Tài liệu Báo cáo khoa học: Insights into and speculations about snake venom metalloproteinase (SVMP) synthesis, folding and disulfide bond formation and their contribution to venom complexity pdf

REVIEW ARTICLE
Insights into and speculations about snake venom
metalloproteinase (SVMP) synthesis, folding and disulfide
bond formation and their contribution to venom
complexity
Jay W. Fox
1
and Solange M. T. Serrano
2
1 Department of Microbiology, University of Virginia, Charlottesville, VA, USA
2 Laborato
´
rio Especial de Toxinologia Aplicada-CAT ⁄ CEPID, Instituto Butantan, Sao Paulo, Brazil
Introduction
Since the first discovery of zinc-dependent proteinases
in viperid snake venom, investigators have intensively
studied the structure and function of these proteinases
in order to understand their role in envenomation
pathologies [1]. With the advent of the first complete
sequence determination of these proteinases, it was
thought that they belonged to the matrix metallo-
proteinase family of proteinases [2]. However, it soon
became obvious that they in fact comprised a novel
family of metalloproteinases, the M12 family, to which
the ‘a disintegrin and metalloproteinase’ (ADAM)
proteins also belong [3]. As studies progressed, the
snake venom metalloproteinases (SVMPs), as this
group of proteinases is now named, were further cate-
gorized into the PI, PIIa and PIIb, PIIIa and PIIIb,
and PIV classes [4,5]. The criterion for this differential
classification essentially was based on the presence or

absence of various nonproteinase domains as observed
via mRNA transcripts and proteins isolated in the
venom. To date, no PIV mRNA transcript has been
observed, and thus it very well may be that the PIV
structure simply represents another post-translational
modification of the canonical PIII structure; therefore,
in our new classification scheme, we have collapsed the
Keywords
autolysis; disintegrin; disulfide bond;
metalloproteinase; post-translational
processing; proteome snake venom;
structure; SVMP transcriptome
Correspondence
J. W. Fox, Department of Microbiology,
University of Virginia, PO Box 800734,
Charlottesville, VA 229080734, USA
Fax: +1 434 982 2514
Tel: +1 434 924 0050
E-mail:
(Received 4 February 2008, revised 27
March 2008, accepted 15 April 2008)
doi:10.1111/j.1742-4658.2008.06466.x
As more data are generated from proteome and transcriptome analyses of
snake venoms, we are gaining an appreciation of the complexity of the
venoms and, to some degree, the various sources of such complexity. How-
ever, our knowledge is still far from complete. The translation of genetic
information from the snake genome to the transcriptome and ultimately
the proteome is only beginning to be appreciated, and will require signifi-
cantly more investigation of the snake venom genomic structure prior to a
complete understanding of the genesis of venom composition. Venom com-

plexity, however, is derived not only from the venom genomic structure
but also from transcriptome generation and translation and, perhaps most
importantly, post-translation modification of the nascent venom proteome.
In this review, we examine the snake venom metalloproteinases, some of
the predominant components in viperid venoms, with regard to possible
synthesis and post-translational mechanisms that contribute to venom com-
plexity. The aim of this review is to highlight the state of our knowledge
on snake venom metalloproteinase post-translational processing and to
suggest testable hypotheses regarding the cellular mechanisms associated
with snake venom metalloproteinase complexity in venoms.
Abbreviations
ER, endoplasmic reticulum; MMP, matrix metalloproteinase.
3016 FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS
PIV class into the PIII class pending the finding of a
transcript in venom glands representing a PIV struc-
ture containing lectin-like domains. In Fig. 1, we show
a modified classification scheme that reflects the nas-
cent P structural classes and well as the observed
products following post-translational modification and
processing. Functionally, the SVMPs display a wide
array of biological activities, many of which are toxic,
and this scope of activities reflects the multitude of
products derived from the four SVMP classes [5].
Over the past several years, as the result of a variety
of proteomic and transcriptomic investigations of
snake venom and venom glands respectively, numerous
databases have been generated that illuminate the com-
plexity of snake venoms, particularly the viperid
venoms. Hundreds of proteins comprise viperid venom
proteomes, and it is estimated that most viperid

venoms are composed of at least 32% SVMPs, which
suggests the potential for a significant role for the
SVMPs in the pathologies associated with viperid
evenomation [6–8]. Given the apparent complexity of
venoms and the SVMP components in venoms, one
must question the molecular mechanisms responsible
for this complexity [9]. Superficially, one could simply
attribute venom complexity to genomic ⁄ transcriptomic
complexity, and to a certain degree this is the case.
However, over the past several years, there have been
features, both structural and functional, reported in
the literature, which suggest that there may be other
factors involved in generating the observed complexity
of the SVMPs in venom. In this short review, we will
discuss biosynthetic features of the SVMPs that we
believe may be involved in SVMP structural and func-
tional complexity, with the aim of generating renewed
interest in understanding the molecular and cellular
biology of the SVMPs. Perhaps, as in the past, this
review may also help guide us to the development of
enhanced therapeutics for viperid snake envenomation
and novel toxin-based drug discovery.
Venom protein biosynthesis
Snake venoms are the products of specialized secretory
glands located above the upper jawbone in venomous
snakes. These glands are considered to be specialized
secretory glands evolved for the biosynthesis of ven-
oms. Investigations on venom production in vivo have
demonstrated that, like most secretory proteins, venom
Fig. 1. Schematic of SVMP classes. Question marks (?) in the figure indicate that the processed product has not been identified in the

