Crystal structure of enterotoxigenic Escherichia coli colonization factor CS6 reveals a novel type of functional assembly


  • Saumendra P. Roy,

    1. Department of Molecular Biology, Uppsala BioCenter, Swedish University of Agricultural Sciences, Uppsala, Sweden
    2. Department of Chemistry, University of Turku, Turku, Finland
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Mohammad M. Rahman,

    1. Department of Chemistry, University of Turku, Turku, Finland
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Xiao Di Yu,

    1. Department of Molecular Biology, Uppsala BioCenter, Swedish University of Agricultural Sciences, Uppsala, Sweden
    Current affiliation:
    1. The F. M. Kirby Neurobiology Center, Boston Children's Hospital, Boston, MA, USA
    2. Department of Neurology, Harvard Medical School, Boston, MA, USA
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Minna Tuittila,

    1. Department of Chemistry, University of Turku, Turku, Finland
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Stefan D. Knight,

    1. Department of Molecular Biology, Uppsala BioCenter, Swedish University of Agricultural Sciences, Uppsala, Sweden
    Search for more papers by this author
  • Anton V. Zavialov

    Corresponding author
    1. Department of Chemistry, University of Turku, Turku, Finland
    • Department of Molecular Biology, Uppsala BioCenter, Swedish University of Agricultural Sciences, Uppsala, Sweden
    Search for more papers by this author

For correspondence. E-mail or; Tel. (+46) 18 47 14 047; Fax (+46) 18 53 69 71.


Coli surface antigen 6 (CS6) is a widely expressed enterotoxigenic Escherichia coli (ETEC) colonization factor that mediates bacterial attachment to the small intestinal epithelium. CS6 is a polymer of two protein subunits CssA and CssB, which are secreted and assembled on the cell surface via the CssC/CssD chaperone usher (CU) pathway. Here, we present an atomic resolution model for the structure of CS6 based on the results of X-ray crystallographic, spectroscopic and biochemical studies, and suggest a mechanism for CS6-mediated adhesion. We show that the CssA and CssB subunits are assembled alternately in linear fibres by the principle of donor strand complementation. This type of fibre assembly is novel for CU assembled adhesins. We also show that both subunits in the fibre bind to receptors on epithelial cells, and that CssB, but not CssA, specifically recognizes the extracellular matrix protein fibronectin. Taken together, structural and functional results suggest that CS6 is an adhesive organelle of a novel type, a hetero-polyadhesin that is capable of polyvalent attachment to different receptors.


Enterotoxigenic Escherichia coli (ETEC) infection is the leading cause of diarrhoea in the developing world, accounting for 780–900 million diarrhoeal episodes and 300 000–500 000 deaths annually, most of which occur in children under the age of five (WHO, 2006). In addition, ETEC is the most common cause of traveller's diarrhoea, being responsible for one-third to one-half of all diarrhoeal episodes in travellers to Africa, Asia and Latin America (WHO, 2006).

Adhesion of ETEC to the small intestine is an essential step in the initiation of infection. ETEC binds to the epithelium by means of colonization factors (CFs), a heterogeneous group of proteinaceous surface structures (Gaastra and Svennerholm, 1996; Qadri et al., 2005). More than 25 distinct CFs have been identified, of which coli surface antigen 6 (CS6) is the most commonly detected (Gaastra and Svennerholm, 1996; Qadri et al., 2005). Many ETEC strains express only CS6 or coexpress it with coli surface antigens CS4 or CS5 (Gaastra and Svennerholm, 1996; Qadri et al., 2005; Ghosal et al., 2007). For all these strains, CS6 was shown to be essential for intestine colonization (Svennerholm et al., 1988). The role of CS6 in intestinal adherence was demonstrated in vitro using bacteria only expressing CS6 and intestinal cell lines (Tobias et al., 2008; Ghosal et al., 2009). Recently, two receptors for CS6 were identified: the extracellular matrix protein fibronectin (Ghosal et al., 2009) and glycosphingolipid sulphatide (Jansson et al., 2009).

Many CFs have the typical rod-like morphology of fimbriae. High-resolution structural analysis of colonization factor antigen CFA/I (Li et al., 2009) revealed a structural organization similar to that of adhesive pili assembled via the classical chaperone/usher (CU) pathway (Waksman and Hultgren, 2009). CS6 has a different morphology and is generally believed to be non-fimbrial (i.e. lacking a rod-like architecture) (Gaastra and Svennerholm, 1996; Wolf et al., 1997; Qadri et al., 2005; Ludi et al., 2006). CS6 is assembled from two distinct subunits, CssA and CssB (Wolf et al., 1997). Analysis of native CS6 purified from a clinical isolate of ETEC demonstrated that the CssA and CssB subunits are present in equal stoichiometry in CS6 (Ghosal et al., 2009). The relative expression of the cssA and cssB genes analysed by semi-quantitative RT-PCR suggested that equal amounts of transcripts are made, also indicating equal production of both subunits (Ghosal et al., 2009). Such a subunit composition is unusual for CU fimbriae, which are typically assembled from a major structural subunit that constitutes most of the fibre and minor subunits that are present in low amounts and required for special tasks such as receptor binding (Zavialov et al., 2007; Zav'yalov et al., 2010).

The structure of the gene cluster expressing CS6 is nevertheless typical for fimbriae (Wolf et al., 1997). In addition to the genes encoding two CS6 structural subunits, it contains genes for the CssC chaperone and CssD usher. It was demonstrated that the CssC chaperone is required for CS6 assembly (Tobias et al., 2008). However, assembly of CS6 appeared to be independent of the CssD usher (Tobias et al., 2008), which is surprising in view of recent studies (Remaut et al., 2008; Phan et al., 2011) demonstrating the key role of the usher in fimbriae assembly. The expression and assembly of CssA are significantly reduced (Tobias et al., 2008) or abolished (Wolf et al., 1997) in the absence of the second subunit, CssB. The CssB subunit however was expressed in the absence of CssA in laboratory strains (Tobias et al., 2008). Recently, Wajima et al. (2011) demonstrated that expression of each of four proteins (CssA, CssB, CssC and CssD) is essential for assembly of functional (adhesion competent) CS6 in native ETEC and suggested that CssA and CssB subunits assemble together.

