Open Journal of Genetics, 2012, 2, 18-22 OJGen
Published Online December 2012 (h ttp://www.SciRP.org/journal/ojgen/)
Published Online March 2012 in SciRes. http://www.scirp.org/journal/ojgen
A bio-informatics study of the c25 cysteine protease family
K.J. Cross, N.L. Huq, E.C. Reynolds
Bio21 and Oral Health CRC, University of Melbourne, Melbourne 3010, Australia
Received 2012
ABSTRACT
The oral pathogen Porphyromonas gingivalis is recog-
nized as one of the major aetiological agents of
chronic periodontitis. The gingipains, which are the
principal virulence factors of P. gingivalis, are mul-
ti-domain proteins containing an N-terminal C25
cysteine protease domain. We have conducted a
bio-informatics study of the C25 cysteine protease
domains and have identified related domains in over
two thousand proteins from 739 organisms in 35 dis-
tinct phyla. Proteins having significant similarity to
the gingipain C25 cysteine protease domain are also
found in Gram +ve bacteria, Archaea, algae, higher
fungi, and a wide variety of Eukaryotic species.
Keywords: C25 Cysteine Proteases; Evolution;
Substrate Preference
1. INTRODUCTION
Gingivitis is an inflammatory disease of the gum tissue.
If not checked, the disease can progress to periodontitis
leading to inflammation of the soft tissues surrounding
the teeth, resorption of bone, and eventual loss of teeth.
Porphyromonas gingivalis, is a major pathogen
associated with chronic periodontitis in adults.
Gingipains were identified as the outer membrane,
multi-domain virulence factors of the oral pathogen
Porphyromonas gingivalis [1]. The N-terminal domain
of the gingipains is a C25 cysteine protease domain. The
evolution of the C25 cysteine protease family has been
difficult to elucidate due to both the limited number of
family members identified and their narrow distribution
by species. The aim of this study was to undertake a
detailed search for C25 protease-like sequences in public
genome databases.
The MEROPS database [2] defines clan CD as
containing families of proteases with either a protein fold
or a sequence motif similar to those found in the caspase
family (C14) and a histidine-cysteine catalytic dyad with
the histidine located N-terminal to the cysteine. The
catalytic histidine is usually in a histidine-glycine motif
and is preceded by a block of hydrophobic residues. The
catalytic cysteine is found predominantly in an
alanine-cysteine motif and is preceded by a second block
of hydrophobic residues.
Enzyme specificity in clan CD is determined primarily
by the P1 residue of the substrate, which is normally an
asparagine in family C13 (legumains), an aspartate in
family C14 (caspases), and either an arginine or lysine in
C11 (clostripains), C25 (gingipains), and C50 (separases),
and a leucine in C80 (RTX toxin) [2]. The C25
(gingipain) specificity preference is based on the limited
experimental data available for three proteins (RgpA,
RgpB, and Kgp) from Porphyromonas gingivalis: RgpA
and RgpB share greater than 97% sequence identity
through their respective catalytic domains and display
specificity toward arginine residues, while the more
divergent Kgp displays specificity toward lysine
resid ues.
Tertiary structures are only available for members of
families C14 [3], C25 [4] and C80 [5]. These show
α/β-proteins with a fold consisting of an α/β/α sandwich.
The β-sheet contains six strands (in the order 213456)
with strand 6 anti-parallel to the rest. The fold is believed
to be unique to members of clan CD [2]. Other protein
families are included in clan CD because of the
conservation of motifs around the catalytic residues [6].
2. MATERIALS AND METHODOLOGY
2.1. Fugue Runs
We used a “pattern initiated hit and search” strategy to
identify possible C25 domain sequences in the ‘nr’
database [7]. In brief, a preliminary alignment of C25
cysteine protease domain sequences was used to develop
a regular expression that described the key features of the
alignment. The regular expression was then used to pare
down the number of sequences to be considered in the
second step of the process.
Sequences that passed the regular expression were
aligned against the RgpB sequence using Fugue [8]
taking into account the structural preferences of residues.
Sequences with a Z-score greater than 5.94 were
accepted for further study. Fugue alignments were

K.J. Cross et al. / Open Journal of Genetics 2 (2012) 18-22
performed using the structurally annotated sequence
(hs1cvra) for RgpB available from the HOMSTRAD
database [9] and the Fugue program [8]. The sequence
and structure of RgpB used throughout this paper is that
of Eichinger et al. [4].