venom.
J. W. Fox and S. M. T. Serrano Snake venom metalloproteinases and venom complexity
FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS 3017
proteins are synthesized in the cytoplasm of secretory
cells in the gland and transferred to the rough endo-
plasmic reticulum, and then the Golgi apparatus, and
finally transported via secretory granules to the lumen
of the venom gland [10]. On the basis of cDNA studies,
all venom proteins have a signal sequence that probably
targets the nascent protein to a signal recognition parti-
cle on the endoplasmic reticulum (ER). During trans-
port into the ER, the signal sequence is removed.
Presumably, like most secretory proteins, it is in the
rough ER that the nascent venom proteins fold, and
undergo disulfide bond oxidation ⁄ formation, glycosyla-
tion, and, in the specialized instances of some venom
proteins, multimerization, such as is the case with
dimeric disintegrins [PIId and PIIe, dimeric PIII class
SVMPs (PIIIc)] and the multimeric PIIId class SVMPs
(formerly PIV). As in other eukaryotic cells, in the
venom gland secretory cells, incorrectly folded proteins
would not be expected to enter into the Golgi network.
Proteolytic processing of the proforms of venom pro-
teins probably occurs, as with most latent protein forms
in the trans-Golgi network and nascent secretory vesi-
cles, and is completed by the time that the vesicles are
released into the venom gland lumen. In only a few
instances have sequences associated with the prodomain
of SVMPs been detected in the venom [11] (Serrano,
unpublished results), suggesting that, in fact, the bulk

of the proteolytic processing of the proforms of the
SVMPs occurs in advance of release of mature secre-
tory vesicle contents into the lumen of the venom gland.
In the case of the SVMPs, activation may best occur in
an environment in which widespread, random proteo-
lysis of venom components is minimized. Studies have
suggested that the acidic pH of the venom gland lumen,
in addition to pyrol-glutamate containing tripeptides,
contributes to the lack of proteolytic activity of SVMPs
in the gland [12,13]. Likewise, the acidification of the
secretory vesicles as they mature may give rise to an
environment that could limit the proteolytic activity of
processed SVMPs. Therefore, the maturing secretory
vesicle is probably the best environment for SVMP acti-
vation, and could preclude wholesale degradation of
the venom by activated SVMPs. A hypothetical sche-
matic of venom SVMP biosynthesis is presented in
Fig. 2, which typifies the protein biosynthetic pathway
of most eukaryotic cells.
Fig. 2. Schematic representing hypothetical biosynthetic pathways from transcription at the ER surface, through the endoplasmic reticulum
to the Golgi, and release of the secretory vesicles into the venom lumen for the production of the three SVMP venom classes. Parentheses
in the figure indicate that the processed product has not been observed in the venom. P, prodomain; M, metalloproteinase domain; S,
spacer; D, disintegrin domain; DL, disintegrin-like domain; Cys, cysteine-rich domain; L, lectin-like.
Snake venom metalloproteinases and venom complexity J. W. Fox and S. M. T. Serrano
3018 FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS
Structural features of the SVMPs that
contribute to venom complexity
Role of disulfide bond patterns in post-transla-
tional proteolytic processing of the SVMPs
Venom proteins must maintain structural integrity in

the oxidative extracellular world; hence, most have
evolved to contain a number of disulfide bonds that
stabilize the particular molecular scaffold that is criti-
cal for toxic action [14] as well as participate in the
generation of multimeric venom proteins. On the basis
of studies of venoms using two-dimensional PAGE
under reducing and nonreducing conditions, Serrano
et al. [15] clearly demonstrated that disulfide bond
formation among venom proteins, leading to multimeric
structures, has a profound effect on venom complexity.
Reducing gels of Crotalus atrox and Bothrops jararaca
venom, as compared to nonreducing gels, indicated
that apparent venom complexity is reduced by 60%
due to disulfide bond crosslinking of venom proteins.
The biological synthesis of toxins, particularly the
SVMPs, must entail a somewhat complex phenomenon
that probably involves a variety of chaperones and
other proteins to help ensure appropriate folding and
disulfide bond pairing, as shown by the fact that heat-
shock proteins, protein disulfide isomerases and pept-
idyl-prolyl cis ⁄ trans isomerases have been identified in
snake venoms and venom glands [6,16]. The need for
ancillary proteins for appropriate post-translational
modifications is further substantiated by the fact that
it has proven to be relatively difficult for investigators
to produce recombinant SVMPs or, for that matter,
their various domains in heterologous in vitro
expression systems.
Cysteinyl residues in PI SVMPs range in number
from four, to six, to seven, with two to three disulfide

bonds being reported (Fig. 3) (see also table 2 of [5]).
However, the PI adamalysin contains five cysteinyl res-
idues, four of which are involved in disulfide bonds,
leaving one free cysteinyl residue in the N-terminal
region of the domain. Given the presence of a sulfhy-
dryl group in the protein, one would suspect that
disulfide bond shuffling could be promoted by the resi-
due; however, a consistent disulfide bond pattern is
observed for specific venom proteins. Most likely,
incorrectly folded proteins and hence unusual disulfide
bond patterns are removed from the post-translational
process. The crystal structures for eight PI SVMPs
have been reported, and although there are different
disulfide bond patterns observed among these PI
SVMPs, they do not appear to significantly affect the
crystal structures, in that all are very similar with
regard to secondary structure and folding [17–25].
Thus, at least superficially, it seems that similar folding
motifs and structures can give rise to different disulfide
bonding patterns in the PI SVMPs. Furthermore, there
does not appear to be a correlation between disulfide
bond pattern, folding and biological activity.
The PII SVMPs are distinguished by having a
canonical disintegrin domain present in the nascent
protein that, in most cases, is post-translationally, pro-
teolytically processed from the proteinase domain. It is
important to note that the proteolytic products, the
proteinase domain and the disintegrin domain, seem to
be stable, in that they have been isolated from the
crude venom [26,27]. Thus, this represents one form of

post-translational proteolytic processing that contrib-
utes to the SVMP-derived venom complexity, and will
be further discussed below. The proteinase domains
are observed to have five, six or seven cysteinyl resi-
dues, most having three disulfide bonds paired in a
similar way to that observed for the PI SVMPs. As
was the case with PI adamalysin, the unpaired, free
cysteinyl residue, such as in the PIIa trigramin, is
located in the N-terminal portion of the domain and
does not seem to affect folding or disulfide bond pair-
ing (Fig. 3). This may explain why the proteinase
domains, when processed from the nascent PII struc-
ture, are stable like the PI SVMPs. An interesting
exception to the typical PII SVMP processing scheme
that gives rise to a proteinase and a disintegrin in
venom is jerdonitin, a PIIb proteinase isolated from
the venom of Trimeresurus jerdonii. In jerdonitin, the
processing of these domains does not appear to occur
[28]. The structural analysis of jerdonitin reveals two
additional cysteinyl residues, one located in the spacer
region and one in the disintegrin domain, which could
promote a more compact structure that may preclude
the proteolytic processing between the two domains
(Fig. 1). Here again, we observe that the generation of
an additional SVMP structure based on variations in
post-translational proteolytic processing contributes to
SVMP-based venom complexity.
Some disintegrins, as processed from the nascent PII
SVMPs, contain four, six (four intrachain and two
interchain), six, seven or eight disulfide bonds [29]. The