The CssC chaperone belongs to the subfamily of FGL periplasmic chaperones involved in assembly of ultra-thin fimbriae with a simple subunit composition (Zavialov et al., 2007). The organelles of this type often have an atypical morphology, caused by aggregation of fibres on the cell surface. The F1 antigen of Yersinia pestis, assembled from one subunit, Caf1, is the best-studied FGL CU assembled organelle (Zavialov et al., 2007). Details of the mechanism of biogenesis of F1 antigen have been obtained from high-resolution structural studies of chaperone–subunit complexes (Zavialov et al., 2003; 2005). The monomeric subunit fold has only six β-strands and lacks the last (G) β-strand of a seven-stranded immunoglobulin (Ig)-like fold. This incomplete fold exposes a hydrophobic cleft between the two β-sheets of the Caf1 subunit, corresponding to the hydrophobic core of an Ig domain. The instability of this structure accounts for the rapid degradation and aggregation of Caf1 when expressed in the absence of functional chaperone (Chapman et al., 1999; MacIntyre et al., 2001). The structure of the minimal Y. pestis F1-antigen fibre revealed that fibre subunits are linked together by donor strand complementation (DSC) (Choudhury et al., 1999; Sauer et al., 1999; 2002; Zavialov et al., 2002) with an N-terminal Gd donor strand segment of one subunit inserted into the hydrophobic cleft of a neighbouring subunit (Zavialov et al., 2003; 2005). The resulting linear fibre is composed of globular modules –[(Gd)n:(Caf1)n+1]– each having an intact immunoglobulin (Ig) topology generated by donor strand complementation. Later, two other representatives of FGL CU assembled organelles, Salmonella Saf (Remaut et al., 2006) and E. coli AfaIII fimbriae (Anderson et al., 2004), were shown to be similarly constructed.

Theoretically, for any fimbrial subunit capable of DSC polymerization it is possible to create a circularly permuted construct (Barnhart et al., 2000; Anderson et al., 2004; Zavialov et al., 2005). In this construct, the Gd strand is moved to the C-terminus of the subunit, enabling self-complementation of the subunit and formation of a monomeric fibre module with a classical Ig-fold. Such self-complemented (SC) constructs have been designed for several subunits from different CU assembled fimbriae, including Caf1 (Zavialov et al., 2005). High-resolution studies showed that SC constructs and fibre assembled subunits have nearly identical structures. SC subunits were shown to have unusually high stability, as is typical for assembled CU polymers (Zavialov et al., 2005; Puorger et al., 2008; Piatek et al., 2009).

Here we report X-ray crystallographic, biophysical and biochemical studies of designed SC subunits that provide an atomic resolution model for the structure of CS6 and reveal the role of individual subunits in receptor binding. Collectively, our results suggest that CS6 is a polyadhesin of a novel type that is capable of polyvalent binding to different receptors on human epithelial cells.


Models of CS6 assembly and design of self-complemented CS6 subunits

CssA and CssB display no sequence similarity to any other subunit of CU-assembled fimbriae. Nonetheless, threading of their sequences through known 3D structures of fimbrial subunits revealed sequence-structure compatibility, suggesting that the subunits might have the typical six-stranded incomplete Ig-like fold of other CU subunits. Moreover, a pattern of alternating hydrophobic and hydrophilic residues with the characteristic signature of the Gd donor strand motif (Zavialov et al., 2003) was detected at the N-terminus of both subunits (Fig. 1). Hence, we hypothesized that both CssA and CssB are able to form inter-subunit interactions by donating and accepting N-terminal Gd donor strands.

Figure 1.

Sequence alignment of CssA and CssB subunits. The beginning of mature sequences was identified by mass spectrometry analysis of purified proteins (data not shown). Secondary structure elements of CssA and CssB core structures are shown in magenta and cyan respectively. The GdA and GdB strands are shown in red and blue respectively. Residues in the alignment are coloured by the type of side-chain: yellow, hydrophobic; cyan, positively charged; red, negatively charged; green, polar; magenta, aromatic; orange, Gly and Pro. Absolutely conserved residues are shown in boxes. The donor residues, occupying pockets 1–5 of the acceptor cleft and residues engaged in non-classical donor strand binding in additional pockets −1 and −2 are indicated by the pocket number. clustalw alignment of sequences was modified based on superposition of the corresponding crystal structures.

Two principle modes of polymerization are consistent with this hypothesis (Fig. 2A and B). Insertion of donor strands into acceptor clefts of subunits of the same type would result in formation of homo-polymeric fibres (CssAn and CssBn). Insertion of donor strands into acceptor clefts of subunits of a different type would result in hetero-polymeric fibres with CssA and CssB alternating along the fibre (…–CssA–CssB–CssA–CssB–…). If both modes of polymerization occurred simultaneously, CS6 fibres with a stochastic distribution of subunits reflecting the probability for each type of linkage would result.

Figure 2.

Donor strand complementation-based models of CS6.

A. Schematic diagram of non-complemented subunits CssA (red) and CssB (blue). The N-terminal donor sequences (ds, donor strand, or Gd, G donor) are shown with wavy arrows. The six-stranded (6st) core structure is shown with incomplete oval to indicate its incomplete immunoglobulin like fold (Ig-like fold missing G strand).

B. Three possible versions of donor strand complementation-based CS6 polymers: CssAn and CssBn homo-polymers and …–CssA–CssB–CssA–CssB–… hetero-polymer. The numbers indicate possible donor strand complementation contacts.

C. Versions of monomeric self-complemented (SC) subunits corresponding to the four possible donor strand complementation contacts in CS6 polymers (indicated). Sequences of designed constructs are shown in Supplementary Fig. S1.

To elucidate which mode of polymerization is used to form CS6 fibres, we engineered four self-complemented subunits, based on the four possible types of acceptor cleft-donor strand contacts: CssA6st–GdA (CssAdsA), CssA6st–GdB (CssAdsB), CssB6st–GdB (CssBdsB) and CssB6st–GdA (CssBdsA) (Fig. 2C and Supplementary Fig. S1). To ensure that the donor strand has sufficient conformational freedom to insert correctly into the acceptor cleft, we introduced a four-residue linker between the subunit's last residue and the first residue in the donor strand. To facilitate discussion of results we have retained the numbering of residues in donor strands according to their position in wild-type subunits. Using native signal sequences, the SC subunits were expressed in the E. coli periplasm. All four subunits were detected in the periplasmic extracts. However, the hetero-constructs CssAdsB and CssBdsA accumulated at significantly higher levels than the homo-complemented subunits CssAdsA and CssBdsB (Fig. 3A and Table 1), suggesting that they secrete and fold more efficiently and/or are more resistant to proteases.

Figure 3.

Stability analysis of self-complemented (SC) subunits.

A. SDS-PAGE of periplasmic extracts of E. coli expressing four different SC subunits (labelled). Proteins were extracted from equal amounts of cells.

