-PTH–Cystine in Cycle 7, confirming the presence of a disulfide bridge. Furthermore, reductive cleavage of this fragment with TCEP yielded two expected peptides, residues 1–19; m/z = 1990.8 and residues 51–67; m/z = 1169.5, providing additional supporting evidence for the C7–C55-linked heterodimer. Since pepsin split the protein into two domains, and since the disulfide C7–C55 was identified, in agreement with Ref. 1, it follows that the remaining eight half–Cystines were confined to two cystine “clusters” within each domain, where they could be connected in three different modes [Fig. 1(A)].

Figure 2. Peptide maps of SVN. (A) RP-HPLC peptide map of SVN at pH 6. The peptides absorbing at 214 nm were manually collected and analyzed by MS and N-terminally sequenced (see Material and Methods for conditions). The structures of the peptides in each fraction are indicated (see Table I for more details). Cleaved peptide bonds are spaced. (B) RP-HPLC peptide map of SVN at pH 8 obtained under chromatographic conditions optimized to separate the newly emerging species of m/z = 1318.6 (fraction 3) and m/z = 1553.6 (fraction 7). Note that tryptic cleavage at pH 6 was much slower compared to pH 8 and that the pH 6 peptide map represents the ∼50% time point of total SVN cleavage. Fraction 6 contains partially separated QCD derivative of a fragment isolated in fraction 4. Heterodimers of m/z = 2511.0 and m/z = 2719.1 were present as minor components in the chromatogram. Insets indicate the full scale of absorbance. Peptide numbers correspond to Figure 1(B) and Tables I and II.

Table II. Molecular Characteristics of Cysteine–Containing Peptides of SVN Obtained by Digestion at pH 6 and pH 8

Sequence	Residues	Cystine bridge	Figure 2 Fraction No.	[M + H] Observed (monoisotopic)	[M + H] Theoretical (monoisotopic)	Confirmation method	Conditions
SD1
GSGPTYCWNEANNPG	1–19	C7–C55	10	3157.460	3157.317	Edman	Trypsin, pH 6, pH 8
GPNRGPTYCWDEAK	51–60
NPGGPNR	61–67	N/A	1	711.305	711.353	MS/MS	Trypsin, pH 6, pH 8
KPDPGPK	44–50	N/A	2	738.350	738.414	M/MS	Trypsin, pH 6, pH 8
CSNNK	20–24	C20–C32	4	2529.039	2528.997	Edman	Trypsin, pH 6, pH 8
TCSSSGFCQGTSR	31–43	C26–C38
QCDGAR	25–30
CSNNK	20–24	C20–C32	N/A	1103.434	1103.409	Edman	Thermolysin of 2529, pH 6
TCSSSG	31–36
FCQGTSR	37–43	C26–C38	N/A	1444.656	1444.605	Edman	Thermolysin of 2529, pH 6
QCDGAR	25–30
CSNNK	20–24	C20–C32	6	2244.841	2244.837	Edman	Trypsin, pH 6, pH 8
TCSSSGFCQGTSR	31–43	C26–C38
QCD	25–27
CSNNK	20–24	C20–C32	N/A	2227.809	2227.837	Edman	Trypsin, pH 6, pH 8
TCSSSGFCQGTSR	31–43	C26–C38
pECD	25–27
TCSSSGFCQGTSR	31–43	C32–C38	3	1318.589	1318.517	Edman	Trypsin, pH 6, pH 8
CSNNKQCDGAR	20–30	C20–C32	5	2511.006	2510.986	Edman	Trypsin, pH 6, pH 8
TCSSSGFCQGTSR	31–43	C26–C38
CSNNKQCDGAR	20–30	Reduced	N/A	1195.511	1195.494	MS	2511, TCEP
TCSSSGFCQGTSR	31–43	Reduced	N/A	1320.540	1320.530	MS	2511, TCEP
SD2
CSNSK	68–72	C68–C80	8	2737.106	2737.081	Edman	Trypsin, pH 6, pH 8
TCSSSGFCQGTAGHAAA	79–95	C74–C86
QCDGAR	73–78
CSNSK	68–72	C68–C80	N/A	1076.412	1076.398	Edman	Thermolysin of 2737, pH 6
TCSSSG	79–84
FCQGTAGHA	85–93	C74–C86	N/A	1537.693	1537.627	Edman	Thermolysin of 2737, pH 6
QCDGAR	73–78
TCSSSGFCQGTAGHAAA	79–95	C80–C86	7	1553.609	1553.611	Edman	Trypsin, pH 6, pH 8
CSNSKQCDGAR	68–78	C68–C80	9	2719.083	2719.071	Edman	Trypsin, pH 6, pH 8
TCSSSGFCQGTAGHAAA	79–95	C74–C86
CSNSKQCDGAR	68–78	Reduced	N/A	1168.533	1168.483	MS	2719, TCEP
TCSSSGFCQGTAGHAAA	79–95	Reduced	N/A	1555.641	1555.626	MS	2719, TCEP

