Long tail fibres of the novel broad-host-range T-even bacteriophage S16 specifically recognize Salmonella OmpC

Authors


For correspondence. E-mail jklumpp@ethz.ch; Tel. (+41) 446325378; Fax (+41) 446321266.

Summary

We report isolation and characterization of the novel T4-like Salmonella bacteriophage vB_SenM-S16. S16 features a T-even morphology and a highly modified 160 kbp dsDNA genome with 36.9 mol % G+C, containing 269 putative coding sequences and three tRNA genes. S16 is a virulent phage, and exhibits a maximally broad host range within the genus Salmonella, but does not infect other bacteria. Synthesis of functional S16 full-length long tail fibre (LTF) in Escherichia coli was possible by coexpression of gp37 and gp38. Surface plasmon resonance analysis revealed nanomolar equilibrium affinity of the LTF to its receptor on Salmonella cells. We show that OmpC serves as primary binding ligand, and that S16 adsorption can be transferred to E. coli by substitution of ompC with the Salmonella homologue. S16 also infects ‘rough’ Salmonella strains which are defective in lipopolysaccharide synthesis and/or its carbohydrate substitution, indicating that this interaction does not require an intact LPS structure. Altogether, its virulent nature, broad host range and apparent lack of host DNA transduction render S16 highly suitable for biocontrol of Salmonella in foods and animal production. The S16 LTF represents a highly specific affinity reagent useful for cell decoration and labelling, as well as bacterial immobilization and separation.

Introduction

Members of Salmonella enterica ssp. enterica are the primary causative agents of human salmonellosis, one of the leading food-borne illnesses worldwide. This subspecies is highly diverse, with more than 2500 recognized serovars (Grimont and Weill, 2007). In 2009, the European Food Safety Agency (EFSA) reported 108 614 cases (23.7 per 100 000), with a mortality rate of 0.08% (EFSA, 2011). In the same year, a total of 40 828 laboratory-confirmed cases of salmonellosis were reported in the USA (CDC, 2011a). Still, this represents only a fraction of the estimated annual number of cases, which may be as high as 1.2 million, with more than 23 000 hospitalizations and 450 deaths (CDC, 2011b). Although usually self-limiting and with case numbers slightly declining over the past years, the impact of this food-borne illness remains significant.

The emergence of antibiotic-resistant pathogens creates a need for alternative antimicrobial treatments, which should also be suitable for veterinary applications. Bacteriophages represent a natural and safe treatment option, and are ideally suited to control pathogens in the food chain (Hanlon, 2007). However, every researcher familiar with Salmonella soon realizes that, due to the large number of different serovars and strains, biocontrol of S. enterica ssp. enterica with phages is a challenging task. Broad-host-range phages should not be limited to certain serovars only, i.e. their receptors should be more universally distributed within the genus. Many recent studies illustrate the usefulness of Salmonella-specific bacteriophages in diverse applications (Atterbury et al., 2007; Borie et al., 2008; Wall et al., 2010; Hooton et al., 2011; Lim et al., 2011; Guenther et al., 2012). Many of these studies employed the well-studied broad-host range Salmonella phage Felix-O1 (Felix and Callow, 1943; Whichard et al., 2010).

The recognition of potential host cells via the highly specific tail fibres represents the crucial initial step of phage infection. The long tail fibres (LTF) of T4-like phages mediate the first, yet reversible binding to the host cell surface. This interaction is more selective in nature than the subsequent binding of short tail fibres, which in the case of T4 irreversibly associate with the highly conserved inner core structure of the LPS molecule. This event triggers sheath contraction and DNA transfer from phage to the host cell. In T4, gp34 to gp37 constitute the LTF from proximal to distal segments. This hetero-oligomeric structure has been shown to occur and be active in a trimeric state (Cerritelli et al., 1996). It has been proposed that the gp37 homologues of T4-like phages actually occur as trimers in their functional conformation (Hashemolhosseini et al., 1996). Additional accessory proteins are required for full function of the LTF. However, gp38 apparently plays different roles in T4 and T2. In the former, gp38 assumes the function of a chaperone required for trimerization of gp37, is itself not present in the mature LTF, and the C-terminal portion of gp37 mediates binding to the phage receptor (Tetart et al., 1996). The situation in T2-like phages is different. An unrelated gp38 product attaches to the tip of the folded gp37 and acts as the actual adhesin (Riede et al., 1985, 1987). Moreover, a chaperone gp57A has also been reported to be required for trimerization of all segments of the LTF (Dickson, 1973).

