Cytophysiologic effects and molecular inhibition of a functional actin specific ADP ribosyltransferase CDT from clostridium difficile 4

114
Chapter 5
Peptide antibiotic and actin-binding protein as mixed-type inhibitors of
Clostridium difficile CDT toxin activities

5.1. Introduction
Aside from ATP phosphorylation and GTP-protein binding, ADP-ribosylation is a post-
translational, protein-modification process involved in many cellular functions (Obara et al.,
1991; Sooki-Toth et al., 1987). Eukaryotes produce nuclear poly (ADP-ribosyl) synthetase which
mediates substrate linkage of several ADP-ribose moieties and mono (ADP-ribosyl) transferase
which is likewise produced by prokaryotes. The catalytic process involves N-glycosidic bond
hydrolysis of NAD
+
resulting in the release of nicotinamide and attachment of adenosine-
diphosphoribose group to an acceptor protein (Moss and Vaughan, 1988). The bulky ADP-
ribose group is attached to arginine-177 of actin subdomain III forming bulk hindrance that
prevents actin polymerization (Vandekerckhove et al., 1988). The acceptor amino acid is
embedded in the filamentous actin form, thus explaining F-actin role as a poor CDT substrate. A
variety of acceptor molecules has been identified such as histones, enzymes, regulatory and
structural proteins (Ding and Smulson, 1994; Nakajima et al., 2004; Obara et al., 1991).
Having demonstrated the actin-specific ADPRT activity of CDT, the abundance and
equimolar concentration of G- and F-actin must be protected against modifying agents through
the development of effective inhibitors. Moreso, there has been an increase in the isolation of
cdt-encoding pathogenic C. difficile, an organism implicated in mammalian enterotoxemia,
antibiotic-associated diarrhea and colitis (Borriello and Carman, 1983; Goncalves et al., 2004;
Hatheway, 1990). Thus, the urgency of discovering potential molecules which can neutralize the
activities of the toxin.
Previous studies have focused largely on compounds against eukaryotic poly (ADP-
ribosyl) synthetases whose inhibitory actions were found to be highly specific (Banasik et al.,
1992; Rankin et al., 1989). In fact, differences in inhibitory capability yielded a broad range of
115

ordered efficacy. While novobiocin, vitamin K
1
and vitamin K
3
were reported most potent for
eukaryotic mono (ADP-ribosyl) transferase from hen heterophil, preventive efficacy of the
vitamin derivative nicotinamide has differed depending on ADPRT source (Banasik et al., 1990;
Rankin et al., 1989). Here we report in vitro effects of several groups of natural and synthetic
compounds such as metal ions, nucleotides, antimicrobials, vitamins and actin-binding proteins
(ABPs) on bacterial mono ADPRT activities. The most potent inhibitors were α-actinin and
striated muscle thin filament constituents, antagonistic to specific CDT actions.

5.2. Results
5.2.1. Nucleotide binding in CDT
On testing direct CDTa action on actin, ADP-ribosylation proceeded in a time-dependent
manner having peaked at 1 h mark (197.32) (Fig. 5.1A). Proportional rise in CDTa dosage and
actin labeling (Fig. 5.1B) supported substrate specificity of CDT action. NAD binding to actin
was then determined via UV cross-linking. Photoinsertion into actin was proportional to
[
32
P]NAD concentration with saturation effects observed at 46 µM (Fig. 5.1C). Competition for
binding by increasing unlabeled parent nucleotide resulted in corresponding decrease in
radiolabeling (Fig. 5.1D) with half maximal prevention obtained at 42.7 µM. These demonstrate
specificity of [
32
P]NAD interaction with actin nucleotide binding site. Since attachment of NAD
to CDT catalytic site was requisite in ADP-ribosylation, we next compared prevention of
[
32
P]NAD photoinsertion into CDT by increasing NAD and ATP. Both nucleotides prevented

photoinsertion with ATP causing 16% while NAD had 29% prevention (Fig. 5.1E), indicating
higher NAD affinity to CDT nucleotide binding site. For control reactions, non-irradiated
reaction mixtures and those with CDT or actin alone showed no detectable band indicating
absence of autologous phosphorylation and endogenous or contaminant radiolabels.


116



Figure 5.1. Kinetics of recombinant CDT. A, time-course analysis of actin-directed
ADP-ribosylation arrested at indicated times. Purified SK1203 (0.5 µg) was substituted
for CDTa in the (-) lane. B, ADP-ribosylation with increasing CDTa in increasing
concentrations (ng/ml). C, saturation of actin nucleotide binding site by photolabeling
with increasing [
32
P]NAD concentrations (mCi/mmol). No actin was added for the (-)
lane. D, prevention of [
32
P]NAD photolabeling of actin with increasing NAD
concentrations (µM). E. comparative prevention of [
32
P]NAD photoinsertion into CDT
by NAD and ATP at increasing concentrations (µM).
117
5.2.2. Divalent metal ions enhanced CDT-NAD binding
We then quantified the effects of ionic compounds on NAD photolabeling of CDT to
detect cofactors which could hasten or inhibit biochemical reactions. Relative to MgCl
2

