Molecular Insights on the Recognition of a Lactococcus lactis Cell Wall Pellicle by the Phage 1358 Receptor Binding Protein

ABSTRACT

The Gram-positive bacterium Lactococcus lactis is used for the production of cheeses and other fermented dairy products. Accidental infection of L. lactis cells by virulent lactococcal tailed phages is one of the major risks of fermentation failures in industrial dairy factories. Lactococcal phage 1358 possesses a host range limited to a few L. lactis strains and strong genomic similarities to Listeria phages. We report here the X-ray structures of phage 1358 receptor binding protein (RBP) in complex with monosaccharides. Each monomer of its trimeric RBP is formed of two domains: a “shoulder” domain linking the RBP to the rest of the phage and a jelly roll fold “head/host recognition” domain. This domain harbors a saccharide binding crevice located in the middle of a monomer. Crystal structures identified two sites at the RBP surface, ∼8 Å from each other, one accommodating a GlcNAc monosaccharide and the other accommodating a GlcNAc or a glucose 1-phosphate (Glc1P) monosaccharide. GlcNAc and GlcNAc1P are components of the polysaccharide pellicle that we identified at the cell surface of L. lactis SMQ-388, the host of phage 1358. We therefore modeled a galactofuranose (Galf) sugar bridging the two GlcNAc saccharides, suggesting that the trisaccharidic motif GlcNAc-Galf-GlcNAc (or Glc1P) might be common to receptors of genetically distinct lactococcal phages p2, TP091-1, and 1358. Strain specificity might therefore be elicited by steric clashes induced by the remaining components of the pellicle hexasaccharide. Taken together, these results provide a first insight into the molecular mechanism of host receptor recognition by lactococcal phages.

IMPORTANCE Siphophages infecting the Gram-positive bacterium Lactococcus lactis are sources of milk fermentation failures in the dairy industry. We report here the structure of the pellicle polysaccharide from L. lactis SMQ-388, the specific host strain of phage 1358. We determined the X-ray structures of the lytic lactococcal phage 1358 receptor binding protein (RBP) in complex with monosaccharides. The positions and nature of monosaccharides bound to the RBP are in agreement with the pellicle structure and suggest a general binding mode of lactococcal phages to their pellicle saccharidic receptor.

INTRODUCTION

Phages are now recognized as the most abundant biological entities on our planet, and they play major roles in the ecological balance of microbial life. Understanding the complex dynamics of phage-bacteria interactions is thus important for ecological systems and also for industrial applications. For example, cheese is manufactured through large-scale bacterial fermentation of milk, and this process is susceptible to phage infection (1). Lactococcus lactis is the most important bacterial species for the production of cheese, and, consequently, its phages have been intensively studied worldwide, mainly with the aim of developing phage-resistant strains.

Most phages described to date belong to the Caudovirales order as they possess a double-stranded DNA (dsDNA) genome packaged into a capsid connected to a tail (2). Tailed phages are generally classified into three families based on their tail features: members of the Myoviridae family have a contractile tail (3), Podoviridae members have a short residual tail (4), and the seemingly most abundant Siphoviridae phages have a long, noncontractile tail (5, 6). While useful for phage classification purposes, the tail is, above all, an essential structure for host recognition. The tip of the phage tail possesses a protein device, with complexities that will vary from one phage to another, that is used to recognize different types of bacterial host receptors, proteins and saccharides (7–11). Phages that recognize and infect L. lactis strains are largely classified within the Siphoviridae family (siphophages), but a few of them are in the Podoviridae family (12).

For practical applications, lactococcal phages were further classified into 10 genetically distinct groups (12). This grouping is based on electron microscopy and comparative genomics (13). Phages belonging to the same group have similar morphologies and share a high level of nucleotide identity, whereas members of distinct groups share limited, if any, DNA identity (12). International studies have identified members of three groups of lactococcal siphophages, namely, 936, c2, and P335, as mostly responsible for milk fermentation collapses. Phages can be differentiated through genomics and also by their host range, i.e., the specific L. lactis strains they infect.

In the past few years, the protein structures and infection patterns of members of the lactococcal 936 and P335 groups, namely, phages p2 (936) (14–17) as well as TP091-1 and Tuc2009 (P335) (6, 18–23), have been extensively studied. Structures have been solved for their receptor binding proteins (RBPs) (14, 15, 18, 19, 24), which are located within the baseplate (BP), a large protein complex and host recognition device. The large structure of the BP was also solved for a few lactococcal phages (16, 21, 23). Among others, lactococcal RBPs of phages from the 936 and P335 groups were found to harbor a saccharide-binding site at the C-terminal end that could be blocked by a camelid nanobody or a designed ankyrin repeat protein (DARPin), thereby preventing host recognition and infection (14, 15, 25–27) (19).