Assignment of C20–C32/C26–C38 (SD1) and C68–C80/C74–C86 (SD2) in major cluster forms

As shown in Figure 2(A) and Table II, each cystine cluster was obtained in chromatographically distinct major and minor forms. The two major forms (m/z = 2529.0, fraction 4, and m/z = 2737.1, fraction 8) were identified as heterotrimeric peptides, each yielding three N-termini upon Edman degradation and corresponding to disulfide-linked fragments of [20–24]-[25–30]-[31–43] (SD1) and [68–72]-[73–78]-[79–95] (SD2). This finding is in agreement with the pairing modes B and C but is in a disagreement with the pairing mode A1 as this mode does not permit the existence of tryptic heterotrimers.

The Gly-Phe bonds in the major cystine clusters were readily cleaved by thermolysin at pH 6 and the resulting disulfide-linked heterodimers, residues [25–30]-[37–43] (m/z = 1444.6; SD1) and [73–78]-[85–93] (m/z = 1537.6; SD2), were isolated and their identities were confirmed by N-terminal sequencing. Hence, disulfide bridges C26–C38 and C74–C86 were unambiguously assigned in these species. The other two thermolytic sub-fragments, residues [20–24]-[31–36] and [68–72]-[79–84], that could not be retained on C18-packings due to their high polarity, were purified in solvents containing 0.1% HFBA (see Materials and Methods), ionizing as species of m/z = 1103.4 and m/z = 1076.4 (Table II). Sequencing of these sub-fragments confirmed the presence of disulfide-linked peptides (Table II), providing unambiguous assignment of the remaining two disulfide bridges as C20–C32 and C68–C80, respectively. The SVN disulfide pairing pattern for the major cluster forms was thus determined as C20–C32/C26–C38 and C68–C80/C74–C86.

Assignment of C20–C32, C26–C38, C68–C80, and C74–C86 in minor cluster forms

The minor cystine clusters, heterodimeric fragments of m/z = 2511.0 (fraction 5) and m/z = 2719.1 (fraction 9), that were isolated from the pH 6 digest [Fig. 2(A)], each yielded two N-terminal sequences assigning them as heterodimers, residues [20–30]-[31–43] and [68–78]-[79–95], respectively. Hence, the Lys24-Gln25 and Lys72-Gln73 peptide bonds remained intact in either heterodimer at pH 6. These findings are in disagreement with the previous report1 that was based on the process of elimination, but did not experimentally prove, that the two detected heterodimers of the same corresponding m/z were composed of residues [20–24]-[25–43] (SD1) and [68–72]-[73–95] (SD2), assuming that the K-Q bonds were cleaved, and that the Arg31-Thr32 and Arg78-Thr79 bonds remained intact (compatible with mode A). The reductive cleavage of the heterodimers with TCEP followed by MALDI-TOF MS yielded the expected reduced polypeptide species, confirming the presence of disulfide linkages in each (Table II). Finally, upon sequencing, the heterodimers yielded di-PTH–Cystine in cycles 2 and 8, proving that their cystine pairing is that of the mode B (C20–C32/C26–C38 and C68–C80/C74–C86), as determined for the major species, and not that of the other theoretically possible mode, mode C. Remarkably, the heterodimers proved completely resistant to sub-fragmentation (in Gly-Phe bonds) by thermolysin under the conditions that were sufficient for a complete cleavage of the heterotrimers (see above), implying folding differences between the two types of clusters.

