The cellulosome is an intricate multi-enzyme complex, known for its efficient degradation of recalcitrant cellulosic substrates. Its supramolecular architecture is determined by the high-affinity intermodular cohesin–dockerin interaction. The dockerin module comprises a calcium-binding, duplicated ‘F-hand’ loop–helix motif that bears striking similarity to the EF-hand loop–helix–loop motif of eukaryotic calcium-binding proteins. In the present study, we demonstrate by progressive truncation and alanine scanning of a representative type-I dockerin module from Clostridium thermocellum, that only one of the repeated motifs is critical for high-affinity cohesin binding. The results suggest that the near-symmetry in sequence and structure of the repeated elements of the dockerin is not essential to cohesin binding. The first calcium-binding loop can be deleted entirely, with almost full retention of binding. Likewise, significant deletion of the second repeated segment can be achieved, provided that its calcium-binding loop remains intact. Essentially the same conclusion was verified by systematically mutating the highly conserved residues in the calcium-binding loop. Mutations in one of the calcium-binding loops failed to disrupt cohesin recognition and binding, whereas a single mutation in both loops served to reduce the affinity significantly. The results are mutually compatible with recent crystal structures of the type-I cohesin–dockerin heterodimer, which demonstrate that the dockerin can bind in an equivalent manner to its cohesin counterpart through either its first or second repeated motif. The observed plasticity in cohesin–dockerin binding may facilitate cellulosome assembly in vivo or, alternatively, provide a conformational switch that promotes access of the tethered cellulosomal enzymes to their polysaccharide substrates.
Aerobic and anaerobic bacteria can transform plant cell wall polysaccharides into a source of carbon and energy by secreting glycoside hydrolases that break them down into simple sugars . Some cellulolytic bacteria produce extracellular free cellulases [2,3]. In others, individual cellulases are attached directly to the cell surface, and still others produce a multi-enzyme cell-surface cellulosome complex that binds to and efficiently degrades crystalline cellulose and associated plant cell wall polysaccharides [4–9]. Maintaining the multi-enzyme complex in close proximity to the cell surface appears to be advantageous for efficient sugar assimilation. The most extensively studied cellulosome system belongs to Clostridium thermocellum, which comprises a variety of cellulases and hemi-cellulases, attached to a non-catalytic scaffoldin subunit (CipA) [10,11]. Each of the nine highly conserved type-I cohesins on the primary scaffoldin subunit binds to an enzyme-associated type-I dockerin module through a high-affinity non-covalent interaction (Figure 1A). In early studies on the cellulosomes from two clostridial species, C. thermocellum and Clostridium cellulolyticum, the interaction between cohesins and dockerins was found to be generally species-specific; experiments carried out with isolated modules from the two species revealed that cohesins from the scaffoldin of one species bind to the dockerins of its own enzymatic subunits with high affinity but fail to recognize those of the other species, despite the relatively high sequence homology among the analogous components [12,13].
Simplified schematic view of the cellulosome from Clostridium thermocellum
In efforts to determine molecular and structural elements that dictate binding and species specificity, crystal studies have been performed to solve the three-dimensional structure of the type-I cohesins from C. thermocellum [14,15] and C. cellulolyticum . Both cohesin species exhibit an identical nine-stranded β-sandwich with jelly-roll topology. Later an NMR solution structure of the dockerin module  was reported, which revealed a novel symmetrically oriented duplicated ‘F-hand motif’, comprising a calcium-binding loop and adjacent α-helix.
Previously the crystal structure of a type-I cohesin–dockerin complex from C. thermocellum has been published . The interface between the two modules of the heterodimer was found to occur mainly between one surface of the cohesin and the second loop–helix motif of the dockerin (Figure 1B). However, since the crystallization of the complex represents only one possible binding orientation between the cohesin and the dockerin, the internal sequence symmetry of dockerin could imply theoretically a second mode of binding to cohesin. Indeed the plasticity in dockerin binding has been documented recently  by crystallization of the cohesin in complex with a dockerin mutated in the recognition residues of the second loop–helix motif. In this case, the dockerin module was oriented 180° relative to its position in the original crystal, interacting with cohesin mainly through the first F-hand motif, thus exemplifying the dual-binding capacity of the type-I dockerin in C. thermocellum.
