The structure of Clostridium difficile toxin A glucosyltransferase domain bound to Mn2+ and UDP provides insights into glucosyltransferase activity and product release


M. Martinelli, Microbial Molecular Biology, Novartis Vaccines and Diagnostics, Via Fiorentina 1, 53100 Siena, Italy
Fax: +39 0577 243564
Tel: +39 0577 243876


Clostridium difficile toxin A (TcdA) is a member of the large clostridial toxin family, and is responsible, together with C. difficile toxin B (TcdB), for many clinical symptoms during human infections. Like other large clostridial toxins, TcdA catalyzes the glucosylation of GTPases, and is able to inactivate small GTPases within the host cell. Here, we report the crystal structures of the TcdA glucosyltransferase domain (TcdA-GT) in the apo form and in the presence of Mn2+ and hydrolyzed UDP-glucose. These structures, together with the recently reported crystal structure of TcdA-GT bound to UDP-glucose, provide a detailed understanding of the conformational changes of TcdA that occur during the catalytic cycle. Indeed, we present a new intermediate conformation of a so-called ‘lid’ loop (residues 510–522 in TcdA), concomitant with the absence of glucose in the catalytic domain. The recombinant TcdA was expressed in Brevibacillus in the inactive apo form. High thermal stability of wild-type TcdA was observed only after the addition of both Mn2+ and UDP-glucose. The glucosylhydrolase activity, which is readily restored after reconstitution with both these cofactors, was similar to that reported for TcdB. Interestingly, we found that ammonium, like K+, is able to activate the UDP-glucose hydrolase activities of TcdA. Consequently, the presence of ammonium in the crystallization buffer enabled us to obtain the first crystal structure of TcdA-GT bound to the hydrolysis product UDP.


• Coordinates of apo-TcdA-GT and Mn2+–UDP–TcdA-GT are available in the Protein Data Bank under the accession numbers 4DMV and 4DMW, respectively


differential scanning calorimetry


differential scanning fluorimetry




large clostridial toxin


Clostridium sordellii lethal toxin


Protein Data Bank


Clostridium difficile toxin A


Clostridium difficile toxin A glucosyltransferase domain


Clostridium difficile toxin B


Clostridium difficile toxin B glucosyltransferase domain

T m

melting temperature




Clostridium difficile, a Gram-positive spore-forming anaerobic bacterium, is a human pathogen responsible for pathological conditions such as antibiotic-associated diarrhea and fulminant pseudomembranous colitis, particularly in hospitalized patients. Since 2003, the incidence and severity of C. difficile-associated disease has increased in North America and Europe, with the emergence of a hypervirulent strain (NAP1/027) characterized by increased toxin production and antibiotic resistance [1–4].

The main virulence factors of C. difficile are two large exotoxins, designated toxin A (TcdA) and toxin B (TcdB). Both toxins are UDP-glucose hydrolases and glucosyltransferases (GTs) [5,6], and they belong to the family of large clostridial toxins (LCTs). The LCT family also includes the hemorrhagic toxin and the ‘lethal toxin’ (LT) from Clostridium sordellii [7,8], and α-toxin from Clostridium novyi [9,10]. The family of LCTs share a conserved four-domain organization: an N-terminal GT domain that precedes a cysteine protease domain; a central hydrophobic region (transmembrane domain) that is thought to be involved in pore formation [11,12]; and a C-terminal ‘binding’ domain composed of a variable number of well-defined repeats, responsible for highly specific binding to target cells. The most widely accepted hypothesis regarding the mechanism of action of the LCTs can be summarized in four steps: (a) highly specific binding of the C-terminal domain to receptors on the intestinal epithelial cell surface, followed by receptor-mediated endocytosis of the toxin into the endosomal compartment; (b) a pH decrease within the endosomal compartment, which causes conformational changes and pore formation, with subsequent translocation of the GT and cysteine protease domains of the toxin into the cytosol [12,13]; (c) autoproteolysis of the toxin, which releases the enzymatic N-terminal domain into the cytosol [13–16]; and (d) glucosylation of a specific threonine of the small GTPase proteins, which leads to depolymerization of actin filaments, disruption of the cytoskeleton, and, eventually, cell rounding and cell death [5,17].

Target substrates of α-toxin from C. novyi, hemorrhagic toxin, TcdA and TcdB are the small GTPases RhoA, Rac, and Cdc-42. LT from C. sordellii and TcdB-1470 are also able to inactivate Ral, Rap, and Ras, but not RhoA [7,18]. Although the LCTs are homologous GTs, significant differences exist between their in vivo and in vitro effects. For example, TcdA and TcdB are cytotoxic to cultured cells, but, for most cell lines, TcdB has been reported to be ∼ 100–1000-fold more potent than TcdA [19,20]. Several factors, such as the different rates of cellular binding and uptake, different glucosylhydrolase activities [18], and different specificities and affinities for the small GTPases, have been proposed as the causes of the different toxicity of the LCTs.