Recombinant SD1 domain assumes natural SVN fold

We determined the disulfide pairing in oxidatively folded rSD1. Upon tryptic digestion at pH 6, rSD1 yielded a disulfide pattern identical to the one found for native SVN and SD1B. The m/z = 2529.0 species, as well as its thermolytic fragment, residues [25–30]-[37–43] (m/z = 1444.6), were sequenced. Under these conditions, the heterodimer of m/z = 2511.0 was not found in the digest. These findings support the view that this fold is the thermodynamically most favored motif of SVN. Several additional pieces of evidence support the assumption that rSD1 contains the same pattern of disulfides as synthetic SD1B. First, rSD1 co-eluted with synthetic SD1B on analytical reversed-phase HPLC while being chromatographically distinct from synthetic SD1A [Fig. 3(A)]. Second, a stronger retention of synthetic SD1A on HPLC compared with SD1B suggests that the conformation of the latter is more compact under denaturing conditions, consistent with their respective disulfide topologies. Third, rSD1 and synthetic SD1B showed nearly identical trypsin digestion profiles that significantly differed from the profile of synthetic SD1A [Fig. 3(B)]. Taken together, our results are not in agreement with the conclusions of the previous study2 that reported mode A of disulfide pairing for rSD1.

Figure 3. HPLC analysis of SVN and its fragments. (A) Synthetic scytovirin (SD1A: Cys20–Cys26/Cys32–Cys38 in blue; SD1B: Cys20–Cys32/Cys26–Cys38 in red) and recombinant rSD1 (in green) analyzed at 40°C by reversed-phase HPLC. (B) Trypsin-digested rSD1 (green), synthetic SD1A (blue), and synthetic SD1B (red) analyzed by LC-MS. The determined molecular masses (Da) of major fragments are shown, and the 4943.1 Da peak is intact SD1B. The y-axis value corresponds to percent total ion count relative to the most abundant peak. See Materials and Methods for chromatographic and mass spectrometric conditions.

Disulfide interchange within cystine clusters at alkaline pH

When SVN was digested at pH 8, a more complex pattern emerged [Fig. 2(B)] as compared to the pH 6 profile [Fig. 2(A)], the most notable difference being the lower abundance or the complete disappearance (depending on time of incubation) of the major cystine clusters and the appearance of two new species, m/z = 1318.6 (fraction 3) and m/z = 1553.6 (fraction 7). The heterodimeric species of m/z = 2511.0 and m/z = 2719.1 were only present in trace amounts in pH 8 digests, as confirmed by nanobore LC-ESI-MS (data not shown). The two new species were sequenced as residues [31–43] and [79–95], respectively, each containing a single intramolecular disulfide bridge (Table II). The native and recombinant SVN were indistinguishable with respect to the emergence of these new tryptic peptide species. This demonstrates that an extensive disulfide interchange took place within each cluster under alkaline conditions, leading to the accumulation of these new species. This phenomenon was confirmed using synthetic SD1 derivatives with predetermined pairing as a reference. Indeed, upon tryptic digestion at pH 6, SD1B yielded only a species of m/z = 2529.0, whereas at alkaline pH it yielded both species, m/z = 2529.0 and m/z = 1318.6. In contrast, SD1A yielded upon trypsin digestion only a species of m/z = 1318.6, at either pH 6 or 8. The same behavior was also observed for rSD1, which shares its disulfide pairing with SD1B, as can be seen in the chromatographic (Fig. 4) and MALDI-TOF MS (Fig. 5) profiles as the disappearance of the heterotrimeric species of m/z = 2529.0 and the corresponding accumulation of the single–Chain fragment, species of m/z = 1318.6, residues 31–43. It should be noted that the other two short peptide fragments, residues 20–24 and 25–30, expected to be released upon dissociation of the heterotrimeric cluster of m/z = 2529, were not recovered under standard HPLC conditions, most likely because of their very polar character. To ascertain that the interchange was independent of other components of the digestion mixtures, HPLC-purified heterotrimeric clusters were exposed to alkaline pH and their conversion was monitored by MALDI-TOF MS (Fig. 5). It can be seen in Figure 5 that both purified heterotrimers dissociated into products with a similar half-life of ∼6–7 h. Next, we compared the conversion rates for HPLC-purified heterodimeric and heterotrimeric clusters isolated from both domains. As demonstrated in Figure 6, the conversion of heterotrimers was much faster than that of heterodimers. These experiments demonstrate that after an “overnight” (∼16 h) digestion, only about 10% of heterotrimers remained, as compared to heterodimers that could be mostly recovered (∼70–90%). This is likely so because heterotrimers possess a higher degree of internal freedom as compared to heterodimers. It is perhaps not surprising that heterotrimers were not previously reported in the pH 8 tryptic digests of SVN.1, 5 To ascertain whether disulfide interchange could also be induced within the clusters in their native fold, rSD1 was incubated at pH 8 (3 h and 18 h; 37°C) or at pH 10 (3 h; 37°C). The alkaline pH-treated rSD1 was then subjected to tryptic mapping at pH 6, essentially yielding a “pH 6 profile” characterized by the sole presence of species of m/z = 2529.0 (data not shown). Therefore, no evidence was obtained for the interchange in the folded domain, supporting the view that the reaction is accelerated by a collapse of the native fold upon proteolysis at alkaline pH.

