Cohesin-dockerin interactions within and between Clostridium josui and Clostridium thermocellum: binding selectivity between cognate dockerin and cohesin domains and species specificity.

The cellulosome components are assembled into the cellulosome complex by the interaction between one of the repeated cohesin domains of a scaffolding protein and the dockerin domain of an enzyme component. We prepared five recombinant cohesin polypeptides of the Clostridium thermocellum scaffolding protein CipA, two dockerin polypeptides of C. thermocellum Xyn11A and Xyn10C, four cohesin polypeptides of Clostridium josui CipA, and two dockerin polypeptides of C. josui Aga27A and Cel8A, and qualitatively and quantitatively examined the cohesin-dockerin interactions within C. thermocellum and C. josui, respectively, and the species specificity of the cohesin-dockerin interactions between these two bacteria. Surface plasmon resonance (SPR) analysis indicated that there was a certain selectivity, with a maximal 34-fold difference in the K(D) values, in the cohesin-dockerin interactions within a combination of C. josui, although this was not detected by qualitative analysis. Affinity blotting analysis suggested that there was at least one exception to the species specificity in the cohesin-dockerin interactions, although species specificity was generally conserved among the cohesin and dockerin polypeptides from C. thermocellum and C. josui, i.e. the dockerin polypeptides of C. thermocellum Xyn11A exceptionally bound to the cohesin polypeptides from C. josui CipA. SPR analysis confirmed this exceptional binding. We discuss the relationship between the species specificity of the cohesin-dockerin binding and the conserved amino acid residues in the dockerin domains.

Cellulosomes, which have been identified and characterized in cellulolytic clostridia such as Clostridium thermocellum (1), C. cellulolyticum (2), C. cellulovorans (3), and C. josui (4), ruminal bacteria (5), and anaerobic fungi (6), are defined as multienzyme complexes having high activity against crystalline cellulose and related plant cell wall polysaccharides such as xylan, mannan, and pectin. A common feature of the clostridial cellulosomes is that they consist of a large number of catalytic components arranged around noncatalytic scaffolding proteins. The scaffolding proteins have been identified as CipA (or scaffoldin) in C. thermocellum (7), CipC in C. cellulolyticum (8), CbpA in C. cellulovorans (9), and CipA in C. josui (4). These proteins fundamentally consist of repetitive noncatalytic domains of about 140 residues, termed cohesin domains, and a carbohydrate-binding module (CBM), 1 e.g. C. josui CipA is composed of an N-terminal family 3 CBM followed by a hydrophilic domain and six cohesin domains (see Fig. 1A); and C. thermocellum CipA contains a CBM between the second and third of nine repeated cohesin domains and a type II dockerin domain at its C terminus (Fig. 1A). The amino acid sequences of all the cohesin domains from these bacteria are strikingly similar to each other, especially within the same species (Fig. 1, B and C). Each cohesin domain is a subunit-binding domain that interacts with a docking domain, called dockerin, of each catalytic component. Pagès et al. (10) found that C. thermocellum Cel48A (formerly CelS) did not interact with the first cohesin domain of C. cellulolyticum CipC and C. cellulolyticum Cel5A (formerly CelA) did not recognize the second cohesin domain of C. thermocellum CipA. This phenomenon was recognized as species specificity of cohesin-dockerin interactions. Dockerin domains consist of a pair of well conserved 22-residue repeats spaced by a linker of 8 -18 residues (see Fig. 2A). In all cases, the 11th and 12th residues of both segments are conserved within the same type of dockerin domain but are different between different types, i.e. the 11th and l2th residues are a serine residue in the first position and either a serine or a threonine in the second for C. thermocellum dockerin domains, and an alanine and a hydrophobic residue (usually leucine or isoleucine) for C. cellulolyticum dockerin domains, suggesting that they are involved in determining the binding specificity (11) between C. thermocellum and C. cellulolyticum. This was demonstrated by swapping these residues between the dockerin domains of C. thermocellum Cel48A and C. cellulolyticum Cel5A: each of the mutated dockerins acquired the ability to bind to its noncognate cohesin domain but did not lose the affinity for its cognate cohesin domain, suggesting that the conserved 11th and 12th residues do indeed play a role in biorecognition, whereas additional residues may also contribute to the specificity of the interaction (10).