A novel cell wall polysaccharide (pellicle) located at the surface of L. lactis cells has been implicated as a receptor of 936-type phages. These surface polysaccharides were shown to be composed of repeating units of hexasaccharide phosphate that are distinct from other bacterial polysaccharides (11, 28). A comparative analysis of the gene cluster of the pellicle revealed diversity among L. lactis strains but also an apparent correlation between the lactococcal pellicle genotype of a given strain and the host range of tested 936-type phages (29). This correlation was recently confirmed through the characterization of the pellicle of L. lactis 3107, the host of phage TP091-1 (28).

In contrast to the three lactococcal phage groups mentioned above, the seven other groups are far less represented, and for some so far, they are represented by a unique member. It is unclear why these seven phage groups are not as predominant in industrial settings. In this context, we became interested in the virulent siphophage 1358 (the reference member of the lactococcal phage group that bears its name) (30). First, many genes of phage 1358 surprisingly share sequence similarities with some Listeria phages, e.g., P35 and P40 (31). It was even suggested that phage 1358 might originate from a Listeria phage which adapted to L. lactis (30). Thus, the study of phage 1358 may give insights into the requirements for host adaptation. Second, phage 1358 has a different host range than model lactococcal phages p2, Tuc2009, and TP091-1. Consequently, studying its receptor binding protein may shed further light on the specificities of host recognition.

We report here the X-ray structure of the RBP of phage 1358 in complex with monosaccharides found at the cell surface. We analyzed the pellicle of its host L. lactis SMQ-388 and determined that it possesses a phospho-hexasaccharidic repeat, differing from the L. lactis hosts of phage p2 (11) and TP091-1 (28). These data led us to model a trisaccharidic moiety, GlcNAc-galactofuranose (Galf)-GlcNAc/glucose 1-phosphate (Glc1P), that is surprisingly common to the receptors of phages p2, TP091-1, and 1358. We suggest that, in addition to recognition through a common central moiety, strain-specific recognition by lactococcal phages might be elicited by the remaining components of the pellicle hexasaccharide of their hosts.

MATERIALS AND METHODS

Phage and host.

Phage 1358 and its host L. lactis SMQ-388 (HER1205) were obtained from the Félix d'Hérelle Reference Center for Bacterial Viruses (www.phage.ulaval.ca). They were grown in GM17 (M17 broth supplemented with glucose) medium as described previously (30). For phage amplification, CaCl₂ was added to the medium at a final concentration of 10 mM.

RBP production and crystallization.

The orf20 of phage 1358, coding for the RBP (see Results section), was cloned into the Gateway destination vector pETG-20A for protein production in Escherichia coli BL21 and purified by Ni affinity and gel filtration chromatography according to standard procedures (32, 33). ORF20 was concentrated to 13 mg/ml and subjected to crystallization screening with a Mosquito Crystal (TTP LabTech) in Greiner Bio-One CrystalQuick plates. Crystals were obtained at 20°C by mixing 300 nl of protein (10 mM HEPES, pH 7.5, 150 mM NaCl) with 100 nl of precipitant solution. Two initials hits were obtained: one under condition 14 of the Wizard II screen (Emerald Biosystem) and one under condition 27 of the Structure screen II (Molecular Dimension). Conditions were optimized by varying pH and precipitant concentration. Condition 14 of Wizard screen II gave a crystal exploited for native data collection which was grown in 3 days in a solution containing 20% (wt/vol) polyethylene glycol 1000, 100 mM sodium/potassium phosphate buffer (pH 6.2), and 200 mM NaCl. To obtain phase information, native crystals were soaked in well solution supplemented with CsI. Crystals were cryoprotected with the mother liquor supplemented with 15% polypropylene glycol and immediately flash-frozen under a stream of nitrogen. Condition 27 from the Structure screen II gave crystals in a solution containing 10 mM zinc sulfate heptahydrate, 100 mM morpholineethanesulfonic acid (MES) buffer (pH 6.5), and 25% (vol/vol) polyethylene glycol monomethyl ether 550. Crystals were cryoprotected as indicated above. Crystals with GlcNAc and Glc1P were obtained by mixing the protein at 13 mg/ml in 10 mM HEPES, pH 7.5, and 150 mM NaCl with 100 mM ligand and then grown under condition 27 from the Structure screen II as described above.

X-ray crystallography.

Crystals were grown by hanging-drop vapor diffusion, and data sets were collected at the synchrotron Soleil (PROXIMA-1). Data were treated by XDS and XSCALE (34). The structure was determined by single-wavelength anomalous dispersion (SAD) at a wavelength of 1.5498 Å. Form 1 crystals belong to the cubic space group P2₁3 with unit cell dimensions of a = b = c = 107.9 Å and diffracted to a resolution of 1.75 Å for the CsI derivative (Table 1). There is one monomer in the asymmetric unit, with a Matthews coefficient (V_m) of 2.6 ų/Da and 53% solvent. Twelve initial derivative sites were located with the SHELXC/D programs (35). Using the program Phaser (36), the heavy atom model was then completed up to 50 sites, and phases were refined before being improved by density modification using Parrot (37). The first steps of model building were performed automatically using Buccaneer (38) and completed manually with Coot (39). Refinement was performed with AutoBUSTER (40). Form 2 crystals belong also to the cubic space group P2₁3 with unit cell dimensions of a = b = c = 165.6 to 166.2 Å and a V_m of 4.7 with 74% solvent for two monomers in the asymmetric unit. Each of the two monomers in the asymmetric unit is part of a trimer reconstituted by the 3-fold axis. The structures of the complexes with the three saccharides were solved by molecular replacement using the form 1 structure with Molrep (41) and refined with AutoBUSTER (Table 1).