2.2. Sequence Alignments and Analysis
T-Coffee [10] was used in the ‘espresso’ mode to
generate an initial (master) alignment of the C25 domain
sequences identified using Fugue. MAFFT [11] was used
to generate sequence alignments of the cysteine protease
domains identified using PSI-BLAST [12]. To ensure
consistency, all alignments were performed using the fft
option and ‘re-treeing’ twice. The quality of the hits
identified as possible C25 cysteine protease domains was
assessed using sequence alignments and the Shannon
information content of those alignments [13]. A Tcl/Tk
script was developed that populated the EPS template
used by WebLogo and allowed selected columns from
the sequnce alignments to be plotted.
2.3. PSI-BLAST Runs
Preliminary runs of PSI-BLAST were performed using a
single C25 domain sequence (RgpB) and varying the
‘inclusion threshold’ and maximum ‘E value’. Two runs
were performed with a maximum E-value of 1e-3 and
inclusion threshold values of 5e-4 and 5e-3. A further
two runs were performed with an inclusion threshold of
5e-4 and maximum E-values of 1e-2 and 1e-3.
Two production runs were performed. The first using a
master sequence alignment of 103 protein sequences
(MSA) identified by Fugue as being C25 domains. The
inclusion threshold was set at 5e-4, the maximum
E-value of 1e-4, and 6 iterations of the PSI-BLAST algo-
rithm were performed. The master sequence index was
incremented in consecutive runs so that a PSI-BLAST was
performed for each of the C25 sequences identified by
Fugue. The PSI-BLAST data was aggregated into an sqlite
database for further analysis. The second run used a cu-
rated subset of the sequences identified in the third itera-
tion of the first run with 336 sequences (MSA) and three
PSI-BLAST iterations were performed and analysed as for
the first run.
3. RESULTS AND DISCUSSION
Identification of Seed Sequences
A total of 103 sequences were identified as possible C25
cysteine protease domains using the Fugue-filter
criterion. They constitute a subset of the 185 sequences
identified in the Pfam database [14]. However, at least
six of the sequences identified by Pfam as C25 cysteine
protease sequences appear to lack the catalytic cysteines.
These non-cysteine protease sequences are
A9B8P8_HERA2,A4BYP5_9FLAO,E1K598_9EURY,
P96966_PORGI, E1K599_9EURY, and A7BN31_9GAMM.
PSI-BLAST uses two parameters to control the search for
related sequences. The ‘expectation value’ (E-value)
threshold determines whether a hit is accepted or not,
smaller values are associated with increased significance.
The ‘inclusion threshold’ is the maximum expectation
value for a hit to be used to calculate the PSSM (Position
Sensitive Substitution Matrix), the matrix describing the
probability of particular mutations occurring at specific
locations in the protein sequence.
In preliminary experiments, we demonstrated that
changing the E-value threshold from 1.0e-3 to 1.0e-2,
while keeping the inclusion threshold fixed at 5.0e-4,
increased the number of hits from 327 to 332 after five
iterations starting from the RgpB C25-domain sequence.
Changing the ‘inclusion threshold’ had no effect on the
number of hits located. Having demonstrated that the
PSI-BLAST algorithm was fairly insensitive to these
parameters at these levels (results not shown), we used a
significance level of 5.0e-4 for the inclusion threshold
and 1.0e-4 for the E-value threshold to minimize the risk
of incorporating misassigned sequences.Figure 1 shows
Figure 1: Schematic representation of the ‘BLAST connectivity’
of sequences in the ‘seed’ set of C25 sequences. Heavier lines
represent smaller, more significant expectation values in the
range 1e-10 to 1e-420. Note the location of the archetypical RgpA,
RgpB, and Kgp C25 sequences on the periphery of the cluster,
and the tight cluster of sequences primarily associated with
bacteria from the Prevotella phylum.
the ‘BLAST connectivity’ network for the C25 domains
19

K.J. Cross et al. / Open Journal of Genetics 2 (2012) 18-22
identified by the Fugue-filter process: the plot was
prepared using Cytoscape [15]. Heavier lines indicate
smaller expectation values for the blast connectivity
between the C25 domains (i.e. higher significance). All
the connectivities shown in Figure 1 have expectation
values of 1e-10 or less and can be considered as highly
significant. Figure 1 emphasizes the fact that the C25
domains of RgpA, RgpB, and Kgp (the archetypical
C25-proteins) are in fact on the edge of a cluster of
proteins that should be thought of as having typical C25
domain sequences.