On the other hand, a lack of specificity in the cohesin-dock-erin interactions within the same species was shown by nondenaturing PAGE, e.g. C. thermocellum Cel9A (formerly CelD) and Cel5C (formerly CelE) bound to the purified CipA at molar ratios up to 9:1 (12,13), and C. thermocellum Cel48A bound nonselectively to four cohesin polypeptides from C. thermocellum CipA, which were separately produced by Escherichia coli recombinants. These observations suggested that the dockerin domains of C thermocellum Cel9A, Cel5C, and Cel48A could bind to any cohesin domains in the cognate CipA with a similar affinity. On the other hand, the second and third cohesin domains of C. thermocellum CipA interacted with all the catalytic subunits, except for the largest catalytic subunit now known as Cel9D-Cel44A (formerly CelJ) (14), of the cellulosome in an affinity blotting analysis (14), suggesting that generally there is no selectivity in the interactions among cohesin and dockerin domains within the same species, but that there may be at least one exception. Because Cel9D-Cel44A is known as a component of the cellulosome, the cohesin domains, except for the second and third ones, should associate with the dockerin domain of Cel9D-Cel44A, i.e. it is possible that the dockerin domains preferentially bind to particular cohesin domains of CipA. However, there have been no reports containing a quantitative analysis of the interaction between a dockerin domain and different cohesin domains within the same species other than C. cellulolyticum (8) or describing an exception to the species specificity of the cohesin-dockerin interactions between different bacteria. C. josui is closely related to C. cellulolyticum taxonomically (4), although their optimum growth temperatures are different, i.e. 45°C for the former but 32-35°C for the latter. The gene cluster of the former, including cipA and a number of genes responsible for catalytic components, is identical to that of the latter with respect to the gene order, and the corresponding genes are highly homologous with each other between the two bacteria. Furthermore, the dockerin domains of C. josui are specified by the two pairs, including an alanine in the first position and a hydrophobic residue (usually leucine or isoleucine) in the second ( Fig. 2A) like the dockerin domains of C. cellulolyticum. Therefore, we predicted that the interactions among cohesins and dockerins between C. thermocellum and C. josui would be species-specific. In this study, however, we unexpectedly found that the dockerin domain of C. thermocellum Xyn11A exceptionally interacted with all the tested cohesin domains of C. josui CipA, whereas that of C. thermocellum Xyn10C did not, and furthermore those of C. josui Aga27A and Cel8A (formerly CelB) showed no affinity for any of the cohesin domains of C. thermocellum CipA.
Here, we report the results of qualitative and quantitative analyses of the cohesin-dockerin interactions within C. thermocellum and C. josui, respectively, and describe an exception to the species specificity of the cohesin-dockerin interactions between these two bacteria, i.e. the binding of the dockerin domain of C. thermocellum Xyn11A to the cohesin domains of C. josui CipA.
