Domain Architecture of Collagen-like Proteins in EHEC Genomes

Several open reading frames potentially encoding collagen-like proteins have been identified by automatic sequence annotation in the genomes of EHEC strains. Those from the Sakai and EDL933 genomes will be discussed here, but many related sequences have been identified in other strains. Their primary structures show one or more collagen-like domains (Col) with the repeating collagen signature sequence (Gly-X-Y)_n, flanked at both ends by a series of non-collagenous, conserved domains (Figure 1 and Table 1). Domains PfN, Pf2 and PfC have been described on the basis of sequence conservation and are associated to fibre tail proteins from phages. They appear in automatic annotation of EPclPs. Figure 1 shows the different protein architectures and the nomenclature used here to refer to them, plus two representative sequences. The most common architecture (EPclA) appears in multiple copies in each genome, with more than 90% amino acid sequence identity across copies. Table 2 gives the complete list of EPclP sequences from the Sakai and EDL933 genomes, whereas representative examples of other architectures and strains are given in Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Collagen-like proteins from prophages embedded in the genomes of E. coli O157:H7 and other EHEC strains, referred here as EPclA to EPclD (EHEC Prophage collagen-like A to D).

(A) Domain architectures. The collagen triple helical domains are labelled “Col”, and domains predicted to adopt an α-helical coiled-coil conformation (see text) are labelled “PCoil” (for phage coiled-coils). Key to other domain labels (Table 1): PfN, phage fibre N-terminal domain; PfC, phage fibre C-terminal domain; PfC2, phage fibre C-terminal domain, variant 2; Pf2, phage fibre repeat 2. (B) Sequence of a representative collagen-like protein with EPclA architecture (ECs2717), from the genome of E. coli O157:H7 Sakai. (C) Sequence of a representative collagen-like protein with EPclB architecture (Z1483), from the genome of E. coli O157:H7 EDL933. Amino acid sequences corresponding to the different predicted domains are colour-coded as in (A).

https://doi.org/10.1371/journal.pone.0037872.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Domains observed in collagen-like proteins from the genomes of EHEC strains.

https://doi.org/10.1371/journal.pone.0037872.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Collagen-like proteins from the genomes of E. coli O157:H7 EDL933 and Sakai strains, and their corresponding prophage locations.

https://doi.org/10.1371/journal.pone.0037872.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Examples of collagen-like proteins from other E. coli and Shigella strains, and their prophage locations.

https://doi.org/10.1371/journal.pone.0037872.t003

The EPclA architecture shows a single collagen triple helical sequence capped by PfN and PfC domains at the N- and C-termini, respectively. Between the PfN domain and the Col domain there is a region of low-complexity. Analysis of its amino acid sequence suggests a coiled-coil conformation (see below), and thus will be referred here as PCoil domain. The EPclB architecture shares the presence of PfN, PCoil and PfC non-collagenous domains, and contains two Col domains separated by a Pf2-type repeat. Protein sequences within each type of architecture show variable lengths of their Col and PCoil domains (Tables 2 and 3). Differences in length are typically multiples of three for Col domains and multiples of seven for PCoil domains, which is consistent with the lengths of the repetitive motifs in collagen and coiled-coil sequences, respectively. In sequences with two Col domains it is common that the first one contains a single interruption of the (Gly-X-Y)_n repeating pattern, with a conserved Gly-X-Pro-Gly-Gly-Pro-X-Gly sequence.

Only a few sequences conform to the EPclC and EPclD architectures (Table 3), which are characterized by different C-terminal regions with no sequence homology to the PfC domains of EPclA or EPclB architectures. Also, the PCoil domain is often missing in EPclC architectures. The EPclD sequence Stx2-86_gp21, from the Stx2-86 prophage in the Shiga toxin-producing E. coli strain O86:H– (accession codes Q08J84, YP_794068), has a 322-amino acid Col domain (Table 3), by far the longest collagen-like sequence of all EPclPs. This long domain also shows a single Gly-X-Pro-Gly-Gly-Pro-X-Gly interruption. No examples of EPclC or EPclD sequences are found in the EDL933 or Sakai genomes.

The PfN and Pf2 domains are not exclusive to collagen-like proteins from E. coli prophages and were first identified in the side tail fibre protein coded by the stf gene from λ bacteriophages [35], [36] (accession code P03764). Virions with a functional stf gene show jointed tail fibres, expanded receptor specificity, and adsorb more rapidly to E. coli cells. Homologous proteins have been identified in embedded prophages of many E. coli strains, including the laboratory reference strain K12 (protein stfR/ynaB, accession code P76072).

Other than their consistent presence in prophage tail fibre proteins, little is known about the structure and function of the PfN and Pf2 domains. Some sequence similarity between PfN and a regulatory domain of eukaryotic carboxypeptidases may be indicative of a proteolysis-related function, but to date there is no experimental evidence for this. Due to the presence of these domains, EPclP sequences are automatically annotated as putative tail fibre proteins.

Amino Acid Composition and Positional Preference in the Collagen Domains of EPclPs

Collagen domains in EPclPs show amino acid preferences in the X and Y positions that differ from those seen in other collagen proteins. By far the most common residue in the X position is proline (Pro, P), which occurs there close to half of the time (Table 4). By contrast, Pro is relatively infrequent in the Y position. Both X and Y positions also show a strong preference for charged amino acids, aspartate/glutamate (Asp/Glu, D/E) in the X position and lysine/arginine (Lys/Arg, K/R) in the Y position. Alanine (Ala, A) is also relatively frequent at both X and Y positions, and both glutamine (Gln, Q) and threonine (Thr, T) show a clear preference for the Y position. Interestingly, cysteine, phenylalanine, histidine, tryptophan and tyrosine are absent in collagen domains from EPclPs. At the triplet level, the most common are GPK (15%), GPQ (12%), GPA (12%), GER (8%) and GET (8%). The triplet pattern GP(Q/P)–GPK–G(D/E) is repeatedly observed in the collagen domains of EPclPs.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Position-specific amino acid preferences in collagen triple-helical domains of EPclPs, human collagens, and collagen-like proteins from different groups of organisms.