Whereas the structure of the TcdB GT domain (TcdB-GT) has been available since 2005 [21], the crystal structure of the TcdA GT domain (TcdA-GT) has been only recently reported [22]. These structures reveal that the active site organization of the two toxins is nearly identical, although the previously reported differences in glucose hydrolysis [19,22,23] and GT activity [19] suggested otherwise. Also, comparisons of the structures of TcdA-GT and TcdB-GT with other LCTs revealed the mobile nature of a so-called ‘lid’ loop (residues 510–522 in TcdA), which assumes different conformations depending on the presence or absence of bound cosubstrate (UDP-glucose).

In order to contribute to a better understanding of the TcdA-GT catalytic mechanism, we present here the first crystal structure of TcdA-GT bound to the hydrolysis product UDP at 2.5 Å resolution, and the apo structure of TcdA-GT at 1.5 Å resolution. These structures, together with the recently reported Mn2+-bound and UDP-glucose-bound forms [22], provide a detailed picture of the conformational changes that take place during the catalytic cycle of TcdA. In addition, the role of Mn2+ was investigated with differential scanning fluorimetry (DSF) and differential scanning calorimetry (DSC), and a glucosylhydrolase activity assay was employed in order to characterize the enzymatic properties of TcdA-GT.


UDP-glucose and Mn2+ induce a large increase in the stability of TcdA-GT

Recombinant wild-type (WT) TcdA-GT, and the Y283A/D285A/D287A mutant (mutant TcdA-GT) were expressed by use of a Brevibacillus choshinensis expression system. The triple mutant was designed in order to abolish UDP-glucose binding, using a structure-based sequence alignment with TcdB-GT and on the basis of results reported for a TcdB-GT mutant [24]. The thermal stability of both WT TcdA-GT and mutant TcdA-GT were measured by DSF, which revealed highly similar melting temperatures (Tm) of these apo forms of 39.0 ± 1.0 °C (Fig. 1A,B). However, incubation of WT TcdA-GT with either Mn2+ or UDP-glucose resulted in a moderate Tm increase (2–3 °C), whereas a large Tm increase (11 °C) was observed upon the simultaneous addition of both UDP-glucose and Mn2+. In contrast, no significant effect was observed for mutant TcdA-GT under these conditions, confirming that these mutations also target the binding site in TcdA-GT (Fig. 1A,B). The moderate Tm increases (3–4 °C) observed upon addition of glucose and/or UDP (Fig. S1) to TcdA in the presence of Mn2+ suggest that cofactor and intact substrate have a cooperative effect in stabilizing the WT protein. The large Tm increase, measured by DSF, for WT TcdA-GT in presence of a large excess of Mn2+ and UDP-glucose was also confirmed by DSC (ΔTm of 10 °C) (Fig. S2). The addition of EDTA to purified WT TcdA-GT did not induce significant changes in Tm, suggesting that the purified protein was obtained in the inactive apo form (Fig. 1C). Finally, we tested other bivalent metal cations (Ni2+, Mg2+, and Ca2+) as substitutes for Mn2+, and obtained a similar Tm shift in the absence of UDP-glucose but a reduced Tm increase (≤ 6 °C) in the presence of UDP-glucose (Fig. S3), suggesting that Mn2+ is the most potent stabilizer of the enzyme–substrate complex.

Figure 1.

 DSF analysis. (A) Fluorimetric thermal unfolding of WT TcdA-GT in 25 mm Tris and 150 mm NaCl (pH 8.0) buffer (blue line), in the presence of 10 mm UDP-glucose (yellow line), in the presence of 10 mm Mn2+ (green line), and in the presence of 10 mm UDP-glucose/Mn2+ (red line). (B) Correlation of TcdA-GT melting temperature with the concentration of UDP-glucose/Mn2+. WT TcdA-GT and mutant TcdA-GT are shown in red and in violet, respectively. (C) Correlation of TcdA-GT melting temperature with the concentration of UDP-glucose/Mn2+ (red line), Mn2+ (green line), and EDTA (blue line).

Ammonium activates the UDP-glucose hydrolase activities of TcdA-GT

A new colorimetric assay [25], developed to determine the glucosylhydrolase activity of TcdB, was used to determine the Km and Vmax values of recombinant WT TcdA-GT and mutant TcdA-GT. The ability of purified WT TcdA-GT to hydrolyze UDP-glucose in vitro was dependent on the addition of Mn2+ (Fig. S4A), whereas mutant TcdA-GT did not show any significant hydrolase activity under these conditions (Fig. 2A). This is in agreement with the previously reported kinetic data for the corresponding triple mutant of TcdB-GT [24]. WT TcdA-GT showed a Km of 43.9 ± 5.6 μm and a Vmax of 252.2 ± 7.4 pmol·min−1 per μg in the presence of K+ (standard conditions). The specific activity was determined from an enzyme dose curve obtained with a concentration of UDP-glucose of 1 mm (198.5 ± 7.4 pmol·min−1 per μg) (Fig. 2B). These kinetic properties (Vmax and Km) are similar to those reported for TcdB-GT [25].

Figure 2.