TABLE 1.

Data collection and refinement statistics of the phage 1358 RBP crystals

Parameter	Value for the ORF20 crystal
	CsI (form 1)	Glycerol	GlcNAc	GlcP
Data collection
PDB code	4L9B	4L99	4L92	4L97
Source	Soleil PX 1	Soleil PX 1	Soleil PX 1	ESRF-1D29
Space group	P213	P213	P213	P213
Unit cell dimensions, a = b = c (Å)	107.6	166.2	165.6	165.9
Resolution limits (Å)a	50.0–1.75 (1.86–1.75)	50.0–2.20 (2.33–2.20)	50–2.10 (2.23–2.10)	50–2.61 (2.68–2.61)
Rmerge (%)a	4.6 (83.0)	15.4 (79)	9.9 (70)	10.6 (70)
No. of observationsa	723,892 (60,665)	792,229 (77,486)	891,068 (142,602)	243,747 (18,129)
No. of unique reflectionsa	81,007 (12,967)	127,119 (12,354)	88,185 (14,123)	45,762 (3,338)
Avg I/SD(I)a	24.3 (2.0)	9.3 (2.15)	14.8 (2.3)	10.6 (2.4)
Completeness (%)a	99.7 (98.2)	99.9 (99.6)	99.9 (99.7)	99.4 (98.6)
Multiplicitya	9 (4.7)	6.2 (6.5)	10 (10)	5.3 (5.0)
Refinement
Resolution (Å)a	25.2–1.75 (1.79–1.75)	32.21–2.20 (2.26–2.20)	30.75–2.10 (2.15–2.10)	47.79–2.61 (2.68–2.61)
No. of reflectionsa	41,981 (2,868)	77,461 (5,644)	88,079 (6,486)	45,761 (3,252)
No. of atoms
Protein	5,828	6,088	6,088	6,068
Water	937	814	841	472
Ligand		12	51	32
No. of test set reflections	2,100	3,873	4,404	2,289
Rwork/Rfree (%)a	17.0/20.4 (22.1/23.9)	18.8/20.3 (19.7/21.9)	17.6/19.3 (20.2/20.2)	17.9/21.1 (20.6/22.9)
RMSD
Bond length (Å)	0.01	0.009	0.01	0.01
Bond angle (°)	1.10	1.12	1.12	1.17
B-Wilson/B-avg	32.0/43.0	51.9/56.5	42.1/49.2	63.1/57.3
Ligand S1/S3			42.6/49.5	64.5
Ramachandran plot (%)
Preferred regions	96.3	96.9	96.5	96.5
Allowed regions	3.7	2.4	2.9	2.9
Outlier regions	0	0.7	0.6	0.6

^aNumbers in parentheses refer to the highest-resolution shell.

Pellicle polysaccharide analysis. (i) Cell growth and PSP preparation.

The polysaccharide pellicle (PSP) of L. lactis SMQ-388 was isolated essentially as described earlier (11). Cells were grown in M17 medium (4 liters), supplemented with 0.5% glucose (Glc). Cells were collected and washed with deionized water four times. Washed cells (∼4 g) were suspended in 5% trichloroacetic acid (TCA; 60 ml) and stirred for 48 h at 5°C. The suspension was centrifuged, and the supernatant was dialyzed and lyophilized to give a crude TCA extract (87 mg). Then, 13 mg of crude preparation was fractionated on a Q-Sepharose ion exchange column, eluted with a 0 to 0.5 M NaCl gradient. The main fraction, eluted at ∼0.1 M NaCl, was desalted on a G-50 column and further purified by ion exchange chromatography on a HiTrap Q column.

(ii) Chromatographic methods.

Ion exchange chromatography was performed on a Q-Sepharose Fast Flow column (1 by 10 cm) and HiTrap Q column (GE Healthcare). The column was washed with water for 10 min and then eluted with a linear gradient of 0 to 1 M NaCl over 60 min, with a flow rate of 3 ml min⁻¹ and UV detection at 220 nm. Gel filtration chromatography was performed on a Sephadex G-50 column (1 by 40 cm), eluted with 0.01% acetic acid (AcOH). Aliquots of each fraction were assayed for neutral sugars and, if necessary, for amino sugars. Gas chromatography-mass spectrometry (GC-MS) was performed on a Varian Saturn 2000 ion trap instrument, equipped with a capillary column DB-17, using a temperature gradient of 160 to 260°C at 4°C min⁻¹.