(maximal activity) with mean pixel value of 62, ZnCl
2
at 112 and MnCl
2
at 79 units enhanced
photoinsertion by 81% and 27 % respectively, while Ca
2+
and other monovalent ions weakened
reactions by 2-7% indicating inability to substitute for Mg
2+
(Fig. 5.2A-C). This was reflected in
phosphorimages showing distinctly stronger intensities for Mn
2+
and Zn
2+
ions particularly at 100
µM and still evident at higher concentration. When chelators were added, differential reduction
in labeling was observed (Fig. 5.2D), implicating the influence of metal ions in nucleotide
binding. At 50 µM concentrations, the effect of ZnCl
2
was lowered upon EGTA (62%), EDTA
(43%) and calmodulin (37%) preincubation. MgCl
2
and MnCl
2
effects were diminished except
by calmodulin which enhanced photoinsertion by 14-27%. EDTA lowered Zn
2+
, Mn
2+

and Mg
2+

effects by 43%, 38% and 34% respectively. EGTA caused 49% reduction on Mg
2+
and only 26%
on Mn
2+
effects. Such divergent influence signifies a pattern of hierarchy for functional
specificity of ionic species on NAD attachment to CDT. Overall, Zn
2+
and Mn
2+
appeared to
have fulfilled the metal-ion requirement more efficiently than Mg
2+
.

5.2.3. Effects of various compounds on CDT activities
Several groups of compounds proved inhibitory to CDT actions (Table 5.1). The more
potent among them were ATP and nitrogenous compounds. Consistent with its direct
competition with the labeled counterpart (Fig. 5.1D,E), NAD expectedly posted a high significant
reduction of 36% and 58% on ARTase and NADse activities respectively (p<0.01). This was
reflected as decrease in actin radiolabeling (Fig. 5.2G) and free nicotinamide that corresponded to
higher percentage inhibition of transferase and glycohydrolase activities (Table 5.1). Similar
trends were observed on assay results performed at 50, 200 and 400 µM inhibitor concentrations
(data not shown).
118

119









Figure 5.2. Representative phosphorimages showing effects of various compounds on
CDT actions. Effects of cations and chelators on [
32
P]NAD photoinsertion into CDT. A-
C, CDTa incubated with CaCl
2
, CsCl
2
, RbCl
2
, MnCl
2
, MgCl
2
, and ZnCl
2

(lanes 1-6,
respectively) at indicated concentrations. D, CDT incubated with 100 µM MnCl
2
,
MgCl
2
, and ZnCl
2
(lanes 1-3, 4-6, 7-9, respectively) after preincubation with no chelator,
50 µM EDTA and EGTA (lanes 1,4,7; 2,5,8; and 3,6,9 respectively). Effects of chemical
compounds on ADP-ribosylation. E, preincubated with amoxycillin to cyclosporin A
(lanes 1-11, respectively); F, preincubated with erythromycin to thymine (lanes 12-26,
respectively); G, preincubated with thymidine to nicotinic acid (lanes 27-40,
respectively). Note that numerical assignment for each lane corresponds to a compound
of the same numerical designation in Table 1. C1-C3, control lanes containing no
inhibitor incubated with DMSO, without DMSO and with SK1203 in DMSO,
respectively. M, protein ladder. Effects of actin binding proteins on ADP-ribosylation.
H, preincubated with no inhibitor (lane 1), with 6 µg (lanes 2-5) and 12 µg (lanes 6-9) of
α-actinin, myosin, tropomyosin and troponin (lanes 2,3,4,5 and 6,7,8,9, respectively). I,
photoaffinity labeling of actin with no inhibitor (lane 1), no actin (lane 2), with 1 µg α-
actinin, myosin, and tropomyosin (lanes 3,4,5, respectively). Asterisk represents
consistent inhibition in all 3 trials at varied concentrations (50, 100, 200, 400 µM).
Compound concentrations are indicated at the end of each lane series.

120


Table 5.1. Effects of various compounds on CDT enzymatic functions

Transferase Activity
a
Glycohydrolase Activity
a


Compound
Chemical Nature Mol. Wt. (kDa)
Mean %
inhibition
b

IC
50
(µM)
Mean %
inhibition
c

IC
50
(µM)
01. Amoxycillin β-lactam 365.4 0 n 12 833
02. Ampicillin β-lactam 349.4 -6 n 11 909
03. Bacitracin cyclic heptapeptide 1422.69 11 455 37 270
04. Cefotaxime β-lactam 477.45 8 625 38 263
05. Cefuroxime β-lactam 446.37 8 625 16 625
06. Cephradine β-lactam 349.4 6 833 38 263
07. Chloramphenicol nitrophenyl propanediol 323.13 2 2500 -12 n
08. Clindamycin lincosamide 424.93 -10 n -19 n
09. Cloxacillin β-lactam 475.9 14 357 -25 n
10. Cycloheximide organophosphorus 281.3 8 625 17 588
11. Cyclosporin A cyclic oligopeptide 1202.64 6 833 35 286
12. Erythromycin macrolide 733.93 0 n -2 n
13. Kanamycin aminoglycoside 483.93 0 n 18 556
14. Methicillin β-lactam 379.41 -1 n -6 n
15. Nalidixic Acid quinolone 232.24 -3 n 13 769
16. Nystatin polyene 926.13 -18 n 3 n
17. Penicillin G β-lactam 333.5 12 417 28 357
18. Polymyxin B basic cyclic decapeptide 1188 21 238 45 222
19. Streptomycin aminoglycoside 581.58 1 5000 11 909
20. Tetracycline tetracycline 443.44 0 n 6 1802
21. Trimethoprim sulfonamide 290.32 5 1000 5 2165
22. Vancomycin glycopeptide 1413 17 294 6 1686
23. Adenine nitrogenous base 135.13 3 1667 16 625
24. Guanine nitrogenous base 151.13 18 278 47 211
25. Cytosine nitrogenous base 111 5 1000 16 625
26. Thymine nitrogenous base 126 1 5000 15 646
27. Thymidine nucleoside 242.2 -12 n 16 608