B. Temperature stability of CssAdsA and CssAdsB (upper panel) and CssBdsA and CssBdsB (bottom panel). Temperature dependence of the CD signal was recorded at a heating rate of 1°C min−1 at 208 nm for CssAdsA and CssAdsB and at 215 nm for CssBdsA and CssBdsB.

C. GdmHCl-induced denaturation. The CD signal was recorded at different concentrations of GdmHCl at 230 nm for CssAdsA (upper panel), at 220 nm for CssAdsB (middle panel), and 223 nm for CssBdsA and CssBdsB (bottom panel). All measurements were performed in 10 mM phosphate buffer, pH 7.0.

Far UV spectra of SC subunits are shown in Supplementary Fig. S2.

Table 1. Expression level and stability of donor strand complemented subunits
ConstructLevel of expressionaTm (oC)b[GdmHCl]m (M)cΔGU-N (kJ mol−1)d
  1. aEstimated from SDS-PAGE gel (Fig. 3A).
  2. bMelting temperature (Fig. 3B).
  3. cTransition concentration of GdmHCl (Fig. 3C).
  4. dΔGU-N, free energy of unfolding.
  5. eIn presence of 4.3 M GdmHCl.
  6. fUnfolding not detected at GdmHCl concentrations up to 4 M at 21°C.
CssAdsA+55.8 ± 0.20.67 ± 0.018.4 ± 0.2
CssAdsB++++78.0 ± 0.22.12 ± 0.0444.6 ± 0.5
CssBdsB++58.9 ± 0.32.30 ± 0.0538.4 ± 0.3
CssBdsA+++++71.1 ± 0.3> 4Ndf
61.3 ± 0.8e

Stability analysis of SC subunits supports a hetero-assembly model for CS6

All four SC subunits were purified to high homogeneity and analysed by circular dichroism (CD) spectroscopy (Supplementary Fig. S2). The subunits showed CD spectra typical for proteins with a high content of β-sheets, with minima in the 208–215 nm region. Heating dramatically changed the spectra, suggesting loss of native structure. To study temperature induced conformational transitions, the temperature-dependence of the CD signal was recorded for each SC subunit at a characteristic wavelength (Fig. 3B). The resulting thermograms had the characteristic sigmoidal shape of a cooperative melting transition. The hetero-constructs melted at high temperatures typical for highly stable proteins. The homo-complemented subunits melted at significantly lower temperatures than the hetero-constructs, with transition temperature differences (ΔTm) of 22.2°C between CssAdsB and CssAdsA and 12.2°C between CssBdsA and CssBdsB (Table 1).

Addition of guanidinium hydrochloride (GdmHCl) to SC subunits caused large changes in the CD spectra of CssAdsA, CssAdsB and CssBdsB, but not CssBdsA in the far UV region (Supplementary Fig. S2). To study GdmHCl-induced unfolding transitions, CD signals at characteristic wavelengths were recorded at different concentrations of GdmHCl (Fig. 3C). Evaluation of the data according to the two-state model of folding/unfolding provided the free energy of unfolding for the first three constructs (Table 1). CssAdsA unfolded at low concentrations of GdmHCl and was the least stable. The corresponding hetero-construct CssAdsB showed five times higher stability. CssBdsB was significantly more stable than CssAdsA, but less stable than CssAdsB. CssBdsA showed no denaturation transition at GdmHCl concentrations up to 4 M. Melting of CssBdsA in the presence of 4.3 M GdmHCl revealed a cooperative transition at a melting temperature of 61.3±0.8°C (Table 1, Fig. S3), demonstrating that this protein still has a cooperative structure in 4.3 M GdmHCl. Several structural subunits of CU assembled organelles have been shown to possess such unusually high stability (Zavialov et al., 2005; Puorger et al., 2008; 2011; Yu et al., 2012), which can be used as a reliable indicator of a correctly donor strand complemented subunit.

Hence, the analysis of thermal and chemical denaturation of SC subunits clearly demonstrates that the hetero-constructs are significantly more stable than the homo-constructs. The low stability of the homo-constructs suggests that the CssA6st–GdA and CssB6st–GdB interactions are unlikely to be native, supporting the hetero-polymerization model for CS6.

Crystal structures of CssAdsB and CssBdsA

To elucidate the structure and assembly mechanism of CS6 at the atomic level, we determined high-resolution crystal structures of CssAdsB (1.5 Å) and CssBdsA (1.0 Å) (Fig. 4, Table 2 and Supplementary Fig. S4).

Figure 4.

Crystal structures of CssAdsB and CssBdsA.

A. Cartoon diagrams of 3D structures of CssAdsB and CssBdsA. The core structures of CssA and CssB subunits are coloured in magenta and cyan respectively. The GdA and GdB strands are painted in red and blue respectively. The β-strands are labelled.

B. Structural superposition of CssAdsB and CssBdsA (stereo view).

C. DSC contacts in CssAdsB and CssBdsA. Molecular surfaces of core structures are shown in beige with hydrophobic residues painted in orange. Donor strands are shown with sticks and coloured by the type of element. Carbon, oxygen and nitrogen atoms are shown in cyan, red and blue respectively. The donor residues and pockets in the acceptor cleft are labelled.

Table 2. Data and refinement statistics
Data collection
Data setSeMetNativeSeMetNativeNative
  1. aBased on maximum likelihood.
  2. Values for high-resolution shells in parenthesis. Rmerge = ΣhΣiI(h)i − (I)|/ΣhΣiI(h)i, where I(h) is the intensity of a reflection h, Σh is the sum over all reflections and Σi is the sum over i measurements of reflection h. Rmeas = Σh[nh/(nh − 1)]0.5Σi|I(h)i – (I)|/ΣhΣiI(h)i, where nh is multiplicity. Rwork = Σ|FOFC|/ΣFO where FO and FC are the observed and calculated structure factors respectively. Rfree is calculated for a test set of reflections randomly excluded from refinement. R.m.s.d. stereochemistry is the deviation from ideal values. R.m.s.d. B-factors is deviation between bonded atoms.
Wavelength (Å)0.9790.8730.9780.8730.933
Resolution range (Å)56.77–2.20 (2.32–2.20)44.43–1.50 (1.58–1.50)41.82–1.50 (1.58–1.50)27.29–1.04 (1.11–1.04)43.00–2.54 (2.68–2.54)
Space group and unit cell dimensions (a, b, c, Å)P21212P21212P212121P212121P21212
44.38, 49.06, 56.7744.45, 49.15, 56.7351.38, 69.16, 71.9851.96, 69.23, 72.3543.00, 48.95, 56.96
No. unique reflections6699 (943)20 523 (736)41 811 (1455)124 197 (2985)4190 (150)
Multiplicity12.9 (12.3)4.5 (4.0)29.2 (29.1)4.0 (4.0)3.2 (3.2)
Completeness100.0 (100.0)100.0 (100.0)100.0 (100.0)98.2 (97.8)98.8 (98.3)
Rmerge0.118 (0.387)0.059 (0.434)0.054 (0.146)0.067 (0.426)0.027 (0.068)
Rmeas0.130 (0.423)0.065 (0.502)0.055 (0.151)0.078 (0.490)0.032 (0.081)
 <I/σ(I)>14.0 (6.0)13.5 (3.2)46.8 (22.3)11.3 (3.1)33.9 (16.8)
Resolution (Å)35–1.5027.29–1.0437.13–2.54
Number of reflections   
Total19 98712 41974124
Test set1 0356 238191
Number of atoms   
Protein9862 226949
R.m.s.d. stereochemistry   
Bonds (Å)0.0070.0200.005
Angles (°)1.0731.7220.834
Co-ordinate errora (Å)
Ramachandran plot   
Favoured (%)97.499.097.4
Outliers (%)000