In the present study, we have challenged experimentally the relevance of different regions of the type-I dockerin module for cohesin recognition by two independent approaches. In the first, we created a series of truncated derivatives progressively deleted at the N- and C-termini of the dockerin module. The resultant truncated derivatives were then examined for their binding to an appropriate type-I cohesin using an ELISA-based system . In the second approach we have generated by alanine scanning a series of dockerins mutated in the calcium-ligating residues of the first and second calcium-binding loops. The resultant proteins were then examined as to whether the designated mutations serve to destabilize calcium affinity and impair the dockerin interaction with the cohesin.
Escherichia coli XL-1-Blue and Escherichia coli BL21(λDE3) were purchased from Stratagene.
Truncated dockerin-containing constructs
A construct of Geobacilus stearothermophilus xylanase T-6 with a His-tag and BspHI site at the 5′-terminus and a KpnI site at the 3′-terminus was produced using PCR [21,22]. This construct was ligated at the KpnI site with the PCR product of a C. thermocellum Cel48S dockerin (recombinant cellulosomal Family-48 cellulase from C. thermocellum) , containing a 5′-terminal KpnI site and a 3′-terminal BamHI site and inserted into the pET-9d vector at the NcoI and BamHI sites. This plasmid allows facile replacement of the Cel48S dockerin (termed herein DocS) with any other desired dockerin by digestion with KpnI and BamHI, and the resulting expressed product constitutes a His-tagged xylanase T-6 fusion protein bearing a dockerin at the C-terminus (termed in this case XynDocS). The desired truncated dockerins were generated by PCR, with a sense primer that introduced a KpnI site and an antisense primer that introduced a stop codon and a BamHI site, utilizing wild-type XynDocS as a template.
CBM (cellulose-binding module)–cohesin construct
All mutants were prepared using the QuikChange® site-directed mutagenesis kit (Stratagene). The DNA template for introducing a single mutation was wild-type XynDocS. To verify that only the designated mutations were inserted by the Pfu-Turbo DNA polymerase, the full dockerin module was sequenced. In order to exclude the possibility that more mutations were generated outside of the sequenced region, the mutated dockerin module was digested with KpnI and BamHI and then re-cloned into the XynDocS cassette. Mutants containing several mutations were produced in a sequential manner, in which one mutant served as a template for the subsequent one.
Expression of recombinant proteins
E. coli BL21(λDE3) strain was used for overexpression of the recombinant proteins. Host cells were grown in 500 ml of LB (Luria–Bertani) medium, supplemented with 50 μg/ml kanamycin, at 37 °C until the culture reached D600>0.6. IPTG (isopropyl β-D-thiogalactoside) was added at final concentration of 0.5 mM for induction of protein expression. Culture growth was continued for another 3 h, 5 h or overnight at 37 °C, 30 °C or 16 °C respectively, according to predetermined optimization experiments. Cells were harvested by centrifugation (6000 g for 20 min at 4 °C). Immediately before purification, cells were resuspended in 15 ml of TBS (Tris-buffered saline, 137 mM NaCl, 2.7 mM KCl, 25 mM Tris/HCl, pH 7.4) and lysed by sonication. The lysate was centrifuged (20000 g for 30 min at 4 °C), and the supernatants were used for protein purification. The purity of all proteins was then estimated by SDS/PAGE, and the protein concentration was determined by spectrophotometeric absorbance (280 nm) using the calculated molar absorption coefficient of the protein as determined using the Vector NTI suite (InforMax).
Purification of His-tagged XynDocS
The XynDoc-containing supernatants were loaded on to a 5 ml Ni-NTA (Ni2+-nitrilotriacetate) column (AKTA-prime System, Amersham Pharmacia Biotech), pre-equilibrated with TBS containing 1 mM CaCl2 (TBS-Ca). The column was then washed with 40 ml of TBS-Ca supplemented with 10 mM imidazole. Proteins were eluted in TBS using a linear gradient of 20–250 mM imidazole over 30 ml (with variations in gradient linearity and range between proteins). Fractions were collected and analysed on SDS/PAGE. The fractions containing relatively pure protein were pooled and dialysed against the appropriate TBS-Ca overnight at 4 °C.
