The ability to bind to collagens is expressed by 70% of Lactobacillus
), and it appears that lactobacilli express multiple adhesin types interacting with these abundant tissue proteins. Roos et al. (34
) recently described Cnb, a 29-kDa protein of L. reuteri
with an affinity to solubilized type I collagen. Cnb is a putative ABC transporter protein and has no significant sequence homology to CbsA. Our results confirm the function of the S-layer protein CbsA of L. crispatus
as a collagen-binding protein and show that adhesion to collagens is not expressed by all S-layer proteins of Lactobacillus
isolates in the Acidophilus group. Despite its common occurrence, the biological functions of collagen binding by lactobacilli remain open. Our finding that the S-layer-protein-expressing cells of JCM 5810 adhered to collagen-rich regions in the colon tissue of chicken, the natural host for JCM 5810, supports the notion that collagen binding represents a true tissue-binding property of L. crispatus
. Adhesiveness to collagens has been proposed to enhance Lactobacillus
colonization at tooth surfaces and to promote the spread of lactobacilli from the oral cavity to cause infective endocarditis (14
). However, lactobacilli are most numerous in the intestine, and the role of collagen binding in colonization by intestinal lactobacilli remains to be determined.
The ability of CbsA to bind collagens was reduced by ca. 20% by the N-terminal fusion to the 6-mer His-tag peptide, as indicated by the collagen-binding efficiency of CbsA from lactobacilli and His-CbsA from recombinant E. coli
. Both CbsA and His-CbsA were solubilized by guanidine hydrochloride during their isolation and, when dialyzed into a physiological buffer, formed morphologically highly similar S-layer-like structures, which also resembled those seen by electron microscopy on the JCM 5810 cell surface. Refolding into a morphologically or functionally correct form in a detergent- or chaotropic agent-free medium has also been observed with other lactobacillus S-layer proteins (21
), as well as with the collagen-binding S-layer-like protein YadA of Y. enterocolitica
). Binding of radiolabeled collagen by His-CbsA from E. coli
and by CbsA from L. crispatus
and on the other hand by JCM 5810 cells (43
) was similarly inhibited by unlabeled extracellular matrix proteins. These findings indicate that the N-terminal fusion and expression in a heterologous host species did not significantly change the collagen-binding specificity or polymerization of CbsA into a crystalline layer.
It appears that CbsA contains two peptide regions binding to collagens, the major one at 1 to 287 reacting with type I and IV collagens, and another, weaker type I collagen-binding site within the C-terminal region. The His-CbsA288–410 peptide bound to immobilized type I collagen but not to type IV collagen, and its binding to solubilized type I collagen was barely detectable. It could be that the peptide recognizes the immobilized form of collagen or, more probably, due to low affinity it needs a high receptor density in the solid-phase assay. The C-terminal regions in Lactobacillus S-layer proteins are conserved, but the binding levels of His-SlpnB, His-CbsB, and His-SlpA to immobilized or solubilized collagens are very low. The C-terminal His-CbsA288–401 peptide did not form an S-layer, and it remains open whether this low binding activity represents a cryptic binding site exposed in the peptide or whether it is indeed another collagen-binding site within the CbsA S-protein.
It appears that the major collagen-binding region within CbsA is large and that a crystalline structure with sheet morphology is important for optimal binding. This is suggested by the behavior of the His-CbsA peptides and mutated proteins as well as the His-CbsA/SlpA and His-CbsA/SlpnB hybrid molecules; i.e., those that did not form an S-layer also failed to bind collagens. This is different to what has been found with the laminin- and fibronectin-binding S-layer protein A of A. salmonicida
, where a soluble 38-kDa N-terminal peptide from the 51-kDa A-layer protein was found to express adhesiveness but did not assemble into a tetragonal array (11
). Similarly, the 302-amino-acid collagen-binding region in the S-layer-like protein YadA of Y. enterocolitica
has been functionally expressed as an internal peptide fusion in flagella of E. coli
). Collagen binding and crystalline-layer formation were exhibited by the His-CbsA1–287 peptide but not by His-CbsA1–212, His-CbsA1–250, or His-CbsA42–287, indicating that the region from 1 to 287 in CbsA contains information for both processes. It is interesting that the substitution mutants KSDV257TANN and V260N failed to adhere to collagens and predominantly exhibited a crystalline layer with a cylinder-like morphology. This finding also indicates the importance of the sheet-like crystalline structure in correct expression of the CbsA residues interacting with collagens. The assembled structures of CbsA and mutated proteins must be analyzed in more detail. Our ongoing work by Fourier analyses of the electron microscopic images has suggested that the CbsA crystalline layers have oblique structure and that the KSDV256TANN substitution mutant indeed forms a crystal structure different from that of CbsA.