(iii) NMR spectroscopy analysis.

Nuclear magnetic resonance (NMR) experiments were carried out on a Varian INOVA 500 MHz (¹H) spectrometer with a 3-mm gradient probe at 25°C with an acetone internal reference (2.225 ppm for ¹H and 31.45 ppm for ¹³C) using standard pulse sequences for gradient correlation spectroscopy (gCOSY), total correlation spectroscopy (TOCSY; mixing time, 120 ms), rotating-frame nuclear Overhauser effect spectroscopy (ROESY; mixing time 500 ms), gradient heteronuclear single-quantum coherence (gHSQC), and gradient heteronuclear multiple-bond correlation (gHMBC; 100-ms long-range transfer delay). Acquisition time (AQ) was kept at 0.8 to 1 s for H-H correlations and at 0.25 s for HSQC; 256 increments were acquired for T1 relaxation time (512 for gCOSY). Assignment of spectra was performed using the Top Spin 2 (Bruker Biospin) program for spectrum visualization and overlap.

Molecular modeling.

The trisaccharide was modeled by extracting the corresponding coordinates from high-resolution structures of protein-saccharide complexes in the Protein Data Bank (PDB). A unique Galf structure was found from the entry 2VK2. The residues were linked in silico according to standard bond lengths and angles, and the resulting structure was docked into the p2 RBP glycerol-binding groove. The resulting complexes were idealized using REFMAC (42) to obtain a reasonable and clashless model, and the best model in terms of favorable contacts was selected.

Protein structure accession numbers.

X-ray structures and structure factors have been deposited in the Protein Data Bank under PDB codes 4L9B (CsI derivative), 4L99 (glycerol complex), 4L92 (GlcNAc complex), and 4L97 (Glc1P complex).

RESULTS

Crystal structure of ORF20, the RBP of phage 1358.

The published genome and deduced proteome of phage 1358 (30) were reanalyzed using the secondary-structure software retrieval program HHpred (43). Analysis of phage 1358 ORF20 (393 amino acids) revealed a striking similarity between its first ∼170 residues and the N-terminal region of the RBP of the lactococcal p2 (936 group) (14–17). The rest of ORF20 (∼200 residues) did not match with phage proteins or domains. As the C-terminal region of other lactococcal RBPs have been shown to be involved in host recognition (15), this may also be the case for ORF20 of phage 1358, and it would match its unique host range (30).

In order to confirm the RBP of phage 1358, we cloned and expressed ORF20 (393 residues) using standard procedures (33). Using multiangle-laser light scattering (MALS)/size exclusion chromatography (SEC)/UV absorption, ORF20 was found to be trimeric, consistent with other lactococcal RBPs (14, 18). Then, two forms of crystals were analyzed. For the form 1, we collected a data set of CsI-derived crystals at the PX1 beamline (Soleil, Saint Aubin, France) and solved the structure with SAD methods. The structure was further refined to 1.75-Å resolution (Table 1). The structure of form 2 was solved by molecular replacement and then refined (Table 1). Each of the two monomers in the asymmetric unit is part of a trimer reconstituted by the 3-fold axis. Although the diffraction extended to lower resolution, the map was of excellent quality, and the chain could be traced from residues 2 to 393 (the last one). A glycerol molecule (from the cryoprotectant solution) could be located in the structure of form 2, a feature also fortuitously observed in the crystal structures of RBP of lactococcal phages p2 and TP091-1 (14, 18).

In the trimeric ORF20, each monomer is formed of two domains (Fig. 1A). The N-terminal domain (residues 17 to 173, for a total of 156 residues) forms a β-sandwich of two β-sheets (Fig. 1B). A Dali search with this domain retrieved the N-terminal domain of phage p2 RBP as the best hit (14) (Fig. 2), with a Z-value of 12.7 and a root mean square deviation (RMSD) of 2.5 Å calculated with 125 out of 156 residues, as predicted by HHpred. This region in phage p2 was previously referred to as the “shoulder” domain, which is linked to other proteins in the phage structure (14). After an α-helix (residues 17 to 29), β-sheet 1 comprises four β-strands (residues 30 to 40, 47 to 57, 60 to 67, and 73 to 78) and the residues 2 to 7 swapped from another monomer. β-Sheet 2 also comprises four β-strands (residues 84 to 94, 115 to 122, 145 to 156, and 161 to 169). An α-helix (residues 123 to 134) and a β-hairpin (β-strands 94 to 104 and 108 to 114) are plugged and decorate the β-sandwich structure (Fig. 1B).

FIG 1.

X-ray structure of the phage 1358 RBP (ORF20). (A) Ribbon view of the overall structure of the RBP trimer. Two monomers are displayed in gray; the third one is shown in rainbow mode. (B) Ribbon view of the N terminus segment (residues 1 to 16) and the shoulder domain (residues 17 to 173) colored in rainbow mode. (C) Stereo ribbon view of the head domain (residues 176 to 393) colored in rainbow mode. Nt, N terminus; Ct, C terminus.