As shown in Figure 2 there is a rapid increase in the
number of sequences identified after the third iteration of
PSI-BLAST accompanied by a rapid decline in the infor-
mation content of the alignments whether considered as a
sum over all positions in the alignment, or as information
per position as shown. This behaviour could be consis-
tent with ‘contamination’ of the PSSM used by PSI-BLAST.
The observation that the rapid increase in the number of
hits occurred even after the removal of suspected
non-gingipain sequences in the second, production
PSI-BLAST run suggests that many of these additional hits
are to significantly related sequences.
Figure 2. Number of unique sequences identified and the in-
formation content per position in the sequence alignments as a
function of PSI-BLAST iteration using the 103 C25-sequences
identified by Fugue as the ‘seed’ set.
The putative C25 cysteine protease domains were
curated by inspection of the alignments. For example,
inspection of the alignment of 362 sequences identified
at iteration 3 of the first PSI-BLAST run identified 44
sequences that either lacked a conserved cysteine at the
catalytic site, or had residues such as valine or proline
aligned with the catalytic histidine, or had a cluster of
three bulky residues aligned with the ‘GHG’-motif.
These sequences were removed from the alignment and
their GI numbers added to a ‘black list’ of sequences to
be excluded from future PSI-BLAST runs. The information
per position increased from 0.45954 ± 0.00029 to
0.49367 ± 0.00034 after removal of the ‘bad’ sequences
(compared to 0.50987 ± 0.00038 at the first iteration).
Figure 3 shows a ‘WebLogo’-style representation [16]
of those columns that have fewer than 40% gaps in the
alignment of the 2,333 proteins identified as being re-
lated to C25 domains in this work. The gaps between
conserved portions (note the indices are discontinuous)
are consistent with a family of proteins having a con-
served core connected by more variable regions.
Table 1 summarizes the progress of the search for C25
cysteine protease domains. The number of Archaeal and
Bacterial proteins identified increases at each stage of the
search process, as does the number of Eukaryotic species.
This suggests the families of proteins within Clan CD are
significantly overlapped in sequence space.
Figure 3. A ‘WebLogo’-style representation of the mafft
alignment of the 2,333 proteins identified in the final, curated
psi-blast data set. To compress the graphic, only those columns
with at most 40% gaps are represented as a consequence the
column indices labelling the x-axis are discontinuous.
Figure 4 emphasizes the structural similarities between
RgpB (C25), Yca1 a meta-caspase from yeast (C14B),
the RTX toxin from V. cholerae (C80), and human cas-
pase-7 (C14). Not only is there a strong topological
20

K.J. Cross et al. / Open Journal of Genetics 2 (2012) 18-22
Table 1. Summary of the number of different species and
proteins organized by phyla at various stages of the search for
C25 cysteine protease domains. The ‘Seed Set’ consists of the
103 proteins confirmed by Fugue alignment to be ‘C25 cysteine
protease domains’. The ‘1st Round’ proteins are those identified
in three iterations of PSI-BLAST starting from the ‘Seed Set’, but
excluding sequences as described in the text. The ‘Final’ set
represents the proteins identified after a further three iterations
from the “1st Round” set. The figure in brackets represents the
number of proteins.