Figure 4. Monitoring of disulfide interchange of heterotrimeric cluster of m/z = 2529.0 by RP-HPLC peptide mapping. (A) Time course of tryptic digestion of rSD1 at pH 8 documenting appearance of a new species of m/z = 1318.6. Asterisk designated peak appeared with time but did not yield any peptide signal on MS. (B) pH dependence of tryptic digestion (3 h, 37°C) at pH 6, pH 8, and pH 10. Heterodimer of m/z = 2511.0 was not present in the chromatograms.

Figure 5. Conversion of purified heterotrimeric tryptic peptide clusters of SVN monitored by MALDI-TOF MS. Purified peptides were incubated at pH 8 and aliquots of the reaction mixture were desalted and analyzed (see Materials and Methods for details). (A) Representative mass spectra of purified fragment of m/z = 2737.1 at τ = 0 and τ = 16 h at pH 8. The m/z = 2720.105 species is the pGu derivative of the m/z = 2737.136 species. (B) Stability plot for m/z = 2529.0. (C) Stability plot for m/z = 2737.1.

Figure 6. Stability plot for purified heterodimeric and heterotrimeric species obtained from SVN domain 1 (A) and SVN domain 2 (B). The conditions were the same as described in legend for Figure 5.

Discussion

Proper formation of disulfides during nascent polypeptide folding often critically affects the biological activity of proteins. Misconnected disulfides are enzymatically repairable in vivo via a process of thiol-disulfide interchange that involves nucleophilic scission of the aberrant sulfur–sulfur bond by an enzyme thiolate.7 However, protein disulfide interchange also proceeds non-enzymatically under close to physiological conditions with the participation of extraneous (e.g., low-MW) thiolate, via thiyl radical mechanism, or with direct participation of the hydroxide ion (see e.g., Refs.8 and9). At alkaline pH, cystine beta-elimination, and desulfurization with subsequent formation of alternative cross-links (such as the lanthionine thioether links) are common8, 10, 11 and observable by modern mass spectrometry techniques.12, 13 However, unlike cystine beta-elimination/desulfurization, cystine interchange is more difficult to trace by current analytical approaches and has long been recognized as the confounding factor in protein structural analyses; if not controlled or minimized, artifactual assignments may ensue.6

The presence of different species of the same protein with different topology of disulfide bridges has been reported only very rarely, and in most cases, the putative differences were later found to be artifacts of experimental procedures. Examples include the α–Chain of murine CD8, where the presence of a non–Canonical disulfide bond between Cys26 and Cys36 was reported on the basis of limited proteolysis data followed by analysis of amino acid sequences of the resulting peptides,14 only to be disproved by the crystal structure of human CD815 in which only the expected disulfide bond between Cys22 and Cys94 was visible. An example of a discrepancy in the assignment of disulfide bonds by NMR, crystallography, and biochemical methods is provided by the structure of the somatomedin B domain of vitronectin. NMR structure of this small domain indicated the presence of four disulfide bonds connecting residues 5–9, 19–21, 25–31, and 32–39,16, 17 whereas enzymatic digestion followed by mass spectrometry identified the pairings 5–9, 19–31, 21–32, and 25–39. By contrast, the 2.3-Å crystal structure of the complex of this domain with plasminogen activator inhibitor-1 indicated the connectivity 5–21, 9–39, 19–32, and 25–31.18 The controversy was finally resolved by analysis of synthetic peptides with predetermined disulfide connectivity.19