Purification of CBM–cohesin
The supernatants containing CBM–cohesin were mixed with 15 ml of amorphous cellulose in a 50 ml tube for 1 h, on a rotator at 4 °C. The amorphous cellulose was pelleted by centrifugation (4000 g for 5 min at 4 °C). The pellet was washed three times with 45 ml of TBS containing 1 M NaCl and three times with 45 ml of TBS. Protein was eluted from the amorphous cellulose pellet using 12 ml (two successive batches) of 1% (v/v) triethylamine. Eluted fractions were quickly neutralized to pH 7 with 1.5 ml of 1 M Mes at pH 5.5, and the purity of each fraction was estimated by SDS/PAGE (15% gels). To remove the triethylamine and change buffer the protein was dialysed against the appropriate buffer overnight at 4 °C.
ELISA binding/affinity assay
The ELISA-based cohesin–dockerin binding assay was performed essentially according to the method of Barak et al.  with minor modifications. MaxiSorp 96 well ELISA plates (Nunc A/S) were coated overnight at 4 °C with 0.03–100 ng of the desired protein (CBM-Coh-Ct) in 0.1 M Na2CO3 (pH 9.0; 100 μl/well). The following steps (unless otherwise stated) were all performed at a volume of 100 μl/well at room temperature (25 °C). The coating solution was discarded and blocking buffer (TBS containing 1 mM CaCl2 and 1% BSA) was added, and the plates were incubated for 1 h, after which time the blocking buffer was discarded. After the incubation period, the plates were washed three times with washing buffer (blocking buffer supplemented with 0.05% Tween 20 without BSA; 200 μl/well per wash). Wild-type or mutant forms of XynDocS (0.03–100 ng), diluted in blocking buffer or, where appropriate, in blocking buffer supplemented with 1 mM EDTA, were added. Plates were again incubated for 1 h and washed three times. Primary antibody (rabbit anti-XynT6, 1:10000), diluted in blocking buffer was added and incubated for 1 h. Plates were washed, and secondary antibody [HRP (horseradish peroxidase)-labelled-goat-anti-rabbit] diluted in blocking buffer (1:10000) was added for the final incubation of 1 h, and plates were again washed three times. The colour reaction was obtained by addition of 100 μl/well TMB (tetramethylbenzidine) substrate-chromogen and terminated upon addition of 50 μl/well of 1 M H2SO4, 0.5–5 min after the addition of TMB. Absorbance was measured at 450 nm using a tunable microplate reader (OPTImax, Molecular Devices).
Progressive truncation of the dockerin module
A series of homologous deletions in both the N- and C-terminal portions of the xylanase-fused dockerin modules was performed (Figure 2). For the present studies, the dockerin of the family-48 C. thermocellum cellulase, Cel48S (termed DocS), was chosen as a representative type-I dockerin. The resultant truncated dockerin derivatives were examined for their binding to cohesin, compared with the wild-type dockerin (Doc WT), using an efficient ELISA-based approach, together with a matching fusion protein system . The method enabled the comparative binding analysis of numerous dockerin samples to cohesin under standardized conditions.
Sequence alignment of progressively truncated xylanase-fused dockerin derivatives
Previous studies have suggested that the linker sequence preceding the dockerin module is important for its interaction with cohesin [26,27]. It was thus of interest to evaluate its contribution by incremental deletion. The eight-residue linker segment, connecting the dockerin to the xylanase T6 carrier protein, was progressively removed (Figure 2A). The results indicated that all of the latter truncated XynDocS proteins bound to cohesin in a manner very similar to that of the wild-type dockerin (Figure 3A), even after subsequent removal of the three extraneous residues (Val-Val-Pro), artificially inserted for cloning purposes, B(Δ′8), and the highly conserved glycine residue that precedes the first calcium-binding loop, B(Δ′9). In contrast with previous indications [26–28], the present results suggest that the linker is not a crucial element in cohesin–dockerin binding, although within the context of the intact native cellulosome, both the linker and the conserved glycine residue are likely to have a significant contribution to cellulosome action.
Relative binding of truncated dockerins
The internal 2-fold symmetry and distinctive sequence homology between the two anti-parallel ‘F-hand’ motifs [17,18] implies that both dockerin motifs could theoretically interact with the cohesin in an identical manner. Nevertheless, the crystal structure of the cohesin–dockerin heterodimer indicated that the interaction with cohesin is asymmetric. Indeed, Carvalho et al.  have recently crystallized a variant heterodimer by mutating critical recognition residues in the second repeat. In the latter complex, the mutated dockerin is rotated 180° relative to the wild-type dockerin, such that the first duplicated segment participates in most of the same interactions with its protein partner. It was therefore of interest whether we could delete more of the residues in the first repeat without severely compromising cohesin recognition (Figure 2B). Indeed, the first calcium-binding loop, including recognition residues Ser10 and Thr11, could be entirely deleted with near-full retention of cohesin binding (Figure 3B). Additional deletions of residues within helix-1 resulted in a marked reduction in cohesin binding.