predominance of disintegrins in venom in conjunction
with the oxidation of monomeric disintegrins to form
homodimeric and heterodimeric structures certainly
contributes to the complexity of SVMP-based venom
complexity. Another interesting observation associated
with the dimeric disintegrins is that it seems that many
are formed by one subunit derived from a processed
PII SVMP and another from a translated gene pro-
duct representing the disintegrin domain alone [30].
J. W. Fox and S. M. T. Serrano Snake venom metalloproteinases and venom complexity
FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS 3019
Fig. 3. Sequence alignment of SVMPs (PI, green; PII, blue; PIII, red; PIIId, brown) by the program CLUSTALW. Cysteine residues and putative
N-glycosylation sites are highlighted in gray and green, respectively. Disintegrin motifs (MGD; RGD; MVD; VGD; KGD) are shown in red. The
hypervariable region is highlighted in yellow. Cysteine residues are numbered according to the VAP1 sequence [32].
Snake venom metalloproteinases and venom complexity J. W. Fox and S. M. T. Serrano
3020 FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS
Fig. 3. Continued.
J. W. Fox and S. M. T. Serrano Snake venom metalloproteinases and venom complexity
FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS 3021
Fig. 3. Continued.
Snake venom metalloproteinases and venom complexity J. W. Fox and S. M. T. Serrano
3022 FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS
Presumably, the post-translational processing machin-
ery in snake venom gland secretory cells must be pre-
disposed to this form of multimerization as compared
to dimerization between either disintegrin gene prod-
ucts or between two disintegrins processed from the
PIId structure. As mentioned, disulfide bond formation
typically occurs prior to proteolytic processing in the
trans-Golgi or secretory vesicles. Thus, if disulfide

bond pairing to form a heterodimeric disintegrin
occurs, it presumably happens between the nascent dis-
integrin gene product and a PIIe SVMP. This suggests
some structural or mechanistic advantage to having
one disintegrin domain residing in the structural envi-
ronment of the unprocessed PIIe for the formation of
the heterodimer later in the post-translational process.
The PIII SVMPs are the most intriguing of the
SVMP categories in terms of their contribution to
venom complexity and function. One of the first
sequences determined for a PIII SVMP was atroly-
sin A from C. atrox. The mature protein was observed
to be composed of three domains: a metalloproteinase
domain, a disintegrin-like domain, and a cysteine-rich
domain [31]. The metalloproteinase domains of the
PIII SVMP class have either six or seven cysteinyl resi-
dues with three disulfide bonds. The disulfide bond
pattern of the PIII proteinase domains appears to be
different from that observed in the PI or PII protein-
ase domains (Fig. 4). Most of the PIIIs have an odd
(seventh) cysteinyl residue in the proteinase domain.
The positions of the seventh cysteinyl residues in the
PIIIs appear to segregate to either the region near the
cysteinyl cluster of residues 350, 352 and 357 or
between cysteinyl residues 374 and 390 (Fig. 4).
Recently, crystal structures for three PIII SVMPs
have been determined: VAP1, a dimeric PIIIc, catro-
collastatin ⁄ VAP2B, a PIIIb, both from the venom of
C. atrox, and RVV-X, a PIIId from Vipera russelli
[32–34]. Many important observations regarding PIII

structure and function were obtained from those
studies. The first is that although the disulfide bond
patterns are different in these PIIIs from that observed
in the PI SVMPs, the crystal structures of the metallo-
proteinase domains are very similar. This underscores
the point that similarly folded domains can support
different disulfide bond patterns.
Second, and perhaps most interesting, is the observa-
tion that the disulfide bond pattern of VAP2B ⁄ catrocol-
lastatin as observed from the crystal structure is rather
different from that determined by MS for catrocollasta-
tin C, the processed disintegrin-like and cysteine-rich
domain from catrocollastatin (Fig. 5). This observation
will be further discussed in the next section.
In 1994, Usami et al. isolated from the venom of
B. jararaca a protein that was composed of a disinte-
grin-like and a cysteine-rich domain [35]. Sequence
analysis of the protein indicated that it was the
processed disintegrin-like and cysteine-rich domains
from the PIIIb hemorrhagic toxin jararhagin [36].
Since then, several other processed disintegrin-like and
cysteine-rich domains from PIII SVMPs have been iso-
lated from viperid venoms [37,38], suggesting that it is
not an isolated event. As noted above, most PIII pro-
teinase domains have an unpaired cysteinyl residue.
Upon examination of the proteinase domains of two
of the PIIIbs that appear to undergo post-translational
proteolytic processing to yield a disintegrin-like ⁄ cyste-
ine-rich domain product in the venom, there does seem
to be some structural similarity. The unpaired cysteinyl

residues of both jararhagin and catrocollastatin are
located at position 379 in the loop between cystei-
nyls 374 and 390 (Fig. 4; VAP1 numbering). Two pro-
teins, which to date have not been observed to be
processed in the venom gland, atrolysin A and VAP1,
have their unpaired cysteinyl residues located at posi-
tions 360 and 365 respectively, N-terminal to the
unpaired cysteinyl residues observed in jararhagin and
catrocollastatin. Jarahagin and catrocollastatin are
known to undergo post-translational processing to
yield a disintegrin-like ⁄ cysteine-rich domain product in
the venom. Furthermore, it is interesting to note that
the PII atrolysin E, which is readily processed to yield
a stable proteinase domain and free disintegrin in the
venom, also has an unpaired cysteinyl residue located
at position 379, the same position as the unpaired
Fig. 3. Continued.
J. W. Fox and S. M. T. Serrano Snake venom metalloproteinases and venom complexity
FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS 3023
cysteinyl residue in the PIIIbs jararhagin and catrocol-
lastatin. This leads one to speculate that the presence
of an unpaired cysteinyl residue at that region of the
proteinase domain may be important for subsequent
proteolytic processing of the disintegrin-like ⁄ cysteine-
rich domains from the PIII SVMP proteinase domain.
Fig. 4. Sequence alignment of SVMP catalytic domains. Disulfide bonds of adamalysin II ( ) [17] and VAP1 (—) [32] revealed by
crystal structure analysis are depicted. Cyteine residues are numbered according to VAP1 [32]. Cysteine residues not involved in bonding in
adamalysin II and catrocollastatin ⁄ VAP2B are shown in bold; Cys365 of VAP1 is underlined.
Fig. 5. Disulfide bonds of catrocollastatin ⁄
VAP2B disintegrin-like (underlined) and