The propensity to interchange disulfides seems to be of a particular concern within Cys-rich small proteins, domains, or fragments. In such cases, multiple conformations having different cystine topologies but similar energies may exist due to limited hydrophobic packing.16 The extent to which a particular disulfide is prone to scrambling would therefore be expected to substantially vary. This was exemplified by preferential formation of disulfide isomers of α-lactalbumin and lysozyme in the presence of denaturants20, 21 and the existence of a scrambled form of sillucin, a cysteine-rich peptide.22 In this study, we have re-examined disulfide connectivity of SVN based on isolation, N-terminal sequencing, and accurate mass determination of its purified fragments obtained under acidic conditions known to suppress the interchange. We have demonstrated that at pH 8–10 the native fold SVN disulfides were fairly resistant to the interchange or other types of Cys modification. At the fragment level, we found that the inter-domain disulfide, Cys7–Cys55, was also comparatively very stable. However, we found that the four–cysteine clusters from both SVN domains readily exchanged and dissociated into fragments on the time scale of a typical alkaline pH tryptic digestion experiment. Additionally, we demonstrated using highly purified material that the heterotrimers interchanged significantly faster as compared to their corresponding heterodimers. This differential in susceptibility of closely related species to the interchange could be rationalized by differences in their packing, the heterodimers conceivably retaining some of the SVN fold, thus having their half cystines less accessible to either β-elimination or direct sulfur–sulfur scission by OH⁻.8, 9, 11 This view is supported by our finding that heterodimers also proved resistant to thermolysin under conditions that led to complete cleavage of heterotrimers (see Results). Thus we surmise that under the previously utilized conditions,1 the heterodimers largely remained intact while the less stable heterotrimers completely vanished, leading to accumulation of the artifactual fragments at alkaline pH (m/z = 1318.6 and m/z = 1553.6) which were used for the assignments. Their accumulation is likely driven by the formation of an intramolecular disulfide bond within each peptide [Fig. 1(B)]. While we have not analyzed in greater detail the fate of other fragments constituting the parent cystine clusters, work of others,6 most recently exemplified on conversions of single–Cystine containing peptides,12, 13 points out to the complex phenomena.

The results presented here resonate with those recently published12, 13, 23, 24 and demonstrate that sample preparation steps adopted in the “classical” protein chemistry approaches, that include protein fragmentation combined with mass spectrometry and/or Edman sequencing, can lead to erroneous conclusions concerning protein structure and that detailed characterization of each step is necessary to assure that the results reflect the true protein structure. They may also help in clarifying some of the reasons for the difference between the previously reported NMR and crystal structures of SVN.3, 4 Though the elucidation of the solution structure of SVN by NMR was not dependent on the previously-published disulfide bond topology, it is clear that solution structure identified a different form of SVN than the one described here or in the crystal structure. The reason for these differences may now be easier to discern. This is particularly true for a small, disulfide-bridged protein, as NMR calculations, due to limited hydrophobic packing in the small protein, can easily result in multiple conformations with different disulfide topologies but similar energies.16

Simultaneous presence of two distinct disulfide-bonded species of the same cysteine-rich peptide was recently shown for a mutant of a cysteine-rich domain of cnidarian nematocyst protein.25 Although only a single structure of the native peptide was determined by NMR, two species were simultaneously present in a mutant. However, despite the fact that the thermodynamic stability of the two forms was not the same, there was no indication of their interconversion, unless forced by thiol-disulfide interchange in the presence of oxidized and reduced glutathione.25 In view of the results presented here, it is highly unlikely that SVN exists in nature in two forms with differing disulfide patterns. The fact that both synthetically determined disulfide bond topologies render biologically active molecules (SD1A and SD1B) is an unusual finding that may allow further refinement of the minimal functional unit of the antiviral protein SVN. Additional structural studies on these domains bound to SVN's target oligosaccharides would be integral to that effort.