 Glucosylhydrolase activity determination. (A) Specific activity (SA) versus UDP-glucose. Km and Vmax were obtained by fitting the data to the Michaelis–Menten equation. The lines for WT TcdA-GT and mutant TcdA-GT are shown in black and orange, respectively. WT TcdA-GT and mutant TcdA-GT were incubated in presence of 5 mm MnCl2 and 150 mm monovalent ion (K+, Na+, or inline image, as indicated) for 30 min at 37 °C. (B) Glucosylhydrolase activity versus enzyme concentration. Increasing amounts of WT TcdA-GT were incubated in the presence of 5 mm MnCl2, 1 mm UDP-glucose and 150 mm KCl (pH 8.0) for 30 min at 37 °C. The dashed line represents the linear regression line of the data points.

Moreover, we compared the effects of ammonium, K+ and Na+ on TcdA-mediated hydrolase activity. Ammonium enhanced UDP-glucose hydrolase activity by approximately fivefold when an equivalent concentration of NaCl was used (Fig. 2A). TcdA-GT showed a Vmax of 162.8 ± 3.7 pmol·min−1 per μg in the presence of 150 mm ammonium and a Vmax of 36.4 ± 0.9 pmol·min−1 per μg in the presence of NaCl.

TcdA-GT is active in the presence of physiological concentrations of Mn2+ and UDP-glucose

DSF and activity assays were performed to evaluate the minimum amount of cofactor necessary to reach an optimal enzyme activity. The measurements were performed in the presence of 0.3 mm UDP-glucose, which is reported to be the physiological concentration in mammalian cells [26], and increasing amounts of Mn2+ (Fig. S4). The Kd value, obtained from the TcdA-GT activity assay, was 11.4 ± 0.9 μm (Fig. S4A), and the half-maximal Mn2+-induced increase in Tm was reached at 40 μm (Fig. S4B). These values suggest that, in the presence of a physiological concentration of Mn2+ (20–30 μm [27]) and UDP-glucose (300 μm [26]), TcdA may possess significant glucosyltransferase activity.

Overall structure of apo-TcdA-GT and Mn2+–UDP-bound TcdA-GT

We determined the crystal structure of apo-TcdA-GT at 1.5 Å by molecular replacement, using as input coordinates the catalytic domain of LT from C. sordellii [Protein Data Bank (PDB) code 2VKD], with which TcdA-GT shares 51% sequence identity. The asymmetric unit of the crystals of TcdA-GT contains one protein molecule, and electron density maps of excellent quality allowed modeling from residue 2 to residue 538 of the original 541 residues of TcdA-GT.

Although crystals of apo-TcdA-GT were obtained only when a 1 : 1 molar ratio of UDP-glucose and Mn2+ were present in the crystallization solution, electron densities suitable for modeling were not observed in the active site of TcdA-GT, either for UDP-glucose or Mn2+. This could have resulted from the presence in the crystallization buffer of 0.2 m tripotassium citrate, which may compete with Mn2+ binding [28], but possibly also from residual activity of TcdA-GT at the temperature of incubation of the crystallization plates (room temperature). We then sought to obtain cocrystals of TcdA-GT and UDP-glucose–Mn2+ by cocrystallization at a lower temperature (4 °C), at which TcdA-GT is supposedly inactive, and by using an excess (molar ratio of 1 : 50) of UDP-glucose and Mn2+. This resulted in growth of crystals in conditions similar to those used for apo-TcdA-GT (see Experimental procedures), and which diffracted X-rays up to 2.5 Å resolution. The structure was solved by molecular replacement, using as input the coordinates of apo-TcdA-GT previously refined (Fig. 3A,B).

Figure 3.

 Overall structure and mobile loop of apo-TcdA-GT and Mn2+–UDP-bound TcdA-GT. (A) Apo-TcdA-GT and UDP–Mn2+-bound TcdA-GT are depicted as cartoons and colored in blue and orange, respectively. The glucosyltransferase type A (GT-A) fold of both structures is colored in light gray, the bound Mn2+ is shown as a purple sphere, and bound UDP is shown as magenta orange sticks. Residues of helix α17 (M444 and K448) and nearby residues (Glu460, Arg462, and Gly471) that are known to be involved in the interaction with RhoA for TcdB are mapped onto the structure of TcdA-GT, and shown as red spheres. (B) Zoom into the substrate-binding site of TcdA superimposed on the structure of the homologs shown in Fig. 5. Conserved Trp residues (Trp520 in 2BVL and 2VKD, Trp525 in 2VK9, and Trp519 in TcdA) involved in interactions with UDP-glucose in TcdB and LT are shown with sticks and labeled. UDP is shown as orange and cyan sticks for the complexes Mn2+–UDP–TcdA-GT (orange) and Mn2+–UDP-glucose–TcdB-GT (cyan), respectively. Glucose from the complex Mn2+–UDP-glucose–TcdB-GT is also shown as cyan sticks. Manganese ions are shown as spheres for the complexes Mn2+–UDP–TcdA-GT (purple) and Mn2+–UDP-glucose–TcdB-GT (cyan), respectively. Trp101 residues in apo-TcdA-GT and Mn2+–UDP–TcdA-GT are also shown as sticks and labeled. The distance from the open conformation [as observed in α-toxin (2VK9)] and the closed conformation [as observed in LT (2VKD) and TcdB-GT (2BVL)] is labeled.