The initial crystal structure of the cohesin–dockerin complex  showed that the second dockerin repeat (helix-3) dominates cohesin binding. Thus a progressive series of truncated dockerins from the C-terminus (Figure 2C) was examined for cohesin binding. Specifically, it was of interest to assess whether the integrity of the second loop–helix repeat is of preferential importance to cohesin binding over that of the first repeat. The results (Figure 3C) demonstrate that the residues upstream of Lys58 are indispensable for binding. Interestingly, Arg59 (one of the important contact residues responsible for the species-specificity ) appears not to be absolutely essential for the full binding of the truncated dockerin.
Using the information described above, we have tried to combine the N- and C-terminal deletions in an attempt to create a minimized version of an active and viable dockerin. An abbreviated dockerin, D(Δ14), comprising only 37 residues was thus constructed by combining B(Δ21)* with C(Δ14). However, this combined truncated form failed to bind the cohesin probe. Using the same B(Δ21)* as a starting point, we designed progressive truncated derivatives (Figure 2D), all of which displayed only negligible levels of activity (Figure 3D). Even a single deletion of the last residue (asparagine) of D(Δ1) essentially served to abolish cohesin recognition.
Alanine scanning of calcium-co-ordinating residues
Previous studies have indicated that the type-I dockerin module can bind two calcium atoms which are indispensable for cohesin recognition via its calcium-binding loop [19,30]. Alignment of 86 different C. thermocellum dockerin sequences  revealed the extent of conservation within the repeated F-hand motifs (Figure 4). In addition to the established recognition and hydrophobic residues, the five known calcium-co-ordinating residues are remarkably conserved in both motifs: i.e. aspartate, asparagine, aspartate, asparagine and aspartate at positions 1, 3, 5, 9 and 12 respectively (or, in short, DNDND). It was important to assess the contribution of these residues to cohesin recognition. For this purpose, the relevant aspartate and asparagine residues were subjected to alanine scanning, either alone or in combination, using the same representative type-I dockerin (DocS) and an ELISA-based affinity assay.
WebLogo alignment of the F-hand motif
A series of 17 mutated dockerins (as XynDoc fusion proteins) were thus cloned, expressed and purified (Table 1). In group A, single mutations were systematically performed at each position in the two calcium-binding loops. In group B, double mutations were carried out on either the first or second loop, at co-ordinates x and −z. In group C, double mutations, one in each loop on the same co-ordinate, were performed. The designated mutated proteins were subjected to the ELISA-based affinity assay to analyse their binding to cohesin (Figure 5). Single mutations at all co-ordinates in either loop failed to cause any significant change in cohesin recognition (Figure 5A). Likewise, two mutations in the same loop had only a minor effect on cohesin binding (Figure 5B). In contrast, two mutations, one in each loop, significantly weakened the cohesin–dockerin interaction (Figure 5C), except in the case of the two mutated residues that participate in water-mediated calcium-binding (C4). These results suggest that as long as one calcium-binding loop is intact, cohesin recognition is maintained. Impairment of both calcium-binding motifs severely affects the cohesin–dockerin interaction.
|Double mutations in one calcium-binding loop||B1||A||N||D||N||A||D||N||D||N||D|
|Double mutations: one mutation in each calcium-binding loop||C1||A||N||D||N||D||A||N||D||N||D|
|Double mutations in one calcium-binding loop||B1||A||N||D||N||A||D||N||D||N||D|
|Double mutations: one mutation in each calcium-binding loop||C1||A||N||D||N||D||A||N||D||N||D|
Relative binding of dockerins mutated in their calcium-co-ordinating residues
Apo-dockerin recognition of cohesin
It has been shown previously  that EDTA inhibits complex formation between cohesin and dockerin. The assay was therefore repeated whereby the interaction step was supplemented with 1 mM EDTA (Figure 6). EDTA had a striking effect on wild-type dockerin binding to cohesin: the affinity was reduced by over an order of magnitude. The mutated dockerins of Group A, however, were even more affected than the wild-type, all of which essentially failed to bind cohesin. This result is dramatically different from those presented in Figure 5(A), wherein all of the single-mutated dockerins bound to cohesin in a similar manner to the wild-type dockerin. We can conclude from these results that although a single mutation presumably impairs the calcium-binding affinity of the F-hand motif, the relatively low affinity of the mutants is compensated for under conditions of excess (1 mM) CaCl2.