FIG 2.

Comparison of the X-ray structures of the phage 1358 RBP with those of phages p2 and TP091-1. (A) Ribbon view of the phage 1358 RBP trimer (crystal form 2). (B) Ribbon view of the phage p2 RBP trimer. (C) Ribbon view of the phage TP091-1 RBP trimer. Figures are colored in rainbow mode.

The second domain of ORF20 of phage 1358 spans residues 174 to 393 (Fig. 1C) and contains a glycerol-binding site. This C-terminal region was previously referred to as the “head” domain in the RBP of phage p2, which is involved in host recognition (14). A Dali search returned structurally similar proteins, the tumor necrosis factor (TNF) domain of ectodysplasin A (PDB 1RJ7) (Z-value of 11.3; RMSD of 2.4 Å/130 residues), a molecule involved in cell-cell recognition and contact during embryogenesis. The next hit was TNF (Z = 10.6; RMSD = 2.9 Å), a cytokine involved in a wealth of cellular regulation processes. Both proteins are β-sandwiches, and their biological assembly is trimeric. The C-terminal domain of ORF20 is essentially β-stranded, in addition to having a small C-terminal α-helix (residues 384 to 392), and forms an overall β-sandwich structure of two β-sheets composed of 4 and 5 β-stands each (Fig. 1C). The β-sandwich comprises β-strands 5, 1, 13, 8, and 11 (β-sheet 1) and 2, 12, 9, and 10 (β-sheet 2). This β-sandwich core is decorated by β-hairpins (β-strands 3 and 4 and β-strands 6 and 7), and a short two-turn helix completes the domain at its C terminus (Fig. 1C).

The trimerization of ORF20 involves a large buried surface area of 2,300 Ų per monomer, on a total surface of 19,000 Ų. The two domains of ORF20 participate in the trimerization. Taken together, ORF20 of phage 1358 has the structural hallmarks of a lactococcal phage RBP.

The saccharide binding site of the RBP head domain.

As indicated above, a well-defined electron density accounting for a bound glycerol molecule was found in the crystal of form 2, which was cryo-frozen with glycerol. This site is located between a two-turn helix (residues 237 to 244) and β-sheet stretches 287 to 289 and 340 to 343. The glycerol is stacked against the aromatic ring of Phe 240 and establishes hydrogen bonds with Arg 237 and Asn 289. However, in contrast with the RBPs of lactococcal phages p2 and TP091-1 (14, 18), the glycerol binding site of ORF20 is located in the middle of a monomer and not in a crevice between two monomers.

We then further explored the binding capacity of the glycerol binding site. We cocrystallized ORF20 of phage 1358 with a series of monosaccharides, which are constituents of the strain-specific polysaccharide pellicle at the surface of L. lactis cells and also the phage receptor of several lactococcal phages, namely, β-GlcNAc, β-Galp, α-Rha, and α-Glc1P (11, 44). The electron density map of the cocrystallized ORF20 trimer with α-Rha and β-Galp revealed featureless small bulbs of electron density at the glycerol position. In contrast, an excellent density map was observed for β-GlcNAc at the expected site, allowing precise positioning of the saccharide (Fig. 3). Surprisingly, another molecule of β-GlcNAc, ∼8 Å away from the first one, could be modeled with good precision (Fig. 3 and 4A and B) in what we called the S3 site since a potential unoccupied S2 site was located between the S1 and S3 sites. The electron density map of ORF20 crystals soaked with Glc1P showed also the presence of a monomer bound in the same site as for the first β-GlcNAc molecule (Fig. 3), allowing us to position it in the RBP structure (Fig. 4C). Furthermore, the ring structures of β-GlcNAc and Glc1P superimposed well, especially the C-3, C-4, and C-5 atoms, while the upper part of the sugars differ by ∼1.0 Å due to a 20° rotation around the C-4 atom (Fig. 4D).

FIG 3.

Sigma A-weighted 2F_o-F_c (where F_o and F_c are observed and calculated structure factor amplitudes, respectively) electron density maps of the GlcNAc and Glc1P saccharides in the 1358 RBP. The first GlcNAc molecule, in the S1 binding site (A), and the second GlcNAc molecule, in the S3 binding site (B) are shown (2.1-Å resolution, 1σ cutoff). (C) The first Glc1P molecule, in the S1 binding site (2.61-Å resolution, 1σ cutoff).

FIG 4.

The saccharide binding site of phage 1358 RBP.(A) Surface representation of the phage 1358 RBP trimer/GlcNAc complex, with the two GlcNAc molecules bound in the saccharide binding site (sphere representation). (B) Close-up of the two GlcNAc molecules bound in the saccharide binding site (stick representation). Three saccharide binding sites are identified (S1 to S3); a water molecule bridges the two GlcNAc molecules. (C) Close-up of the Glc1P bound in the saccharide binding site (stick representation). (D) Close-up of the superimposed Glc1P and GlcNAc saccharides in the S1 site (stick representation). The sugar residues are shown in stick representation (white, C; red, O; blue, N; orange, P), and the Glc1P carbon atoms in panel D are shown in light-blue.