Electron density maps calculated during rebuilding and refinement of this second structure revealed the presence in the active site of a coordinated Mn2+, and the presence of UDP but not glucose. It is likely that residual activity of the toxin at 4 °C during incubation, or during crystallization (performed at room temperature), allowed the hydrolysis and consequent release of glucose from UDP-glucose. Accordingly, the cocrystal structure of TcdA-GT was fully rebuilt and refined with Mn2+ and UDP, the product of hydrolysis, bound in the active site (Fig. 4 and Fig. S5).

Figure 4.

 Stereoview of the interactions between UDP, Mn2+, and TcdA residues. TcdA is shown as an orange cartoon, and residues involved in interactions with UDP and Mn2+ are shown as sticks, with carbon, nitrogen and oxygen atoms in yellow, blue and red, respectively. Water molecules are shown as red spheres. UDP is shown as sticks, with carbon, nitrogen and oxygen atoms in orange, blue and red, respectively. The Mn2+ is shown as a purple sphere. Polar interactions between UDP and TcdA residues are shown as black dashes, and Mn2+ coordination with UDP and TcdA residues is shown as red dashes.

Structural comparison of apo-TcdA-GT and Mn2+–UDP-bound TcdA-GT

The overall fold of Mn2+–UDP-bound TcdA-GT is identical to the fold of apo-TcdA-GT (Fig. 3A), with an rmsd of 0.5 Å over 534 identical residues. However, one notable difference between the two structures is a slight rotation of the indole side chain of Trp101, which, in the Mn2+–UDP-bound TcdA-GT structure, is visibly closer to the uracil ring of UDP, suggesting the formation of favorable aromatic stacking interactions upon ligand binding (Fig. 3B). The functional implication of this slight rotation of the side chain of Trp101 is in agreement with previous studies showing that this Trp (number 102 in TcdB) is important for UDP binding [29]. The TcdB mutant with Tyr at position 102 shows 100-fold reduced GT activity, and almost no difference in glycohydrolase activity. In contrast, mutation to Ala has an apparently more dramatic effect on the structure of TcdB, resulting in no detectable glycohydrolase activity [29]. Importantly, it has been shown previously that mutation of Trp101 to Ala in TcdA results in an ∼ 400-fold reduction in GT activity and an ∼ 50-fold reduced cytopathic activity in cell culture experiments [30].

Structural comparison between TcdA-GT and other clostridial toxins

The two structures of TcdA-GT described here compare extremely well with the recently reported structures of the same toxin [22], revealing only small differences in the overall fold (rmsd of 1 Å on Cα atoms when all structures are compared). Comparisons with the structures of the catalytic domain of TcdB (PDB code 2BVL), the catalytic domain of LT from C. sordellii (PDB code 2VKD) and the catalytic domain of α-toxin from C. novyi (2VK9) revealed a highly similar, if not identical, overall fold. Superposition of apo-TcdA-GT onto TcdB-GT, LT and α-toxin resulted in rmsd values of 1.7, 1.5, and 2.1 Å, respectively (Fig. 5). Interestingly, a difference was observed in the conformation of the loop (residues 510–523) that covers the substrate-binding pocket of all glucosylating toxins. This loop was previously observed in a closed conformation in the cocrystal structure of TcdB-GT with hydrolyzed UDP and glucose [21], and in the structure of LT from C. sordellii in complex with intact UDP-glucose [31]. In contrast, the substrate-free crystal structure of the catalytic domain of α-toxin from C. novyi [31] showed this loop in an open conformation, with the Cα atom of Trp525 in α-toxin located ∼ 10 Å from Cα of Trp520 of both TcdB and LT (Fig. 3B). Both structures of TcdA-GT presented here show how this loop is partially disordered. Specifically, weak electron density maps for residues 518–520 and residue 519 were observed in apo-TcdA-GT and UDP–Mn2+–TcdA-GT, respectively. To account for the partial electron density maps, this region was refined by the use of partial occupancies, which suggested that the loop in TcdA-GT lies in a somewhat intermediate state between the closed and open conformations, with frequencies of 70–90%. This agrees with the notion of an intrinsic mobility of this loop that, as previously suggested, might play a role in cosubstrate recognition and binding. The highly conserved Trp520 (numbering in TcdB and LT, corresponding to Trp525 in α-toxin, and Trp519 in TcdA) forms direct hydrogen bonds with the glycosidic oxygen of UDP-glucose. The replacement of Trp520 by Ala in TcdB has been shown to severely impair glucosylation activity and prevent toxicity. Other residues known to be important for enzyme activity are Asp269, Arg272, Tyr283, and Asn383 [24]. These residues are conserved between TcdA and TcdB, and they line the bottom wall of the cosubstrate-binding pocket (Fig. S6). Although not directly in contact with UDP-glucose, this region of the active site accommodates the glycosidic moiety of UDP-glucose.