cysteine-rich (double-underlined) domains
(AAC59672) revealed by the crystal struc-
ture analysis (——) [33] and by N-terminal
sequencing and MS ( ) [53]. Cysteine resi-
dues are numbered according to catrocol-
lastatin ⁄ VAP2B [33]. (- - - -) is the only
coincident bond determined by both
methods.
Snake venom metalloproteinases and venom complexity J. W. Fox and S. M. T. Serrano
3024 FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS
However, there must remain significant differences
between the processing of a disintegrin domain from a
nascent PIIa SVMP and that of the disintegrin-
like ⁄ cysteine-rich domain from a PIIIb SVMP. To
date, it seems that the processed proteinase domains
from PIIas are stable and can be isolated from the
venom [39], whereas no processed proteinase domain
from a PIIIb has been isolated from the venom. Possi-
ble explanations could be as simple as that the disul-
fide bond patterns found in PIIa SVMPs allow for a
stable proteinase domain when not in the presence of
the disintegrin domain, whereas the disulfide bond pat-
tern and the odd cysteinyl residue in the PIIIb
domains are such that the domain is unstable and
perhaps susceptible to degradation by the numerous
proteinases in the venom. For example, isolated PIIIbs
such as bothropasin and brevilysin H6 can be induced
to undergo autolysis in vitro, when the proteinase
domain is observed to be degraded after release of
disintegrin-like ⁄ cysteine-rich (DC) domains [40,41].

Another possible reason for these differences, although
perhaps less likely, is that the cellular mechanisms of
post-translation proteolytic processing of the PIIas and
PIIIbs are different.
Several years ago, we became intrigued by the obser-
vation that both the processed and unprocessed forms
of PIIIs could be found in venoms. For example, both
full-length jararhagin and catrocollastatin and their
processed disintegrin-like ⁄ cysteine-rich domains
(jararhagin C and catrocollastatin C) products could
be isolated from their respective crude venoms, but not
the product proteinase domain [37,42]. This led us to
ask the question as to why both forms were found in
the venom; why are not 100% of those particular
PIIIb SVMPs completely processed rather than only a
fraction of them? We proceeded to investigate this by
attempting to manipulate jararhagin in vitro to
undergo autolysis. What we observed during the
course of those experiments was the following:
(a) under most conditions, jararhagin as isolated from
the venom is stable against autolysis, as would be
expected, given that it is present in the venom; and
(b) only under conditions such alkaline pH, low
calcium or the presence of reducing agents did some
low fraction of jararhagin undergo autolysis to pro-
duce a disintegrin-like ⁄ cysteine-rich domain. Interest-
ingly, the N-terminus of the jararhagin C produced
in vitro was different from that observed in the natu-
rally occurring jararhagin C, suggesting several possi-
bilities. Perhaps the jararhagin C found in venom may

not be an autolysis product, but rather a product from
a different proteinase, or perhaps the structure of the
jararhagin that is processed in venom is different from
that of the jararhagin that is not processed. We feel
that the latter scenario is more likely to be correct,
because when the jararhagin is artificially perturbed
in vitro to undergo autolysis, the alternative site
observed for the proteolysis is due to a structural iso-
meric form of jararhagin that is not normally pro-
cessed during biosynthesis.
Several conclusions and ⁄ or questions result from
these experiments. From a single venom pool, we have
isolated 3% of full-length jararhagin and 0.5% of
jararhagin C [42]. Thus, from this example, we can
estimate that approximately only one-quarter of the
jararhagin synthesized is processed to jararhagin C. The
question is why only 25% of the jararhagin synthesized
is processed to jararhagin C. The jararhagin found in
the venom is relatively stable against processing
in vitro, and this suggests to us the possibility that
there are multiple isoforms of nascent jararhagin, such
as might be the case with folding isomers, only a
limited number of which are susceptible to processing
to the end-product of jararhagin C. As seen in Fig. 5,
the disulfide bond patterns of processed catrocollasta-
tin C and those of the full-length catrocollastatin are
different. One disulfide bond pairing is shared between
the two structures. Furthermore, as seen in Fig. 5,
close inspection of the disulfide bond pairing in the
disintegrin-like domain of VAP2b indicates that the

disulfide bond pattern determined for catrocolla-
statin C could not occur without significant structural
rearrangement ⁄ folding. The different disulfide bond
patterns observed between VAP2b ⁄ catrocollastatin and
catrocollastatin C may be explained by disulfide bond
shuffling during experimental determination of cyste-
inyl pairs for catrocollastatin C (an explanation that
we feel to be unlikely). Another possibility is that
during proteolytic processing of the nascent catrocol-
lastatin, disulfide bond rearrangement occurred. Alter-
natively, the two different disulfide bond patterns
representing folding isomers were in place before post-
translational proteolytic processing and only one of
those isomeric forms (the catrocollastatin C form) was
processed. We hypothesize that it is most likely that
the former case is valid. In summary, we propose the
following scenario for PIIIb post-translational
proteolytic processing: In the case of catrocollastatin
and jararhagin, where a certain population of the pro-
teins has been seen to be processed, there are folding
isomers, perhaps promoted by the presence and rela-
tive positions of the unpaired cysteinyl residues in the
proteinase domain. One of the folding isomers is read-
ily processed to give rise to an unstable proteinase
domain and a biologically active disintegrin-like ⁄ cyste-
ine-rich domain. The other folding isomer is refractory
J. W. Fox and S. M. T. Serrano Snake venom metalloproteinases and venom complexity
FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS 3025
to proteolytic processing in the ER ⁄ Golgi system, and
is found in the venom in the form of the canonical PI-