Plasmids and Plasmid Constructions-Plasmids pR1, pR2, and pR3 are derivatives of pRSETB (Invitrogen, San Diego, CA) and contain DNA fragments encoding the first, second, and third cohesin domains (referred to as Coh1-Ct, Coh2-Ct, and Coh3-Ct in this report), respectively, of C. thermocellum CipA (17). Plasmid pCip7 is a derivative of pQE-31 (Qiagen) carrying a DNA fragment encoding the seventh cohesin domain of C. thermocellum CipA (18). Plasmid pET-28a(ϩ) was purchased from Novagen (Madison, WI). The plasmid used to produce the fourth cohesin domain of C. thermocellum CipA was constructed as follows: pCip102 (13) was digested with Sau3AI, and the 984-bp fragment was recovered from an agarose gel using a GeneClean II kit (BIO101, Vista, CA) after agarose gel electrophoresis. The DNA region encoding Coh4-Ct was amplified by PCR from the recovered 984-bp fragment with KOD dash DNA polymerase (Toyobo, Osaka, Japan) and a combination of PCR primers, CtCoh4-F and CtCoh4-R (Fig. 1D). The resulting PCR fragment was cloned into pT7Blue (Novagen) as specified by the supplier. After sequencing the inserted DNA fragment for confirmation of the absence of mutation, the inserted fragment was cleaved from the plasmid by digestion with NdeI and SalI, and transferred to pET-28a(ϩ) (Novagen), yielding pET-R4. Construction of plasmids for the production of the first, second, fifth, and sixth cohesin domains (Coh1-Cj, Coh2-Cj, Coh5-Cj, and Coh6-Cj) of C. josui CipA (4) was carried out using similar procedures to those used for the construction of pET-R4. Different combinations of PCR primers containing an artificial NheI or SalI recognition sequence were used (Fig. 1D), i.e. CjCoh1-F and CjCoh1-R for Coh1-Cj, CjCoh2-F and CjCoh2-R for Coh2-Cj, CjCoh5-F and CjCoh5-R for Coh5-Cj, and CjCoh6-F and CjCoh6-R for Coh6-Cj. PCR was performed with pMK-2 as the template, and the resultant PCR fragments were cloned between the NheI and SalI sites of pET-28a(ϩ), yielding pCoh1-Cj, Coh2-Cj, pCoh5-Cj, and pCoh6-Cj. The DNA regions encoding the dockerin domains of C. thermocellum Xyn11A (19) and Xyn10C (20) and C. josui Aga27A (21) and Cel8A (22) were amplified from pKS101, pKS103, pCj-Aga27A, and pUCJ2E, respectively, with different combinations of PCR primers containing an artificial NheI or SalI recognition sequence, i.e. Ct-Xyn11ADocF and Ct-Xyn11ADocR for the dockerin domain of Xyn11A, Ct-Xyn10CDocF and Ct-Xyn10CDocR for that of Xyn10C, Cj-AgaADocF and Cj-AgaADocR for that of Aga27A, and Cj-Cel8ADocF and Cj-Cel8ADocR for that of Cel8A (Fig. 2B). The resultant PCR fragments were cloned between the NdeI and SalI sites of pET-28a(ϩ), yielding pCt-Xyn11ADoc, pCt-Xyn10CDoc, pCj-Aga27ADoc, and pCj-Cel8ADoc, respectively. All plasmids were checked for PCR fidelity by DNA sequence analysis.
Production and Purification of Recombinant Proteins-All recombinant polypeptides produced by E. coli were referred to as rCoh1-Ct, rCoh2-Cj, rXyn11ADoc, and others in this study. E. coli BL21(DE3)RIL or M15(pREP4) harboring a recombinant plasmid was grown at 37°C to A 600 ϭ 0.6 in Luria-Bertani broth supplemented with kanamycin (50 g/ml) and chloramphenicol (25 g/ml), or ampicillin (50 g/ml), and then isopropyl-␤-D-thiogalactopyranoside was added to the culture to give a final concentration of 0.8 mM. After an additional incubation of 2 h, the cells were centrifuged (3,000 ϫ g, 10 min) and resuspended in lysis buffer (50 mM Na 2 HPO 4 pH 8.0, 0.3 M NaCl, 10 mM imidazole). The cells were disrupted by ultrasonication, and cell debris was removed by centrifugation (15,000 ϫ g, 10 min). The recombinant proteins, containing a His 6 tag at their N termini, were purified as follows. The cell-free extracts were mixed with 0.5 ml of nickel-nitrilotriacetic acid resin (Qiagen) equilibrated with lysis buffer, and the resin was subsequently washed with wash buffer (50 mM Na 2 HPO 4 , pH 8.0, 0.3 M NaCl, 20 mM imidazole). The recombinant protein adsorbed onto the resin was eluted with elution buffer (50 mM Na 2 HPO 4 , pH 8.0, 0.3 M NaCl, 250 mM imidazole), followed by dialysis against 50 mM HEPES buffer (pH 7.5). The protein concentration was routinely estimated with bovine serum albumin (BSA) as the standard using a Micro BCA Protein Assay Reagent Kit (Pierce, Rockford, IL).