Figure 5.

 Structural comparison of TcdA-GT with LCTs. Superposition of substrate-free and Mn2+–UDP-bound TcdA-GT, depicted as in Fig. 3A, on structural homologs: the catalytic domain of TcdB in complex with UDP, glucose and Mn2+ (PDB code 2BVL) is shown in cyan; the catalytic domain of LT of C. sordellii in complex with UDP-glucose and Mn2+ (PDB code 2VKD) is shown in green; and the catalytic domain of α-toxin from C. novyi (PDB code 2VK9) is shown in yellow.

Structure-based and sequence-based conservation

A search in clostridial genomes for sequences similar to the catalytic domain of LCTs revealed 37 sequences (Fig. S7), all containing the residues involved in glucosyltransferase activity, including the DXD motif, and other residues important for glucose and UDP binding (Asp270, Arg273, Trp520, Trp102 and Tyr284 in TcdB-GT). Although the sequences show a high degree of similarity, they can be divided into five subfamilies on the basis of overall sequence conservation, and these include: α-toxin, LT, TcdA-like (TcdA), TcdB-630 (TcdB), and TcdB-1470 like (TcdB-F) sequences. TcdB-1470 is a different TcdB variant present in C. difficile serogroup F [32], which was shown to be a functional hybrid between TcdB and LT. A further alignment between the representative sequences of the five clostridial toxin subfamilies revealed remarkable differences in the region involved in GTPase binding, with residues located near a so-called helix α17 showing the highest amino acid variability (Fig. 6).

Figure 6.

 Sequence conservation among TcdA-GT homologs. Variability from a multiple sequence alignment of TcdA homologs TcdB-1470, TcdB-630, LT from C. sordellii [7,8] and α-toxin from C. novyi [9,10] is mapped on the surface of TcdA-GT, and colored according to the scale shown in the inset, from variable (cyan) to conserved (purple). (A) The back face of TcdA-GT or the opposite face to the substrate-binding pocket and GTPase interaction site is shown. (B) The front face of TcdA-GT, is shown and enclosed in the black rectangle the region of helix α17. (C) Top view to highlight the highly conserved residues of the core of the GT domain. UDP is shown as sticks and colored in cyan.

From mutagenesis studies on TcdB and LTs, it was hypothesized that residues located on helix α17 (Glu449 and Arg455) and in a nearby loop–helix region (Asp461, Lys463, and Glu472) (Fig. 7) play a crucial role in determining different specificities for GTPases [24].

Figure 7.

 Comparison of the GTPase interaction regions in TcdA-GT and TcdB-GT. (A) Surface representation of TcdB-GT, highlighting the region of interactions with GTPases in cartoon form for both TcdB-GT (cyan) and the superimposed TcdA-GT (orange). As reference, the UDP-glucose and the Mn2+ bound to TcdB are shown in the substrate pocket as sticks and spheres, respectively. (B) Zoom into the region enclosed by the square in (A). Residues known to be involved in the interaction with RhoA for TcdB, and equivalent residues of TcdA from a structure-based sequence alignment, are shown as sticks and labeled in cyan (for TcdB) and orange (TcdA).

Despite the high overall identity of the GT domains (∼ 50%), helix α17 has an average residue identity of 25%. Also, whereas the overall sequence identities of TcdA with TcdB and LT are similar, at 51.7% and 52.6%, respectively (Fig. S8A), the primary sequence of helix α17 in TcdA shows 50% sequence identity with LT and only 12.5% identity with TcdB (Fig. S8B). By mapping the amino acid conservation on the structure of TcdA-GT (Fig. 6), it is possible to observe that the back region of the toxins (with respect to the position of the active site) is mostly characterized by average conservation (Fig. 6A). In contrast, the front region, where both the active site and the site of interaction with GTPases (helix α17) are located (Fig. 6B,C), is mostly occupied by highly conserved residues in the catalytic core, and mostly variable residues in the GTPase interaction region. These differences in TcdA and TcdB suggest potentially different modes of recognition of small GTPase targets, although, to date, a degree of overlap in Rho/Ras family substrate modification in vitro has been reported [22].


It has been previously reported that the two main virulence factors of C. difficile, TcdA and TcdB, possess different GT activities, different specificities for GTPases, and different rates of UDP-glucose hydrolase activity [19,22,23]. Several observations implied that such differences could be explained by changes in the three-dimensional structures of the two toxins, although they share high sequence identity (∼ 50%). Although the functional and structural bases for the catalytic mechanisms of TcdB were already elucidated, until very recently the structure of TcdA-GT was unknown, and less information was available regarding its hydrolase and GT activities. However, a recent publication has reported the structure of TcdA-GT and the characterization of its GT activity and GTPase (Rap) inactivation [22]. In this study, we solved the crystal structure of TcdA-GT in two new forms, apo and UDP-bound, which differ from the aforementioned structures of TcdA-GT by the presence of hydrolyzed UDP-glucose and an absence of Mn2+ in the apo structure.