IIa three-domain structure.
As cDNA sequences for various P classes of the
SVMPs became available, it was clear that there was a
short sequence coding for approximately 10 residues or
so immediately C-terminal to the proteinase domain,
which in the case of the PII and PIII classes was fol-
lowed by the disintegrin-like domain [31]. We termed
this region the ‘spacer’, suggesting that it provides a
structural space between the proteinases and disinte-
grin or disintegrin-like domains that could be involved
in post-translational modification of the nascent
SVMPs. As shown in Fig. 1, the SVMPs found in the
venom demonstrate a variety of differences with regard
to the presence of spacer sequences in the mature pro-
teins. For example, in the case of the PIa SVMP atrol-
ysin C, the spacer domain is processed from the
proteinase during post-translational modification,
whereas in the PIIa atrolysin E, the spacer sequence is
still found C-terminal to the proteinase. For the pro-
cessed PIIIb catrocollastatin C, there is a spacer
sequence N-terminal to the disintegrin-like domain.
Thus it appears that for some members in each of the
P classes, a degree of proteolytic processing occurs at
the spacer (Fig. 1; PIa, PIIa, PIId, PIIe, PIIIb). This
suggests a functional role for the spacer domain in
providing accessible proteolytic processing sites to give
those structural classes, whereas in some PIII SVMPs
and long distintegrins, the spacer region is not pro-
cessed, and as such is an integral part of those struc-
tures. From the crystal structures of the monomeric

SVMP catrocollastatin ⁄ VAP2b, the spacer region, or
linker as per their terminology, appears to be accessi-
ble to the solvent and contributes to the flexibility
observed between the proteinases and the disintegrin-
like domains [33]. However, in the case of this specific
molecule, processing did not occur, suggesting that
simple solvent accessibility of the spacer is not suffi-
cient for processing. Overall, spacer accessibility is
probably associated with the structure of the adjacent
proteinase and disintegrin-like domains and the pack-
ing of those two domains together in the structure.
Comparison of the sequences in the spacer domain
among the P classes shows significant sequence con-
servation and leads to no obvious conclusions as to
why some spacer regions are cleaved and some are
not, based on simple peptide bond specificities of
processing proteinases. Thus, at this point we can
only suggest that the overall accessibility of the
spacer to processing enzymes during post-transla-
tional processing is necessary, but not sufficient, for
proteolytic processing.
The PII SVMP class precursor gives rise to a mini-
mum of five product classes in the venom (Fig. 1). In
fact, this precursor class is probably responsible for a
significant amount of the venom complexity associated
with the SVMPs, in that the PII precursor class gives
rise to a proteinase and a disintegrin. As is observed
with the PII SVMP class, the PIII class also signifi-
cantly contributes to venom complexity. In Fig. 1, it
can be seen that there are essentially four protein

product classes found in the venom from the PIII pre-
cursor. Thus, the PIII class is also one of the more
prolific in terms of contributing to both the structural
and functional complexity of the venom.
Interesting structural features in disintegrin,
disintegrin-like and cysteine-rich domains
There are several interesting structural features associ-
ated with SVMP disintegrin, disintegrin-like and cyste-
ine-rich domains that warrant some discussion. The
most obvious and interesting structural difference
between the disintegrin-like domain of the PIII class
and the disintegrin domain of the PII class is the pres-
ence of a cysteinyl residue in the PIII disintegrin-like
domain, residing in what is called the RGD loop in
disintegrin domains (Fig. 6; residue 468, VAP1 num-
bering). The potent integrin-binding inhibitor activity
associated with the disintegrins was proposed to be the
evolutionary result of the loss of the cysteinyl residue
in this region, thereby structurally freeing the loop for
integrin interaction, coupled with the presence of the
integrin-binding RGD motif [29]. Other structural
alterations probably occurred during the evolution of
the disintegrin domains, expanding their repertoire of
possibilities to interact with integrins but without caus-
ing drastic fold changes. Indeed, disintegrin domains
show similar overall structures, although there appear
to be as many as three different disulfide bond patterns
in the disintegrins [43]. As seen in Fig. 6, four disulfide
bond pairs located in the center and in the C-terminal
region of the domains are shared between the disinte-

grin flavoridin and VAP1. The disulfide bond patterns
of the disintegrins and disintegrin-like domains are
different in the N-terminal region, and reflect the
fact that the lack of the first cysteinyl residue (Fig. 6;
residue 406, VAP1 numbering) in the disintegrins
gives rise to the necessity for a different disulfide bond
pattern in this region.
The C-terminal region of the PIII class has been
termed ‘cysteine-rich’, due to the abundance of cyste-
inyl residues (13 residues) in this domain. Interestingly,
neither the disintegrin-like nor the cysteine-rich
domain from the PIII class has been separately
Snake venom metalloproteinases and venom complexity J. W. Fox and S. M. T. Serrano
3026 FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS
isolated from venom; they have only been found as a
single biologically active protein composed of these
domains [35,37]. This suggests that, structurally, these
domains are very closely associated, which may
preclude further processing. Over the past several
years, evidence has been mounting which indicates that
the functionality attributed to the disintegrin-like ⁄
cysteine-rich proteins found in venom resides in the
cysteine-rich domain region [44–49]. The biological
activities attributed to this domain result from its
ability to interact with other proteins, such as FACIT
collagens, von Willebrand factor and integrins. Fur-
thermore, in the case of the PIIId factor X activating
proteinase, RVV-X (formerly classified as a PIV class
member), the hypervariable region in the cysteine-rich
domain has been suggested to manifest the functional-

ity (factor X activation) of the protein by its ability to
interact with the first lectin-like subunit of the multi-
meric structure via a unique cysteinyl residue found in
the hypervariable region [34].
Finally, there are three PIII class SVMPs, HR1B,
from the venom of Trimeresurus flavoviridis [50],
BjussuMP_I, from Bothrops jararacussu [51], and
kaouthiagin, from Naja kaouthia [52], that have the
integrin-binding motifs KGD or RGD in their cyste-
ine-rich domains (Fig. 6). Comparing the sequences of
these SVMPs with those of other PIII SVMPs, one can
observe, in the case of BjussuMP_I, that cysteinyl resi-
dues at positions 406, 425 and 468 (VAP1 numbering)
are absent from the disintegrin-like domain. In kaou-
thiagin, the disintegrin-like domain lacks cysteinyl resi-
dues at positions 443, 448, 449 and 452 (VAP1
numbering). This might reflect a different folding
structure in the N-terminal regions of the disintegrin-
like domains of these two proteins, although the C-ter-
minal regions of the domains are probably similar to
those of the rest of the class. Whether the different
disulfide bond patterns of these two proteinases are in
some way structurally related to the presence of the
integrin-binding motifs in their cysteine-rich domains
and somehow impact on the function of the proteins is
unknown; ultimately, additional functional studies and
crystal structures will be required to fully appreciate
the relevance of these structures in the context of other
PIII SVMPs.
Concluding remarks