A detailed analysis of monosaccharide binding reveals important features of the interactions. In the primary binding site (site S1), shared by GlcNAc and Glc1P, both sugars are stacked against an aromatic residue, Phe 240, a hallmark of saccharide binding sites (45, 46) (Table 2 and Fig. 5A and B). They establish similar hydrogen bonds using their OH₃ and OH₄ hydroxyl groups, interacting with Arg 237 and Asn 341, respectively. GlcNAc S1 establishes an extra hydrogen bond between its NH group (from the N-acetyl moiety) and Ser 291 (Fig. 5A), while the Glc1P OH₆ hydroxyl group is hydrogen bound to Asn 244 (Table 2; Fig. 5B). The latter Glc1P interaction is not observed in the GlcNAc structure because the OH₆ hydroxyl group has rotated by 180°, establishing a hydrogen bond with a water molecule bridging the S1 and S3 GlcNAc molecules (Fig. 5A and C). We suggest that GlcNAc binding to Ser 291 and Glc1P interaction with Asn 244 may explain the 20° rotation of GlcNAc compared to Glc1P.

TABLE 2.

Interactions between bound saccharides and the phage 1358 RBP

Site and residue	GlcNAc			Glc1P
	BSA Å2 (% of ASA)a	Hydrogen bond(s)	Hydrogen bond length(s) (Å)	BSA Å2 (% of ASA)a	Hydrogen bond(s)	Hydrogen bond length(s) (Å)
Site S1
Total	227 (62)			213 (62)
Arg 237	19 (24)	NH1,2—HO3	2.89, 2.9	14 (17)	NH1,2—HO3	2.84, 3.12
Phe 240	36 (50)			36 (52)
Asn 244	6.4 (75)			8.5 (88)	C=O—HO6	3.0
Asn 289	25 (41)			32 (46)
Ser 291	26 (25)	OH—HNAc	3.17	15 (15)
Ala 292	17 (52)			9 (28)
Asn 341	9 (95)	C=O—HO4	2.91	8 (100)	C=O—HO4	2.92
Phe 343	17 (78 ± 16)			17 (57)
HOH 682		HO6	2.55
Site S3	GlcNAc
Total	202 (55)
Val 200	27 (37)
Gln 201	11 (60)
Ser 202	22 (74)	NH—HO3	3.0
Lys 239	17 (10)
Phe 240	24 (35)
Asp 243	25 (80)	C=O—HO6	2.66
		C=O—HO4	3.0
Asn 244	2 (20)
Gln 345	13 (48)	NHε—O=C7	2.8
HOH 682		HO1	2.69

^aBSA, buried surface area; ASA, accessible surface area.

FIG 5.

Detail of the interactions of GlcNAc and Glc1P in the receptor binding site of phage 1358 RBP. (A) GlcNAc in the S1 site. (B) Glc1P in the S1 site. (C) GlcNAc in the S3 site. (D) Details of the S1 site interlaced hydrogen/ionic bonds. The protein side chains and the sugar residues are shown in stick representation (white, C; red, O; blue, N; orange, P). Hydrogen bonds are visualized as thin blue lines. The structure of the protein backbone is visualized as a green ribbon. Distances are in Å.

The GlcNAc saccharide in the S3 site is less efficiently bound than the one in the S1 site, as suggested by its smaller buried surface area (Table 2). In contrast with the former GlcNAc, no stacking is observed with aromatic residues. Three hydrogen bonds are observed between this GlcNAc and the RBP: Asp 243 establishes forked hydrogen bonds with the OH₄ and the OH₆ hydroxyl groups, Ser 202 hydroxyl establishes a hydrogen bond with GlcNAc OH₃, and the carbonyl group from the N-acetyl moiety is hydrogen bound to the NH₂ group of Gln 345 (Table 2; Fig. 5C). As mentioned above, a bridging water is observed, bound to the OH₁ hydroxyl of GlcNAc S3.

The S1 and S3 sites therefore differ significantly as the former one seems more adapted to strong monosaccharide binding. The stronger binding and the specificity of the S1 site are encoded in its rigid structure. The two dominating interactions involve Arg 237 and Asn 341, two residues involved, together with Asp 235, in an intricate and strong network of hydrogen/ionic bonds (Fig. 5D).

Characterization of the pellicle of L. lactis SMQ-388.

The polysaccharide pellicle (PSP) was isolated and purified from cells of L. lactis SMQ-388, the host of phage 1358, as described earlier (11). The PSP was found to contain Glc, Gal, and GalNAc in an approximate molar ratio of 1:2:3, which were shown to be in d-configuration.