Both structures have a classical Ig-like fold that consists of two β-sheets packed against each other in a β sandwich (Figs 4A and B and 7B). Sheet 1 comprises four major strands ABED. Sheet 2 comprises three major strands FCC′ and donor strand Gd, completing the Ig-like fold of the subunit. Despite low sequence similarity between CssA and CssB (23% identical residues, Fig. 1), their structures have only small differences in fold topology. In both structures strand C has a switch in the middle, splitting it into strands C1 and C2. Whereas this region forms only a loop in CssA, in CssB, it contains a loop and a short strand, C′′, running anti-parallel to strand C2. CssA instead contains an additional short strand D′, which runs anti-parallel to strand C1. The regions between strands C1 and C2 and strands D and E, which include the fold topology differences, are closely positioned and represent the most dissimilar part in the structure of CssA and CssB. Hence, this region may contribute to the differences in biological function of the two subunits (see section CS6 subunits recognize different host cell receptors below). The subunits also have significant differences in the structure of the regions flanking stand A. In CssA, these regions are proline rich and strand A is relatively short. In CssB, strand A is longer and follows a short α helix, which is not present in the structure of CssA. The rest of the structures of CssA and CssB are relatively similar: 105 matching Cα atoms are superimposed with a r.m.s.d. of 1.59 Å.

The E. coli DraD invasin and Y. pestis Caf1 subunits, which both belong to the class of FGL CU assembled fimbriae, were identified as the most structurally similar proteins to CssA and CssB by a DALI (Holm and Rosenstrom, 2010) search of the protein data bank. Although these proteins have no statistically valid sequence homology, the structural similarity was relatively high. For example, 112 Cα atoms in CssB and DraD structures were superimposed with a r.m.s.d. of 2.6 Å with only 6% of these residues being identical. The large difference in conservation of structure and sequence of fimbrial subunits is probably a result of a strong pressure to constantly change the surface of these virulence organelles to avoid host immune responses.

Structures of CssAdsB and CssBdsA provide an atomic resolution model for assembly contacts in CS6. CssA–CssB and CssB–CssA donor strand complementation contacts are surprisingly extensive (Fig. 4C). In each case, the donor strand inserts five residues into the acceptor cleft of the respective subunit. The interactive surfaces are mostly hydrophobic and the binding is clearly driven by the hydrophobic effect. In addition to hydrophobic interactions, the side-chain of Thr13 in GdA in CssBdsA forms a hydrogen bond with the main chain of CssB. The number of pockets in the acceptor cleft of CU subunits varies from 3 to 5 (Zav'yalov et al., 2010). In terminology originating from structural studies of Pap and type 1 fimbriae and of F1 antigen (Sauer et al., 2002; Zavialov et al., 2005), the pocket accommodating the last donor residue has number 5 (P5). In both subunits, P5 is occupied by the same large donor residue, Ile. A large donor residue at this position is essential for initiation of donor strand exchange (Remaut et al., 2006; Yu et al., 2012). Large donor residues, Leu8 and Phe9 in GdB and GdA, respectively, also occupy pocket 2. However, due to a twist in the donor strands these residues are not inserted as deeply as those in pockets 3–5, but lay on the surface of a relatively shallow cleft. Similarly, pocket 1 in both structures is shallow and is filled by the aliphatic part of lysine side-chains.

CssA6st–GdB and CssB6st–GdA contacts also include additional ‘non-classical’ interactions. Trp3 and Tyr5 in GdB and Ile4 and Ala5 in GdA bind to the surface of CssA6st and CssB6st, respectively, which is located almost on the opposite side of the molecule to the classical P1–P5 pockets (pockets −2 and −1, Fig. 4C). The N-terminal part of the main chain of the donor strands also provides significant additional contribution to the binding area. In total, these ‘non-classical’ interactions contribute ∼ 30% and 19% of the CssA6st–GdB and CssB6st–GdA interactive surface area respectively. This makes the total binding area large: ∼ 2260 and 2250 Å2 area becomes buried upon the CssA6st–GdB and CssB6st–GdA association respectively. For comparison, the donor strand complementation interface in Caf1 fibres, which also consists of five classical acceptor pocket–donor residue contacts, is about 10% smaller (calculated using structure 1Z9S). Previously, non-classical donor strand interactions were observed in two different fimbrial systems, FGL CU fimbriae Saf (SafA subunit) (Remaut et al., 2006) and FGS CU fimbriae F4 (FaeG subunit) (Van Molle et al., 2009). The extra pockets in SafA and FaeG subunits, as in the case of CssA, are occupied by large aromatic residues (Phe and Trp respectively), suggesting that this is probably a general feature. The non-classical donor strand interactions were found in inter-subunit, but not in chaperone–subunit donor strand complementation (Van Molle et al., 2009) and hence may be of importance for donor strand exchange.