In this review, we have attempted to present, based on
the literature, the current understanding of the struc-
tural features that contribute to the complexity
observed in many snake venoms. The significance of
this resides in the concept that the pathophysiological
effects observed following envenomation can be attrib-
uted to the combination of individual activities of dis-
crete toxins in the venom as well as, in some cases,
synergistic effects among the toxins, and therefore
understanding venom complexity and the source of
such complexity represents an important challenge in
the field. Although recent transcriptome and proteome
studies on venoms have provided us with tremendous
insights into venom complexity, very little is empiri-
cally known about how the biosynthetic process of
nascent venom protein production contributes to the
complexity. It is our hope that this review will provide
a solid foundation on which future investigations can
be conducted to illuminate the biosynthetic processes
that give rise to venom complexity and, as such,
Fig. 6. Sequence alignment of disintegrins (blue) trigramin (CAA35910), flavoridin (1FVLA), contortrostatin (Q9IAB0) and bitistatin (P17497),
and disintegrin-like (blue) ⁄ cysteine-rich domains (red) of the SVMPs atrolysin A (AAA03326), jararhagin (CAA48323), VAP1 (2EROA), Bjus-
suMP_I (ABD73129) and kaouthiagin (P82942)] by the program
CLUSTALW. Cysteine residues are highlighted in gray. Disintegrin and disinte-
grin-like sequences are shown in italic bold type. Disulfide bonds of VAP1 [32] and flavoridin [54,55] are depicted (———). Common disulfide
bonds between VAP1 and flavoridin are shown in red. Cysteine residues are numbered according to the VAP1 sequence [32].
J. W. Fox and S. M. T. Serrano Snake venom metalloproteinases and venom complexity
FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS 3027
lead to a better understanding of venom action and
possibly therapeutic intervention.

References
1 Takahashi T & Osaka A (1970) Purification and some
properties of two hemorrhagic principles (HR2a and
HR2b) in the venom of Trimeresurus flavoviridis;
complete separation of the principles from proteolytic
activity. Biochim Biophys Acta 207, 65–75.
2 Shannon JD, Baramova EN, Bjarnason JB & Fox JW
(1989) Amino acid sequence of a Crotalus atrox venom
metalloproteinase which cleaves type IV collagen and
gelatin. J Biol Chem 264, 11575–11583.
3 Bjarnason JB & Fox JW (1995) Snake venom metallo-
endopeptidases: reprolysins. Methods Enzymol 248, 345–
368.
4 Hite LA, Jia LG, Bjarnason JB & Fox JW (1994)
cDNA sequences for four snake venom metalloprotein-
ases: structure, classification, and their relationship to
mammalian reproductive proteins. Arch Biochem
Biophys 308, 182–191.
5 Fox JW & Serrano SMT (2005) Structural consider-
ations of the snake venom metalloproteinases, key
members of the M12 reprolysin family of metallo-
proteinases. Toxicon 45, 969–985.
6 Junqueira-de-Azevedo I de L & Ho PL (2002) A survey
of gene expression and diversity in the venom glands of
the pitviper snake Bothrops insularis through the genera-
tion of expressed sequence tags (ESTs). Gene 299, 279–
291.
7 Wagstaff SC & Harrison RA (2006) Venom gland
EST analysis of the saw-scaled viper, Echis ocellatus,
reveals novel a9b1 integrin-binding motifs in

venom metalloproteinases and a new group of puta-
tive toxins, renin-like aspartic proteases. Gene 377,
21–32.
8 Calvete JJ, Jua
´
rez P & Sanz L (2007) Snake venomics.
Strategy and applications. J Mass Spectrom 42, 1405–
1414.
9 Fox JW & Serrano SMT (2008) Exploring snake venom
proteomes: multifaceted analyses for complex toxin mix-
tures. Proteomics 8, 909–920.
10 Warshawsky H, Haddad A, Goncalves RP, Valeri V &
De Lucca FL (1973) Fine structure of the venom gland
epithelium of the South American rattlesnake and
radioautographic studies of protein formation by the
secretory cells. Am J Anat 138, 79–119.
11 Cominetti MR, Ribeiro JU, Fox JW & Selistre-de-Ara-
ujo HS (2003) BaG, a new dimeric metalloprotein-
ase ⁄ disintegrin from the Bothrops alternatus snake
venom that interacts with alpha5beta1 integrin. Arch
Biochem Biophys 416, 171–179.
12 Robeva A, Politi V, Shannon JD, Bjarnason JB & Fox
JW (1991) Synthetic and endogenous inhibitors of snake
venom metalloproteinases. Biomed Biochim Acta 50,
769–773.
13 Odell GV, Ferry EC, Vick LM, Fenton AW, Decker
LS, Cowell RL, Ownby CL & Gutierrez JM (1998)
Citrate inhibition of snake venom proteases. Toxicon
36, 1801–1806.
14 Menez A (1998) Functional architectures of animal tox-