Surface Plasmon Resonance (SPR) Analysis-SPR with a BIAcore 2000 (BIAcore AB, Uppsala, Sweden) was used to determine the kinetic and equilibrium constants of the cohesin-dockerin interactions at 25°C. HBS buffer (10 mM HEPES, pH 7.4, 0.15 M NaCl) was used as an immobilization buffer, and 10 mM CaCl 2 , 0.005% surfactant P20 (BIAcore), and 50 mM Tris-maleate buffer, pH 6.5, was used as a running buffer, as described previously for SPR analysis of C. cellulolyticum proteins (23). Each of the cohesin polypeptides tested was immobilized on a dextran matrix with free carboxylic groups (CM5 chip, BIAcore) employing conventional carbodiimide coupling chemistry and subsequent deactivation of excess active esters using ethanolamine (EDC/ NHS coupling kit, BIAcore). The purified cohesin polypeptides were injected for 10 s resulting in ϳ400 -500 resonance units of immobilized protein. Two flow cells were prepared for the kinetic experiments: in one cell, one of the cohesin polypeptides (rCoh1-Ct, rCoh2-Ct, rCoh3-Ct, rCoh4-Ct, rCoh7-Ct, rCoh1-Cj, rCoh2-Cj, rCoh5-Cj, or rCoh6-Cj) was immobilized and an adjacent cell was used as a reference with no treatment. The flow cells were routinely equilibrated with running buffer. The ligands (rAga27ADoc, rCel8ADoc, rXyn11ADoc, and rXyn10CDoc) were diluted in running buffer and allowed to interact with the sensor surface by a 240-s injection. In all cases, at least four different concentrations (2.1-13 nM) of ligands were injected at a flow rate of 30 l per min. Kinetic evaluation was performed with BIAevaluation 3.1 software (BIAcore), fitting the four curves in each concentration series using the global fitting method. The obtained sensor grams were evaluated with the biomolecular interaction analysis evaluation software (BIAcore) to calculate the kinetic constants. The data were interpreted on the basis of the simple model, L ϩ A 7 LA, where L denotes the mobile ligand and A is the immobilized receptor. With the exception of Xyn11A, the surface of the sensor chip was efficiently regenerated by injection of 10 mM HCl between each protein injection.
Other Procedures-The cellulosome fractions of C. thermocellum F1 and C. josui grown on BMC (ball-milled cellulose) were, respectively, prepared from the culture supernatants as described previously (4,24). SDS-PAGE was performed by the method of Laemmli (25). Preparation of the antiserum against C. thermocellum Xyn11A and immunoblotting were carried out as described previously (19).

Production and Purification of Various Cohesin and Dockerin Polypeptides-C. thermocellum
CipA consists of nine cohesin domains, a family 3 CBM, and a type II dockerin domain (Fig. 1A). Among these cohesin domains, Coh3-Ct, Coh4-Ct, Coh5-Ct, Coh6-Ct, Coh7-Ct and Coh8-Ct are highly homologous with each other, and Coh4-Ct, Coh5-Ct, Coh6-Ct, and Coh8-Ct, in particular, show high sequence identity (higher than 99%) among them (Fig. 1B). In contrast, Coh1-Ct, Coh2-Ct, and Coh9-Ct are relatively less homologous with the other cohesin domains. Therefore, we chose Coh1-Ct, Coh2-Ct, Coh3-Ct, Coh4-Ct, Coh7-Ct, and Coh9-Ct to quantitatively measure the cohesin-dockerin interactions, but unfortunately the Coh9-Ct polypeptide could not be purified. On the other hand, C. josui CipA contains six cohesin domains in addition to a family 3 CBM and a hydrophilic domain (Fig. 1A). Among these cohesin domains, Coh2-Cj, Coh3-Cj, Coh4-Cj, and Coh5-Cj are highly homologous with each other, whereas Coh1-Cj and Coh6-Cj are less similar to the other cohesin domains (Fig. 1C). Therefore, rCoh1-Cj, rCoh2-Cj, rCoh5-Cj, and rCoh6-Cj were produced and used for quantitative analysis of the cohesindockerin interactions. The dockerin domains of C. thermocellum Xyn11A and Xyn10C, and C. josui Aga27A and Cel8A have typical amino acid residues in their 11th and 12th positions, i.e. "ST" in the first segment and "ST" in the second segment in Xyn11A, "SS" and "ST" in Xyn10C, "AL" and "AL" in Aga27A, and "AL" and "AI" in Cel8A (Fig. 2A). These dockerin domains were produced with three amino acid residues upstream of the first conserved residue of each dockerin domain. All the recombinant proteins used in this study contained a His tag at their N termini and were found in the soluble fraction of the recombinant E. coli lysate. We obtained homogeneous purified polypeptide preparations by a single step purification using nickel-nitrilotriacetic acid resin, by keeping the amount of resin lower than that necessary to bind all the His-tagged polypeptides.