Structural comparison of hetero- and homo-SC subunits favours the hetero-assembly model for CS6

To examine why GdA cannot stabilize CssA as strongly as GdB (Table 1), we determined a 2.5 Å resolution crystal structure of CssAdsA (Fig. 5, Table 2 and Supplementary Fig. S4) and compared it with those of CssAdsB and CssBdsA (Fig. 4). The six-stranded core structure of CssA is almost identical in CssAdsA and CssAdsB (r.m.s.d. = 0.48 Å for 107 Cα atoms, Fig. 5A). In contrast, the structure of GdA is very different in CssAdsA and CssBdsA (r.m.s.d. = 1.86 Å for 13 Cα atoms, Fig. 5B). In fact, GdA in CssAdsA is structurally closer to GdB in CssAdsB (r.m.s.d. = 0.56 Å for 13 Cα atoms, Fig. 5A). However, a simple visual inspection of CssA6st–GdA and CssA6st–GdB contacts reveals that the GdA donor residues fit significantly poorer to the acceptor pockets of CssA than GdB donor residues do. Several examples of this are shown in Fig. 5.

Figure 5.

Donor strand of CssA fits less well to acceptor cleft of CssA than does the donor strand of CssB.

A and B. Structural superpositions of CssAdsA and CssAdsB (A) and CssAdsA and CssBdsA (B) (stereo views). CssAdsA (magenta) and CssBdsA or CssAdsB (cyan) are shown by ribbons and donor strand side-chains as sticks coloured by the type of element. Carbon atoms in GdA and GdB are painted in cyan and magenta respectively. Oxygen and nitrogen atoms are painted in red and blue, respectively, in both structures. The GdB donor residues are labelled.

C–E. Examples demonstrating the poor fit of the GdA donor residues to the acceptor cleft of CssA. The core structure of CssA is shown as a molecular surface. Surfaces of hydrophobic residues are painted in orange. Surfaces of polar and charged residues are painted in beige. The GdA and GdB strands are shown with lines or sticks to indicate the problem residues. The colour-coding is the same as in (B). Acceptor pockets are indicated.

Pocket 4 in CssA is highly hydrophobic and perfectly adjusted to bind hydrophobic Val12 in GdB (Fig. 5C). However, GdA inserts polar Thr12 in this pocket, where its hydroxyl group has no possibility to form a hydrogen bond. Interactions at the extra pockets provide another striking example of poor fit (Fig. 5D). In CssAdsB pockets −2 and −1 are occupied by large aromatic residues, Trp3 and Tyr5 respectively. GdA in CssAdsA inserts significantly smaller residues in these pockets (Thr and Ile) with insufficient size to fill the pockets. The fit of GdA to the acceptor cleft is poor compared with that of GdB even in the regions where these strands use the same type of donor residue. For example, both GdA and GdB insert an Ile residue into pocket 5 of the acceptor cleft (Fig. 5E). Whereas GdB Ile is deeply inserted in this pocket, GdA Ile has a slightly tilted conformation and hence occupies only the more shallow part of the pocket.

To further examine the geometrical fit between the donor strands and acceptor clefts, we calculated the shape correlation statistic (Sc) (Lawrence and Colman, 1993) for the CssA6st–GdA, CssA6st–GdB and CssB6st–GdA interfaces. This statistic provides a measure of the packing of two protein surfaces. A value of Sc = 0 indicates no geometrical fit, whereas a value of Sc = 1 corresponds to two perfectly packed surfaces. Calculation of the shape correlation statistic for the CssA6st–GdA, CssA6st–GdB and CssB6st–GdA interfaces gave values of Sc = 0.69, 0.75 and 0.8 respectively. The shape correlation statistic of donor strand complementation for the hetero-SC subunits was very high (Lawrence and Colman, 1993) and similar to that calculated for Caf1 (0.784) and SafA (0.791) FGL CU fibres (structures 1Z9S and C2O4 respectively). For homo-SC subunit CssA6st–GdA the shape correlation statistic was considerably lower, suggesting that GdA fits less well to the acceptor cleft of CssA6st than does GdB and poorer than donor strands from other systems fit to corresponding acceptor clefts.

The poor fit between donor strand and acceptor cleft explains the low stability of the homo-SC construct (Table 1). It has previously been demonstrated that complementarity of the donor strand–acceptor cleft contact determines the order of subunits in complex fibres (Rose et al., 2008). Hence, our structural studies strongly support the hetero-assembly model for CS6.

Analysis of the CS6 architecture with glutaraldehyde cross-linking

To verify our hetero-assembly model, we studied cross-linking of CS6 with glutaraldehyde. CS6, purified from the ETEC strain E10703 (a kind gift of Dr F. Cassels and Dr E. Oaks from Walter Reed Army Institute of Research) (Wolf et al., 1997) was incubated with different concentrations of glutaraldehyde, boiled in SDS loading buffer to dissociate CS6 into CssA and CssB monomers, and analysed in Western blots (Fig. 6). SDS electrophoresis of glutaraldehyde untreated CS6 in high-density polyacrylamide gels resulted in well-resolved single bands of CssA and CssB subunits that were specifically recognized by anti-CssAdsB and anti-CssBdsA polyclonal antibodies, respectively, as previously demonstrated in Ghosal et al. (2009). In contrast, CS6 samples treated with relatively low concentrations of glutaraldehyde showed a dominant pattern of bands. The most intensive bands of the cross-linked species corresponded to dimers, trimers and tetramers of CS6 subunits, as suggested from the molecular weight estimation. All these bands were recognized with both anti-CssAdsB and anti-CssBdsA polyclonal antibodies, suggesting that the cross-linked species consists of both types of subunits. This result strongly supports the hetero-assembly model for CS6, because only in this model CssA and CssB subunits are adjacent and form the close contact essential for their efficient cross-linking. In addition to these major cross-linked species, we observed a pattern of faint bands moving slightly slower and visualized by anti-CssAdsB, but not anti-CssBdsA antibodies. Hence, a small fraction of CssA subunits was cross-liked directly. This phenomenon could be explained by at least two different hypotheses. A certain number of CssA subunits from different fibres might form close contacts as a result of a super-structure formed from linear fibres, e.g. fibres might coil to form super-coils. These interacting CssA subunits could be cross-linked. It is possible to suggest that the polymerization process is not always specific, producing stretches of homo-polymers of CssA subunits in addition to the dominant hetero-polymeric structure of alternating CssA and CssB subunits, which also would allow cross-linking of CssA subunits. A similar phenomenon might explain the possible formation of non-functional CssB polymers in the absence of CssA subunits (Wajima et al., 2011).

Figure 6.

Glutaraldehyde cross-linking analysis of the CS6 architecture. Samples of surface purified CS6 untreated and treated with glutaraldehyde at 0.035% and 0.07% concentration were analysed by Western blotting using anti-CssA (left panel) and anti-CssB (right panel) polyclonal antibodies. Protein species detected are indicated by arrowheads and labelled. The species assignment was performed based on the result of immunodetection and estimated molecular weight. Cross-linked hetero-trimers and hetero-tetramers may correspond to several indistinguishable subunit compositions (indicated). Antibodies bound to monomers less efficiently than to the larger, cross-linked species.