https://doi.org/10.1371/journal.pone.0037872.t004

The position-specific amino acid preferences in EPclPs are quite different from those seen in animal collagens, as shown for example by the human sequences (Table 4 and [37]). The most obvious difference is in the Pro distribution: human collagens have a clear preference for Pro residues in both the X and Y positions, close to 30% and 35% respectively. There is some variation between fibrillar and non-fibrillar collagens and with the collagen-like proteins, but Pro residues are invariable more common in the Y position of human collagens. The reason is well known: Pro residues in the Y position are often modified post-translationally to 4-hydroxyproline (Hyp, O), which contributes to the thermal stability of the collagen domains. Charged residues are also frequent in the X and Y positions of human collagens, with the same positional preferences as in collagen domains from EPclPs (Asp/Glu more often in X, Lys/Arg more often in Y). However, they are overall less frequent and their preferential position is less strict (Table 4). Other amino acids significantly contribute to the sequence variability at each position.

The expected average conformational parameters of the triple helical Col domains can be calculated from the distribution of imino acids along their sequences [38]. The expected values are –106° for the average twist and 2.88 Å for the average height, same as those predicted for human fibrillar collagens [38]. Thus, despite the differences in amino acid composition and positional preference, the overall conformation of the triple helical Col domains is expected to be very similar to that of human fibrillar collagens.

Collagen sequences found in other viral proteins are more similar to those from EPclPs, although the preference for Pro in the X position is not that strong. Viral collagens share with EPclPs the low proportion of Pro residues in the Y position, large number of charged amino acids, and relatively common occurrence of Gln and Thr in the Y position. Collagens from gram-positive bacteria, which include the well-studied examples of Bacillus anthracis or Streptococcus pyogenes [22], [24], [39], show a lower presence of Pro residues in the X and Y positions, much lower proportion of charged amino acids, and a higher proportion of Ala residues in the X position and Gln and Thr in the Y position.

The main difference between human collagens and the three groups of non-animal collagens in Table 4 is the lack of preference for Pro in the Y position (as already noted in an earlier analysis of viral and bacterial collagen structural motifs [21]). Bacteria and viruses do not have the prolyl-hydroxylase enzymes required for hydroxylation of Pro in the Y position of a collagen triple helix, and therefore there is no contribution to the stability of their collagen domains by Hyp residues. Collagen domains from EPclPs appear to compensate this lack of prolyl hydroxylation with a larger proportion of Pro residues in the X position. The high ratio of charged amino acids and the relatively high occurrence of Ala, Gln and Thr in EPclPs and bacterial and viral collagens may be indicative of different mechanisms for stability of their collagen domains [40], [41], [42].

Interestingly, the metazoan collagen sequence closest to EPclPs comes from a sea anemone, Nematostella vectensis (NCBI accession code XP_001625905, incomplete sequence), with 56% identity to the collagen domain of the EPclA protein ECs2717 and containing a repetitive [GP(Q/E)-GPK-GDT-GIT]₁₂ sequence, reminiscent of the commonly observed triplet pattern mentioned above.

A Low Complexity Region is Predicted as α-helical Coiled-coil Domain (PCoil)

The region between the predicted PfN and Col domains in the most common architectures, EPclA and EPclB, shows an unusual low-complexity sequence with predominance of Ala (32%), Ser (19%) and Glu (13%) amino acids that often appear in tandems or in stretches of up to four consecutive identical residues (Figure 1). Different coiled-coil predicting algorithms (PCoils, Marcoil, MultiCoil) give high scores for the region between residues 101 and 245 in both EPclA and EPclB (Figure 2). This region includes the low-complexity sequence between the PfN and Col domains plus the last 34 residues of PfN, and shows a loose seven-residue Ala-X-X-Ala/Ser-X-X-Ser periodicity, where residues in the X positions are often charged. On account of the coiled-coil predictions we will refer to the low-complexity region between PfN and Col as the PCoil domain. The MultiCoil and SCORER 2.0 prediction algorithms favour a trimeric rather than dimeric coiled-coil structure for PCoil. Secondary structure prediction by Jpred3 suggests that the PfN domain has mainly a β-sheet structure for the first 80 residues and some α-helical conformation from residues 90 onwards, whereas the PCoil region is predicted to be mainly α-helical. Jpred3 does not predict any secondary structure for the PfC domains (data not shown).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Coiled-coiled predictions for the amino acid sequences of (A) EPclA (ECs2717) and (B) EPclB (ECs1228), using the PCoils [70] and Marcoil [71] algorithms.

The graphs indicate regions of high probability for α-helical coiled-coil formation. Three different sequence window sizes were used with the PCoils algorithm: 14, 21 and 28 residues. Two different matrices were used in Marcoil: 9FAM, and MTK-based.

https://doi.org/10.1371/journal.pone.0037872.g002

EPclA is a Trimeric Protein that Dissociates When Denatured

The quaternary structure of rEPclA was investigated by sedimentation equilibrium analytical ultracentrifugation (AUC) at increasing concentrations of guanidinium chloride (GuHCl) (Figure 3). The relative molar mass of rEPclA at 0 M GuHCl was 138±6 kDa, corresponding to the predicted molecular weight of a trimer of rEPclA molecules (3×47 kDa). As the concentration of GuHCl increased, a transition from trimer to monomer was observed and the relative molar mass of rEPclA at 5 M GuHCl was 43 kDa, which is consistent with the predicted molecular weight of the rEPclA monomer. Thus, rEPclA trimers dissociate into monomers as the GuHCl concentration increases; the trimer-to-monomer transition point was estimated at around 2.5 M GuHCl.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Analysis by analytical ultracentrifugation of the average molar mass of a sample of purified rEPclA as a function of increasing concentration of guanidinium chloride (GuHCl).