HSQC spectra of the polysaccharide (Fig. 6) indicated six anomeric signals. Detailed chemical structure of this compound was elucidated by NMR spectroscopy sets of two-dimensional (2D) NMR spectra (gCOSY, TOCSY, NOESY, ¹H-¹³C gHSQC, ¹H-¹³C gHMBC, and ¹H-³¹P HMQC) that were recorded and interpreted. Monosaccharides in the pyranose form (Glc and GlcNAc) were identified by COSY, TOCSY, and NOESY cross-peak patterns and ¹³C NMR chemical shifts. Amino group location was concluded from the high field signal position of aminated carbons (CH signals at 45 to 60 ppm). β-Galf was identified by ¹³C chemical shifts in comparison with published values (28, 47, 48), based on the known monosaccharide composition. Complete assignment of ¹H and ¹³C spectra showed two α-GlcNAc saccharides (A and E), one β-GlcNAc saccharide (F), two β-Galf saccharides (B and C), and one α-Glc saccharide (D) (Table 3). Connections between monosaccharides were determined from transglycosidic NOE and HMBC correlations: B1:A6, C1:F3, D1:F6, E1:C3, and F1:B2.

FIG 6.

HSQC spectrum of the polysaccharide. Minor signals marked with a prime (′) belong to partly degraded PSP (A-P linkage hydrolyzed).

TABLE 3.

¹H and ¹³C chemical shifts of Lactococcus lactis SMQ-388 polysaccharide

Sugar	Atom	Chemical shift(s) (ppm) at position:
		1	2	3	4	5	6
α-GlcNAc A	H	5.47	3.97	3.81	3.63	3.99	3.82, 3.99
	C	95.1	54.9	71.8	70.7	72.9	67.3
β-Galf B	H	5.15	4.20	4.08	4.00	3.77	3.64, 3.70
	C	107.7	80.8	77.0	83.7	71.8	63.9
β-Galf C	H	5.08	4.26	4.00	4.22	3.75	3.65, 3.65
	C	110.1	90.0	85.3	82.8	71.9	63.9
α-Glc D	H	5.00	3.57	3.76	3.44	3.73	3.77, 3.87
	C	99.1	72.7	74.4	70.7	73.1	61.8
α-GlcNAc E	H	4.96	3.95	3.75	3.55	3.90	4.13
	C	99.2	54.9	72.0	70.9	72.6	65.9
β-GlcNAc F	H	4.81	3.80	3.75	3.70	3.71	3.81, 4.05
	C	101.4	56.3	82.3	69.4	75.4	66.4

A phosphodiester bond between A1 and E6 was confirmed by ¹H-³¹P correlation of A H-1 and E H-6 to the same ³¹P signal at 1.5 ppm. H-1 signal of GlcNAc saccharide A had additional coupling of 8 Hz to phosphorous; C-1 of saccharide A was at 95.1 ppm, as expected for the glycosyl phosphate. C-6 and H-6 signals of GlcNAc molecule E were shifted to the low field (Table 3), compared to the nonphosphorylated signals appearing at ∼3.8 ppm (H) and 61.5 ppm (C). Overall, NMR analysis revealed that the polysaccharide was composed of hexasaccharide repeating units, linked via phosphodiester bonds (Fig. 7).

FIG 7.

Structure of the phospho-hexasaccharide repeating motif of the pellicle of the L. lactis strains MG1363, 3107, and SMQ-388. The three saccharides forming the core are boxed.

Modeling of a pellicle trisaccharide.

The observation that two β-GlcNAc molecules, located ∼8 Å away from each other, bind to the ORF20 of phage 1358 as well as the composition of the hexasaccharide repeating units of L. lactis SMQ-388 pellicle prompted us to generate a model of the trisaccharide GlcNAc-Galf-GlcNAc. We extracted the β-Galf saccharide structure from PDB entry 2VK2 (structure of a galactofuranose binding protein) (49), and we could readily position it in the S2 site, between the β-GlcNAc A (S1 site) and the β-GlcNAc B (S3 site) molecules, with the proper linkages (Fig. 7A). Refinement of the trisaccharide keeping the crystallographic restraint with a low occupancy for β-Galf atoms led to an optimized structure while keeping close positions for both GlcNAc molecules compared with the crystallographic ones (Fig. 8A and B). It is worth mentioning that all efforts to bridge the two crystallographically positioned GlcNAc molecules with a pyranoside sugar were unsuccessful. The same modeling operation was performed with β-GlcNAc B (S3 site) and Glc1P (S1 site) for comparison purposes.

FIG 8.

Model of a trisaccharidic pellicle core bound in the 1358 phage receptor binding site. (A) Structure of the 1358 RBP trimer (beige surface) in complex with the modeled GlcNAc-Galf-GlcNAc trisaccharide (sphere representation). (B) Structure of the refined trisaccharide GlcNAc-Galf-GlcNAc model within the RBP binding site. The three saccharide binding subsites are labeled S1 to S3. The red arrows indicate the path of the next pellicle motif or the path of the remaining saccharidic units. (C) Structure of the refined trisaccharide GlcNAc-Galf-Glc1P model within the RBP binding site.