To complement our structural studies, we investigated the thermal stability of TcdA-GT as influenced by its cofactor (Mn2+) and cosubstrate (UDP-glucose), and the kinetic properties of the hydrolase activity. Therefore, our investigation provides further details for a more comprehensive overall picture of the catalytic mechanism of TcdA. The catalytic pocket of TcdA shows the conserved features of an LCT, including the key residues for coordinating Mn2+ and its ligand, UDP-glucose. Indeed, significant thermal stabilization of TcdA-GT (+11 °C) was observed upon addition of an excess of UDP-glucose and Mn2+ (Fig. 1). The thermal stabilization of TcdA-GT upon substrate binding can be explained by the extensive network of bonds established between several residues of the TcdA catalytic pocket, UDP-glucose, and Mn2+, as already observed in TcdB [24]. The limited thermal stabilization of TcdA-GT (Fig. S1) observed upon separate addition of both UDP and glucose in the presence of Mn2+ suggests that disruption of the bond between UDP and glucose destabilized this network. The small difference in hydrolase activity is in agreement with the identical organization of the active sites of TcdA and TcdB.

It is known that TcdA and TcdB require the presence of K+ but not Na+ as a cofactor for optimal UDP-glucose hydrolysis. In this context, the sodium chloride/Bicine buffer used by Pruitt et al. to crystallize UDP-glucose–Mn2+–TcdA-GT [22] may explain the presence of intact UDP-glucose in the catalytic site. In contrast, we found that ammonium ions are also able to activate the UDP-glucose hydrolase activity (Fig. 2A), which may explain why UDP, as a product of UDP-glucose hydrolysis, was observed in our structure, which was determined following crystal growth in the presence of 0.2 m ammonium tartrate.

As already reported for other glycosyltransferases [33,34], the hydrolase reaction follows a sequential ordered mechanism, whereby Mn2+ binds first to the enzyme (Mn2+–TcdA), followed by UDP-glucose (UDP-glucose–Mn2+–TcdA). Once catalysis occurs, the glucose is released, and the mobile loop in front of the active site moves to the ‘open’ conformation. This state corresponds to the Mn2+–UDP–TcdA crystal structure elucidated in this work, which was not observed in previous reports [22]. Following the release of UDP or UDP–Mn2+, the toxin is again able to bind the cofactor and cosubstrate. Despite the observation that Mn2+ and UDP-binding induce significant thermal stabilization, the comparison of all four available TcdA-GT structures reveals that the overall fold is not affected by the binding of the substrate or Mn2+, except for a slight rotation of the indole side chain of Trp101 (Fig. 3B), which is involved in UDP binding.

The flexible loop located in front of the active site in all LCTs (residues 510–522 in TcdA) has been previously postulated to act as a lid and to regulate UDP-glucose access [35]. In the holo-TcdA-GT structure reported recently [22], the side chain of Trp519 in this loop establishes a direct hydrogen bond interaction with the glycosidic oxygen of UDP-glucose, and the loop appears to be well structured. In contrast, in our structures, which lack the glucose moiety, there is enhanced mobility in this loop, as the observed partial electron density suggests.

On the basis of our DSF and structural data, we hypothesize that UDP-glucose degradation destabilizes the hydrogen bond network involving Trp519. In turn, the loss of these key hydrogen bond interactions allows the loop to assume a more open conformation, as observed here both in the apo form and in the UDP–Mn2+-bound structures of TcdA-GT, thus facilitating the release of the glucose or of the glucosylated substrate. Interestingly, UDP-glucose degradation was also observed previously in the TcdB-GT structure, but, unlike in our structure of the UDP–Mn2+–TcdA-GT complex, in TcdB the glucose was retained in the catalytic site. This difference in product release may result from differences in the crystallization conditions or possibly higher intrinsic mobility of the TcdA loop than that of TcdB.

We have determined the structure of TcdA-GT by cocrystallization with its substrate UDP-glucose, but in the absence of a glucosyl-acceptor target protein. Consequently, the observation of hydrolase activity is a surrogate for the physiologically relevant transferase reaction that would be catalyzed by TcdA in the presence of a real target. Nevertheless, the conformational changes observed upon UDP-glucose hydrolase activity of TcdA-GT provide insights into the GT reactions catalyzed by these toxins. In particular, the intrinsic mobility of the loop observed in TcdA-GT, as predicted for other enzymes [36], could influence enzyme motions and modulate the efficiency of the catalyzed biochemical reaction, in this case the GT reaction.

Herein, we have reported structural and biochemical characterizations that, together with complementary recent reports [22], allow an enhanced description of the catalytic mechanism of TcdA-GT. This domain shares 50% amino acid identity with that of TcdB, and its similarity to the corresponding domains of the clostridial toxin family members ranges from 33% to 53%. The majority of such differences can be mapped to the molecular surface, and this detailed knowledge of subtle variations may provide guidance for the investigatation of mechanisms that modulate the affinity and specificity for distinct cellular partners in the different pathogens.