We report isolation, sequencing and characterization of a novel Salmonella bacteriophage, a T4-like Myovirus termed vB_SenM-S16. To our knowledge, S16 represents the first T-even-like phage which is specific for Salmonella. The host range of the virulent S16 phage is extremely broad; it can infect a larger variety of Salmonella strains than Felix-O1, which has long been used for phage-based identification by plaque formation. Apart from its T4-like properties, we demonstrate that the distal segment of the mature S16 LTF is composed of both gp37 and gp38, similar to phage T2. Although the S16 LTF uses OmpC as primary adsorption ligand, deletion of the LPS shows an additive effect on phage adsorption and efficiency of plating (e.o.p.) in a ΔompC background. The overall properties of S16 render it an ideal candidate for biocontrol of Salmonella in foods and animal production.

Results

Phage S16 is a T4-like myovirus

Transmission electron microscopy confirmed that phage S16 belongs to the order Caudovirales. Its contractile tail is the defining morphological feature of the Myoviridae family (Fig. 1). It features a slightly elongated head of 117.2 ± 4.1 nm length and 91.5 ± 2.8 nm width (n = 10), and its tail averages 120.2 ± 2.8 nm in length (n = 10). It is morphologically similar to Escherichia coli phage T4, where the head is 120 nm long and 85 nm wide, with a tail of 113 nm length (Tetart et al., 2001). S16 therefore falls within the A2 morphological group, members of which constitute approximately 3.2% of all known tailed phages (Ackermann, 1998). Tail sheath disks are best visible in Fig. 1C, while Fig. 1D shows the contracted tail state with an obvious collar structure.

Figure 1.

Morphology of T-even type Salmonella phage S16.

A. Transmission electron micrograph of S16 phage particles.

B. Tail fibres are stowed in a folded position along the tail sheath (arrow).

C. Extended long tail fibres. Note the two-part structure of the fibres, consisting of proximal and distal parts connected by a knee-like hinge (arrow).

D. Empty phage with contracted tail and extended tail fibres. Note the tail tube (arrow) protruding from the contracted tail sheath, a distinctive morphological feature of the Myoviridae.

Bars are 200 nm (A) and 100 nm (B–D) in length. Stain used was 2% uranyl acetate (A, C) or 2% phosphotungstic acid (B, D).

Phage S16 specifically infects Salmonella host strains

Phage S16 infection was tested on members of the genus Salmonella (32 selected strains and 14 S. Typhimurium LT2 LPS mutants), and E. coli (six laboratory strains) (Table 1). S16 was able to lyse all but one of the clinical Salmonella isolates. Not only S. enterica ssp. enterica, but also members of the other subspecies and also Salmonella bongori were all susceptible to S16 infection (Table 1). Moreover, the LPS-deficient mutants of S. Tm. LT2 were also infected, and even the LPS ‘Re’ mutants completely devoid of carbohydrate substitution of the KDO (2-keto-deoxy-d-octanoate) core residues, were sensitive to S16 infection. E. coli were generally insensitive to S16. An additional set of 126 Salmonella strains were tested using the spot-on-the-lawn technique and 96 of them (76%) were susceptible to S16 (not listed). In addition, 25 other strains of E. coli, E. hermannii (1 strain), E. vuneris (1 strain), Cronobacter sakazakii (38 strains), C. turicensis (2 strains), C. dublinensis (1 strain), C. malonaticus (1 strain), C. muytiensii (1 strain), Enterobacter aerogenes (1 strain), E. asburiae (1 strain), E. cloacae (1 strain), E. helveticus (1 strain), Citrobacter freundii (1 strain), Klebsiella pneumonia (1 strain), Vibrio natriegens (1 strain), Campylobacter jejuni (1 strain), Pseudomonas aeruginosa (2 strains) and P. fluorescens (1 strain) were tested and found to be resistant.