Acm, acetamidomethyl; ATZ, 2-anilino-5-thiazolinone; AUFS, absorbance unit full scale; Boc-Gly-OCH₂-PAM, t-butyloxycarbonyl-glycyl-4-(hydroxymethyl)phenylacetamidomethyl–copoly(styrene/1% divinylbenzene) resin; DIEA, diisopropylethylamine; ESI-Q-TOF, electrospray ionization quadrupole-time-of-flight mass spectrometry; HBTU, N-[(1H-benzotriazol-1-yl)(dimethylamino) methylene]-N-methyl-methanaminium hexafluorophosphate N-oxide; HFBA, heptafluorobutyric acid; MALDI-TOF, matrix-assisted laser desorption-ionization time-of-flight; MS/MS, tandem mass spectrometry; m/z, mass-to-charge ratio; pE, pyroglutamyl; PTH, phenylthiohydantoin; RP-HPLC, reversed-phase high pressure liquid chromatography; rSD1, recombinant/oxidatively folded N-terminal domain of SVN (residues 1–48Cys7Ser); R.T., retention time on HPLC; SD1A, synthetic N-terminal domain of SVN [residues 1–48Cys7Ser (Cys20–Cys26, Cys32–Cys38)]; SD1B, synthetic N-terminal domain of SVN [residues 1–48Cys7Ser (Cys20–Cys32, Cys26–Cys38)]; TCEP, tris(2–Carboxyethyl)phosphine; TFA, trifluoroacetic acid.

Materials and Methods

Preparation of proteins

Native SVN used in this study was isolated and characterized as previously described.1 Recombinant SVN, as well as the recombinant SD1 domain (rSD1; residues 1–48), were expressed in E. coli and purified as described previously.2, 5 Briefly, rSD1 was expressed as fusion protein fused to TRX Tag thioredoxin to improve the disulfide bond formation in the recombinant products. The clarified E. coli lysate was then purified by metal chelation chromatography on a Talon resin and eluted by the addition of 150 mM imidazole. Eluted rSD1-Trx-His fractions were pooled and dialyzed into 50 mM Tris-HCl, pH 7.5, and concentrated to a final concentration of ∼5 mg/mL. The fusion protein was then digested by recombinant enterokinase (rEK, 1 unit/0.1 mg protein) overnight at 25°C. The final purification was achieved by C-18 RP-HPLC, eluting bound protein using a 0.05% TFA aqueous acetonitrile gradient from 15% to 60% acetonitrile; rSD1 eluted at ∼25% acetonitrile. To avoid acid hydrolysis of rSD1 at the labile Asp46-Pro47 peptide bond, fractions were collected into tubes containing sufficient Tris-HCl, pH 8.0, to maintain pH > 7.0. The eluant was then collected and pooled, and the acetonitrile was removed by rotary evaporation. The aqueous protein was lyophilized to dryness prior to reconstitution for both crystallization and mass spectroscopy experiments.

Two variants of the SD1 domain with Cys7 mutated to a serine were prepared by total chemical synthesis. One of these constructs (SD1A) had the four remaining cysteines forced into disulfide configurations Cys20–Cys26 and Cys32–Cys38 [Fig. 1(A)], whereas the other one (SD1B) included disulfides Cys20–Cys32 and Cys26–Cys38 [Fig. 1(B)]. Both constructs were synthesized on Boc-Gly-OCH₂-PAM resin using the HBTU activation/DIEA in situ neutralization protocol developed by Kent and colleagues for Boc chemistry.26 After stepwise chain assembly and HF cleavage/deprotection, crude peptides were purified by preparative RP-HPLC and their masses ascertained by ESI MS. For the synthesis of the SD1A peptide, Cys20 and Cys26 were orthogonally protected by Acm groups, whereas Cys20 and Cys32 were Acm-protected for the synthesis of the SD1B peptide. Formation of the first disulfide bridge between two unprotected Cys residues of SD1 was achieved by dissolving the peptide at 0.2 mg/mL in 0.5 M NaHCO₃ solution, followed by overnight stirring in an open-air container. Acm deprotection and simultaneous formation of the second disulfide bridge were done in acidic aqueous solution in the presence of iodine according to the previously published procedures19, 27. The final products were purified to homogeneity, and their molecular masses determined to be [M + H] = 4943.1 ± 0.1 Da, in agreement with the theoretical value of [M + H] = 4943.076 Da calculated on the basis of the monoisotopic compositions of oxidatively folded SD1 (see Table I).