Groups
Phylum
Seed Set
1st Round
Fina l
Archaea
Crenarchaeota 2 (4) 1 (1)
Eurya r chae ota 2 (2) 7 (19) 17 (26)
Korarchaeota
1 (1)
1 (1)
Thaumarchaeota 1 (2)
Bacteria
Acidobacteria
1 (1)
6 (18)
Actinobacteria 4 (4) 89 (208)
Aquificae
1 (1)
Bacteroidetes
56 (65)
131 (210)
149 (333)
Caldiserica 1 (1) 1 (1)
Chl amyd iae
1 (13)
Chlorobi 1 (1) 1 (2) 6 (11)
Chloroflexi 6 (7) 8 (19) 10 (73)
Cyanobacteria
3 (3)
53 (677)
Deinococcus-
The rmus
1 (1)
Firmicutes 1 (1) 12 (13)
Fusobacteria 1 (1)
Gemmatimonadetes
1 (1)
Ignavibacteria 2 (3) 2 (4)
Nitrospirae
2 (7)
Planctomycetes
3 (3)
5 (5)
6 (14)
Proteobacteria 5 (8) 31 (42) 176 (317)
Spirochaetes 13 (13) 18 (19)
Verrucomicrobia
1 (1)
1 (2)
2 (2)
Fungi
Ascomycota
2 (4)
76 (184)
Basidiomycot a
21 (248)
Eukaryota
Arthropoda
3 (3)
Bacillar iop hyt a 2 (3)
Chordata 29 (33)
Cni daria
1 (24)
Ec hinodermata 1 (1)
Platyhelminthes 1 (1)
Porifera
1 (2)
Viridi pl antae
Chlorophyta 2 (2)
Streptophyta
13 (30)
Brown algae
Phaeophyceae 1 (3)
similarity as shown by the order and orientation of the
β-strands forming the core of the proteins, but the cata-
lytic histidine and cysteine are consistently two residues
Figure 4. Cartoon representations of the β-strand core of RgpB
from Porphyromonas gingivalis (PDB:1cvr) a C25 cysteine
protease (A), caspase-7 from Homo sapiens (PD B: 1 k88) a C14
cysteine protease (B), the RTX-toxin from Vibrio cholerae
(PDB:3gcd) a C80 cysteine protease (C), and Yca1 a metacas-
pase from Saccharomyces cerevisiae (PDB:4f6o) a C14B
cysteine protease (D). The location of the catalytic histidine and
cysteine shown in stick form, are always located two residues
C-terminal of the fin al residues in the third and fourth β-strands
respectively.
C-terminal to the last residues of the 3rd and 4th
β-strands respectively. Given this strong structural simi-
larity, it is not surprising that C25 domain proteins dis-
play a marked sequential similarity with other Clan CD
proteins.
The sequence specificity of the C25 proteases is de-
termined by the residues that line the ‘S1’ specificity
pocket. As shown in Figure 5 there is very little se-
quence conservation in this region apart from the
‘GHG’-motif and the catalytic cysteine. For example, in
RgpB the side-chain of Asp163 hydrogen bonds to the
guanidine group of the arginine substrate whereas in Kgp
the analogue of Asp163 is a threonine and it is probably
Asp516 that is the predominant hydrogen-bonding part-
ner to the lysine substrate. Inspection of the alignment of
the putative gingipain sequences identifies three se-
quences that had previously been annotated as either
“propeptide peptidase C25” (GI: 373458037) or “pepti-
dase C25” (GI:326279641 and 307565663) from Cal-
dithrix abyssi, Odoribacter splanchnicus, and Prevotella
amnii [7] respectively, where Thr209 of RgpB is re-
placed by an arginine residue. Thr209 is located near the
bottom of the ‘S1’ specificity pocket, and an arginine
residue would have its side-chain extend to near the top
of the specificity pocket suggesting that these three bac-
terial gingipains have a caspase-like specificity toward
aspartate (or possibly glutamate) residues N-terminal to
the cleavage point. In, for example, RgpB a short-chain
residue near the bottom of the specificity pocket is used
to ‘recognize’ the long-chain of the arginine substrate.
The three proteins identified here, appear to use a
long -chain arginine at a neighbouring location to ‘recog-
nize’ a short-chain substrate. These specific examples
underscore the wide range of substrate specificities sug-
21

K.J. Cross et al. / Open Journal of Genetics 2 (2012) 18-22
gested by Figure 5.
Figure 5. A plot of the frequency of occurrence of various re-
sidues that align with the ‘S1’-pocket residues of RgpB in the
‘seed’ set of 103 C25 cysteine protease sequences. As shown in
Table 1, these are Bacterial or Archaeal proteins. These resi-
dues determine the substrate specificity of the various C25
proteases. The low overall conservation suggests a broad range
of C25 substrate preferences.
4. CONCLUSION
The bacterial C25 cysteine proteases share significant
sequential and structural similarity with other Clan CD
cysteine proteases. The number of identified bacterial
and archaeal sequences increases at each stage of the
search procedure as seen in Table 1, while the number of
sequences associated with other phyla increases dramat-
ically in the final round of the search.
The lack of sequence conservation in the ‘S1’-binding
site argues for a wide-range of substrate specificities
among the C25 cysteine proteases further blurring the
distinctions between the various protease families within
the Clan CD proteases.