This study also provides insights into the flexible loop rearrangements that govern cosubstrate access to and release from the binding pocket. These observations suggest the intriguing possibility that a combination of conformational changes in the ‘lid’ loop and differences in the rotamer orientation of aromatic residues in the active site may exert an important regulatory function on TcdA in vivo, in analogy to what has already been observed for a number of other bacterial virulence factors, such as diphtheria toxin [37], exotoxin A of Pseudomonas aeruginosa [38], and α-toxin of Clostridium perfringens [39]. Moreover, this study paves the way for the design of point mutations that can affect substrate preference by modulating the open-to-closed transition of the catalytic site.

Experimental procedures

Cloning and site-directed mutagenesis of TcdA-GT

The gene fragment encoding TcdA-GT, corresponding to TcdA residues 1–541, was amplified by PCR from C. difficile strain 630 genomic DNA, with the primers TcdAFor and TcdARev (Table S1). The PCR fragment was ligated into the BamHI and XhoI sites of the pNI-His vector (Takara Bio Inc, Shiga, Japan) to express TcdA-GT as a His6-tagged protein. Site-directed mutagenesis with the polymerase incomplete primer extension method was performed to generate mutant (Y283A/D285A/D287A) TcdA-GT [22,40], with the primers listed in Table S1. The resulting plasmids were used to transform the B. choshinensis HPD31-SP3 strain (Takara Bio Inc) by electroporation, according to the manufacturer’s instructions.

Expression and purification of WT TcdA-GT and its mutant

B. choshinensis cells transformed with WT pNI-TcdA-GT and pNI-TcdA-GT mutant were cultured in TMNm medium [1% glucose, 1% polypeptone (Becton Dickinson, Franklin Lakes, NJ, USA), 0.5% meat extract (Becton Dickinson), 0.2% yeast extract, 10 mg·L−1 FeSO4.7H2O, 10 mg·L−1 MnSO4.4H2O, 1 mg·L−1 ZnSO4.7H2O, 10 μg·mL−1 neomycin] at 25 °C in 2-L shake-flasks for 48 h. The harvested cells were lysed by sonication in binding buffer (50 mm Tris/HCl, 300 mm NaCl, and 10 mm imidazole, pH 8.0). The cleared lysate was loaded onto a PD-10 gravity flow empty column (GE Healthcare, Piscataway, NJ, USA) packed with 2 mL of Ni2+–nitrilotriacetic acid FF resin (Qiagen GmbH, Hilden, Germany), equilibrated with the binding buffer, and eluted with the lysis buffer containing 300 mm imidazole. Buffer exchange, with the PD-10 desalting columns (GE Healthcare), was performed to remove the NaCl, and the proteins were further purified by ion exchange chromatography (5-mL Q-FF column; GE Healthcare) and gel filtration (Supedex75 26/60; GE Healthcare). Protein concentrations were estimated with the bicinchoninic acid assay (Pierce, Rockford, IL, USA). The final purity quality of the proteins was checked by SDS/PAGE.

Crystallization and data collection

Crystals of WT apo-TcdA-GT were grown by hanging drop vapor diffusion, manually set up in Linbro plates, from a well solution consisting of 0.2 m tripotassium citrate, and 20% (w/v) poly(ethylene glycol) 3350. The protein solution (14 mg·mL−1 in 50 mm Tris/HCl, pH 8.0, 300 mm NaCl) was preincubated with 233 μm UDP-glucose and 233 μm Mn2+ (pH 8.0), and then mixed with well solution at a ratio (v/v) of 1 : 1. Crystallization of UDP-glucose–Mn2+–TcdA-GT was performed at 4 °C in 96-well low-profile Intelliplate crystallization plates in a nanodroplet sitting drop vapor diffusion format with a Crystal Gryphon liquid handling robot (Art Robbins Instruments, Sunnyvale, CA, USA). The protein solution was incubated with an excess (1 : 50 molar ratio) of UDP-glucose and Mn2+ (1 : 25 molar ratio; pH 8.0), and subsequently mixed at a 1 : 1 ratio (v/v) with reservoir solution composed of 0.2 m diammonium tartrate and 20% (w/v) poly(ethylene glycol) 3350. All crystals were frozen in liquid nitrogen, with 20% glycerol as cryoprotectant.

Crystals of both apo-TcdA-GT and UDP–Mn2+–TcdA-GT belong to space group P21212, with unit cell dimensions of a = 66.0, b = 153.0, and c = 66.2, and a = 66.6, b = 156.7, and c = 65.7, respectively. Diffraction data were collected on beamlines ID14-4 and ID14-1 at the European Synchrotron Radiation Facility (Grenoble, France), and processed with the ccp4 package [41]. Data collection and refinement statistics are shown in Table 1.

Table 1.   Data collection and refinement statistics.
  1. a Highest-resolution shell is shown in parentheses. Rsym = ΣhklΣi|Ii(hkl) − <I(hkl)>|/ΣhklΣiIi(hkl). Rwork = Σ||F(obs)| − |F(calc)||/Σ|F(obs)|. Rfree as for Rwork, but calculated for 5.0% of the total reflections that were chosen at random and omitted from refinement.