28. Uracil nitrogenous base 112 -17 n 24 417
29. ATP nucleotide 507.18 21 238 34 294
30. ADP nucleotide 427.2 4 1250 1 n
31. AMP nucleotide 347.22 20 250 -8 n
32. GTP nucleotide 523.18 -8 n 2 n
33. UTP nucleotide 468.14 -3 n -5 n
34. NAD nucleotide 663.43 36 139 58 172
35. Ascorbic Acid theoascorbic acid 176.12 -5 n 23 435
36. Menadione (Vit K
3
) Methyl naphthoquinone 172.18 19 263 35 286
37. Pantothenic Acid dimethyl oxobutyl alanine 219.24 -3 n 26 385
38. Thiamin aneurine 300.82 1 5000 41 245
39. Riboflavin flavin adenine dinucleotide 376.36 1 5000 28 357
40. Nicotinamide Amide form of VitB3 122.13 12 417 16 625
41. α-actinin actin crosslinker 100 64 78 42 236
42. Myosin actin binding 500 35 143 26 391
43. Tropomyosin myosin blocker 36/42 26 192 23 440
44. Troponin tropomyosin binding 38/24/18 -13 n 3 2961

121
a
assay mixture at 2% final DMSO concentration with 3 µg muscle actin

b
calculated from mean values with reference to control (maximal activity, no inhibitor) from assay trials
performed at least 3 times at 10 µg ABP concentration and 100 µM for all the other inhibitors.

c
calculated from mean values with reference to control (maximal activity, no inhibitor) from assay trials

performed at least 5 times at 10 µg ABP concentration and 200 µM for all the other inhibitors.

n=non-inhibitory or low inhibition

Data interpretation: IC
50
<400=inhibition, 400-500=intermediate inhibition, >500=non-inhibition (ARTase);
IC
50
<300=inhibition, 300-400=intermediate inhibition, >400=non-inhibition (NADse).



ATP caused weaker ARTase and NADse inhibition than NAD, while other nucleotides
including ADP (p=0.39) did not inhibit (Table 5.1) (Fig. 5.2G), except for AMP which lowered
ARTse but not NADse reactions implicating its effect on actin Arg177-ADP-ribose interaction
rather than CDT-NAD. Contrastingly, nitrogenous bases particularly guanine showed prevention
of NAD glycohydrolysis (Table 5.1). ADP, GTP and UTP were not as preventive at
concentrations up to 500 µM. Direct ATP competition for nucleotide binding sites in CDT or
actin seems probable as inhibitory effects were evident in all biochemical assays performed.
In
general however, nucleotide prevention may involve metal ion chelation.

Of the antimicrobials tested, polymyxin B (PMB), penicillin G and bacitracin prevented
both ARTase and NADse activities (Table 5.1, Fig. 5.2). In decreasing order, more potent
ARTase inhibitors were polymyxin B > vancomycin > cloxacillin > penicillin G and bacitracin
while the rest exhibited either low to moderate action or even stimulatory effects (Table 5.1) (Fig.
5.2E,F). On NADse activity, polymyxin B > cephradine and cefotaxime > bacitracin >
cyclosporin A > penicillin G proved most inhibitory. To illustrate the magnitude of prevention,
polymyxin B posed 21% more ARTase inhibition compared to nystatin’s -18% at 0% control

value without inhibitor. In addition, polymyxin concentration for half maximal inhibition is 21-
fold lower than the highest positive value for streptomycin. While vancomycin and cloxacillin
worked against ARTase and not NADse activity indicating interference on CDT-ADP-ribose
+
-
actin interaction, cephradine and cefotaxime showed more effective NADse prevention (Table
122
5.1)(Fig. 5.2E,F) suggesting specific disruption of CDT-NAD interaction. Of the vitamins and
derivatives tested, synthetic vitamin K
3
had strong inhibition on both activities particularly
ARTase (p<0.01) while thiamin showed pronounced NADse inhibition which was weaker in
riboflavin and panthotenic acid (Table 5.1). Nicotinamide exhibited only intermediate ARTase
and low NADse inhibition, reiterating action specificity to particular ADPRT.

5.2.4. Actin binding proteins exhibited diverse modes of CDT inhibition
The most potent inhibitors tested were actin binding proteins (ABPs) including myosin
which was initially more potent than NAD and α-actinin which reduced ARTase activity as about
efficiently as NAD. The gap in myosin potency with α-actinin at 3 µg (p=2.8X10
-6
) was
equalized at 12 µg concentration (p=0.14)(Fig. 5.2H, lanes 2 and 6), suggesting the presence of α-
actinin ligand in colonic lysate whose influence was dissipated at higher ABP concentration. To
test our hypothesis, analytical grade muscle actin was used as substrate in photolabeling assay. α-
actinin and myosin reduced nucleotide insertion by an average of 34% and 16% respectively,
while tropomyosin posted only 4% (Fig. 5.2I). Indeed, there was reversal of potency suggesting
endogenous factor influence which was reduced if not eliminated. On ARTase, tropomyosin was
59% and 28% less efficient than α-actinin and NAD respectively, while troponin was not
inhibitory (Fig. 5.2H). A hierarchy in inhibitor capacity was observed.
To further trace the mechanism of prevention, interruption of NAD hydrolysis by CDT