Binding analyses of the His-CbsA peptides and mutants indicate that the major collagen-binding region is within the 287 N-terminal residues of CbsA; however, the role of the extreme N terminus of CbsA in collagen binding remains uncertain. The His-CbsA-derived peptide from 1 to 287 exhibited adhesiveness and formed an S-layer, whereas the peptide from 42 to 287 failed to adhere and to polymerize into an S-layer. The hybrid S-layer proteins His-SlpnB1–19/CbsA29–410 and His-SlpnB1–72/CbsA82–410 exhibited efficient collagen binding, indicating that the N termini could be exchanged between CbsA and SlpnB without a loss of collagen binding. This can be explained either by complementation of residues in CbsA important for binding by the corresponding residues in SlpnB or by a lack of a role of CbsA residues 1 to 72 in the binding. The former hypothesis is supported by the low binding efficiency of the deletion derivative His-CbsAΔ22–26, and the latter hypothesis is supported by the adhesiveness exhibited by the substitution mutant proteins NNN14INL and F19S.
CbsA resembles the other well-characterized collagen-binding bacterial proteins, CNA of S. aureus
), Ace of E. faecalis
), and the S-layer-like protein YadA of Y. enterocolitica
) in that the collagen-binding region is large. In YadA, the binding region is ca. 300 amino acids and probably discontinuous. The crystal structure of the 168-amino-acid collagen-binding domain CBD of CNA revealed that the binding site is located along a groove on a concave β-sheet with a structure complementary to the helical structure of collagen. A similar structural motif was recently identified in Ace (32
). Secondary-structure predictions for CbsA gave a high content (37%) of β-sheet, as also found in CBD and Ace. However, we found only weak identity of CbsA to the other bacterial collagen-binding proteins, and the highest local identity (22 to 24%) was detected in amino acid overlaps at 92 to 172 in CbsA and 88 to 166 in CNA, as well as 211 to 282 in CbsA and 35 to 103 in CNA. These sequences in CbsA are within the binding region but in CNA correspond to sites that are either in the vicinity of or partially within the region important for collagen binding. Furthermore, some of the residues in CBD found to interact with collagen do not have homologs within the CbsA collagen-binding regions. It therefore seems likely that the structural motifs detected in CNA and Ace are not directly applicable to collagen binding by the rigid CbsA S-layer. We found that replacement of residues D130, N226, S258, V260, TA264, and P268A, which are located within the collagen-binding region in CbsA, decreased the binding without affecting the formation of a crystalline layer, suggesting that these residues are involved in the adherence by CbsA.
Very little is presently known about the incorporation of S-layer subunits into a growing S-layer (for a recent review, see reference 12
). Our results suggest that the region from 250 to 287 in CbsA is important for the assembly of subunits into a crystalline layer. First, the peptide from 1 to 287 was the only His-CbsA peptide that polymerized into an S-layer-like structure and aggregated the JCM 5810 cells; the latter probably resulted from binding of the peptide to S-layer covering the JCM 5810 cell surface. On the other hand, the peptide from 42 to 287 failed to form a crystalline layer and to aggregate JCM 5810 cells, indicating a role for the extreme N terminus of CbsA in the assembly. Second, the CbsA-SlpnB hybrids created at residue 250 in CbsA failed to form a crystalline layer whereas the other CbsA-SlpnB hybrids polymerized into a periodic structure. The substitution mutant proteins His-CbsA KSDV257TANN and His-CbsA V260N formed a crystalline structure with different morphology from that of His-CbsA. A search for protein motifs in lactobacillus S-layer proteins indicated that this region is conserved in Lactobacillus
S-layer proteins, which also contain shorter conserved regions in the N terminus. Such conserved blocks might be important for correct assembly of the rigid S-layer structure. Our results suggest that the extreme C terminus of CbsA plays no significant role in S-layer assembly. This conserved region could be involved in the secretion of the S-layer subunit onto the Lactobacillus
cell surface. It is interesting that the His-CbsA-SlpA hybrids and the His-CbsA-SlpnB hybrids created at site 212 of CbsA behaved differently: the His-CbsA-SlpA hybrids failed to polymerize into a crystalline layer, whereas the His-CbsA-SlpnB hybrids formed a periodic polymer. This suggests that the structural constraints in the S-layer assembly by SlpA and SlpnB are differentially compatible with the CbsA terminal regions, which may reflect the sequence differences at the N-terminal half of these S-proteins.
As noted for many Lactobacillus
), strain JCM 5810 of L. crispatus
has two homologous S-layer genes, cbsA
. We found no evidence for the expression of CbsB under the growth conditions we used, nor was a cbsB
-specific RNA transcript detected, suggesting that cbsB
is a silent gene. His-CbsB from E. coli
polymerized into a crystalline layer, but we could not detect any collagen binding by His-CbsB.
S-layers are expressed by isolates belonging to homology groups A1 to A4 of the Acidophilus group of Lactobacillus
. When using DNA hybridization, we detected a cbsA
homolog only in isolate JCM 5810, which is a member of homology group A2. Furthermore, we have not detected collagen binding by S-layer proteins from 12 additional isolates of L. crispatus
(S. Tankka, and J. Sillanpää, unpublished data). These findings suggest that CbsA is not common in these bacteria and also are in agreement with the observations that Lactobacillus
S-layers exhibit marked sequence variability, even among isolates representing the same Acidophilus homology group (7
). We are currently analyzing how this sequence variability is correlated with adhesive functions of Lactobacillus