Data collection
Space group P21212 P21212
Cell dimensions
 a, b, c (Å)66.0, 153.0, 66.266.6, 156.7, 65.7
 α, β, γ (°)90, 90, 9090, 90, 90
Resolution (Å)1.5 (1.58)a2.5 (2.6)
R sym 8.2 (58.1)14.5 (52.1)
II13.8 (3.4)6.7 (2.7)
Completeness (%)99.9 (100)92.2 (86.8)
Redundancy7.1 (7.2)3.6 (3.8)
Wilson B-factor (Å2)16.235.2
Resolution (Å)1.5–34.52.5–41.1
No. of reflections108 02022 429
R work/Rfree18.9/21.420.1/25.9
No. of atoms
B-factors (Å2)
 Bond lengths (Å)0.0060.008
 Bond angles (°)1.0011.173
Ramachandran plot (%)

Structure solution and refinement

The crystal structure of apo-TcdA-GT was solved by molecular replacement in phaser [42], with the structure of LT from C. sordellii (PDB code 2VKD) as a search model. The final refined model of apo-TcdA-GT was subsequently used as a search model to solve, also by molecular replacement in phaser, the structure of the complex UDP–Mn2+–TcdA-GT. Model building and refinement were carried out in coot [43], phenix [44], and refmac [45]. Validation was performed with molprobity [46] as implemented in phenix. Figures were prepared with pymol ( Refinement statistics are shown in Table 1.

Glucosylhydrolase activity assay

The UDP-glucose hydrolase activity was measured with a nonradioactive glycotransferase assay that has been successfully exploited to determine the kinetic parameters of several GTs, including TcdB-GT [25]. The glucosylhydrolase reactions of WT and mutant TcdA-GT were carried out in 100 μL of reaction buffer (25 mm Tris/HCl, pH 8.0) in a 96-well plate. To determine the kinetic parameters of the proteins, variable amounts of the enzyme (range: 1–14 μg), the cofactor Mn2+ (range: 0–5 mm), and the substrate UDP-glucose (range: 0–1 mm), and 150 mm monovalent ion (KCl or NaCl or NH4Cl), were incubated at 37 °C for 30 min in the presence of fixed amounts of all other components, including a coupling phosphatase CD34L3 (0.2 μg) (R&D Systems, Minneapolis, MN, USA). A well containing all components except for the enzyme served as a blank control. The reactions were initiated by addition of UDP-glucose (Sigma-Aldrich, St. Louis, MO, USA) and CD34L3 to TcdA-GT, and terminated by addition of 20 μL of Malachite reagent A and 20 μL of Malachite reagent B to each well, followed by gentle mixing and incubation at room temperature. After 30 min, the absorbance at 620 nm was determined with an Infinite M200 spectrophotometer microplate reader (Tecan, Mannedorf, Switzerland). A phosphate standard curve (range: 0–2.5 μm phosphate) was made to determine the conversion factor between the absorbance and the inorganic phosphate contents. For determination of Km and Vmax, the results were plotted against substrate concentrations and fitted to the Michaelis–Menten equation with the graphpad program. For estimation of the Kd value, the data were fitted with graphpad by nonlinear regression, using the model for one-site binding – specific binding Y = BmaxX/(Kd + X).


The shift in the melting temperatures of WT TcdA-GT and mutant TcdA-GT as a function of Mn2+ and UDP-glucose concentration were determined with DSF analysis, following the previously reported protocol [47]. In each well, 20 μL of protein in 50 mm Tris and 150 mm NaCl (pH 8.0) was incubated with EDTA (final concentration range: 0–5 mm), Mn2+, UDP-glucose, Mn2+/UDP-glucose, Ca2+, Mg2+, Mn2+/UDP, Mn2+/glucose, and Mn2+/UDP/glucose, respectively, with a range of final concentration of 0–20 mm. The final concentration of the tested proteins was 2 μm. The plate was subjected to a temperature gradient scan (25–95 °C) in a Real Time PCR machine (Agilent Technologies, Santa Clara, CA, USA). The fluorescence intensity was measured by use of fluorescence Sypro Orange dye (×5 final concentration) at different temperatures with excitation/emission wavelengths of 490 and 575 nm, respectively.


In order to assess whether Mn2+ also promotes interaction at lower concentrations, analyses were performed with two different protein/Mn2+ ratios, 1 : 2 and 1 : 500, in the presence of an excess of UDP-glucose (1 : 100), by DSC (Microcal VP-Capillary DSC, GE Healthcare). The final concentration of the tested proteins was 10 μm in 20 mm Tris and 150 mm NaCl (pH 8.0). Thermal denaturation was performed with a scan rate of 180 °C·h−1 in an 8–105 °C range. All raw data were analyzed with origin 7.0.2 for Microcal LLC capDSC (GE Healthcare). Raw data were reference subtracted and normalized to concentration.


We wish to thank C. Brettoni for discussions on Gram-positive expression systems, S. Pileri (GE Healthcare, Munich) for support with DSC experiments, the ESRF for access to beamline ID14, and C. Martinoli and S. Rovida for beamline support during remote collection of diffraction datasets.