Calcium-dependent binding properties of mutated dockerins
Three-dimensional structural analysis of cohesins and dockerins, either alone or in complex, has provided critical information regarding their interaction on the atomic level. The initial crystal structures of individual type-I cohesin modules from the C. thermocellum and C. cellulolyticum scaffoldins [15,16] revealed the overall shape and structural characteristics of the molecule, but the structures failed to provide insight into their dockerin-binding sites. Later, the solution structure of the free C. thermocellum type-I dockerin module from cellobiohydrolase Cel48S  was solved, revealing two duplicated F-hand motifs, each consisting of a calcium-binding loop in a classical pentagonal bi-pyramidal sphere arrangement, followed by an α-helix. The calcium-binding loop comprises 12 residues, of which five of the side chains (aspartate and asparagine residues) serve to co-ordinate the calcium ion inside the loop. Situated on each loop at positions 10 and 11 are two exquisitely conserved residues (serine and threonine/serine respectively), which were proposed to serve as species-specific ‘recognition residues’ [13,32]. Calcium titration indicated that calcium-binding induces folding of the dockerin  and is required for cohesin recognition and stability . Although it was clear that the calcium-binding loops play a pivotal role in complex formation and cohesin recognition, it was still unclear how the dockerin binds selectively to its cohesin counterpart.
A crucial contribution to our understanding of complex formation was achieved when Carvalho et al.  succeeded in crystallizing the type-I cohesin–dockerin complex. The detailed crystal structure of the complex revealed that the dockerin prefers binding to cohesin through its second duplicated segment. This asymmetric binding interaction was incompatible with the sequence and structure symmetry observed for the dockerin module. A dual mode of binding was thus suggested and demonstrated by mutating two of the recognition residues in the second F-hand motif; the resultant mutant preferred binding to cohesin through its first duplicated segment .
Despite the detailed structural information on the cohesin–dockerin interaction, many aspects concerning the binding have yet to be elucidated. Previous reports [29,32] have concentrated on the contribution of the conserved recognition residues, proposed to govern specificity for cohesin binding on the basis of bioinformatic analyses. In the present study, we have addressed the involvement of other regions of the dockerin by progressive truncation and site-directed mutagenesis, in order to assess the importance of the internal dockerin symmetry to normal cohesin binding.
Previous studies have suggested that the linker segment, which connects the parent catalytic module to the dockerin module, is important for cohesin binding [26–28]. However, in the present study the linker could be omitted entirely (including the strictly conserved glycine residue that immediately precedes the calcium-binding motif), without any apparent reduction in cohesin binding. The apparent discrepancy with our results could be ascribed to the fact that the previously described dockerin derivatives were constructed in the free form, emphasizing the significance of their linker, whereas the dockerins in the present study are fused to a stabilizing carrier protein. Additional truncation into the first repeat revealed that the entire calcium-binding loop and several residues of the helix could be deleted (Figure 3B) without drastically affecting cohesin binding. In total, approx. 23 residues could be deleted from the N-terminus of the dockerin module with significant retention of cohesin binding. The continuity of the C-terminus proved more critical to dockerin structure. Deletion of the terminal 15 dockerin residues failed to destroy binding to cohesin, even though one of the strictly conserved contact residues (Arg62) was deleted. Additional truncation of the C-terminus served to eliminate binding, and the dockerin module was vulnerable to combined truncation at both ends. In some respects, these results are rather surprising. Why would the linker that connects the dockerin to the enzyme be unnecessary? Why would certain residues of the dockerin sequence be strictly conserved in over 80 known dockerin sequences of this organism? Why would there be a repeated loop–helix motif if one of them can be deleted without affecting cohesin binding?
In the case of the linker and the conserved glycine residue, these elements may be required for a purpose other than cohesin binding. For example, a flexible linker and the conserved glycine residue may be important for correct orientation of the catalytic module of the parent enzyme in the cellulosome complex [34,35]. Regarding deletion of the loop–helix motif, the near-perfect internal symmetry of the dockerin, with respect both to sequence and structure [13,36,37], enables it, theoretically and practically, to bind to a cohesin in two independent and opposite orientations [18,19,38–41].