Synthetic SD1A and SD1B as well as recombinant rSD1 peptides were analyzed on a Waters XBridge C18 column (4.6 × 150 mm, 3.5 μm) running a linear gradient of 5–65% B (B: acetonitrile + 0.1% TFA; A: water + 0.1% TFA) at a flow rate of 1 mL/min for 30 min [Fig. 3(A)]. For trypsin digestion, the peptides at 1 mg/mL in 50 mM Tris buffer containing 20 mM CaCl₂, pH 8.3, were each incubated at 37°C, overnight, with 10 μg/mL bovine trypsin before LC-MS analysis (Thermo Scientific LXQ linear ion trap mass spectrometer equipped with an Accela high speed LC). The digestion samples were injected onto a Thermo Scientific Hypersil GOLD column (2.1 mm × 50 mm, 1.9 μm) running a linear gradient of 0–25% B (B: acetonitrile + 0.1% formic acid; A: water + 0.1% formic acid) at 40°C at a flow rate of 250 μL/min [Fig. 3(B)].

Mass spectrometry analysis

The masses of the native protein and of its peptide fragments were determined by MALDI-TOF MS analysis using model UltraflexIII (Bruker Daltonics) under our standard conditions.28 Electrospray ionization mass spectrometric analysis was performed in a Bruker Daltonics model maXis ESI-Q-TOF instrument interfaced either by flow-injection analysis using a syringe pump or LC-MS/MS using a Dionex model U3000 HPLC configured for nL/min flows. Flow-injection analysis was performed using a syringe pump flowing at 5 μL/min into a software controlled injection valve fitted with a 5 μL sample loop. The syringe pump delivered the sample from the sample loop into the maXis equipped with the standard Bruker Daltonics ESI source. The source was operated at a spray voltage of 3800 V with a sheath gas of nitrogen at a pressure of 0.4 bar and a drying gas of nitrogen flowing at 4 L/min. The capillary temperature was set to 150°C. The mass spectrometer was set to acquire line spectra of m/z 50 to 2200. MS data was acquired at a scan speed of either 5 or 1 Hz. The Dionex U3000 nanobore HPLC was configured with dual ternary pumps with one pump's flow output split using a calibrated 1:1000 splitter with active flow control. The system used a pulled-loop autosampler configured with a 20 μL sample loop. A desalting trap column (0.3 × 5 mm, 5 μm, C18 PepMap, 120 Å, Dionex) was used and the analytical column used was a C18 PepMap (0.075 mm × 150 mm, 3 μm, 120 Å, Dionex). The solvents used were 0.1% formic acid in water (solvent A) and 80% acetonitrile/0.1% formic acid (solvent B). The gradient was either 2–55% B in 30 min or 2–55% B in 90 min. The effluent from the analytical column was introduced into the maXis using the Bruker Daltonics on-line nanospray source. The source was operated at a spray voltage of 2800 V with a drying gas of nitrogen flowing at 4 L/min. The capillary temperature was set to 130°C. The mass spectrometer was set to acquire line spectra of m/z 50 to 1600 or m/z 50 to 2200.

RP-HPLC purification of SVN, rSD1, and synthetic SD1A and SD1B

Prior to protease digestion, native SVN or purified/oxidatively folded recombinant SVN, synthetic 1–48 fragments of SVN in which Cys7 was replaced with Ser7 (SD1A and SD1B), as well as recombinant SD1, were re–Chromatographed by reversed-phase microbore HPLC (Zorbax SB–C18 silica, 5 μm, 300 Å, 1 mm × 150 mm) using a gradient of acetonitrile in aqueous 0.1% TFA as a counter ion at flow rates of 90 μL/min generated using an Applied Biosystems model 140B/783A HPLC.28 The polypeptides yielded major single peaks eluting at 25–30% of acetonitrile and their masses were confirmed by MALDI-TOF MS or ESI-MS (see Table I and Results section).