was investigated. Parallel trend was observed whereby α-actinin showed 1.6, 1.8 and 14-fold
higher inhibition than myosin, tropomyosin and troponin, respectively (Table 5.1), establishing
inability of the latter to obstruct CDT actions. This suggests adduction of NAD via ABP
nucleotide receptors or the presence of CDT receptor for ABP proximal to NAD binding/catalytic
site or both. Sensing variation in the extent of prevention depending on CDT action, modes of
ABP inhibition were assessed for functional multiplicity. With respect to NAD, kinetic analysis
123
(Fig. 5.3) of α-actinin and myosin exhibited mixed-type inhibition in contrast to principally
competitive action of nucleotides NAD and ATP.
Finally, in vitro observations of ABP effects on CDT-induced actin reorganization were
conducted. Similar to SK1203 and ATP (200 µM/ml)-exposed, SK1203-treated and untreated
cells (Fig. 4.2A), confocal images of cells after 8 h treatment with CDT-α–actinin mixture
showed diffuse red strands (F-actin cables) throughout the cytoplasm particularly around the
submembranous cortical region that totally disappeared on CDT-NAD or CDT treatment alone,
manifested as cytopathic rounding with G-actin zonal green patches throughout the cell body
including the nuclear region (Fig. 4.2B) and creation of protrusion stubs from retracting processes
(Fig. 4.2I) due to actin disassembly and cytolysis. Difference in susceptibility was significant
(p<0.017) between untreated and CDT-treated cells showing 20% rounding as early as the 2nd
hour and progressively thereafter up to the 8th (100% CPE). CD
50
on the 6th hour was 71%
compared to only 8% in CDT-inhibited treatment. Neutralization of CDT action by excess
inhibitor/competitor was apparent in colonic cells whose basal ABP level seemed futile in
counteracting CDT effects. Thus, it would be interesting to explore if contracile muscle cells
would mount a more robust protection. Together, these suggest ABP preventive effects on
CDTa catalysis, the exact mechanism of which awaits investigation. Further studies on the
significance of ABP effects on CDT-disrupted, actin-mediated physiologic processes such as
differentiation and signaling are also underway.

124









Figure 5.3. Lineweaver-Burk plots of initial velocity patterns for inhibited CDT with
respect to labeled NAD. At 3 µg actin, ADPRT assay was performed at indicated
[
32
P]NAD reciprocal concentrations. A, ATP concentrations used were 50 (■), 100 (▲),
150 (●) µM. B, α-actinin at 5 (■), 10 (▲), 15 (●) µg concentrations. C, myosin at 5 (■),
10 (▲), 15 (●) µg concentrations. The reciprocal of the reaction rate (1/V) is expressed
as pixel unit
-1
·h· µg of muscle actin. Each data point was mean + SD of triplicate assays.

125
5.3. Discussion
We have characterized a bacterial ADPRT (CDT) whose actin-specific actions were
challenged with potential inhibitors. Our preliminary studies revealed CDT requirement for

divalent cations for optimum catalysis, possibly through stabilization of folded nucleotide binding
domain (fingers). Such roles as essential cofactor of DNA binding enzymes like in polymerases,
have been demonstrated (Berg and Shi, 1996). Metal ions may also impose an orientating effect
on CDT or actin receptors and electrostatically shield negative charges to minimize electron
repulsion on attacking nucleophiles. Inhibitory effects by chelator adduction proved useful in
justifying these roles with our results suggesting a rank order in metal-ion requirement for CDT
functions.
In our survey of inhibitor compounds, we found that nucleotides compete for binding
sites on CDT as demonstrated by ATP efficiency in NAD substitution. While nucleotide
replacement in actin is possible, our data on inversely proportional reduction in radiolabeling and
nucleotide concentration (Fig. 5.1D,E) illustrated that nucleotide substitution mainly occurred at
the CDT nucleotide binding/catalytic site and demonstrated CDT affinity to NAD over other
nucleotide. Besides, actin has a high affinity binding site for ATP that requires divalent cations to
shift equilibrium towards monomeric (Ca
2+
) or polymer (Mg
2+
) state (Kinosian et al., 1993).
Therefore, vulnerability to nucleotide exchange may be attributed to differences in structural
configuration around the NAD cavity among ADPRTs.
Although crystallographic data for CDT has not been reported, mutagenesis studies have
revealed the presence of conserved amino acid residues for catalysis and NAD binding (Gulke et
al., 2001), that allowed classification of CDT into the cholera toxin (CT) group. They were
reported to possess common motifs including β/α, Glu/Gln-X-Glu and arom-Arg in β-strands and
Ser-Thr-Ser motif at the NAD cleft (Domenighini et al., 1994). Thus, CDT may follow similar
mechanisms of interaction as CT which was indeed shown to have lower affinity to its cofactor
compared to other ADPRTs (Galloway and van Heyningen, 1987). Unlike CT, other binary A:B
126
type toxins such as diphtheria and pertussis mediate stable docking of nicotinamide ring in the
NAD cavity (β/α motif) composed of hydrophobic residues with conserved Tyr-X10-Tyr