Table 1. Host range analysis of bacteriophage S16
StrainPhenotypeOther designationsSpot lysisSingle plaquesSource
  1. aVinogradov et al. (1999).
  2. bMuller-Loennies et al. (2002).
  3. Sources: 1: Lab stock; 2: Cheng-Hsun Chiu (Chang Gung Hospital, Taiwan); 3: University of Würzburg, Germany; 4: National Center for Enterobacteria (NENT), Switzerland; 5: Nicholas R. Thomson (Sanger Institute, UK); 6: Thilo Fuchs (Technical University of Munich, Germany); 7: Strains of the Salmonella Genetic Stock Centre (SGSC, University of Calgary, Canada) kindly provided by Uwe Mamat (Research Center Borstel; Germany); 8: Helmut Brade (Research Center Borstel; Germany); 9: Novagen (Merck Biosciences); 10: Coli Genetic Stock Center (CGSC, Yale University, USA).
  4. Lysis in spots: ++: comparable to S. Typhimurium DT7155; +: greater than 2 log reduced lysis compared with S. Typhimurium DT7155; −: no lysis observed; single plaques: +: plaques observed; −: no plaques observed.
S. Typhimurium DT7155+++1
S. Choleraesuis SC-B67++2
S. Enteritidis A  +++3
S. Enteritidis C  +++3
S. Enteritidis N2939-08+4
S. Enteritidis N90-09++4
S. Enteritidis PT4 P125109+++5
S. Enteritidis N239-07+++4
S. Enteritidis N289-074
S. Gallinarum 287/91+++5
S. Hadar N284-07+++4
S. Hadar WS 2691++1
S. Indiana  +++3
S. Javiana N2427-08+++4
S. Montevideo N2888-08+++4
S. Montevideo N1689-08+4
S. Newport N2932-08+++4
S. Newport N2889-08+++4
S. Senftenberg  +++3
S. Senftenberg N1589-08++4
S. Typhimurium N59-09+4
S. Typhimurium N62-09++4
S. Typhimurium N75-09+4
S. Typhimurium DT104+++1
S. Typhimurium LT2 ATCC 14028+++6
S. Typhimurium LT2SmoothSL3770+++7
S. Typhimurium LT2SmoothSA1355+++7
S. Typhimurium LT2Ra LPS mutantSA1627+++7
S. Typhimurium LT2Ra LPS mutantSL3749+++7
S. Typhimurium LT2Rb1 LPS mutantSL733+++7
S. Typhimurium LT2Rb2 LPS mutantSL3750+++7
S. Typhimurium LT2Rb3 LPS mutantSL3748+++7
S. Typhimurium LT2Rc LPS mutantSL1306+++7
S. Typhimurium LT2Rd1 LPS mutantSL3769+++7
S. Typhimurium LT2Re LPS mutantSL1102+++7
S. Typhimurium LT2Re LPS mutantSA1377+++7
S. Typhimurium LT2Re LPS mutantSL3600+++7
S. Typhimurium LT2rfaP LPS mutantSH7770+++7
S. Virchow  ++3
S. e. ssp. entericaO rough : i : -N2592-08+++4
S. e. ssp. salamae30 : l,z28 : z6N09-2794+++4
S. e. ssp. arizonae56 : z4,z23 : -N09-0860+++4
S. e. ssp. diarizonae61 : c : z35N09-2338+4
S. e. ssp. houtenae38 : z4,z23 : -N09-2589+++4
S. e. ssp. indica6.14.25:a:e,n,xN2576-03+++4
S. bongori48 : z35 : -N268-08++4
E. coliLPS chemotype R1F470a8
E. coliLPS chemotype R2F576a8
E. coliLPS chemotype R3F653b8
E. coliLPS chemotype R4F2513b8
E. coliLPS chemotype BBL21 (DE3)9
E. coliLPS chemotype K-12CGSC440110

S16 does not transduce host DNA

Packaging of host DNA and transduction of genetic material into other bacteria is a major source of horizontal gene transfer (Sternberg and Maurer, 1991). A phage potentially used for biocontrol purposes must not be able to transduce, in particular with respect to unspecific generalized transduction (Hagens and Loessner, 2010). Salmonella phage P22, as an established transducer (HT mutant; Schmieger, 1972), was used as positive control. Colonies resistant to both Cm and Kan were readily observed with P22, indicating successful transduction of the Cm resistance cassette into the Kan-resistant strain. No transfer of drug resistance was found with S16 (detection limit equalled transduction rate of 3 × 10−9), indicating that it is unable to transduce, at least under the conditions used here.

The S16 genome features T-even core genes

The complete genome sequence of phage S16 has been obtained by a pyrosequencing approach and in silico assembly and analyses. Individual read lengths averaged 358 bp, with 864 bp and 36 bp being the longest and shortest reads respectively. The average coverage of the genome was 84-fold, and automated and manually corrected assembly eventually yielded a complete S16 unit genome of 160 221 nt length. S16 features a G+C content of only 36.9 mol %, while its Salmonella host has a G+C content of 52.2 mol %. The S16 DNA was found to be highly restriction-enzyme resistant, with only 4 out of 34 restriction enzymes being able to cleave the molecule (Fig. 2; Table S2). A general overview of the S16 genome and a homology-based alignment to the phage T4 genome are depicted in Fig. 3A. A total of 269 putative coding sequences (CDS), as well as 3 tRNA genes (Met, Gln and Arg) were identified (Table S1), yielding a total coding capacity of 95.7%. Based on the existing but limited similarities of S16 to phage T4, putative functions could be assigned to 45.0% of the S16 gene products.

Figure 2.

Bacteriophage S16 DNA is highly restriction enzyme resistant. Of 34 restriction enzymes tested (see Table S2), only SwaI, TaqI, NdeI and SspI were able to cleave the S16 dsDNA molecule. Marker M1: λ Mix 19 ladder; M2: 1 kb ladder (Fermentas).