Qualitative Analysis of Interactions between Cohesin and Dockerin by Affinity Blotting-It was interesting to determine whether the five cohesin polypeptides from C. thermocellum CipA and the four cohesin polypeptides from C. josui CipA interacted nonselectively with cellulosomal proteins within the same species or whether a given cohesin polypeptide interacted specifically with a single or a small number of cellulosomal proteins. Furthermore, we were interested to learn whether the cohesin polypeptides from either C. thermocellum or C. josui interacted specifically with their cognate dockerin domains or whether they interacted with noncognate dockerin domains beyond the species specificity.
To address these questions, an affinity blotting analysis was carried out using various cohesin and dockerin polypeptides as probes. Total cellulosomal proteins from C. thermocellum and C. josui were subjected to SDS-PAGE and transferred onto PVDF membranes. The separated bands were then challenged with the cognate and noncognate cohesin polypeptides rCoh1-Cj, rCoh2-Cj, rCoh5-Cj, rCoh6-Cj, rCoh1-Ct, rCoh4-Ct, and rCoh7-Ct. As shown in Fig. 3A (lane 1), all the cohesin polypeptides from C. thermocellum appeared to interact with almost all of the C. thermocellum cellulosomal proteins. It is noteworthy that these cohesin polypeptides did not detect the largest catalytic component Cel9D-Cel44A. Affinity blotting analysis using a recombinant dockerin polypeptide derived from Cel9D-Cel44A failed to detect the interaction between the dockerin polypeptide and C. thermocellum CipA (data not shown). In all cases, the labeling pattern was almost identical. Similar phenomena were observed in the interactions between the cohesin polypeptides and cellulosomal proteins from C. josui (Fig. 3B). These observations suggested a lack of selectivity in the interactions between cohesin and dockerin domains within the same species. When the cohesin-dockerin interactions between C. thermocellum and C. josui were examined, none of the cohesin polypeptides from C. josui CipA interacted with any of the C. thermocellum cellulosomal proteins (Fig. 3A), supporting the previous conclusion that there is species specificity in the cohesin-dockerin interactions between the two different species. Unexpectedly, in contrast, the cohesin polypeptides from C. josui CipA interacted with two proteins in the C. thermocellum cellulosome, i.e. 75-and 50-kDa proteins (Fig. 3B). We predicted that these two proteins were xylanases Xyn11A and Xyn11B, because the molecular sizes of Xyn11A and Xyn11B were consistent with those of the two proteins that interacted with the cohesin polypeptides, and Xyn11A and Xyn11B have quite similar dockerin domains ( Fig. 2A), which were expected to have similar characteristics in the cohesin-dockerin interaction.
To prove this prediction, an affinity blotting analysis was performed with rXyn11ADoc as the probe. As shown in Fig. 4, rXyn11ADoc bound to the scaffolding protein CipA of C. josui in addition to C. thermocellum CipA, although rXyn10CDoc, rAga27ADoc, and rCel8ADoc interacted with their cognate scaffolding proteins but not with noncognate scaffolding proteins. Furthermore, the antiserum raised against Xyn11A (21) interacted with two proteins, Xyn11A and Xyn11B, whose sizes were in good agreement with those of the two proteins that interacted with the cohesin polypeptides from C. josui CipA (Fig. 5). The results obtained by these qualitative analyses clearly indicated that species specificity in the cohesin-dockerin interactions was generally conserved between C. thermocellum and C. josui, but that there was at least one exception taking the dockerin domain of C. thermocellum Xyn11A as an example. However, because the affinity blotting analysis includes a step of protein denaturation for SDS-PAGE, the results from this method may not necessarily reflect the cohesin-dockerin interactions in the native state. Therefore, these cohesin and dockerin polypeptides were subjected to quantitative analysis of their interactions by SPR.