consensus whose tyrosine residues flank the nicotinamide ring creating a stable pi interaction
(Bell and Eisenberg, 1996; Carroll and Collier, 1984). The lack of such three-ringed order in CT
and CDT may facilitate nucleotide exchange. In terms of ARTase inhibition by nucleotides
through actin, we can speculate that competition between ADP-ribose
+
and excess nucleotide-
divalent ion
+
complex for Arg177 binding may be involved.
The antimicrobials that proved inhibitory to CDT were peptide antibiotics with modified
amino acids forming motifs as in β-lactams, cyclic structures, and those linked to lipids and
sugars (Hancock and Chapple, 1999). Polymyxin B (PMB) is a cyclic cationic decapeptide that
has amphiphatic features. Disruption in CDT activities may have been caused by initial
electrostatic interaction of its positively charged (net charge +5), arginine and lysine-rich peptide
ring residues with negatively charged interactive components of CDT or actin. This has been the
primary mechanism employed in its disrupting action of bacterial membrane particularly on
neutralization of the lipopolysaccharide component (Oren and Shai, 1998). Since PMB is large
and multicharged, it could displace stabilizing divalent cations or attach on or near essential sites
blocking CDT-actin interaction.
Another cyclic peptide, bacitracin was preventive. It binds divalent metal ions like Zn
2+

through its His10 imidazole ring and Mn
2+
via the thiazoline ring for its biological activity
(Scogin et al., 1983), making contention for metal ions seems imminent. Bacitracin has high
reactivity with biomolecules like nucleotides, proteins, lipids and receptors (Ming and Epperson,
2002). It is not surprising then that bacitracin is effective bacteriocide against C. difficile, as well
as vancomycin which were both inhibitory to CDT-ADP-ribose-actin interaction. The
mechanism adopted by another cell wall active antibiotic group, the β-lactams (penicillin and

cephalosporins) may involve CDT binding and steric blockage owing to structural propensity
towards engagement and inactivation of transpeptidases (Rogers and Forsberg, 1971). Of the
127
vitamins, menadione (vitamin K
3
) was consistent in its strong inhibitory effects on CDT and
eukaryotic poly ADPRTs (Banasik et al., 1990), while nicotinamide showed moderate inhibition
of CDT.
Ultimately, we found it interesting to examine if innate actin conjugants would exhibit
discrepant influences on CDT activities. There are around 162 distinct ABPs known with myriad
of actin-directed functions as receptor, linker, stabilizing, sequestering and capping proteins (For
review (dos Remedios et al., 2003)). ABPs bind to overlapping receptor loci on actin surface
which could interfere with CDT binding or affect catalysis. Our results suggest that sarcomeric
proteins considerably reduced ARTase activity by primarily obstructing the ADP-ribosylation site
of G-actin implying contact site proximity.
Based on the atomic model of actin helix (Holmes et al., 1990; Milligan et al., 1990), the
myosin head (S1) in the rigor state binds to the peripheral face of actin subdomain 1 with a slight
extension between the S1 interface and outer domain of the adjacent long-pitch monomer. These
interaction sites (Fig. 5.4) correspond to actin residues Asp1, Glu2, Asp3, Glu4, Asp24, Asp25,
Arg28, Glu93, Arg95 and Lys336 (Kabsch et al., 1990) and subdomain 3 residues Glu99, Glu100,
Ala144, Ile341, Ile345, Leu349, Phe352, Pro332-333 and Trp79-Asn92 (Schroder et al., 1993).
These regions overlaid by tropomyosin in the absence of Ca
+
and S1 (tropomyosin off position)
are in juxtaposition with Arg177 and may present steric impedance on ADP-ribosylation (Fig.
5.4). During reversible blocking (relaxed state), tropomyosin interacts with 7 actin molecules
along the long-pitch actin helix and in effect, binds common acto-myosin interaction sites
(Milligan et al., 1990). Comparative studies identified residues around Lys 215 and Pro307 to
consist the acto-tropomyosin interface between subdomains 3 and 4 in the on position, while
Arg28, Arg95, Arg147, Cys217-Leu236, Lys238, and Lys326, 328 and 336, were present in the

F-actin-tropomyosin complex (Kabsch and Vandekerckhove, 1992). These contact sites outline
two distinct parallel actin regions comprising the acto-tropomyosin interface in the activated and
relaxed muscle states (Fig. 5.4). Our results showing lower tropomyosin inhibition relative to
128






Figure 5.4. Schematic illustration of G-actin showing representative interface sites with
myosin (residue), tropomyosin “on” position (solid line), tropomyosin “off” position
(dashed line), and troponin (asterisk). ADP-ribose acceptor residue is boxed. Contact
sites with α-actinin are mainly distributed throughout subdomain 1 and was not shown
for image clarity. Residues were localized based on modeling, biochemical, electron
microscopic and atomic data presented in the discussion. Structural model for 1ATN
(Kabsch et al., 1990) was created using Cn3D4.1 program.

129
myosin indicate that our assay conditions favored the “on” position since comparable preventive
effects may have been derived in an “off” state, in which S1 and tropomyosin binding on actin
partly coincide. In the absence of Ca
2+
, the first troponin I inhibitory peptide (inhibit acto-myosin
ATPase) interacts with actin at residues 1-7 and 19-44 anchoring troponin-tropomyosin complex
on the filament (McKay et al., 1999) the actin filament (Levine et al., 1988) while binding weakly
with troponin C (McKay et al., 1999). Thus, acto-troponin contact sites cover a limited region of
the acto-myosin interface which may explain the observed lack of inhibition by troponin (Fig.
5.4).
Alpha-actinin was most potent with myosin effects approaching its values in a number of