Figure 3.

Genome alignment of S16 with phage T4 and T2. Arrows represent annotated coding sequences.

A. Whole-genome comparison of S16 and T4.

B. Comparison of tail fibre genes gp34 through gp37 and tail fibre accessory proteins gp38 and gp57A between bacteriophages S16, T4 and T2. Shadings indicate degree of amino acid sequence identity between the interconnected gene products. Similarities between hypothetical proteins of unknown function are not indicated. Coding sequences encoding fewer than 50 aa are not represented for reasons of clarity.

S16 belongs to the T4-like virus family, which were reported to feature a common set of 39 core genes (Petrov et al., 2010) (Table S3), and were classified into several genome types. Presence of DNA modification functions (dCMP hydroxymethylase, β-glucosyltransferase and β-glucosyl-HMC-α-glucosyl-transferase), as well as the overall genome structure clearly identify S16 as a member of the T-even group (Petrov et al., 2010). Interestingly, no homologue to the T4 dam adenine methyltransferase could be identified.

Identification of S16 long tail fibre genes

In silico analysis was used to identify genes encoding the putative long tail fibre (LTF) components gp34–gp37, and accessory proteins gp57A and gp38 in the S16 genome (Fig. 3B). Sequence identity of S16 gp37 to the corresponding gene products in phages T4 and T2 is very low (T4: 20% and T2: 18%). Because this homology is below the widely accepted threshold for relatedness (Tian and Skolnick, 2003), we conclude that these proteins are to be considered orthologues rather than homologues (Jensen, 2001). The situation is different for S16 gp38, which features high sequence identity to (46%) and glycine islands similar to those described for its T2 homologue. A phylogenetic comparison of gp38 of S16 to homologues of other T4-like phages also clusters the S16 variant closer to T2 than to T4 (Fig. S2). The distal part of the S16 LTF likely has a putative structure closely related to what has been reported for phage T2 (Drexler et al., 1986), where gp38 acts as the actual adhesin while bound to the C-terminal (distal) tip of gp37. This was further substantiated by the peptide mass fingerprinting-based identification of gp38 in SDS-PAGE separated whole phage particles (results not shown). For reasons of clarity, the distal segment of the long tail fibre hetero-oligomer protein is further referred to simply as LTF.

Recombinant full-length and soluble S16 LTF

Synthesis of S16 GFP-gp37 in the presence of gp38 (with or without gp57A) in E. coli resulted in the production of functional GFP–LTF. This is similar to what has been found for T4 (Bartual et al., 2010; Leiman et al., 2010), with the difference that gp57A is not required in the case of S16 (Fig. S1). In order to illustrate its oligomeric structure, purified S16 GFP–LTF was subjected to heat denaturation gradient SDS-PAGE (Fig. 4). A clearly visible, stepwise reduction of higher to lower molecular mass bands was observed, indicating successive disintegration of the oligomer. The full-length monomeric protein has a predicted molecular mass of 108.5 kDa, whereas the protein in the boiled sample located just below the 97.2 kDa marker (Fig. 4).

Figure 4.

Oligomeric nature of the S16 long tail fibre. Recombinant S16 GFP–LTF was incubated at 65°C to monitor its stepwise degradation from oligomer to monomers over a period of 60 min. The boiled sample was heated to 100°C for 10 min before loading (5% Tris-HCl Gel, run at 4°C, 100 V).

S16 LTF features a high equilibrium binding affinity

Surface plasmon resonance (SPR) analyses of the interaction between S16 LTF and S. Tm. DT7155 wild-type cells was possible following immobilization of the bacterial cells on the sensor chip surface. The equilibrium association constant (KA) values for the interaction of the LTF with its ligand on the cell surface were found to be in the nanomolar range (3.55 ± 3.50 × 108 M−1) (Table 2). A variation of the experimental set-up by using amine coupling and immobilization of anti-Salmonella antibody pre-coated cells yielded essentially identical results, with KA values of 8.64 × 108 M−1 and 6.21 × 108 M−1 for 100 and 200 nM S16 LTF concentration respectively.