Model of CS6 fibre

Structures of CssAdsB and CssBdsA were used to model the CS6 fibre (Fig. 7). CssAdsB and CssBdsA, representing the CssA6st:GdB and CssB6st:GdA modules in a CS6 fibre, respectively, were positioned alternately in a line in accordance with the hetero-assembly model. To revert the circular permutation in CssAdsB and CssBdsA, the artificial linker sequences connecting C-termini of the SC subunits with N-termini of donor strands were deleted and the C-termini of donor strands were bridged with N-termini of adjacent SC subunits using native linker sequences (Fig. 7A). The native linkers, consisting of four to five and seven residues in CssA and CssB, respectively (Fig. 1), were unstructured in SC subunits (Fig. 4A and B). They were modelled manually using the program O (Jones et al., 1991), with appropriate stereochemistry ensured by regularization. The relative orientation of subunits in the model was manually adjusted to remove clashes between side-chains in the interface between the neighbouring subunits (by subunit rotation and side-chain rotamer optimization) and to minimize strain in the linker. The topology diagram of the resulting fibre is shown in Fig. 7B. The length of the native linkers would allow significant movement between consecutive subunits in the model, suggesting that CS6 fibres are highly flexible. We failed to detect inter-subunit interactions that would determine a tilt between consecutive subunits. However, the specific geometry of donor strand complementation introduced a twist between subunits along the length of the fibres, causing exposure of different surfaces of CssA and CssB subunits along the same side of the fibre.

Figure 7.

Model of the CS6 polymer.

A. 3D reconstruction of a CS6 polymer based on the crystal structures of the CssAdsB and CssBdsA subunits. CssA and CssB are shown in magenta and cyan respectively. Brackets indicate two distinct two-subunit repeats in the CS6 fibre.

B. Topology diagram of the CS6 fibre. Arrows indicate strands; the single α helix is shown as a rectangle. The strands in the same subunit are shown in slightly different colours to indicate two β-sheets of the β sandwich.

Mini-fibres of CS6 contain structural epitopes of CS6

Two two-subunit repeats in the CS6 fibre, –[CssA6st:GdB]–[CssB6st:GdA]– and –[CssB6st:GdA]–[CssA6st:GdB]–, with different interfaces between subunits are suggested by the hetero-assembly model (Fig. 7A). Collectively, these two repeats contain the entire unique surface of the CS6 fibre. Hence, to expose all epitopes of the CS6 antigen on SC constructs, we designed two mini-fibres, corresponding to the repeats, CssAdsB–CssBdsA and CssBdsA–CssAdsB, which contain the CssAdsB and CssBdsA subunits linked by the native linker sequences between GdB and CssB6st and between GdA and CssA6st respectively (Supplementary Fig. S1). The mini-fibres were expressed in the periplasmic space, where they accumulated at high levels.

Mini-fibres and individual SC subunits were analysed for binding to monoclonal antibodies generated against surface purified CS6 [2a:14 and 20:11:9 (Helander et al., 1997)] using ELISA (Fig. 8A). 2a:14 and 20:11:9 efficiently recognized mini-fibres and CssBdsA, but not CssAdsB, suggesting that they specifically bind to epitopes located on CssB. This result is in agreement with a previous study performed using surface purified CS6 (Helander et al., 1997).

Figure 8.

Binding properties of individual SC subunits and mini-fibres.

A. Binding of SC constructs to anti-CS6 monoclonal antibodies.

B. Binding of SC constructs to Caco-2 epithelial cell line. Integrated density of bands of cell-associated proteins in Western blots was plotted against total concentration of added protein. Error bars indicate standard deviation (n = 3). Saturation curve analysis was performed only for the binding data obtained for the mini-fibres. Cell-associated CssAdsB and mini-fibres were quantified using polyclonal anti-CssA antibodies and CssBdsA using polyclonal anti-CssB antibodies. Fifteen-second and 1 min exposures were used to detect mini-fibres and SC subunits respectively.

C and D. CssBdsA, but not CssAdsB binds to fibronectin. (C) Comparison of binding of CssAdsB and CssBdsA to a Biacore sensor chip with immobilized fibronectin. Increasing concentrations of proteins were successively injected into the chip. The binding was detected by measuring the surface plasmon resonance (SPR) signal. (D) Biacore analysis of CssBdsA–fibronectin interactions. Left panel: SPR sensograms recorded for different concentration of CssBdsA. Right panel: Saturation curve (two binding site models). Insert: Scatchard plot analysis of CssBdsA binding to high and low-affinity sites in fibronectin. Analysis of fibronectin–CssAdscA and fibronectin–CssBdscB interactions is shown in Supplementary Fig. S5. Analysis of binding of the CssAdsB–CssBdsA and CssBdsA–CssAdsB mini-fibres to fibronectin is shown in Supplementary Fig. S6.

E. Schematic diagram illustrating polyvalent binding of a CS6 fibre to multiple receptors of two different types (R1 and R2).

CS6 subunits recognize different host cell receptors

To reveal binding properties of the individual subunits of CS6, we examined binding of our constructs to the Caco-2 intestinal cell line using a saturation binding experiment. In this experiment, cell-associated proteins were analysed by Western blotting using anti-CssA and CssB antibodies. All constructs bound to Caco-2 cells in a saturable manner (Fig. 8B). The reproducibility of experiments with mini-fibres was better than with SC subunits, enabling estimation of apparent dissociation constants Kd, ∼ 4 μM for CssAdsB–CssBdsA and ∼ 9 μM for CssBdsA–CssAdsB.

Cell binding experiments with CS6 showed that it binds to at least two different receptors (Ghosal et al., 2009). Two receptors have been identified for CS6, the extracellular matrix protein fibronectin (Ghosal et al., 2009) and glycosphingolipid sulphatide (Jansson et al., 2009). The same studies also attempted to identify subunits that possess the binding sites for these receptors. However, in both studies, binding experiments were performed using CS6 subunits with compromised 3D structure (non-native high temperature-produced aggregates (Ghosal et al., 2009) or non-complemented subunits fused to a carrier protein that were obtained from inclusion bodies (Tobias et al., 2008; Jansson et al., 2009). Since binding properties of most proteins critically depend on their 3D structure, the results of these studies should be treated with caution. Hence, we decided to re-examine interactions with these receptors using our SC subunits and mini-fibres.