trials. Through αA1-2 binding domain, its mediates lateral binding between two actin monomers
at helical Trp79-Asn92 (subdomain 1and 2 interface), Ser350-Lys359 (subdomain 1 of adjacent
monomer in F-actin) (Lebart et al., 1993) and residues 1-12, 86-117 and 350-375 (McGough et
al., 1994) that coincide with 112-125 and 360-372 identified for plastin (Lebart et al., 2004). We
have traced these reactive points and found extensive occlusion around ADP-ribose acceptor and
nucleotide cleft at the polymerization contact end of the molecule (Fig. 5.4). Analysis of
common binding interfaces suggest that myosin and tropomyosin can compete with α–actinin
binding and thus play roles in regulating the distribution of acto-actinin complex as dense bodies
in muscle cells. Furthermore, gelsolin S1, an α–actinin binding competitor tightly attach to the
cleft between subdomains 1 and 3 (McLaughlin et al., 1993) which corresponds to the F-actin
positive end and CDT binding site, supporting observed inhibition of CDT possibly by
competition. These could partly explain its exceptional potency as CDT inhibitor and support
CDT capacity as a capping protein.
Similar reduction in glycohydrolase activity affirms partial and mixed-type inhibition by
ABPs. Divergence in action and magnitude may be attributed to changes in structural form
according to oxidation state, polymerization and others. These were similarly observed in the
130
enhanced inhibition by the micellar compared to monomeric state of arachidonic acid and in Mg
2+

adduction by reduced α–lipoic acid (Banasik et al., 1990).
Recently, the mechanism of NAD glycohydrolysis in the iota catalytic site has been
proposed (Tsuge et al., 2003). It involves nicotinamide release via Sn1-type cleavage of N-
glycoside bond of the ring-like nicotinamide mononucleotide, resulting in the formation of an
oxocarbenium cation intermediate stabilized by Glu380 and Ser338 which is followed by ADP-
ribose
+
transfer to Arg177. This mechanism for Ia ADP-ribosylation of actin was proposed and
illustrated by Tsuge et al. 2003, reprinted in figure 5.5 below.

Specific molecular interference may be established at several steps in the proposed S
N
1-
type mechanism for iota Ia including the stabilization of the nicotinamide mononucleotide
(NMN) ring moiety of NAD (upon binding of NAD to Ia) through electrical charge-charge
interaction of positively charged atoms from Arginine residues with the AO1, NO1 and NO2
atoms of NAD phosphate and H-bond formation between the OE1 of Glu380 and NO2 of the
ribose ring in order to form a highly folded compact ring-like conformation of the NMN moiety
131
(Fig. 5.5A); the cleavage of NAD by H-bond formation of the nicotinamide carboxyamide and
the amide (NO1) and carbonyl group of Arg296 with the NN7 if nicotinamide followed by
spontaneous withdrawal of electron from NN7 causing cleavage of the N-glycoside bond; the
stabilization of the resulting oxocarbenium cation via charge-charge chelation of the positive NC1
of ADP-ribose by negative carboxylate group of Glu380 and Ser388 side chain (Fig. 5.5B); and
the transfer of ADP-ribose
+
to the guanidyl nitrogen of actin Arg117 (proximal to NC1 of ADP-
ribose
+

) which is recognized and positioned by Glu378 (Fig. 5.5C). This type of enzymatic
mechanism for actin ADP-riboslytion by ADPRTs is well-supported by mutational studies on
conserved residues of Ia (Tsuge et al. 2003), C2I (Barth et al. 1998), VIP2 (Han et al. 1999), C3
(Han et al. 2001) and in comparison with diphtheria toxin structure (Tsuge et al. 2003).
These steps from NAD attachment to substrate recognition can only occur with
accompanying structural dynamics which upon any molecular intervention like chelation by
anion stabilizer moiety in the ABP structure can potentially disrupt the process. The precise
modalities employed in abating CDT catalysis needs further investigation.
In conclusion, we have quantified and compared the extent of inhibition posed by a
number of molecules on various stages of ADP-ribosylation by CDT. Consistently versatile
inhibitors possess reactive molecular signatures (e.g. meromyosin S1) whose application is
important particularly in the design of engineered modular drug antagonists that would aid in
understanding ADPRT biology and neutralize deleterious mono-ADP-ribosyltransferases.







132
Chapter 6
General Discussion

Through several decades, it has become clear that studies on bacterial virulence factors
by medical microbiologists were of valuable significance to the understanding of infectious
disease processes. On this premise, we have used the proven technique of genomic subtraction to
successfully isolate a full complement cdt genes from pathogenic C. difficile CCUG 20309.
Although established however, it is of relevance to mention that genomic subtraction is not
without some limitations. Firstly, designation of identity to subtraction products may be

inconclusive due to the limited size of derived fragments. Secondly, caution must be taken in
doing similarity searches with careful evaluation of parameters including E-value, percentage
identity, and extent and quality of matched coverage between submitted query and subject
sequences. The difficulty in setting a fixed cut-off in defining a conclusive match, complicates
the process and makes prediction arbitrary and case to case. Furthermore, important genes may be
left-out since comparative searches are considerably dependent on the availability of sequences in
databanks. Having mentioned these however, our use of genomic subtraction demonstrated the
feasibility of isolating entire set of genes from “starter” fragments. More importantly, its use
resulted in a rapid isolation of known and putative virulence determinants from the partially
sequenced C. difficile genome. Overall, the technique would be a good adjunct to genome
sequencing as it can selectively isolate genetic components that confer unique characteristic in
one strain not found in another.
Aside from realizing certain properties of cdt regulation and comparative transcription
with major toxins, investigation of a single gene has revealed extensive genetic polymorphism
even between closely related strains. CCUG 20309 was first reported as a highly cytotoxic strain
that did not produce a detectable toxin A (Haslam et al., 1986). It is a strain with the largest
deletion in tcdA, with upstream tcdA insertion (Rupnik et al., 1998; Song et al., 1999) and with a
highly mutated 3’ end of tcdB encoding the receptor binding domain that resulted in altered
133
epitopes detected through the use of monoclonal antibodies and restriction fragment length
polymorphism (RFLP). Despite all these however, CCUG 20309 was found to be cytotoxic and
weakly enterotoxic, capable of causing hemorrhagic diarrhea in hamsters and accumulation of
bloody secretion in rabbits. Therefore, the production of an additional toxic factor is not
surprising since a number of strains carrying 1.7-1.8 kb deletion at the PaLoc were still
pathogenic (Alfa et al., 2000; Limaye et al., 2000; Rupnik et al., 1998; Sambol et al., 2000).
Indeed, our results have experimentally demonstrated the presence of a complete binary cdt in
CCUG 20309 that produces a highly potent cytotoxin which may partially explain its observed
pathogenicity. The discovery of binary CDT from CCUG 20309 is a significant finding since
only a minority of strains were found to produce the toxin including CD196. In fact we have not
isolated a cdt operon-carrying strain from our clinical survey in Singapore and most isolates