Table 2. Surface plasmon resonance (SPR) analysis of S16 GFP–LTF and S. Tm. DT7155 wild-type interaction
Concentration (nM)Association rate constant ka (M−1 s−1)Dissociation rate constant kd (s−1)Equilibrium association constant KA = ka/kd (M−1)Equilibrium dissociation constant KD = kd/ka (M)
1001.48 × 1051.96 × 10−47.55 × 1081.32 × 10−9
2004.63 × 1042.24 × 10−42.06 × 1084.85 × 10−9
5004.26 × 1044.04 × 10−41.05 × 1089.49 × 10−9
Mean ± SD7.90 ± 5.98 × 1042.75 ± 1.13 × 10−43.55 ± 3.50 × 1085.22 ± 4.10 × 10−9

Purified S16 LTF binds to OmpC

Because outer membrane proteins are known to frequently serve as phage receptors and because of the intrinsic relationship of S16 to T2 and T4, OmpF and OmpC (Hantke, 1978; Yu and Mizushima, 1982) were considered as potential candidates for this function. We therefore created appropriate null mutants in a S. Tm. DT7155 background, and assessed binding of purified S16 GFP–LTF fusion protein to the mutant cells. The S. Tm. DT7155 wild-type revealed an even decoration of the bacterial surface by the fluorescent LTF (Fig. 5A and B). Removal of OmpF, the binding ligand for phage T2 (Hantke, 1978), did not influence targeted binding of GFP–LTF. In contrast, deletion of OmpC completely abolished decoration by S16 GFP–LTF (Fig. 5C and D). The defective phenotype could be restored by in-trans complementation using ompC from pRM2 (Table 3, Fig. 5E and F). Besides OmpC and OmpF, several other mutants featuring deletion of genes encoding BtuB, FadL, OmpA, OmpX, TonB, and Tsx were also created and tested. However, none of them affected either LTF binding or susceptibility to infection by S16. A shortened version of the S16 GFP–LTF, containing only the C-terminal 183 aa of gp37 was also able to decorate S. Tm. DT7155 cells (results not shown). Altogether, these findings demonstrate that OmpC is necessary and sufficient for recognition and binding of S16 LTF to Salmonella host cells.

Figure 5.

Binding of S16 LTF to Salmonella cells. Salmonella cells resuspended in SM buffer were incubated with S16 GFP–LTF. After washing, samples were observed by phase contrast (A, C and E) and fluorescence microscopy (B, D and F). (A, B) S. Tm. DT7155 wt, (C, D) S. Tm. DT7155 ΔompC, (E, F) S. Tm. DT7155 ΔompC::ompC(DT) induced with 10 mM arabinose.

Table 3. Strains, phages and plasmids used in this study
BacteriaOther designations and featuresSource
Salmonella TyphimuriumDT7155Laboratory stock
S. Tm. DT ΔbtuBDT7155, ΔbtuB ::KanrThis study
S. Tm. DT ΔfadLDT7155, ΔfadL ::KanrThis study
S. Tm. DT lpsReDT7155, lpsRe::Cmr; all LPS synthesis genes for regions more distal than KDO residues were deleted [∼ 13.5 kb region between and including gmhD (rfaD) and waaQ (rfaQ)]This study
S. Tm. DT ΔompADT7155, ΔompA::KanrThis study
S. Tm. DT ΔompCDT7155, ΔompC::KanrThis study
S. Tm. DT ΔompFDT7155, ΔompF::KanrThis study
S. Tm. DT ΔompXDT7155, ΔompX::KanrThis study
S. Tm. DT ΔtonBDT7155, ΔtonB::KanrThis study
S. Tm. DT ΔtsxDT7155, Δtsx::KanrThis study
S. Tm. DT ΔompC, lpsReDT7155, ΔompC::Kanr; lpsRe::CmrThis study
S. Tm. DT ΔompC, ΔfadLDT7155, ΔompC::Kanr; ΔfadL::CmrThis study
S. Tm. DT Δ1493DT7155, Δ1493::CmrThis study
S. Tm. DT ΔphoNDT7155, ΔphoN::KanrThis study
S. Tm. DT Cmr, KanrDT7155, Δ1493::Cmr, ΔphoN::KanrThis study
Salmonella EnteritidisN289-07NENTa
S. E. CmrN289-07, Δ1493::CmrThis study
Salmonella NewportN2932-08NENTa
S. N. KanrN2932-08, Δ1493::KanrThis study
Escherichia coliXL1 Blue MRF′, TetrStratagene, Santa Clara, CA, USA
E. coli K-12 wild-typeCGSC4401CGSCb
E. coli ΔompCCGSC4401, ΔompC::KanrThis study
PhagesOther designations and featuresSource
S16Salmonella Phage S16, vB-SenM-S16This study
FO1-E2Salmonella Phage, Felix-likeLaboratory stock
PlasmidsOther designations and featuresSource
  1. aNational Center for Enterobacteria (NENT), Switzerland.
  2. bColi Genetic Stock Center (CGSC), Yale University, USA.
pBAD18 AmprAmpr; Arabinose-inducible transcriptionGuzman et al. (1995)
pBAD18 CmrCmr; Arabinose-inducible expressionGuzman et al. (1995)
pHGFPAmpr; IPTG-inducible expression; 6xHis-tag (N-terminal)Loessner et al. (2002)
pRM1pHGFP backbone; S16 gp37 cloned under SacI/SalI; S16 gp38 under SalI/PstIThis study
pRM2pBAD18 Ampr backbone; ompC(DT) cloned under EcoRI/SalIThis study
pRM3pBAD18 Ampr backbone; ompC(K-12) cloned under SacI/SalIThis study
pRM4pBAD18 Cmr backbone; gp57A cloned under SacI/SalIThis study
pRM5pHGFP backbone; truncated S16 gp37 cloned under SacI/SalI; S16 gp38 under SalI/PstIThis study
pKD3Ampr; Cmr cassette flanked by P1 and P2Datsenko and Wanner (2000)
pKD4Ampr; Kanr cassette flanked by P1 and P2Datsenko and Wanner (2000)
pKD46Ampr, red recombinase, temp. sensitive, permissive: 30°CDatsenko and Wanner (2000)