To quantitatively characterize binding between our constructs and human fibronectin, surface plasmon resonance (SPR) was used. Human plasma fibronectin was covalently immobilized to a sensor chip, and the binding of the constructs was measured by recording the SPR signal after addition of purified constructs in the liquid phase. Whereas CssAdsB showed no binding (Fig. 8C), CssBdsA exhibited saturation binding (Fig. 8C and D). Analysis of the CssBdsA–fibronectin reaction revealed two binding sites of different affinity with Kd values of 9.1 ± 2.0 and 340 ± 50 μM respectively. CssBdsB bound to fibronectin with a similar affinity, while, as CssAdsB, CssAdsA did not bind to fibronectin (Supplementary Fig. S5). Hence, binding experiments with SC subunits demonstrated that the fibronectin binding site is located on the surface of CssB and not CssA, in contrast to what was suggested by Ghosal et al. (2009). As CssA does not bind fibronectin, but binds to cells, it must recognize a different receptor.

Both mini-fibres bound to fibronectin via a single binding site (Fig. S6). The absence of unspecific fibronectin binding sites on mini-fibres suggests that the low-affinity binding site found for CssBdsA is not native and probably is buried in the interface between subunits. CssBdsA–CssAdsB bound fibronectin with an affinity similar to that of the high-affinity site of CssBdsA (8.5 ± 0.7 μM). CssAdsB–CssBdsA bound to fibronectin notably weaker (30 ± 5 μM), suggesting that the interface between subunits may affect the binding site for fibronectin. The CssAdsB–CssBdsA–fibronectin binding was also weaker than interactions of this mini-fibre with cells (Fig. 8B). Presumably, the tighter binding of CssAdsB–CssBdsA to cells reflects binding to a different receptor via the CssA part of the molecule.

Fibronectin–CssB binding was about 100 times weaker than that between fibronectin and CS6 purified from ETEC (Ghosal et al., 2009), highlighting the important role of avidity for fibronectin–CS6 binding. Polyvalent binding of CS6 to fibronectin molecules is consistent with our model for the CS6 fibre (Fig. 8E).

We failed to detect binding between our SC constructs and sulphatide using an ELISA assay (Supplementary Fig. S7). It is possible that the binding between sulphatide and an individual subunit is not tight, and, as in the case of fibronectin, high avidity is necessary to establish firm interactions between full-length CS6 fibres and surface-exposed sulphatide molecules (Fig. 8E).


A linear polymer of subunits joined ‘head-to-tail’ by donor strand complementation is the basic structure of all fimbriae assembled via the chaperone/usher (CU) pathway so far characterized. We show that despite the non-fimbrial morphology (Ludi et al., 2006), CS6 assembly also involves donor strand complementation. Nevertheless, the structure of CS6 fibres is principally different from that of other CU fimbriae described to date. All previously studied CU fimbriae consist of long stretches of homo-polymers made of subunits of the same type. Often, these homo-polymers are joined into composite structures via specialized ‘linker’ subunits [e.g. type 1 or Pap fimbriae (Waksman and Hultgren, 2009)] and/or decorated by a single specialized subunit at the tip of the fibre. In contrast, the CS6 fibre is built as a hetero-polymer with two different subunits alternating along the entire fibre. What are the functional implications of this structure?

Our cellular studies suggest that both CssA and CssB possess adhesive properties. We also show that CssB, but not CssA specifically binds to the extracellular matrix protein fibronectin, suggesting that CS6 subunits bind to different receptors. This is in accord with the observation of large differences in the structure of the loop region between the C and D β-strands of the two subunits, which is likely to be responsible for their binding properties. Since CssA and CssB alternate along the fibre, the different binding surfaces must appear at a certain periodicity along the fibre. Such an arrangement would enable binding of the CS6 fibre simultaneously to multiple receptors of two (or more) different types (Fig. 8E). Polyvalent binding to cell surface receptors is consistent with the observation of considerably tighter binding of full-length CS6 fibres than mini-fibres to epithelial cells. The high flexibility of the CS6 fibre predicted from our modelling would help to establish such multiple contacts.

CU fimbriae can be grouped by functional properties into monovalent and polyvalent adhesins (Zav'yalov et al., 2010). Our results suggest that CS6 is a polyvalent adhesin of a novel type: a hetero-polyadhesin. Hetero-polyadhesins might be relatively common. A number of FGL CU operons are predicted to assemble surface organelles from two different subunits (Zavialov et al., 2007). These organelles might have a structure and function similar to those of CS6. Hetero-polyadhesins are also likely to be found among the highly diverse class of FGS CU fimbriae, the polyadhesive properties of which are increasingly recognized (Zav'yalov et al., 2010).

CS6 positive strains account for up to 30% of all clinical ETEC isolates and have been isolated with increased frequency in recent studies (Qadri et al., 2005; Sack et al., 2007). Owing to this fact and the recent demonstration of an efficient immune response against CS6 (Harro et al., 2011; Tobias et al., 2011), there is strong interest in trying to use CS6 alone or in combination with other antigens in an ETEC vaccine. The knowledge of the structure of CS6 antigen may significantly facilitate development of such a vaccine. The mini-fibres created in this study, which contain all epitopes of CS6, might represent useful components of an anti-ETEC vaccine.

Experimental procedures

Protein expression and purification

To produce native proteins E. coli BL21(DE3) transformants were grown in Luria–Bertani (LB) medium containing 100 μg ml−1 ampicillin at 37°C. Exponentially growing cells were induced for protein expression with 1 mM isopropyl β-d-1-thiogalactopyranoside. Periplasmic proteins were extracted by an osmotic shock procedure as described in Chapman et al. (1999). Target proteins were purified successively by anion exchange chromatography in 20 mM Tris-HCl buffer, pH 7.4 (CssBdsA, CssBdsB) or pH 7.8 (CssAdsA, CssAdsB) and cation exchange chromatography in 50 mM sodium acetate buffer, pH 4.1, using Mono Q and Mono S HR 10/10 columns (GE healthcare) respectively. A 0–200 mM gradient of NaCl was used to elute proteins in both steps. To obtain highest purity samples, proteins were subjected to gel filtration on a Superdex-75 16/60 HiLoad column (GE Healthcare) equilibrated with 50 mM HEPES, pH 7.5 and 150 mM NaCl. Proteins were concentrated to 2–5 g l−1 for analysis or to 20 g l−1 for crystallization experiments on a Vivaspin device (GE healthcare) with molecular weight cut-off of 5 kDa. Selenomethionine (SeMet)-labelled CssAdsB and CssBdsA were produced by SeMet incorporation with inhibition of Met biosynthesis as described in Supplementary Experimental procedures.