Systematic site-directed mutagenesis of individual calcium-binding residues in the calcium-binding loop of the dockerin also supports functional asymmetry of binding with the type-I cohesin counterpart. Single mutations in the dockerin motif cause an apparent reduction in calcium affinity, which is compensated for by conditions of excess calcium. Surprisingly single substitutions (aspartate or asparagine to alanine) did not alter cohesin–dockerin binding under these conditions. These findings contrast with studies on eukaryotic EF-hand proteins (e.g. calmodulin or troponin), wherein a single substitution in the EF-hand calcium-binding motif was found to abolish their biological activity [42,43].
It is interesting to consider why two mutations on the same loop failed to disturb the cohesin–dockerin interaction, whereas one on each of the two loops drastically reduced the binding. This indicates that only one fully functional calcium-binding loop is required to sustain dockerin structure and recognition of its cohesin counterpart. This finding is also supported by the results of the truncation experiments, in which we were able to delete the entire first calcium-binding loop or delete critical residues from the second F-hand motif, while retaining full binding activity. If we assume that the mutations of the calcium-binding loop affect the binding residues in its vicinity on the dockerin module, it thus follows that the interaction is symmetrical, since either loop can be subjected to the double mutation with full retention of binding. The results from both truncation and mutation experiments thus suggest that the major elements responsible for cohesin binding are found in both duplicated segments of the dockerin. We can conclude from these experiments that dockerin symmetry is not a prerequisite for cohesin recognition, as long as one segment is fully functional.
Indeed, these results are mutually compatible with recent structures of the type-I cohesin–dockerin complexes, which support a dual mode of binding . It should be noted that the status of the individual dockerins in the two studies are different. The crystal structures comprised a dockerin in the free state, as opposed to the dockerin-fused enzyme used herein for the biochemical studies, thus indicating that the dual mode of binding is maintained in the presence of a carrier protein. Moreover, two different type-I dockerins were used in the two studies. The Cel48S dockerin is more typically positioned at the C-terminus of the parent protein as opposed to the internal position of the Xyn10B dockerin . Furthermore, the latter dockerin possesses a few uncharacteristic residues in normally conserved positions. The functional biochemical findings of the present study therefore strengthen the general claims for asymmetry in binding, despite the inherent sequence and structural symmetry of the type-I dockerin module.
Taken together, the results support the previous contention that the plasticity in the type-I cohesin–dockerin interaction may be suggestive of optional steric states required to incorporate in situ a dockerin-bearing enzyme into the scaffoldin during de novo assembly of the cellulosome by the bacterial cell . Alternatively, the observed redundancy in the cohesin–dockerin interaction may enable a conformational switch to occur upon binding of the cellulosome complex to its substrate, thereby allowing the enzymes more freedom to distribute on the plant cell wall and act on their preferred substrate in the natural setting [19,40,41].
Finally, the observed plasticity in binding is a direct consequence of the type-I dockerin symmetry. Clearly this is not necessarily a universal phenomenon. Indeed, the sequence of the type-II dockerin of the C. thermocellum CipA scaffoldin is asymmetric, and its interaction with a matching type-II cohesin involves both dockerin helices over their entire length. In this respect, the type-II heterodimer exhibits a different mode of binding  than that observed for its type-I counterpart. However, dockerin symmetry is not necessarily type-dependent, since the other known type-II dockerins (from Acetivibrio cellulolyticus and Bacteriodes cellulosolvens) appear to be symmetrical in their sequences [45,46]. Moreover, many of the type-III dockerin sequences from the rumen bacterium, Ruminococcus flavefaciens, are clearly asymmetrical, particularly in their presumed recognition residues [47,48]. One would expect that asymmetry in the dockerin sequence would preclude a dual mode of binding. In the final analysis, crystal and solution structures of additional examples of cohesin and dockerins and their complexes will continue to provide an insight into the diversity of these complementary families of interacting protein modules.
We thank Professor Meir Wilchek, Dr Ely Morag and Mrs Rachel Haimovitz (The Weizmann Institute of Sciences, Rehovot, Israel) for critical reading of the manuscript and Dr Ilya Borovok (Tel Aviv University, Ramat Aviv, Israel) for identification of the type-I dockerin sequences in the C. thermocellum genome. This research was supported by the Israel Science Foundation (grant numbers 422/05 and 159/07) and by grants from the United States-Israel BSF (Binational Science Foundation), Jerusalem, Israel.