Phage S16 binds to bacterial cells displaying Salmonella OmpC

Phage adsorption was assessed by pull-down assay, where centrifugation removes phage particles bound to the bacterial cells. In line with the above, our results demonstrate that adsorption of intact phage is dependent on OmpC (Fig. 6A). The ability of S16 to attach to the surface of S. Tm. DT7155 ΔompC cells was drastically reduced compared with the wild-type (47% versus 98%, P-value: 0.01). Complementation with ompC(DT) from pRM2 completely restored wild-type phage adsorption (98%). Pre-incubation of the bacterial cells with GFP–LTF did also affect phage S16 binding (67%, P-value: 0.05).

Figure 6.

Recognition of host cells by phage S16 requires Salmonella OmpC. Salmonella cells were mixed with phage, incubated for 10 min and collected by centrifugation. Infectious phage remaining in the supernatant was quantified by determination of PFU, and the adsorption ratio calculated.

A. Adsorption of S16 is dependent on presence or absence of ompC, and can be reduced by S16 GFP–LTF. Bars are as follows: Aa: S. Tm. DT7155 wt; Ab: S. Tm. DT7155 ΔompC; Ac: S. Tm. DT7155 ΔompC::ompC(DT); Ad: S. Tm. DT7155 wt pre-incubated with GFP–LTF; Ae: S. Tm. DT7155 wt pre-incubated with GFP.

B. Expression of ompC(DT) in E. coli K-12 ΔompC increases adsorption of phage S16, while display of ompC(K-12) does not. Bars are as follows: Ba: S. Tm. DT7155 wt; Bb: E. coli K-12 wt; Bc: E. coli K-12 ΔompC::ompC(DT); Bd: E. coli K-12 ΔompC::ompC(K-12).

C. Removal of the outer LPS core reduces adsorption of S16. An additive effect can be observed if this mutation is combined with the deletion of ompC. Bars are as follows: Ca: S. Tm. DT7155 wt; Cb: S. Tm. DT7155 ΔompC; Cc: S. Tm. DT7155 ΔfadL; Cd: S. Tm. DT7155 ΔompC ΔfadL; Ce: S. Tm. DT7155 lpsRe; Cf: S. Tm. DT7155 ΔompC lpsRe.

All values represent mean values (averages) from three independent experiments; the error bars indicate standard deviations. Asterisks above bars indicate significant differences to S. Tm. DT7155 wt (*P < 0.05, **P < 0.01).

Even though E. coli K-12 does not support productive infection by S16, a low level of binding of the Salmonella phage to the E. coli cells was observed (28%, Fig. 6B). Interestingly, E. coli K-12 ΔompC (CGSC4401) displaying the S. Tm. DT7155 ompC showed Salmonella-like adsorption of S16 (92% versus 98%, P-value: 0.02) (Fig. 6B). As expected, complementation of the E. coli knockout with its own ompC (K-12) yielded no increase in adsorption, which excluded possible plasmid-related effects through different intracellular OmpC levels. Taken together, these findings clearly demonstrate that binding of both LTF and intact phage particles is strongly dependent on Salmonella OmpC.