Weight-averaged molar mass was determined using a single ideal species model (see Methods). Mean value masses for the upper and lower plateaux were 138±6 kDa and 43±1 kDa respectively (averages of three measures). The molecular mass of native rEPclA (0 M GuHCl) is consistent with three times that of denatured rEPclA (see text). The transition midpoint concentration is 2.38±0.09 M GuHCl.

https://doi.org/10.1371/journal.pone.0037872.g003

An independent measurement of the molecular weight of rEPclA was carried out by size exclusion chromatography followed by multiangle laser light scattering (SEC/MALLS) (Figure S1). The molecular weight obtained from MALLS is consistent with a trimer of rEPclA (Table 5). A proteolytic fragment from rEPclA that included only the Col and PfC domains (Col–PfC fragment, Figures S2 and S4) could be produced in enough amounts for biophysical characterization. Analysis by SEC/MALLS of fractions containing the Col–PfC fragment (Figure S1) showed it to be trimeric as well (Table 5), with a molecular weight of 64 kDa consistent with three times the molecular weight of monomer Col–PfC (21–22 kDa, predicted from the peptide fingerprinting data obtained from mass spectrometry, Figure S2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Average molecular weights of different recombinant fragments, calculated from the SEC/MALLS data.

https://doi.org/10.1371/journal.pone.0037872.t005

The molecular weights obtained from AUC and MALLS experiments were consistent with the predicted values for non-glycosylated rEPclA trimers and monomers. Lack of glycosylation of rEPclA was confirmed by periodic acid-Schiff staining analysis (data not shown).

Domains PCoil and PfC from EPclA are Trimerization Modules

Molecular weights of several recombinant fragments containing different combinations of domains were determined by SEC/MALLS (Figure S3 and Table 5). The data indicates that PfC is a trimerization domain, forming trimeric assemblies both when fused to a thioredoxin tag (Trx–PfC) or after removal of thioredoxin by thrombin digestion. The PfN–PCoil fragments were also trimeric, whereas the PfN domains were mainly in the monomer state (Figure S3). This data suggests that PCoil is also a trimerization domain.

EPclA Shows a CD Spectrum Consistent with Collagen and α-helical Conformations

The secondary structures of rEPclA and the Col–PfC fragment were analysed by circular dichroism (CD). Interpretation of the results is easier if the Col–PfC CD data is considered first. A sample of Col–PfC was purified from a preparation of full-length rEPclA by nickel-affinity and size exclusion chromatographies. Its CD spectrum was measured at different temperatures between 195 and 260 nm. The concentration of the Col–PfC sample was calculated as 0.2 mg/ml from its UV absorption at 280 nm and an estimated molar extinction coefficient ε = 11000 M^-1 cm^-1. The CD spectrum of Col–PfC at 4°C (Figure 4A) shows the characteristic features of triple helical collagen: a small maximum of positive ellipticity at 220 nm and a deep minimum of negative ellipticity at around 199 nm [43]. Both these features are associated with the polyproline II conformation [44] characteristic of the collagen triple helix. The Col–PfC fragment includes mainly the collagen domain (Col) of EPclA and the C-terminal PfC domain, and thus its CD spectrum suggests that the Col domain adopts indeed a collagen-like, triple helical structure. The triple helical features disappeared from the CD spectrum upon increase of temperature, as shown by the CD curve at 55°C (Figure 4A). Interestingly, immediate cooling of the same sample back to 4°C recovered completely the triple helical structure, with a CD spectrum practically indistinguishable from the initial one (Figure 4A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Far-UV CD analysis of the Col–PfC fragment after purification from rEPclA by SEC.

(A) CD spectra at 4°C, 55°C and 4°C after immediately cooling back the sample (see text). The vertical axis measures mean residue ellipticity Θ in degrees cm² dmol^-1. The CD data was collected between 195 and 260 nm, with a protein concentration of 0.2 mg/ml in 10 mM Tris, 150 mM NaCl, pH 7.4. Measurements were taken in a 0.5 mm path length cell. (B) Thermal denaturation of the Col–PfC fragment, monitored by CD at 220 nm as a function of increasing temperature between 4°C and 60°C, with a protein concentration of 0.2 mg/ml in 10 mM Tris, 150 mM NaCl, pH 7.4, and a heating rate of 0.33°C/min.

https://doi.org/10.1371/journal.pone.0037872.g004

To examine the thermal denaturation of the collagen triple helix in the Col–PfC domain, the CD of another sample of purified Col–PfC fragment (also 0.2 mg/ml) was monitored at 220 nm as a function of continuously increasing temperature, from 4°C to 60°C. The thermal curve showed a single sharp transition at 42°C, which typically corresponds to the decrease of ellipticity at 220 nm and loss of collagen triple helical structure (Figure 4B).

The CD spectrum of rEPclA is different. A diluted sample of rEPclA was purified by nickel-affinity and size exclusion chromatographies and its CD spectrum was measured between 195 and 260 nm at 4°C (Figure 5A). The concentration of the rEPclA sample was calculated as 0.04 mg/ml from its UV absorption at 280 nm and a molar extinction coefficient ε = 17000 M^-1 cm^-1, deduced from the amino acid sequence of rEPclA. The CD spectrum showed two minima of negative ellipticity at 205 nm and 224 nm, the first one being deeper, and a small local maximum between the two minima, at 216 nm. To investigate this region in more detail a second sample with higher concentration, 0.3 mg/ml, was analyzed at different temperatures. When the sample was heated to 45°C, the height of the 216 nm maximum decreased significantly, the intensities of the two minima became more similar to each other, and their positions shifted to 210 nm and 222 nm, respectively (Figure 5A). This spectrum resembled more that of an α-helical coiled-coil conformation. Upon further increase of the temperature the overall ellipticity became less negative, and the two minima started to disappear and vanished completely when reaching 55°C (Figure 5A). The spectrum did not change upon further increase of temperature to 65°C. The slight decrease in ellipticity at 216 nm around 45°C and the more similar intensities of the two minima at that temperature suggest changes in the secondary structure that are consistent with the loss of the triple helical conformation in the Col domain while maintaining an α-helical conformation (Figures 4 and 5). Such α-helical structure appears to be more stable and does not disappear completely until 55°C. Immediate cooling of the same sample from 65°C back to 4°C recovered approximately half of the initial CD spectrum (Figure 5A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Far-UV CD analysis of rEPclA after purification by SEC.