ins: a clue to drug design? Toxicon 36, 1557–1572.
15 Serrano SMT, Shannon JD, Wang D, Camargo AC &
Fox JW (2005) A multifaceted analysis of viperid snake
venoms by two-dimensional gel electrophoresis: an
approach to understanding venom proteomics. Proteo-
mics 5, 501–510.
16 Rioux V, Gerbod MC, Bouet F, Me
´
nez A & Galat A
(1998) Divergent and common groups of proteins in
glands of venomous snakes. Electrophoresis 19, 788–
796.
17 Gomis-Ruth FX, Kress LF & Bode W (1993) First
structure of a snake venom metalloproteinase: a proto-
type for matrix metalloproteinases ⁄ collagenases. EMBO
J 12, 4151–4157.
18 Gomis-Ruth FX, Kress LF, Kellermann J, Mayr I, Lee
X, Huber R & Bode W (1994) Refined 2.0 A X-ray
crystal structure of the snake venom zinc-endopeptidase
adamalysin II. Primary and tertiary structure determi-
nation, refinement, molecular structure ad comparison
with astacin, collagenase and thermolysin. J Mol Biol
239, 513–544.
19 Zhang D, Botos I, Gomis-Ruth FX, Doll R, Blood C,
Njoroge FG, Fox JW, Bode W & Meyer EF (1994)
Structural interaction of natural and synthetic inhibitors
with the venom metalloproteinase atrolysin C (form d).
Proc Natl Acad Sci USA 91, 8447–8451.
20 Kumasaka T, Yamamoto M, Moriyama H, Tanaka N,
Sato M, Katsube Y, Yamakawa Y, Omori-Satoh T,

Iwanaga S & Ueki T (1996) Crystal structure of H2-
proteinase from the venom of Trimeresurus flavoviridis.
J Biochem 119, 49–57.
21 Gong W, Zhu X, Liu S, Teng M & Niu L (1998) Crys-
tal structures of acutolysin A, a three-disulfide hemor-
rhagic zinc metalloproteinase from the snake venom of
Agkistrodon acutus. J Mol Biol 283, 657–668.
22 Zhu X, Teng M & Niu L (1999) Structure of acutoly-
sin-C, a haemorrhagic toxin from the venom of Agkis-
trodon acutus, providing further evidence for the
mechanism of the pH-dependent proteolytic reaction of
zinc metalloproteinases. Acta Crystallogr D Biol Crys-
tallogr 55, 1834–1841.
23 Huang KF, Chiou SH, Ko TP, Yuann JM & Wang
AH (2002) The 1.35 A structure of cadmium-substituted
TM-3, a snakevenom metalloproteinase from Taiwan
habu: elucidation of a TNFa-converting enzyme-like
active-site structure with a distorted octahedral geome-
try of cadium. Acta Crystallogr D Biol Crystallogr 58,
1118–1128.
Snake venom metalloproteinases and venom complexity J. W. Fox and S. M. T. Serrano
3028 FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS
24 Watanabe L, Shannon JD, Valente RH, Rucavado A,
Alape-Giron A, Kamiguti AS, Theakston RD, Fox JW,
Gutierrez JM & Arni RK (2003) Amino acid sequence
and crystal structure of BaP1, a metalloproteinase from
Bothrops asper snake venom that exerts multiple tissue-
damaging activities. Protein Sci 12, 2273–2281.
25 Lou Z, Hou J, Liang X, Chen J, Qiu P, Liu Y, Li M,
Rao Z & Yan G (2005) Crystal structure of a non-hem-

orrhagic fibrin(ogen)olytic metalloproteinase complexed
with a novel natural tri-peptide inhibitor from venom
of Agkistrodon acutus. J Struct Biol 152, 195–203.
26 Shimokawa K, Jia LG, Wang XM & Fox JW (1996)
Expression, activation and processing of the recombi-
nant snake venom metalloproteinase, pro-atrolysin E.
Arch Biochem Biophys 335, 283–294.
27 Modesto JC, Junqueira-de-Azevedo I de L, Neves-
Ferreira AG, Fritzen M, Oliva ML, Ho PL, Perales J &
Chudzinski-Tavassi AM (2005) Insularinase A, a pro-
thrombin activator from Bothrops insularis venom, is a
metalloprotease derived from a gene encoding protease
and disintegrin domains. Biol Chem 386, 589–600.
28 Chen RQ, Jin Y, Wu JB, Zhou XD, Lu QM, Wang
WY & Xiong YL (2003) A new protein structure of
P-II class snake venom metalloproteinases: it comprises
metalloproteinase and disintegrin domains. Biochem
Biophys Res Commun 310, 182–187.
29 Calvete JJ, Marcinkiewicz C, Monleo
´
n D, Esteve V,
Celda B, Jua
´
rez P & Sanz L (2005) Snake venom disin-
tegrins: evolution of structure and function. Toxicon 45,
1063–1074.
30 Calvete JJ (2005) Structure–function correlations of
snake venom disintegrins. Curr Pharm Des 11, 829–835.
31 Hite LA, Fox JW & Bjarnason JB (1992) A new family
of proteinases is defined by several snake venom metal-

loproteinases. Biol Chem Hoppe Seyler 373, 381–385.
32 Takeda S, Igarashi T, Mori H & Araki S (2006) Crystal
structures of VAP1 reveal ADAMs’ MDC domain
architecture and its unique C-shaped scaffold. EMBO J
25, 2388–2396.
33 Igarashi T, Araki S, Mori H & Takeda S (2007) Crystal
structures of catrocollastatin ⁄ VAP2B reveal a dynamic,
modular architecture of ADAM ⁄ adamalysin ⁄ reprolysin
family proteins. FEBS Lett 581, 2416–2422.
34 Takeda S, Igarashi T & Mori H (2007) Crystal structure
of RVV-X: an example of evolutionary gain of specif-
city by ADAM proteinases. FEBS Lett 581, 5859–5864.
35 Usami Y, Fujimura Y, Miura S, Shima H, Yoshida
E, Yoshioka A, Hirano K, Suzuki M & Titani K
(1994) A 28 kDa-protein with disintegrin-like structure
(jararhagin-C) purified from Bothrops jararaca venom
inhibits collagen- and ADP-induced platelet
aggregation. Biochem Biophys Res Commun 201, 331–
339.
36 Paine MJ, Desmond HP, Theakston RD & Crampton
JM (1992) Purification, cloning, and molecular charac-
terization of a high molecular weight hemorrhagic
metalloprotease, jararhagin, from Bothrops jararaca
venom. Insights into the disintegrin gene family. J Biol
Chem 267, 22869–22876.
37 Shimokawa K, Shannon JD, Jia LG & Fox JW (1997)
Sequence and biological activity of catrocollastatin-C: a
disintegrinlike ⁄
cysteine-rich two-domain protein from
Crotalus atrox venom. Arch Biochem Biophys 343,