Role of LPS in S16 attachment and infection

Although the principal binding ligand for S16 LTF is not present, we found that virus particles were still able to attach to S. Tm. DT7155 ΔompC (47%) and to E. coli K-12 wild-type (28%), although to a much lesser degree. This suggested that other surface structures might also contribute to S16 recognition and binding, or serve as secondary receptors. Such phenomena exist for T4 (strong binding to the LPS of E. coli B, and to OmpC of E. coli K-12 versus slower and weaker binding to LPS of E. coli K-12), and in T2 (OmpF versus FadL as receptors) (Hantke, 1978; Yu and Mizushima, 1982; Black, 1988; Trojet et al., 2011). In order to determine the possible roles of FadL and the LPS macromolecule in S16 reception, we created an additional set of Salmonella mutants: S. Tm. DT7155 ΔfadL::Cmr, S. Tm. DT7155 ΔompC::Kanr, ΔfadL::Cmr, S. Tm. DT7155 lpsRe::Cmr and S. Tm. DT7155 ΔompC::Kanr lpsRe::Cmr. The first two mutants enabled us to assess the possible role of FadL alone or in conjunction with OmpC in phage binding and infection. ‘lpsRe’ stands for the deletion of all genes specifying enzymes required for carbohydrate substitution of the LPS inner core 2-keto-deoxy-d-octanoate (KDO) residues, yielding a ‘deep rough’ Re LPS mutant (Heinrichs et al., 1998). This phenotype allowed us to investigate whether the lack of OmpC in absence of the LPS core has an even stronger effect on phage adsorption and infection. Pull-down experiments (Fig. 6C) revealed that removal of FadL alone did not affect phage adsorption, and the ΔfadL ΔompC double null mutant did behave similar to the ΔompC deletion alone. However, removal of the LPS core in the lpsRe mutant (Fig. 6C) significantly attenuated phage adsorption (72% versus 98%; P-value: 0.01). Binding of S16 to the double mutant lpsRe ΔompC was further reduced, as compared with the ΔompC deletion alone (13% versus 47%, P-value: 0.01). As a conclusion, these data suggest an additive, if not a synergistic effect of OmpC and the Salmonella LPS core in S16 attachment.

Deletion of both OmpC and the LPS outer core confers complete immunity to S16

The efficiency of plating (e.o.p.) of phage S16 on the various mutant strains was investigated (Fig. 7). The ΔompC mutant was approximately 200 000-fold less susceptible to S16 phage infection (plaque formation) than the wild-type (Fig. 7B). Deletion of FadL did not influence the e.o.p. in either the wild-type or the ΔompC background. As expected from the pull-down assays, the lpsRe mutation by itself did not significantly affect the e.o.p. in the wild-type background. However, the double null mutant ΔompC lpsRe was found completely resistant to infection by S16 (Fig. 7F). These data are in perfect agreement with our findings from LTF binding and pull-down assays, and suggest a possible role of the LPS core structure as additional or alternative receptor for the reversible binding of S16.

Figure 7.

Efficiency of plating of S16 on S. Tm. DT7155 is affected by host mutations. Quantification of phage plaques (PFU ml−1) was done by plating of various concentrations of phage S16 on S. Tm., followed by calculation of the e.o.p. as relative values compared with the control DT7155 wild-type host (A), as well as ΔompC (B), ΔfadL (C), ΔompC ΔfadL (D), lpsRe (E) and ΔompC lpsRe (F) mutants of S. Tm. DT7155. All values represent mean values (averages) from three independent experiments; error bars indicate standard deviations. N.d. stands for not detectable. Asterisks above bars indicate significant differences to S. Tm. DT7155 wt (**P < 0.01).

S16 effectively controls Salmonella in contaminated food

As proof-of-concept experiment for the biocontrol ability of phages S16 and FO1-E2 against Salmonella contaminations in food, both phage preparations were used alone and in combination to reduce Salmonella contamination of chocolate milk at 6.5°C over a period of 6 days (Fig. 8). Samples were inoculated with a mixture of three strains: S. Tm. Cmr, Kanr (sensitive to both phages), S. E. Cmr (sensitive to FO1-E2 only) and S. N. Kanr (sensitive to S16 only). This three strain mixture of different S. enterica serovars was chosen in order to illustrate the specificity of phage killing in foods, and the complementation of host ranges in a two-phage cocktail consisting of S16 and the previously described Felix-like phage FO1-E2 (Guenther et al., 2012). The results clearly demonstrate the specific reduction of Salmonella by the respective phage and unhindered survival of the insensitive strain. If reduction occurred, it was complete: no Salmonella were detectable after the storage period by direct plating of the samples (detection limit 1 cfu ml−1). The complementation of the two phage host ranges could be clearly demonstrated by an elimination of all three strains in samples containing both phages.

Figure 8.

Phages S16 and FO1-E2 can control growth of a three strain mixture of Salmonella. Phages S16 and FO1-E2 were used alone and in combination to reduce Salmonella contamination of chocolate milk at 6.5°C over a period of 6 days. All three strains were present in each sample. Results are presented in separate graphs for increased clarity. Strains were chosen to be sensitive to either both phages (A: S. Tm. Cmr, Kanr), FO1-E2 only (B: S. E. Cmr) or S16 only (C: S. N. Kanr).