(A) CD spectra at 4°C, 45°C, 55°C and 4°C after immediately cooling back the sample (see text). The vertical axis measures mean residue ellipticity Θ in degrees cm² dmol^-1. The CD data was collected between 195 and 260 nm, with a protein concentration of 0.04 mg/ml (4°C) or 0.3 mg/ml (the rest) in 10 mM Tris, 150 mM NaCl, pH 7.4. Measurements were taken in a 0.5 mm path length cell. (B) Thermal denaturation of rEPclA monitored by CD at 216 nm (the maximum between the two minima at 208 and 224 nm). The CD was measured as a function of increasing temperature between 20°C and 75°C, with a protein concentration of 0.3 mg/ml in 10 mM Tris, 150 mM NaCl, pH 7.4, and a heating rate of 0.33°C/min.

https://doi.org/10.1371/journal.pone.0037872.g005

To examine the thermal denaturation of rEPclA, the CD of another sample of purified rEPclA (concentration 0.3 mg/ml) was monitored at 216 nm as a function of increasing temperature from 20°C to 75°C. Two transitions were observed: a first transition at 42°C showed a sharp decrease in ellipticity, consistent with the loss of collagen triple-helical structure seen previously for the Col–PfC fragment at the same temperature; a second transition at 52°C showed a sharp and pronounced increase in ellipticity, corresponding to the loss of α-helical structure of the PCoil and PfN domains (Figure 5B). Thus, the α-helical structure of the PCoil and PfN domains is more stable than the collagen triple helix of the Col domain. The transition temperature of the Col domain is the same in rEPclA and its Col–PfC fragment, and seems unaffected by the presence of the PCoil and PfN domains. The melting transitions of the α-helical and collagen structures therefore appear to be largely independent.

The PfN–PCoil Region is Clearly α-helical and is Consistent with a Coiled-coil Structure

The secondary structures of the recombinant fragments PfN–PCoil and PfN were also studied by CD spectroscopy. Recombinant PfN–PCoil and PfN fragments were each purified with nickel-affinity and size-exclusion chromatographies. Concentrations of the PfN–PCoil and PfN samples were measured as 0.2 mg/ml and 0.3 mg/ml respectively, from their absorption at 280 nm and molar extinction coefficients ε = 7000 M^-1 cm^-1, initially calculated from the amino acid sequences of the PfN–PCoil and PfN recombinant fragments and adjusted using the observed UV absorption of samples in 8 M urea (see Materials and Methods). The CD spectrum of PfN–PCoil at 4°C in phosphate buffer (Figure 6A) shows the characteristic features of an α-helical protein, with two minima at 208 nm and 222 nm and a maximum at 195 nm. This spectrum is consistent with the prediction of an α-helical coiled-coil conformation for the PCoil region. The α-helical features were still present in a spectrum measured at 45°C, although the signal intensities at the two minima started to decrease (data not shown). These features disappeared when reaching 60°C (Figure 6B), and were mostly recovered upon cooling the sample back to 20°C (data not shown). The spectrum of the PfN domain (Figure 6A) was also consistent with an α-helical structure but the intensity of the two minima at 208 nm and 222 nm was much lower than in the PfN–PCoil spectrum, indicating a lower α-helical content in this domain. This spectrum also vanished at 60°C (Figure 6C), but it was not recovered upon cooling back to 20°C (data not shown). Thus, the main contribution to the CD signal comes from the PCoil domain, most likely through the formation of a trimeric α-helical coiled-coil structure that disappears at a temperature of 60°C but is regained when the temperature is lowered again.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Far-UV CD spectra of the PfN–PCoil and PfN fragments after purification by SEC:

(A) PfN and PfN–PCoil at 20°C; (B) PfN–PCoil at 20°C and 60°C; (C) PfN at 20°C and 60°C. In all panels the vertical axis measures mean residue ellipticity Θ in degrees cm² dmol^-1. The CD data was collected between 195 and 260 nm, with protein concentrations of 0.2 mg/ml (PfN–PCoil) or 0.3 mg/ml (PfN), in 20 mM phosphate buffer (Na₂HPO_4/NaH₂PO₄), 100 mM NaCl, pH 7.4. Measurements were taken in a 0.5 mm path length cell.

https://doi.org/10.1371/journal.pone.0037872.g006

To examine the thermal transitions of the PfN–PCoil domain, the CD of two more samples of purified recombinant PfN–PCoil (0.35 mg/ml concentration) and PfN (0.1 mg/ml) were monitored at 222 nm as a function of continuously increasing temperature, from 5°C to 95°C and then cooling back to 5°C at the same speed (1°C per minute). The heating thermal curve showed a single sharp transition at around 49°C (Figure 7), corresponding to the loss of the strong α-helical CD spectrum. The cooling thermal curve also showed a single sharp transition at around 45°C indicating partial re-gaining of the α-helical structure (the baseline in Figure 7 does not recover its initial value). This data suggests reversibility for the thermal transition of PfN–PCoil, mainly for the formation of the trimeric α-helical coiled-coil structure in the PCoil domain; the PfN domain does not recover completely after thermal denaturation and slow cooling. The actual value of the temperature of the transition is sensitive to the heating or cooling rate. A repeat of the heating experiment with a slower speed (0.33°C per minute) gave a transition temperature of 52°C for the PfN–PCoil fragment (data not shown).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. Thermal denaturation and renaturation of recombinant PfN–PCoil (orange) and PfN (red) monitored by CD at 222 nm (corresponding to a minimum in both CD spectra).