Interactions between Cognate Cohesin and Dockerin Polypeptides-The SPR method was used to study the real-time interactions of cohesin polypeptides from C. thermocellum and C. josui with the cognate dockerin polypeptides. All the cohesin polypeptides were covalently immobilized to carboxymethyl groups of the dextran on the surface of the CM5 sensor chip via the amine coupling method. The running buffer (23) contained 10 mM CaCl 2 , because calcium ions are required for the binding of dockerin domains to cohesin domains (26). The surface of the sensor chip could be efficiently regenerated by injection of 10 mM HCl without any apparent damage to the immobilized polypeptides. Tables I and II show the affinity constants for the binding of cohesin and dockerin polypeptides from C. josui and C. thermocellum, respectively. Sensorgrams from which the affinity constants were calculated for the binding of cohesin and dockerin polypeptides from C. josui are shown in Fig. 6.
In the case of the cohesin-dockerin interactions of C. josui, rAga27ADoc, which specifically recognized C. josui CipA in the affinity blotting analysis, associated with rCoh1-Cj, rCoh2-Cj, rCoh5-Cj, and rCoh6-Cj with k on values of 1.8 ϫ 10 5 to 6.9 ϫ   affinity for rCoh2-Cj (K D , 2.4 ϫ 10 Ϫ10 ) and the lowest affinity for rCoh1-Cj (1.9 ϫ 10 Ϫ9 ), meaning that there is about an 8-fold difference in the affinity between the cohesin domains in CipA and the dockerin domain from Aga27A. rCelBDoc also showed preference of rCoh2-Cj over rCoh1-Cj, i.e. rCelBDoc associated with rCoh2-Cj with the K D value of 1.3 ϫ 10 Ϫ10 and rCoh1-Cj with the K D value of 4.4 ϫ 10 Ϫ9 , meaning 34-fold difference in the affinity between the cohesin domains in CipA and the dockerin domain from Cel8A.
On the other hand, when SPR analysis was carried out for the binding of rXyn10Cdoc and the cohesin polypeptides (rCoh1-Ct, rCoh2-Ct, rCoh3-Ct, rCoh4-Ct, and rCoh7-Ct) from C. thermocellum, k on values for these interactions were obtained from their sensorgrams (data not shown), i.e. 1.8 ϫ 10 5 to 6.9 ϫ 10 5 , but k off values exceeded the quantitatively meas-urable range (10 Ϫ4 ) of the BIAcore system (Table II). The affinities of the interactions between rXyn10Cdoc and the cohesin polypeptides appear to be comparable to those obtained for the combinations of cohesin and dockerin polypeptides from C. josui.
Interactions of rXyn11ADoc and the Cohesin Polypeptides from C. josui-Because the affinity blotting analysis indicated that rXyn11ADoc bound to C. josui CipA as described above, the affinity constants for the binding of rXyn11ADoc to the C. josui cohesin polypeptides were determined by SPR (Fig. 6). As shown in Table III, rXyn11ADoc interacted with all the cohesin polypeptides with affinities high enough to detect the interactions (K D , 6.4 ϫ 10 Ϫ9 to 1.7 ϫ 10 Ϫ8 ) but with relatively low affinities compared with those observed in combinations of cognate cohesin and dockerin domains. Although the k on values of rXyn11ADoc for the cognate and noncognate cohesin polypeptides were similar to each other, the k off values of rXyn11ADoc for the C. josui cohesin polypeptides were larger than those for the C. thermocellum cohesin polypeptides, resulting in the increased K D values. On the other hand, SPR analysis confirmed that rXyn10CDoc from C. thermocellum and rAga27ADoc and rCel8ADoc from C. josui did not interact with their noncognate cohesin polypeptides.