5. ACKNOWLEDGEMENTS
We acknowledge funding from the Oral Health CRC and NH&MRC.
REFERENCES
[1] Carlsson, J., B.F. Herrmann, J.F. Hofling, and G.K.
Sundqvist. (1984) Degradation of the human proteinase
inhibitors alpha-1-antitrypsin and alpha-2-macroglobulin
by Bacteroides gingivalis. Infection and Immunity. 43,
644-648.
[2] Rawlings, N.D., A.J. Barrett, and A. Bateman. (2010)
MEROPS: the peptidase database. Nucleic Acids
Research. 38, D227-D233.
[3] Walker, N.P.C., R.V. Talanian, K.D. Brady, L.C. Dang,
N.J. Bump, et al. (1994) Crystal structure of the cysteine
protease interleukin-1 beta-converting enzyme: a
(p20/p10)2 homodimer. Cell. 78,
doi:10.1093/nar/gkp971
343-352.
[4] Eichinger, A., H.G. Beisel, U. Jacob, R. Huber, F.J.
Medrano, et al. (1999) Crystal structure of gingipain R:
an Arg-specific bacterial cysteine proteinase with a
caspase-like fold. EMBO Journal. 18,
5453-5462.
doi:10.1016/0092-8674(94)90303-4
[5] Lupardus, P.J., A. Shen, M. Bogyo, and K.C. Garcia.
(2008) Small molecule-induced allosteric activation of
the Vibrio cholerae RTX cysteine protease domain.
Science. 322, 265-8.
doi:10.1093/emboj/18.20.5453
[6] Chen, J.M., N.D. Rawlings, R.A. Stevens, and A.J.
Barrett. (1998) Identification of the active site of
legumain links it to caspases, clostripain and gingipains
in a new clan of cysteine endopeptidases. FEBS Letters.
441, 361-365.
doi:10.1126/science.1162403
[7] nr database. [cited 2012 September 3]; Available
from:
doi:10.1016/S0014-5793(98)01574-9
ftp://ftp.ncbi.nih.gov/blast/db.
[8] Shi, J., T.L. Blundell, and K. Mizuguchi. (2001) FUGUE:
sequence-structure homology recognition using
environment-specific substitution tables and
structure-dependent gap penalties. Journal of Molecular
Biology. 310, 243-257.
[9] Mizuguchi, K., C.M. Deane, T.L. Blundell, and J.P.
Overington. (1998) HOMSTRAD: a database of protein
structure alignments for homologous families. Protein
Science. 7, 2469-2471.
doi:10.1006/jmbi.2001.4762
[10] Notredame, C., D.G. Higgins, and J. Heringa. (2000)
T-Coffee: A novel method for multiple sequence
alignments. Journal of Molecular Biology. 302, 205-217.
doi:10.1002/pro.5560071126
[11] Katoh, K., K. Misawa, K. Kuma, and T. Miyata. (2002)
MAFFT: a novel method for rapid multiple sequence
alignment based on fast Fourier transform. Nucleic Acid
Research. 30, 3059-3066
[12] Camacho, C., G. Coulouris, V. Avagyan, N. Ma, J.
Papadopoulos, et al. (2009) BLAST+: architecture and
applications. BMC Bioinformatics. 10,
421.
[13] Shannon, C.E. (1948) A Mathematical Theory of
Communication. Bell System Technical Journal. 27,
379–423.
doi:10.1186/1471-2105-10-421
[14] [14] Côté, R.G., P. Jones, L. Martens, S. Kerrien, F.
Reisinger, et al. (2007) The Protein Identifier
Cross-Referencing (PICR) service: reconciling protein
identifiers across multiple source databases. BMC
Bioinformatics. 8, 401.
[15] [15] Smoot, M.E., K. Ono, J. Ruscheinski, P.L. Wang,
and T. Ideker. (2010) Cytoscape 2.8: new features for
data integration and network visualization.
Bioinformatics. 27,
431-2.
doi:10.1186/1471-2105-8-401
[16] Crooks, G.E., G. Hon, J.M. Chandonia, and S.E. Brenner.
(2004) WebLogo: a sequence logo generator. Genome
Research. 14, 1188-1190.
doi:10.1093/bioinformatics/btq675
doi:10.1101/gr.849004
22