The CD was measured in a 1 mm path length cell as a function of increasing temperature between 5°C and 95°C (left) and then decreasing temperature between 95°C and 5°C (right). The temperature was changed at a rate of 1°C per minute. Both PfN–PCoil (0.35 mg/ml) and PfN (0.1 mg/ml) were in 10 mM Tris, 150 mM NaCl, pH 7.4. PfN–PCoil showed a sharp transition at around 49°C corresponding to the loss of α-helical coiled-coil structure. The CD signal was almost completely recovered upon cooling, with a sharp transition about 45°C. This behaviour is indicative of a reversible structural transition for the α-helical coiled-coil. The PfN fragment gradually lost its CD signal with a transition midpoint of about ∼52°C. The gradual nature of this transition suggests denaturation rather than a cooperative unfolding. The PfN CD signal was not regained upon cooling.

https://doi.org/10.1371/journal.pone.0037872.g007

Structural Organization of rEPclA and its Fragments Col–PfC and PfN–PCoil

Full-length rEPclA was analyzed by rotary shadowing electron microscopy. Examination of the electron micrographs of rEPclA showed a “dumbbell-shaped” structure with two globular particles joined by a semi-flexible stalk, in which it is possible to distinguish two regions of different thickness (Figures 8 and 9). Sequence analysis of the PCoil region and the CD spectrum of PfN–PCoil both suggest that the PCoil domain is a trimeric α-helical coiled-coil structure, and the observed thicker region of the stalk is consistent with such coiled-coil helical structure, which is known to have a larger cross-section than a collagen triple helix [45]. The remaining thinner region corresponds to the collagen triple-helical domain.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 8. Rotary shadowing electron microscopy of rEPclA.

(A, B) Different micrographs showing dumbbell-shaped structures corresponding to rEPclA trimers. The globular shapes correspond to the PfN and PfC terminal domains, presumably forming trimeric structures themselves. Flexible stalks connecting these globular structures contain the trimeric collagen triple-helical region (Col) and the trimeric α-helical coiled-coil region (PCoil) of rEPclA. The concentration of rEPclA was 5 µg/ml. Scale bar  = 100 nm. (C) Detailed view of an rEPclA trimer. The arrows indicate the globular terminal domains and the thin (Col) and thick (PCoil) regions of the stalk, respectively. The globular domains can be identified as N- or C-terminal by their position with respect to the two stalk zones.

https://doi.org/10.1371/journal.pone.0037872.g008

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 9. Molecular dimensions of rEPclA and its domains.

(A) Magnified view of a representative rEPclA molecule from a rotary shadowing electron micrograph. (B) The same molecule with the background masked out showing the different molecular dimensions analyzed below. (C) Average dimensions obtained from multiple measures on electron micrographs: N···C and <Coil···Col are averages of 224 measures on eight rEPclA micrographs; D_C, L_Col and T_Col are averages of 76, 35 and 74 measures, respectively, on three Col–PfC micrographs (Figure 10A); D_N, L_Coil and T_Coil are averages of 200 measures on one PfN–PCoil micrograph (Figure 10B). (D) Histograms showing the distribution of N···C and <Coil···Col values on the sample of 224 rEPclA molecules.

https://doi.org/10.1371/journal.pone.0037872.g009

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 10. Rotary shadowing electron microscopy of Col–PfC and PfN–PCoil fragments.

(A) Micrograph showing the morphology of Col–PfC fragments (concentration 1 µg/ml). (B) Detailed view of one Col–PfC fragment. The globular shape corresponds to a trimer of PfC domains and the stalk corresponds to the trimeric collagen triple-helical domain (Col). The N-terminal end of the stalk shows a short, unravelled tail, where the PfN and PCoil domains have been removed. (C) Micrograph showing the morphology of PfN–PCoil fragments (concentration 5 µg/ml). (D) Detailed view of three PfN–PCoil fragments. The globular shapes correspond to trimers of PfN domains and the short tails correspond to the PCoil domains.

https://doi.org/10.1371/journal.pone.0037872.g010

Inspection of rEPclA molecules from different micrographs suggests a hinge between the two regions of the stalk (collagen and coiled coil) that results in variable angles between the PCoil and Col domains and variable distances between the two globular domains (Figure 9). For most rEPclA molecules the distance between the centres of the PfN and PfC domains (N···C in Figure 9) has values around 50 nm, reaching 55 nm for the most extended, linear ones. The average N···C distance (48±5 nm) is probably less significant than the values shown by the extended molecules. The N-terminal globular structure, made of three PfN domains, has an approximate diameter of 8 nm and is slightly bigger than the C-terminal globular structure made of three PfC domains (approximate diameter 7 nm) (Figure 9). Taking into account the radii of the globular domains, rEPclA molecules can reach a length of ∼62 nm when totally extended. However, most rEPclA molecules appear slightly bent at the hinge between the PCoil and Col domains and are, overall, a bit shorter (55–60 nm, including the globular domains).

Rotary shadow electron micrographs of Col–PfC fragments showed molecules very reminiscent of those of rEPclA but with only one globular domain (PfC) connected to a stalk (Col) (Figure 10A,B). The stalk region appears slightly unravelled where the α-helical coiled coil and the PfN domain have been removed by endogenous proteolysis of full-length rEPclA. The observed morphology confirms the previous assignment of N- and C-terminal domains for rEPclA images and that the thin region of the stalk corresponds to the collagen triple helix.

Rotary shadowing electron micrographs of PfN–PCoil fragments identified molecular shapes consistent with the N-terminal half of rEPclA molecules (Figure 10C,D). PfN domains (approximately the first 134 residues of EPclA) form a trimeric globular structure attached to an elongated stalk containing the PCoil domain. The micrographs show several instances of PfN–PCoil fragments apparently joined at the tails of their PCoil domains (Figure 10C). This association may result from some interaction between partially unravelled or unfolded chains at the terminal end of the PCoil domains.

From measures on the electron micrographs of the Col–PfC and PfN–PCoil fragments it is possible to estimate the lengths of the PCoil and Col regions as approximately 16 nm and 30 nm respectively (Figure 9). The length of the collagen domain is consistent with the predicted length of a collagen triple helix of 111 residues (111×2.9 Å = 32 nm, where 2.9 Å is an approximate measure of the height of an individual residue in a collagen triple helix [38]. Similar estimates can be obtained from measures on the thick and thin regions of the stalk in the rEPclA micrographs (14 nm for the PCoil region and 28 nm for the Col domain). The slightly shorter values suggest some overlap between domains in the rEPclA stalk that cannot be resolved in the rotary shadowing micrographs. The apparent thickness of the PCoil and Col domains in the electron micrographs over-estimates the true cross-section dimensions of these domains, as it includes the thickness of the shadowing replica. Nevertheless, a slight difference is observed between these estimates for the PCoil (3.3±0.5 nm) and Col (2.7±0.3 nm) domains. Cross-section values seen in high-resolution structures of these domains are closer to 1.3–1.7 nm for a collagen triple helix and 2.2–2.7 nm for a trimeric α-helical coiled coil.