Crystallization and structure determination

Crystallization was performed by the hanging-drop vapour-diffusion method at 293 K. For CssAdsA and CssAdsB, crystals were obtained in drops with 24–30% PEG 4000 in 0.2 M ammonium acetate, 0.1 M Na acetate, pH 4.6. For CssBdsA, crystals were obtained in drops with 30% PEG 4000 in 0.2 M Na acetate, 0.1 M Tris-HCl, pH 8.5. Crystals were soaked for 30–60 s in cryoprotection solution prepared by mixing two parts of precipitant solution with one part of 50% PEG 400. Diffraction data were collected under liquid-nitrogen cryoconditions at 100 K on beam-lines ID29 and BM14 at ESRF, France. Data were collected, processed and reduced using MOSFLM and SCALA (CCP4, 1994). Using PHENIX (Adams et al., 2002), heavy-atom parameters were obtained and refined, and initial phases calculated. The initial models were constructed using PHENIX and O (Jones et al., 1991). Positional, bulk solvent and isotropic (CssAdsA and CssAdsB) or anisotropic (CssBdsA) factor refinement was performed using PHENIX. No TLS refinement was used. Hydrogen atoms were included and refined in the final model of CssBdsA. Data collection and refinement statistics are given in Table 2.

GdmCl- and temperature-dependent folding transitions

To study unfolding, CssAdsA (3.2 g l−1), CssAdsB, (2.7 g l−1), CssBdsB (1.8 g l−1), CssBdsA (2.0 g l−1) were diluted 10 times with 20 mM phosphate buffer (pH 6.4) (buffer B) containing different concentrations of GdmCl and incubated at 21°C for 16 h. The GdmCl-induced unfolding transition was measured by following the change of ellipticity at characteristic wavelengths (Supplementary Fig. S2) and evaluated according to the two-state model of folding using a six-parameter fit (Santoro and Bolen, 1988). Free energy for unfolding of proteins in the absence of GdmCl was determined by linear extrapolation of the dependence of the free energy of unfolding on GdmCl concentration to 0 M (Santoro and Bolen, 1988). To study temperature denaturation, CssAdsA (0.26 g l−1), CssAdsB (0.28 g l−1) and CssBdsB (0.17 g l−1) and CssBdsA (0.12 g l−1) in buffer B were heated from 20°C to 90°C at a heating rate of 1°C min−1, and the CD signal at a characteristic wavelength was continuously recorded. No precipitation was observed.

Fibronectin binding assay

A Biacore X100 system (GE Healthcare) was used for all biosensor experiments. Fibronectin (Sigma) [approximately 1800 resonance units (RU)] was immobilized on flow cell 2 of a CM5 Sensor Chip by amine coupling using an Amine Coupling Kit (GE Healthcare). To record the association and dissociation curves, samples of SC subunits or mini-fibres at varying concentrations were injected into flow cell 2 of the chip for 3 min followed by flushing of the cell with 10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% Tween 20 (HBS-EP) for 3 min at a flow rate of 10 μl min−1. Identical samples were injected over a control flow cell to determine non-specific binding, which was subtracted from the experimental curves. After each data acquisition cycle, the chip was fully regenerated with 10 mM NaOH in HBS-EP containing 1 M NaCl. The equilibrium constants were determined by applying a one- or two-receptor binding model (CssBdsA), using the Biacore X100 evaluation software and Simfit/HLFIT program.

Cell binding assay

Caco-2 cells obtained from the American Type Culture Collection (ATCC, Manasas, VA, Item No. HTB-37) were grown in high glucose Dulbecco's modified Eagle's medium (DMEM, Sigma) supplemented with 20% inactivated fetal bovine serum (FBS), 1%, MEM non-essential amino acids, 4 mM l-glutamine, 10 000 units ml−1 penicillin and 10 mg ml−1 streptomycin at 37°C in a humidified atmosphere containing 5% CO2. For immunoblot analyses, cells were seeded at a density of 8000 cells per well in 96-well tissue culture plates (Greiner) and grown 48 h to reach confluence. Cells were washed twice with plain medium without supplements. SC constructs in plain medium at different concentrations were added to cells in triplicate and cells were incubated at 37°C and 5% CO2 for 3 h. Cells were washed three times with PBS and lysed in boiling lysis buffer (286 mM Tris, pH 6.8, 2.86% SDS). Lysates were resolved by SDS electrophoresis in 15% polyacrylamide gels, proteins transferred to a polyvinylidene fluoride membrane, and the blot developed by an enhanced chemiluminescence detection method (Pierce) using polyclonal rabbit antibodies generated against CssAdsB and CssBdsA (Innovagen AB, Lund) and horseradish peroxidase (HRP)-conjugated anti-rabbit antibody (Sigma).

Monoclonal antibody binding assay

Ninety-six-well EIA/RIA plates (Corning) were coated with 25–200 ng of SC construct per well. After blocking the plates with 2% BSA, the wells were incubated with a monoclonal antibody at a concentration of ∼10−4 mg ml−1 for 1 h, followed by addition of HRP-conjugated anti-mouse antibody for 1 h. Colour reactions were developed with ABTS substrate (Thermo Scientific) and absorbance at 405 nm was measured in a VICTOR plate reader (Wallac).

Glutaraldehyde cross-linking

CS6 (1 g l−1) was incubated with different dilutions of glutaraldehyde in 50 mM phosphate buffer, pH 7.4 for 0.5 h at 20°C. The reaction was stopped by adding Glycine buffer, pH 8.0, to a concentration of 0.15 M. The resulting samples were separated by SDS electrophoresis in a 17% polyacrylamide gel, followed by immunoblotting as described under ‘Cell binding assay’.

Accession numbers

The atomic co-ordinates and structure factors have been deposited in the RCSB Protein Data Bank, (PDB ID codes 4B9J, 4B9I and 4B9G for CssAdsA, CssAdsB and CssBdsA respectively). The model of CS6 fibre was deposited to protein model database, (PMDB id: PM0078450).


This work was supported by Grants FORMAS-221-2007-1057, FA-136333 and FA-140959 to A.V.Z. and VR-2009-5334 to S.D.K. We would like to thank Taha Ghawanmeh for his assistance with protein preparation, Dr Tobias and Prof. Svennerholm for their kind gift of monoclonal antibodies, Dr F. Cassels and Dr E. Oaks for their kind gift of the CS6 antigen, and staff of beam-lines ID29 and BM14 at the ESRF (Grenoble, France) for their kind help during data collection.