contained the fused, truncated form also possessed by the highly toxigenic VPI 10463. More
extensive epidemiologic studies will be necessary to clearly gauge the true contribution of this
emerging etiologic agent to human infectious diseases.
Virulence gene polymorphism is not unique to CCUG 20309 as it has also been observed
in the tox and dtxR of C. diphtheriae (Nakao et al., 1996) or Helicobacter pylori urease genes
(Foxall et al., 1992). Using PCR amplification, DNA segments for the catalytic N-terminal, the
central translocation domain and the C-terminal repetitive ligand of toxin A (A1, A2 and A3
respectively) and toxin B genes (B1, B2 and B3 respectively) from 17 strains showed pronounced
polymorphism (Rupnik et al. 1997). This was supported by data on restriction fragment length
polymorphism generating characteristic variations especially at the 5’ third of tcdB (B1 fragment)
and the 3’ third of tcdA (A3 fragment), later on proposed as a novel approach for toxinotyping.
This propensity to mutation could be attributable to mobile DNA elements like insertion
sequences, transposons and phage integration genes. However, although further investigations on
gene variations among C. difficile toxin genes is interesting and must be pursued further, it was
not our objective to explore the cause or mechanisms involved in such heterogeneity. Instead, we
134
focused on the functional properties of both CDTa and CDTb and their effects on cellular
physiology.
CDTa exhibited modification of actin monomer that caused dramatic F-actin
depolymerization. We have followed and quantified the dynamics of actin rearrangement
through flow cytometry using actin-specific fluorescent reagents. Other toxins like jasplakinolide
and cytochalasin whose actions against actin are established were used in synchrony or as
controls which increased the validity of results. CDT effect was synergistic with cytochalasin’s
increasing the critical concentration of the monomeric form and skewing the equilibrium shift
towards nucleation and polymerization. F-actin waned as early as the first hour of intoxication
and disappeared within 6-8 h. Actin depolymerization is the bulk source of monomeric actin
(Mitchison and Cramer, 1996; Ono, 2003). Thus, ADP-ribosylation can be considered as an
innate process whereby cells maintain the optimal G-actin pool, an event similar to G-actin
engagement and regulation by actin-destabilizing or severing actin binding proteins (ABPs) like
profilin, cofilin or thymosin β4 (Safer et al., 1991) to maintain isoform equilibrium.

We have shown the lethality of CDT action involving actin ADP-ribosylation, actin
disruption and cell rounding. However, it was not clear whether cell death was due entirely to
physical cytoskeletal collapse or an outcome aggravated by the disturbance of essential
components in cellular processes such as those in cell signaling or perhaps the activation of
apoptotic factors. Since most cellular processes from proliferation to death involve
communication, cytoskeletal reorganization may likely affect the expression and activities of
signal proteins. One may also argue on the prospect of a feedback mechanism after CDT-induced
actin disruption that could trigger the stress related and SOS response signal effectors. More so,
the presence of predisposing survival signals as early as CDTb-receptor contact to counter
cellular deterioration must be addressed.
To test our hypotheses, we have quantified the levels of such factors through
immunoassay and found that CDT is an inducer of stress-related MAP kinase pathway and
135
apoptosis. The sustained presence of CDT hindered the development of integrin aggregates into
more contractile adhesions which briefly delayed Rho activation to allow actin
nucleation/polymerization associated with Rac and Cdc42 activities. This is reflective of the fact
that immediate establishment of focal adhesion and spreading was prioritized under stress
conditions over that of network connections which occurs later in the developmental process.
Upregulation of talin and cAMP gave preliminary indication that CDT attachment to
receptor may be linked to a transductional process promoting actin nucleation and
polymerization. CDT may have likewise intensified cAMP/PKA-enhanced actin reformation by
Rho GTPases which are known controller of signals that link membrane receptor to the
cytoskeleton and MAPK pathway (Hall, 1998). Furthermore, cAMP has been previously shown
to directly activate MAPKs (Pomerance et al., 2000). Thus, cAMP via MAPKK stimulation, may
have amplified the CDT-imposed signals for MAPK activation. Although our results
demonstrated CDT stimulation of stress-related MAPKs and not the opposing ras-mitogenic ERK
MAPK (as evidenced by non-activation of downstream substrate Stat3), this seemingly
oversimplified process of strict lineage activation must be considered with caution since cross-
talk among signal components has been observed. At the MAPK level alone for example, MNK
and MSKs can both be activated by ERK and p38 (Roux et al., 2002). Ultimately, we have