DISCUSSION
Although some earlier qualitative analyses using affinity blotting (10,26) and nondenaturing PAGE (11,17) techniques in C. thermocellum suggested a lack of selectivity in the binding of cohesin domains to dockerin domains (26), this suggestion has not been tested by quantitative analysis until now. Before the quantitative analysis, we carried out an affinity blotting analysis and confirmed at a qualitative level that there was no selectivity in the cohesin-dockerin interactions within a given species, in both C. thermocellum and C. josui (Fig. 3, A and B). Simultaneously, however, we found that rXyn11ADoc bound to C. josui CipA and the recombinant cohesin polypeptides (Figs. 3B and 4) beyond the species specificity, which was observed in the cohesin-dockerin interactions between the dockerin polypeptides from C. thermocellum Cel48A and C. cellulolyticum Cel5A, and one cohesin polypeptide from each of C. thermocellum CipA and C. cellulolyticum CipC.
In this study, we used cohesin and dockerin polypeptides   FIG. 5. Identification of Xyn11A amd Xyn11B in the cellulosomal proteins from C. thermocellum using an anti-Xyn11A antiserum. Xyn11A and Xyn11B proteins in Western blot analysis were detected with the polyclonal mouse antiserum raised against the purified Xyn11A (19). Lane 1, the cellulosomal proteins from C. thermocellum detected with the anti-Xyn11A antiserum; lane 2, the purified recombinant Xyn11A detected with the anti-Xyn11A antiserum; lane 3, the cellulosomal proteins from C. thermocellum probed with rCoh2-Cj as described in Fig. 3B. isolated from other functional domains such as the CBM, hydrophilic domain, and catalytic domain, to examine the cohesin-dockerin interactions. Previously, Fierobe et al. (23) reported that the CBM, hydrophilic domain, and linker sequence adjacent to the first cohesin domain of C. cellulolyticum CipC were not critical for binding of the cohesin domain to the dockerin domain of C. cellulolyticum Cel5A but that removal of the linker sequence from the dockerin domain reduced the affinity for a cohesin domain by ϳ600-fold compared with the dockerin containing a linker of 11 amino acids. All the dockerin polypeptides used in this study contained a linker sequence of only three amino acids. All the cohesin polypeptides were cou-pled to a carboxymethyl-dextran layer on the surface of the sensor chip in this study, whereas biotinylated recombinant proteins derived from C. cellulolyticum CipA were used for coupling to a streptavidin-dextran layer in the previous experiments (23). Notwithstanding the differences in the experimental conditions for the SPR analysis, the K D values (1.3 ϫ 10 Ϫ10 to 4.4 ϫ 10 Ϫ9 ) for the interactions between cohesin and dockerin polypeptides from C. josui were comparable with the value (2.5 ϫ 10 Ϫ10 ) for the interaction between the first cohesin domain of CipC and the dockerin domain of Cel5A from C. cellulolyticum, suggesting that the experimental conditions adopted here were suitable for SPR analysis.
The SPR analysis of the cohesin-dockerin interactions revealed the following facts: there is a certain selectivity, with a maximal 34-fold difference in the K D values, in the cohesindockerin interactions within a combination of C. josui, although this was not detected by qualitative analysis; there is at least one exception to the species specificity in the cohesindockerin interactions.
First, when the K D values were compared in C. josui, a 34-fold difference was observed in the combination of rCel-8ADoc and the cognate cohesin polypeptides (Table I). Both the two dockerin polypeptides, rCel8ADoc and rAga27A, from C. josui showed the highest affinity for rCoh2-Cj and the lowest affinity for rCoh1-Cj, suggesting that a particular cohesin domain has a higher affinity for the cognate dockerin domains FIG. 6. Sensorgrams showing interactions between the cohesin polypeptides from C. josui CipA and rAga27ADoc (A) and between he cohesin polypeptides from C. josui CipA and rXyn11ADoc (B). Each cohesin polypeptide, Coh1-Cj (a), Coh2-Cj (b), Coh5-Cj (c), or Coh6-Cj (d), was immobilized on sensor chip CM5 as described under "Experimental Procedures." Sensor chip surfaces were subsequently exposed with different concentrations (shown in this figure) of rAga27ADoc (A) or rXyn11ADoc (B). Surface plasmon resonance was determined as described in the text. than the other cohesin domains. Although the extreme difference in the k on and k off values was not observed between rAga27ADoc and rCel8BDoc of C. josui toward respective cohesin polypeptides, the product of relatively small differences in both of the k on and k off values resulted in the 34-fold difference. This phenomenon is not ascribable to artifacts, e.g. artifactual change in the tertiary structure of the cohesin domains by chemical immobilization to the sensor chips, because rXyn11ADoc showed a higher affinity for rCoh1-Cj than for rCoh2-Cj and the lowest affinity for rCoh6-Cj (Table III). In the combinations of C. thermocellum, the k on values of rXyn11A toward the respective cohesin polypeptides were 5-to 14-fold larger than those of rXyn10C toward the corresponding cohesin polypeptides, suggesting the possibility that there is the binding selectivity of cohesin and dockerin domains within C. thermocellum. In C. josui CipA, five cohesin domains, Coh1-Cj to Coh5-Cj, are highly homologous with each other, but Coh6-Cj is less homologous with the others (Fig. 1). It is interesting that the less homologous cohesin polypeptide, rCoh6-Cj, did not show the lowest affinity for their cognate dockerin polypeptides.