Structural Organization of rEPclB High-molecular Weight Aggregates

Expression of rEPclB both by IPTG induction or auto-induction produced a good yield of soluble protein. However, purification of rEPclB from the soluble fraction by nickel-affinity and size exclusion chromatography showed that all the protein went to form soluble, high-molecular weight aggregates that eluted in the void volume of the size exclusion columns (data not shown). The protein aggregated to such an extent that it was not possible to identify any additional peak or shoulder suitable for molecular mass determination by MALLS. Aggregation was worse in samples produced by auto-induction (presumably due to the increased protein production). Attempts to reduce the degree of aggregation by lowering the protein concentration, adding EDTA (to rule out His₆-mediated metal chelation aggregation), adding glucose in high concentration, or changing the ionic strength of the buffers, were all unsuccessful: it was not possible to obtain enough non-aggregated rEPclB for molecular weight determination or for CD studies. Interestingly, the high molecular weight aggregates were soluble and the protein did not precipitate out of solution, even after high-speed centrifugation. The bands observed in SDS-PAGE experiments suggest that SDS treatment extracts monomeric rEPclB from the high molecular weight aggregates.

To investigate a possible structural organization of these aggregates, a sample of IPTG-induced rEPclB was used for rotary shadowing electron microscopy. The sample contained exclusively high-molecular weight aggregates that appeared in the electron micrographs as large masses of protein that, nevertheless, appeared to have a relatively narrow size distribution (300–500 nm in diameter, data not shown). Close inspection of the smallest aggregates (probably at an early stage of formation) revealed an internal structure that could be reconciled with entangled, multiple flexible linear beaded molecules (Figure 11). In the vicinity of these aggregates it was possible to discover individual features reminiscent of the rEPclA molecular morphology, but with three globular “domains” instead of two, connected by two flexible stalks (Figure 11). The terminal globular structures would correspond to the PfN and PfC domains, and the internal one would include the Pf2 domains. The flexible stalks would correspond to the two Col domains and the PCoil domain predicted in the rEPclB sequence. All these structures (and rEPclB) are presumed to be trimeric due to the presence of the PfC, Col and PCoil domains, all shown to trimerize in rEPclA. The structural organization for rEPclB would therefore be similar to that seen for rEPclA. However, an effective protocol to increase the amount of non-aggregated protein will be necessary to demonstrate these assumptions and to properly characterize the molecular architecture of rEPclB (work in progress).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 11. Rotary shadowing electron microscopy of rEPclB.

(A) Internal structure of a small aggregate of rEPclB molecules. The micrograph suggests multiple flexible molecules, reminiscent of those observed for rEPclA, with dark globular structures (presumably globular domains of PfN, PfC and Pf2), and poorly defined linear structures (presumably stalks containing the PCoil and Col domains). The rEPclB molecules seem to aggregate heavily through one of the globular domains. (B) Possible examples of individual rEPclB molecules observed, isolated from the large aggregates, in some electron micrographs. (C) Interpretation of the observed in terms of three globular domains (PfN, Pf2 and PfC) connected by two stalk regions. Approximate molecular dimensions are shown for comparison purposes with rEPclA.

https://doi.org/10.1371/journal.pone.0037872.g011

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 12. Typical morphologies of podoviridae and siphoviridae particles (reproduced with permission from ViralZone, Swiss Institute of Bioinformatics:

www.expasy.org/viralzone, [79]). The representative 933W phage and most field isolates show a podoviridae morphology, with isometric capsids of about 60–70 nm in diameter and short tails of 10–30 nm in length [12], [56]. EPclPs would be the main components on the side tail fibres.

https://doi.org/10.1371/journal.pone.0037872.g012

Sequence Retrieval and Analysis

Sequences of prophage collagen-like proteins from EHEC strains (EPclPs) were retrieved from the UniProt database [67]. These sequences were classified into different domain architectures (EPclA, EPclB, EPclC and EPclD, Figure 1) that were defined according to the occurrence and relative location of several conserved non-collagenous domains described in the InterPro [68] and Pfam databases [69] (Table 1). The probability of coiled-coil conformation and its oligomerization state in EPclP sequences were calculated with different prediction algorithms: PCoils [70], Marcoil [71], MultiCoil [72] and SCORER 2.0 [73]. Secondary structure predictions for EPclP sequences were obtained from the Jpred3 prediction server [74]. Default settings were used for all prediction algorithms.

Cloning of EPclA and EPclB Sequences from E. coli O157:H7

Recombinant EPclA and EPclB proteins were produced using laboratory E. coli strains. Two DNA fragments coding for EPclA and EPclB were amplified from a sample of genomic DNA of Escherichia coli O157:H7 strain RIMD 0509952 (Sakai), which was a gift from Charles W. Penn (School of Biosciences, University of Birmingham, UK). Forward and reverse primers were designed from the nucleotide sequences of the open reading frames ECs2717 (EPclA) and ECs1228 (EPclB) of the E. coli Sakai strain genome [5], with accession numbers Q7ACX5 and Q8XAX7 (Uniprot), or NP_310744 and NP_309255 (NCBI), respectively. Appropriate Nde I and Xho I restriction sites were incorporated into the primer designs, with the resulting oligonucleotide sequences 5′-CAT ATG ATG GCA GTA AAG ATT TCA GGT GTA CTG-3′ (EpclA forward), 5′-CTC GAG TTC TCC TGT TCT GCC TGT ATC ACT GCC-3′ (EpclA reverse), 5′-CAT ATG ATG ACG ATG GAT CCG GGG GAG TAT GCG-3′ (EPclB forward), and 5′-CTC GAG TCA TTC TCC TGT TCT GCC TGT ATC ACT -3′ (EPclB reverse). The ECs2717 sequence was chosen amongst the six EPclA open reading frames from the Sakai genome as it had a putative promoter sequence (TGTTATGAC) 38 nucleotides upstream of the predicted start of the coding sequence. The products of PCR amplification were ligated into the pET-28a(+) expression vector (Novagen), and the correct frame and ligation of the EPclA and EPclB clones were confirmed by sequencing. The recombinant EPclA construct (rEPclA) was designed with hexahistidine tags both at the N- and C-terminus, whereas the recombinant EPclB construct (rEPclB) was designed with only an N-terminal hexahistidine tag (Figure S4). The entire nucleotide sequence for the rEPclA fragment could be obtained from the sequencing data and it was shown to contain twelve changes with respect to the closest deposited sequence (ECs2717). These changes resulted in four changes in the amino acid sequence (Figure S5). Each of these amino acid changes is a common substitution in other EPclA sequences from the Sakai and other O157:H7 strains. Thus, these changes correspond to normal sequence variability amongst EPclA proteins and are not artefacts introduced during PCR amplification.