shown apoptotic induction by CDT via caspase-3 and ATF-2 activation. CDT via the stress
MAPKs may have activated ATF2 to form Jun:ATF2 and ATF2:ATF2 dimers which possess
anti-apoptotic actions that promote cell proliferation, survival and transformation. Therefore, we
propose that CDT-imposed actin restructuring and cell shape alteration can trigger cellular
responses against stress.
Previous reports have likewise illustrated the direct interaction and activation of actin by
signal proteins. For instance, MAPK-actin interaction was observed after microfilament
disruption in human intestinal epithelial cells (Khurana and Dey, 2003; Nemeth et al., 2004).
Activated mekk1 was also demonstrated to co-localize with elements of the cytoskeleton
136
mediating c-JNK activation near or within the locale of the cytoskeletal network (English et al.,
1999). This is a clear illustration of the specificity of the transduction process towards a certain
molecular target, in this case, the F-actin. Moreso, cytoskeletal rearrangement has been shown to
stimulate mekk1-c-JNK activity for modulation of cell motility (Xia et al., 2000). This is another
proof of the intensification of CDT-induced disruption of the cytoskeleton by signal components
at different stages of transduction. Thus, it is becoming clear that cell death was not a
manifestation of direct CDTa action on actin exclusively but of combined effects with CDT-
stimulated signal transduction and apoptosis. Whether signaling was triggered as prompt as
CDTb attachment on the membrane or after CDTa-induced cytoskeletal collapse (whereby
increased G-actin relayed a stress signal to nuclear factors via MAPKs) remains to be clarified
through further experimentation.
As presented in this study, individual components of CDT possess unique properties
which could have harmful effects. CDTa could impose life-threatening sequelae on its own or
together with LCTs, due to its ability to extensively degrade microfilament networks and induce
cell death. Equally alarming is the wide-spread damage that could occur if CDT is able to
traverse tissue membranes such as the blood-brain barrier after toxin infiltration of the vascular
channels. Further to the cytotoxic effects of CDTa and TcdB toxins in CCUG 20309, two
enterotoxins are potentially secreted as well if both the truncated TcdA and CDTb are functional.
More experiments on intestinal cell lines or rabbit ileal loop could expose its enterotoxic
properties observed in other ADPRT binding components. Studies on CDT receptors are likewise

of utmost urgency. Enzymatic susceptibility, flow cytometry or electron microscopy of
immunolabeled target, mutations via complementation or knock-outs, or yeast-two-hybrid
coupled with GST pull down and two dimensional gel electrophoresis are just some approaches
which could be employed to explore CDT receptor properties.
Elaboration of such toxic factors inspired us to seek for compounds that could inhibit
most of its activities as increasing resistance to drugs such as vancomycin has been noted. It was
137
revealed that several peptide antibiotics and endogenous proteins as actin binding proteins
(ABPs) can be employed to neutralize CDT. Inhibition of CDT by various ABPs provided
evidence on the location of Arg177 which is adjacent to the nucleotide cleft, plus its designation
as the CDT binding site on actin. Through molecular engineering, promising creation of viable
submolecular ABP modules such as the meromyosin (myosin head S1) which binds to
subdomains 1 and 3 of G-actin or the αA1-2 domain of α-actinin which mediates lateral binding
between 2 actin monomers, could give rise to an effective molecule against CDT. This would be
of much benefit as hypersentivity towards foreign antigens will be downplayed if not entirely
evaded. Endogenous ABPs would be much tolerated.
In terms of beneficial applications, CDTa as an actin-freezing molecule and inducer of
cAMP, MAPKs and other components of signal transduction pathways, is a promising
pharmacologic agent that can be utilized for physiological research at the cellular to organ level.
It may also be tapped as an anti-tumor drug owing to its actin-capping and apoptotic actions that
will mostly affect actively-proliferating cells, a process which will be most advantageous through
targeted delivery via CDTb or phage protein. Indeed, the CDTb homologue C2II has been shown
to act as a carrier for intra-endosomal or cytosolic delivery of a heterologous fusion protein
containing C2II-interacting domain of C2I. This could be partly due to the observed ubiquity of
cell receptors for CII, reflective of our observations for CDTb. Moreover, molecular information
on CDT may be used as model which can be extrapolated as parallel properties for other binary
toxins such as edema and lethal toxins of B. anthracis, due their conserved structure and
similarity in intoxication process. Therefore, in light of heightened concerns involving B.
anthracis bioterrorism and biowarfare development, CDT or other

related binary toxins produced
by other gram-positive, spore formers like Clostridium or Bacillus species, could represent
effective vaccine targets for diseases

like anthrax. Such information

may aid not only in the
creation of efficacious vaccines but also in the development of therapeutics against the toxins.


138
Overall, we have demonstrated the extent of microfilament disassembly by CDTa, its
ability to stimulate stress and apoptotic proteins and CDTb’s invasive propensity into various cell
types. Certainly, the inclusion of CDT into the virulence armoury of C. difficile is a boost to its
pathogenicity. Through decades, knowledge on genetics, epidemiology and molecular
pathogenesis of C. difficile have been documented with particular emphasis on the major
clostridial toxin A and toxin B. Etiology of diarrheal and other intestinal diseases caused by C.
difficile has been instantaneously associated with the presence of these major clostridial toxins.
Based on our study however, it would be safe to suggest that toxigenicity should be assessed not
only based on the possession of PaLoc toxins but also of another similarly deleterious factor,
CDT which for a while, may have fortified C. difficile virulence under the radar.