Second, we found for the first time that rXyn11ADoc as a wild-type dockerin bound to noncognate cohesin domains. In the affinity blotting analysis, we confirmed that full-length Xyn11A interacted with C. josui CipA (data not shown), suggesting that the binding of rXyn11ADoc to C. josui CipA is not an artifact. The species specificity in cohesin-dockerin interactions has been qualitatively and quantitatively studied using the dockerin domains of C. thermocellum Cel48A and C. cellulolyticum Cel5A (10). Most dockerin domains, including Cel48A, from C. thermocellum contain an "ST" or "SS" motif in each of the duplicated segments, and most dockerin domains from C. cellulolyticum contain an "AL" or "AI" motif. The importance of these conserved amino acids in the species-specific recognition between the cohesin and dockerin domains was shown by site-directed mutagenesis, e.g. mutation of a single residue from threonine to leucine at the second segment of the Cel48A dockerin changed it from nonrecognition to high affinity recognition (K D , 5 ϫ 10 Ϫ9 ) by a cohesin from C. cellulolyticum, suggesting that the presence or absence of a single decisive hydroxyl group is critical to the observed biorecognition. However, complete displacement of these two amino acid residues in each segment, i.e. "ST" and "ST" to "AL" and "AL" in Cel48A and "AL" and "AF" to "ST" and "ST" in Cel5A, bestowed new affinities for noncognate cohesin domains but did not abolish their original preference for the cognate cohesins, suggesting that additional residues are also important as secondary determinants in the specificity of interaction. These observations are consistent with the phenomenon that rXyn11ADoc containing the "ST" motif can interact with the cohesin domains of C. josui CipA in addition to C. thermocellum CipA, although it is unclear why only the dockerin domain of Xyn11A can recognize the noncognate cohesin domains as counterparts. Before starting this study, we expected that the dockerin domain of Cel9D-Cel44A would interact with the cohesin domains of C. josui CipA, because Cel9D-Cel44A was shown not to react with the second and third cohesin polypeptides from C. thermocellum CipA (26) and the amino acid sequence of its dockerin domain was unusual (14), i.e. it contained A and V in the conserved positions of the first segment. However, in the affin-ity blotting analysis, the cohesin polypeptides from C. josui CipA reacted with only two proteins in the C. thermocellum cellulosome, Xyn11A and Xyn11B, but not Cel9D-Cel44A under the conditions used. In fact, none of the cohesin polypeptides from C. thermocellum CipA reacted with Cel9D-Cel44A in the cellulosome fraction, suggesting that Cel9D-Cel44A is incorporated into the cellulosome by interaction between its dockerin domain and cohesin domain(s) other than those tested in this study or a novel scaffolding protein. From the observations for Xyn11A and Cel9D-Cel44A, it is likely that the species specificity of the cohesin-dockerin interaction is caused by subtle differences in their tertiary structures rather than a small number of conserved residues.
In conclusion, this study indicated that there is a certain binding selectivity in the cohesin-dockerin interactions within a given species and species specificity between C. thermocellum and C. josui but with an apparent exception. The precise molecular determinants responsible for the binding selectivity and species specificity of the cohesin-dockerin interactions should be resolved by tertiary structure analysis of cohesin-dockerin complexes from a given species and from different species.