35–43.
38 Souza DH, Iemma MR, Ferreira LL, Faria JP, Oliva
ML, Zingali RB, Niewiarowski S & Selistre-de-Araujo
HS (2000) The disintegrin-like domain of the snake
venom metalloprotease alternagin inhibits alpha2beta1
integrin-mediated cell adhesion. Arch Biochem Biophys
384, 341–350.
39 Shimokawa K, Jia LG, Shannon JD & Fox JW (1998)
Isolation, sequence analysis, and biological activity of
atrolysin E ⁄ D, the non-RGD disintegrin domain from
Crotalus atrox venom. Arch Biochem Biophys 354, 239–
246.
40 Assakura MT, Silva CA, Mentele R, Camargo AC &
Serrano SMT (2003) Molecular cloning and expression
of structural domains of bothropasin, a P-III metallo-
proteinase from the venom of Bothrops jararaca. Tox-
icon 41, 217–227.
41 Fujimura S, Oshikawa K, Terada S & Kimoto E (2000)
Primary structure and autoproteolysis of brevilysin H6
from the venom of Gloydius halys brevicaudus. J Bio-
chem (Tokyo) 128, 167–173.
42 Moura-da-Silva AM, Della-Casa MS, David AS,
Assakura MT, Butera D, Lebrun I, Shannon JD,
Serrano SMT & Fox JW (2003) Evidence for heteroge-
neous forms of the snake venom metalloproteinase
jararhagin: a factor contributing to snake venom
variability. Arch Biochem Biophys 409, 395–401.
43 Calvete JJ, Ju
¨
rgens M, Marcinkiewicz C, Romero A,

Schrader M & Niewiarowski S (2000) Disulphide-bond
pattern and molecular modelling of the dimeric disinte-
grin EMF-10, a potent and selective integrin alpha5-
beta1 antagonist from Eristocophis macmahoni venom.
Biochem J 345, 573–581.
44 Jia LG, Wang XM, Shannon JD, Bjarnason JB & Fox
JW (2000) Function of the cysteine-rich domain of the
hemorrhagic metalloproteinase atrolysin A: collagen
targeting and inhibition of platelet aggregation. Arch
Biochem Biophys 373, 281–286.
45 Kamiguti AS, Gallagher P, Marcinkiewicz C, Theak-
ston RD, Zuzel M & Fox JW (2003) Identification of
sites in the cysteine-rich domain of the class P-III snake
venom metalloproteinases responsible for inhibition of
platelet function. FEBS Lett 549, 129–134.
46 Pinto AF, Terra RM, Guimaraes JA & Fox JW (2007)
Mapping von Willebrand factor A domain binding sites
on a snake venom metalloproteinase cysteine-rich
domain. Arch Biochem Biophys 457, 41–46.
J. W. Fox and S. M. T. Serrano Snake venom metalloproteinases and venom complexity
FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS 3029
47 Serrano SMT, Jia LG, Wang D, Shannon JD & Fox
JW (2005) Function of the cysteine-rich domain of the
haemorrhagic metalloproteinase atrolysin A: targeting
adhesion proteins collagen I and von Willebrand factor.
Biochem J 391, 69–76.
48 Serrano SMT, Kim J, Wang D, Dragulev B, Shannon
JD, Mann HH, Veit G, Wagener R, Koch M & Fox
JW (2006) The cysteine-rich domain of snake venom
metalloproteinases is a ligand for von Willebrand factor

A domains: role in substrate targeting. J Biol Chem
281, 39746–39756.
49 Serrano SMT, Wang D, Shannon JD, Pinto AFM,
Polanowska-Grabowska RK & Fox JW (2007) Interac-
tion of the cysteine-rich domain of snake venom metal-
loproteinases with the A1 domain of von Willebrand
factor promotes site-specific proteolysis of von Wille-
brand factor and inhibition of von Willebrand factor-
mediated platelet aggregation. FEBS J 274, 3611–3621.
50 Kishimoto M & Takahashi T (2002) Molecular cloning
of HR1a and HR1b, high molecular hemorrhagic fac-
tors, from Trimeresurus flavoviridis venom. Toxicon 40,
1369–1375.
51 Mazzi MV, Magro AJ, Amui SF, Oliveira CZ, Ticli
FK, Sta
´
beli RG, Fuly AL, Rosa JC, Braz ASK,
Fontes MRM et al. (2007) Molecular characterization
and phylogenetic analysis of BjussuMP-I: a RGD-P-
III class hemorrhagic metalloprotease from Bothrops
jararacussu snake venom. J Mol Graph Model 26,
69–85.
52 Ito M, Hamako J, Sakurai Y, Matsumoto M, Fuji-
mura Y, Suzuki M, Hashimoto K, Titani K &
Matsui T (2001) Complete amino acid sequence of
kaouthiagin, a novel cobra venom metalloproteinase
with two disintegrin-like sequences. Biochemistry 40,
4503–4511.
53 Calvete JJ, Moreno-Murciano MP, Sanz L, Ju
¨

rgens M,
Schrader M, Raida M, Benjamin DC & Fox JW (2000)
The disulfide bond pattern of catrocollastatin C, a dis-
integrin-like ⁄ cysteine-rich protein isolated from Crotalus
atrox venom. Protein Sci 9, 1365–1373.
54 Calvete JJ, Wang Y, Mann K, Scha
¨
ffer W, Niewiarow-
ski S & Stewart GJ (1992) The disulfide bridge pattern
of snake venom disintegrins, flavoridin and echistatin.
FEBS Lett 309, 316–320.
55 Senn H & Klaus W (1993) The nuclear magnetic reso-
nance solution structure of flavoridin, an antagonist of
the platelet GP Iib–IIIa receptor. J Mol Biol 232, 907–
925.
Snake venom metalloproteinases and venom complexity J. W. Fox and S. M. T. Serrano
3030 FEBS Journal 275 (2008) 3016–3030 ª 2008 The Authors Journal compilation ª 2008 FEBS