Fragmentation of scytovirin

Purified SVN polypeptides were concentrated to near dryness in a SpeedVac rotary evaporator and reconstituted in digestion buffers of selected acidity. Tryptic digestion was performed at 37°C in the presence of 20 mM ammonium acetate, pH 6, or in 10 mM ammonium hydrocarbonate, pH 8–10, using Promega sequencing grade trypsin (E:S = 1:50 or 1:100) for variable periods of time depending on a particular application. Peptic digestion using crystalline pepsin A (Sigma) was performed in 0.01% aqueous TFA or in 1 mM HCl (E:S = 1:200; 2 h, 37°C). Digestion with thermolysin (Boehringer-Mannheim) was performed in 10 mM ammonium hydrocarbonate, 5 mM in CaCl₂, adjusted to pH 6 using diluted TFA (E:S = 1:50; 37°C, 30 min). Aliquots of the reaction mixture were screened by MALDI-TOF MS and microbore RP-HPLC (using either 0.1% TFA or 0.1% HFBA as counter ions) to determine the optimum time for preparative fragment isolation.

Microbore RP-HPLC of SVN digests

Peptic, tryptic, or thermolytic fragments of native or recombinant SVN, synthetic SD1A and SD1B, and rSD1 were fractionated by microbore RP-HPLC on Zorbax SB–C18 microbore columns (0.5 × 150 mm or 1 × 150 mm; 5 μm, 300 Å) equilibrated at 25°C in 0.1% aqueous TFA and eluted using linear gradients of acetonitrile.28, 29 For pH 6 SVN digest, gradient slope of 0.6%/min of acetonitrile was used,28, 29 whereas gradient slope of 0.5%/min was used for a more complex, pH 8 digest to improve separation of the peptides. The peptides absorbing at 214 nm were manually collected and subjected to MALDI-TOF MS or to flow injection analysis on ESI-Q-TOF MS, to determine their monoisotopic masses, and to Edman degradation to determine their N-terminal sequences. Two highly polar thermolytic peptides (obtained from sub-digestion of purified tryptic cystine clusters) did not sufficiently retain on C18 reversed-phase and were purified by RP-HPLC in solvents containing 0.1% HFBA as the counter ion in place of TFA.

Edman degradation

Edman degradation was used to determine the N-terminal sequences of purified peptides using model cLC-Procise sequencer (Applied Biosystems) as described.28, 29 The cystine-linked heterotrimers or heterodimers yielded three or two PTH-residues, respectively, in each cycle. The positions of the cystine residues linking the peptides were determined based on the known chromatographic behavior of di-PTH–Cystine and/or its degradation products. A characteristic “fingerprint” of PTH derivatives is obtained in the cycle if cystine is cleaved off/extracted in the form of its di-ATZ derivative within the first 10 cycles of degradation. Namely, a peak co-eluting with PTH-Tyr (RT: ∼13.3 min), a unique peak eluting at RT: ∼14.2 min characteristic of the presence of cystine, and traces of PTH-anhydroSer (RT: ∼12.30 min) and PTH-Ser (RT: ∼7.30 min) that most likely result from beta-elimination/rehydration of di-PTH–Cystine during Edman degradation.29–31 To increase the UV (λ = 269 nm) signal yield of di-PTH–Cystine, the sequencer reagent, R4A, 25% TFA/water, that contains 0.01% DTT as supplied by the manufacturer, was replaced by 25%, v/v, HPLC-grade TFA/water to eliminate DTT from the system.30

Alkaline stability of disulfide-linked peptides

SVN tryptic peptides of m/z = 2511.0, m/z = 2529.0, m/z = 2719.1, and m/z = 2737.1, isolated from pH 6 tryptic digests, were incubated at 37°C at pH 6 or pH 8 for various periods of time. Aliquots of the reaction mixture were acidified with 0.1% TFA and analyzed by RP-HPLC mapping and MALDI-TOF MS to monitor the disappearance of the original species and the appearance of new species of lower m/z, presumably taking place due to extensive disulfide interchange. To monitor disulfide interchange induced within the folded SD domain, purified rSD1 was exposed to 50 mM Tris-HCl, pH 8, for 18 h at 37°C. Following alkaline pH treatment, the protein was desalted by RP-HPLC and the tryptic peptides obtained at pH 6 were analyzed by RP-HPLC and MALDI-TOF MS as described above.