Of note, the hydroxyl group of GlcNAc S1 or the phosphate moiety of Glc1P points toward the solvent, making it possible for the rest of the poly-phospho-saccharidic chain to deploy freely (Fig. 7B and C). At the other end of the trisaccharide, the hydroxyl group at position C3 points in the direction of a small pocket available to accommodate the second Galf saccharide from the pellicle motif of L. lactis strain SMQ-388 (Fig. 8B and C). Only small differences occur, especially concerning the orientation of GlcNAc B, probably due to the slightly different orientations observed between GlcNAc and Glc1P at position S1. However, both trisaccharides seem compatible with RBP binding, within the limits of this modeling.

DISCUSSION

The structure of ORF20 of siphophage 1358 (1358 group) displays a few conserved features with the RBP of siphophage p2 (936 group). The N-terminal domain (residues 17 to 173) of ORF20 is structurally related to the shoulder domain of p2 RBP as the domains share the same β-sandwich structure as well as an N-terminal helix that establishes most of the contacts of this domain within the trimer (Fig. 2). The core of the large C-terminal β-sandwich domain (∼200 residues) of ORF20 shares a jelly roll fold with p2 and TP091-1 RBP (Fig. 2). Some differences are, however, noticeable. Large decorations are inserted in the shoulder domain of 1358 RBP, the triple interlaced β-helix forming the shoulder domain is absent, and the C-terminal head domain is twice as large, supporting the distinct host range of this phage. The structural RBP domains of lactococcal phages seem to be assembled in different ways in each phage group as well as within each group. This structural flexibility strongly suggests that they are remarkably adapted to rapidly evolve in response to the common dairy industrial practice of rotating different L. lactis strains.

The receptor binding site of phage 1358 is also different from that of phage p2 or phage TP091-1. It is located in the middle of the RBP head domains, in a cavity formed by a helical decoration plugged onto the jelly roll fold core, while it is located in a deep crevice formed between two RBP head domains in phages p2 and TP091-1 (P335 group) (Fig. 3). On the other hand, most saccharide binding side chains have conserved features, such as the presence of an aromatic ring on which the hydrophobic face of the sugar can stack as well as the possibility to establish several hydrogen bonds with polar or semipolar residues. All of these characteristics are found in the S1 site but not in the S3 site, which lacks a stacking aromatic side chain.

We analyzed the structure of the polysaccharide pellicle from the L. lactis SMQ-388 strain in order to compare it to the two other pellicles identified to date from L. lactis strains MG1363 (11) and 3107 (host of phage TP091-1) (28). The L. lactis strain SMQ-388 pellicle monomer possesses a trisaccharidic motif (Fig. 7, top, A, B, and F) identical to that of L. lactis strain 3107 (Fig. 7, middle). This trisaccharidic core motif is only partially conserved in the MG1363 strain pellicle as GlcNAc-1P (Fig. 7, bottom, A) is replaced by a Glc1P.

Flanking the core motif, the L. lactis SMQ-388 pellicle fourth position (Fig. 7, sugar C) is identical to that of strain 3107 but differs from that of strain MG1363, where a Rha pyrano-saccharide is found. In contrast, the fifth position (Fig. 7, sugar E) is a GlcNAc in MG1363 and a Glc in strain 3107. Finally, the sixth position (Fig. 7, sugar D) is a Glc for strains SMQ-388 and MG1363, while the pellicle of strain 3107 is not substituted.

Our modeling experiments showed that a Galf saccharide at position 2 of the core motif could be readily linked to the two GlcNAc molecules observed in our structure, with this model accounting for the core trisaccharide of the pellicle motif. As far as our modeling shows, both trisaccharides (with GlcNAc or Glc1P at position S1) are able to fit in the RBP receptor crevice. We also noticed that OH1 of GlcNAc S1 (or Glc1P) is orientated toward the bulk solvent, making it possible for the rest of the pellicle polymer to deploy freely. On the other side, position OH3 from GlcNAc S3 (Fig. 8B, the red arrow at position S3) is only compatible with binding to a small furanosic ring, thus excluding a good affinity for a Rha-substituted pellicle, like that of MG1363. The pellicle of L. lactis strain 3107 (28) also has a Galf attached at position 3 of GlcNAc S3, and consequently specificity differences between phages 1358 and TP091-1 toward their strains should be explained by other substitutions, i.e., those of saccharides E or F (Fig. 7).

Taken together, these results suggest that the lactococcal pellicles recognized by phages p2, TP091-1, and 1358 possess a semiconserved core trisaccharidic motif. This motif might promote initial binding, while different substitutions of the rest of the pellicle oligosaccharide may provide a total or partial phage-strain specificity based on exclusion by steric clashes.