Cloning of PfN, PfN–PCoil, and PfC Fragments from EPclA

Separate recombinant constructs were prepared for three fragments of EPclA containing different predicted domains: PfN (residues 1-140), PfN–PCoil (residues 1-250) and PfC (residues 363-426) (Figure S4). All three fragments were amplified by PCR from the rEPclA clone using designed forward and reverse primers containing appropriate restriction sites: Nhe I and Xho I (PfN and PfN–PCoil) or BamH I and EcoR I (PfC). Sequences used for the different primers were 5′-CTC GTC GCT AGC ATG GCA GTA AAG ATT TCA-3′ (PfN and PfN–PCoil forward), 5′-CTC GTC CTC GAG TCA CTG ACT GGC TGA-3′ (PfN reverse), 5′-CTC GTC CTC GAG TCA CAC CAC GGT GGG-3′ (PfN–PCoil reverse), 5′-CTC GTC GGA TCC ATC CGT TTT CGT CTG GGGC-3′ (PfC forward), and 5′-CTC GTC GAA TTC CTA ATC CAG CCC CTT AAC ATC-3′ (PfC reverse). The products of PCR amplification were ligated into the pET-28a(+) expression vector (Novagen) and the correct frame and ligation of the clones were confirmed by sequencing. The PfC construct failed to produce any detectable expression and the PfC sequence was subsequently cloned into the fusion protein expression vector pHisTrx, an in house derivative of pET-32a (Novagen) encoding E. coli thioredoxin (trx) with an N-terminal hexahistidine tag and a thrombin cleavage site, followed by a unique multiple cloning site [75] (Figure S4).

Expression and Purification of Recombinant Proteins

Recombinant proteins and fragments were produced in E. coli BL21(DE3) cells (rEPclA and rEPclB) or JM109(DE3) cells (PfN–PCoil, PfN and Trx–PfC), using both IPTG induction (small-scale test expression and large-scale expression) and auto-induction (only large-scale expression) [76]. Best expression conditions for IPTG induction were achieved by inoculating 5 ml cultures with single bacterial colonies expressing the recombinant proteins, followed by incubation at 37°C overnight, and then inoculating 500 ml cultures in 2-litre flasks with 1% overnight pre-culture, followed by incubation at 37°C until log phase, and induction with 1 mM IPTG followed by further incubation for 4 hours at 30°C. Best expression using auto-induction was achieved by inoculation with 1% overnight pre-culture of 500 ml cultures supplemented with auto-induction solutions (1 M each of Na₂HPO₄, KH₂PO₄, NH₄Cl, Na₂SO₄ and MgSO₄, 20 mM CaCl₂, 50% glycerol, 1 M glucose and 20% lactose) followed by incubation overnight at 37°C. All recombinant products mainly localised to the soluble fraction, auto-induction expression being more effective in producing larger amounts of soluble protein. Recombinant proteins were purified by nickel-affinity chromatography (QIAGEN) using a 5 ml column and following the manufacturer protocols, followed by size-exclusion chromatography using a 120-ml HiLoad 16/60 Superdex 200 column (GE Healthcare), with 10 mM Tris, 150 mM NaCl, pH 7.4 as elution buffer. SDS-PAGE analysis showed that the bands corresponding to rEPclA and rEPclB migrated with apparent molecular weights of ∼66 kDa and ∼100 kDa, respectively (Figure S2 and data not shown), which are higher than their predicted molecular weights of 47 kDa and 66 kDa. Glycosylation was ruled out by a negative periodic acid-Schiff stain analysis (not shown), and thus the anomalous gel migration of purified rEPclA or rEPclB must relate to their collagen-like sequences, a behaviour commonly observed in collagens and collagen-like proteins. Overproduction of rEPclA by auto-induction produced significant amounts of two endogenous proteolytic fragments whose partial sequences were identified by peptide mass spectrometry (see below and Figure S2). A fragment containing only the Col and PfC domains from rEPclA (Col–PfC fragment, Figures S2 and S4) could be isolated in enough quantities from full-length rEPclA for subsequent biophysical characterisation. Conditions for consistent production of the two proteolytic fragments could not be established, and most samples of rEPclA produced by auto-induction and IPTG induction did show one major band in SDS-PAGE analyses of purified samples, corresponding to full length rEPclA (Figure S2A). In these cases, levels of the proteolytic fragments were too low for characterization and isolation.

Gel electrophoresis, Immunoblot Analysis and Protein Identification

Protein samples were denatured by heating at 90°C for 10 minutes and electrophoretically separated in 0.1% SDS, 4–12% gradient NuPAGE Bis-Tris polyacrylamide gels (Invitrogen). Proteins were visualised via Coomassie Brilliant Blue staining, and their electrophoretic migrations compared to those of prestained molecular weight markers (Precision Plus All Blue standards, BioRad). All proteins and fragments containing the Col or PCoil domains showed a slower than normal electrophoretic migration and higher than expected apparent molecular weights in the gels. Thus, identities of individual protein bands had to be confirmed by in-gel trypsin digestion followed by reverse phase chromatography of the tryptic peptides and sequence identification with mass spectrometry in the Biomolecular Analysis Facility of the Faculty of Life Sciences, University of Manchester (BAF-FLS). Bands of interest were excised from the gels, reduced with 10 mM dithiothreitol and alkylated with 55 mm iodoacetamide. Samples were digested overnight with trypsin at 37°C and then analysed using a CapLC (Waters) nanoLC system coupled to a Q-TOF Micro Mass Spectrometer (Waters). Peptides were separated by reverse phase chromatography using a 0.075×150 mm PepMap column (Dionex) and an acetonitrile gradient in 0.1% formic acid. Peptides eluting from the column were selected automatically for fragmentation. Data were searched against the UniProt database using the Mascot engine (Matrix Science). Further confirmation by immunoblotting was performed via transfer of SDS-PAGE separated proteins onto nitrocellulose membranes (Whatman) followed by incubation with a commercial anti-His₆ tag antibody. Horseradish peroxidase-coupled Fab-specific anti-mouse IgG (Sigma) was used as secondary antibody, and detection was achieved with SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific).

Protein Concentration

Protein concentration was determined by measuring UV absorption at 280 nm with a NanoDrop ND-1000 spectrophotometer (Labtech International), using molar extinction coefficients calculated from the amino acid sequences of the different recombinant proteins and fragments [77]. For recombinant fragments without tryptophan residues in their sequences, a corrected value of the molar extinction coefficient was derived from the UV absorption at 280 nm of equal concentration samples in aqueous buffers and in 8 M urea. In these cases the extinction coefficient calculated from sequence was considered accurate enough for determination of the protein concentration in 8 M urea [77], and an adjusted extinction coefficient was derived for the protein in water-based buffers.

Analytical Ultracentrifugation

Sedimentation equilibrium analysis of rEPclA was performed using an Optima XL-A ultracentrifuge (Beckman Instruments) from the BAF-FLS. The protein was in a 10 mM Tris, 150 mM NaCl, pH 7.4 buffer that was supplemented with increasing concentrations of guanidinium chloride (0–6 M). Sample volumes of 110 µl were used in six-sector Epon-filled centrepiece cells equipped with quartz windows. All experiments were conducted at 20°C and sedimentation equilibrium data were collected at 10,000, 18,000 and 28,000 rpm. Weight-averaged molar mass (M_w) was determined using Hetero analysis (developed by J. Cole and J. Lary at the University of Connecticut), using a single ideal species model. A value of  = 0.7139 was used for the partial specific volume of rEPclA, calculated from its amino acid sequence using Sednterp, version 1.09.

Size Exclusion Chromatography and Multiangle Laser-Light Scattering

Molecular weights of the recombinant proteins were determined by size exclusion chromatography coupled to multiangle laser-light scattering (SEC/MALLS) in the BAF-FLS. Depending on their size, proteins were chromatographed in Superdex 200 10/300 GL (10 to 600 kDa), Superose 6 10/300 GL (5 to 5000 kDa), or Superdex 75 5/150 GL (3 to 70 kDa) columns, all from GE Healthcare, using 10 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA as elution buffer. Elution from the column was continuously analysed in-line with a light scattering (LS) detector Dawn Heleos II (Wyatt Technology) and an Optilab rEX refractometer (Wyatt Technology). The LS intensity and eluant RI were analyzed using ASTRA software (version 5.21, Wyatt) to give a weight-averaged molar mass (M_w). Fractions of 0.5 ml were collected for further analysis.

Circular Dichroism (CD) Spectroscopy

Secondary structures and thermal denaturation of the recombinant proteins were analyzed by CD spectroscopy. Samples were equilibrated in 10 mM Tris, 150 mM NaCl, pH 7.4 and purified via size-exclusion chromatography before analysis. For the PfN–PCoil and PfN recombinant proteins a phosphate buffer with 20 mM Na₂HPO₄/NaH₂PO₄, 100 mM NaCl pH 7.4, was also used for CD analysis. CD spectra were recorded with a Jasco J-810 spectrometer equipped with a peltier temperature controller. Wavelength scans were performed using a 0.5 mm-path-length Starna quartz cell of acceptable birefringence for CD, and data were collected every 0.5 nm with a 1 nm bandwidth. Each spectrum was obtained from the accumulation of data from 10 scans; baseline was corrected using the spectrum of a Tris/NaCl buffer blank. Spectra were obtained at different temperatures, and thermal transition profiles were recorded between 4°C and 90°C at 220, 216 or 222 nm (depending on the protein being examined) with a data pitch of 0.5 nm, bandwidth of 1 nm, detector response time of 32 sec and temperature slope of 20°C/hr. Samples were cooled back to 4°C after the different transitions and final spectra were recorded at that temperature. Ellipticities in millidegrees were converted to mean residue molar ellipticities (degree cm² dmol^-1) by normalizing for the number of residues on each protein or fragment.

Rotary Shadowing Electron Microscopy

The structural organization of the different recombinant proteins was investigated under an electron microscope after rotary shadowing using the mica sandwich technique [78]. Five µl of sample (5 µg/ml) were adsorbed onto freshly cleaved mica. Another freshly cleaved mica disc was then gently placed on top, causing the drop to spread evenly between the two mica discs. The sample was allowed to adsorb for 1 min, after which it was washed with water and then freeze-dried at –80°C in a Cressington CFE-50 freeze-fracture apparatus (Cressington Scientific Instruments). The dried sample was maintained under vacuum and its temperature was dropped to –190°C. Rotary shadowing was then performed at an angle of 5° relative to the mica surface by electron-beam evaporation of platinum-carbon. The platinum film thickness in the plane of the substrate was measured by a quartz-crystal film-thickness monitor as 1.8 Å; this film was subsequently coated with a 7 nm carbon backing layer. Replicas were floated from mica on distilled water and placed on 400 mesh copper grids. Photomicrographs were taken with an FEI Tecnai Twin transmission electron microscope (FEI, Eindhoven, Netherlands) operated at 120 kV. Digital images (1024×1024 pixels) were recorded on a TVIPS F214 cooled CCD camera. Magnification was calibrated using a diffraction grating replica (2160 lines/min; Agar Scientific, Stansted, UK).