Discussion

Salmonella myovirus S16 appears to be the first T-even phage specific for the genus Salmonella. The overall genome structure of S16 resembles that of phage T4 (Fig. 3). Out of 39 genes defined as a T4-like core genome (Petrov et al., 2010), only two are missing in S16 (Table S3). Instead of a full-length uvsW gene, specifying a recombination DNA–RNA helicase and DNA-dependent ATPase, S16 features two separate, shorter reading frames designated as uvsW1 and uvsW2. Because their putative products are highly homologues to UvsW of T4, it is likely that they assume a similar function. The other missing gene is orf49 encoding endonuclease VII, which is essential for T4 replication. However, the enzyme may be substituted by endonucleases with catalytic domains similar to I-TevII, such as those found in E. coli phage RB16 and Aeromonas phage 65 (Petrov et al., 2006). Again, S16 features two genes specifying proteins highly similar to I-TevIII (a homing endonuclease), which is non-functional in T4 (Robbins et al., 2007). Therefore, it seems possible that the S16 I-TevIII may be able to functionally compensate for the absence of a T4 gp49 homologue.

In silico analysis indicated that S16 replaces cytosine residues in its DNA with hydroxymethylcytosine (HMC), employing a dCMP hydroxymethylase. This modification renders DNA resistant to cleavage by most restriction enzymes (Fig. 2, Table S2). Some T4-like phages encode glucosyl-transferases, which link glucose molecules to HMC in either α- or β-configuration. S16 also features a β-glucosyl-transferase and a β-glucosyl-HMC-α-glucosyl-transferase. It is reasonable to assume that S16 features di-, mono- but also non-glucosylated HMC, similar to phage T2, which also possesses β- but no α-glucosyltransferase activity (Lehman and Pratt, 1960). This modification would explain the resistance of S16 DNA to restriction enzyme digestion. However, some methylation-insensitive enzymes (Table S2) were able to cleave S16 DNA, indicating that probably not all HMC residues are glucosylated. Alternatively, some endonucleases may be insensitive to glucosylation. There are, however, only few such enzymes described, e.g. GmrSD (Bair and Black, 2007). Other endonucleases able to target HMC residues, such as Mrr and homologues, are blocked by glucosylation (Zheng et al., 2010).

The phage tail-associated receptor-binding proteins (RBPs or adhesins), which in the case of the tailed phages are physically present as long tail fibres and/or short tail spikes, are required for host cell recognition, attachment, and initiation of infection and DNA injection. In S16, we identified T-even like long tail fibre genes encoding gp34 through gp38. However, S16 gp37 was found to be more closely related to its T2 orthologue rather than the T4 component, because of two reasons. First, it does not feature His-boxes as present in T4 gp37 (Tetart et al., 1996). Second, both S16 gp37 and T2 gp37 show strong similarity in their C-terminal regions to an intramolecular chaperone domain (IMC) of the phage K1F endo-N-acetylneuraminidase (Scholl and Merril, 2005). Such phage-encoded endosialidases have been found to undergo autoproteolytic processing to yield functional trimers (Muhlenhoff et al., 2003). The C-terminal domain of these proteins acts as a class II IMC, mediating correct quaternary folding of proteins (Schulz et al., 2010). Several key amino acids, including a serine residue at the putative cleavage site, are conserved in S16 as compared with T2 (Muhlenhoff et al., 2003). Moreover, it has been shown that T2 gp37 undergoes autolytic processing, resulting in removal of 120 amino acids from the C-terminal end, which permits binding to the gp38 adhesin (Drexler et al., 1986). With respect to S16, we were able to produce full-length S16 gp37 (including an N-terminal GFP tag) in E. coli, using an approach similar to what has been described for the T4 protein (Bartual et al., 2010). Interestingly, the predicted molecular mass for full-length monomeric S16 gp37 is 108.5 kDa, whereas the corresponding SDS gel band in the boiled sample located just below the 97.2 kDa marker (Fig. 4). Autoproteolytic processing at the conserved serine residue within the class II IMC domain as described above would remove 111 residues from the C-terminus of S16 gp37 and result in a 95.6 kDa protein, which would explain the observed divergence.

The S16 LTF was made in four different variations, i.e. GFP-gp37 and GFP-gp37 together with gp38, both in presence and in absence of gp57A. Synthesis of GPF-gp37 alone yielded non-functional protein, whereas production of GFP-gp37 and gp38 together yielded a product able to bind to Salmonella cells, irrespective of whether orf57A was coexpressed or not. These observations suggest that gp57A, a putative general tail fibre trimerization chaperone, is not required for synthesis of functional LTF in S16. This result is in line with a situation as discussed above, where trimerization is mediated by a class II intramolecular chaperone in gp37. Altogether, our findings demonstrate the necessity of gp38 for functionality of S16 LTF, whereas gp57A is not required. Gp38 was also detected by peptide mass fingerprinting in the trimeric bands of non-denatured LTF produced in E. coli (Fig. S1). In conclusion, the native S16 LTF protein apparently represents a trimer of gp37 with gp38 as adhesin attached to its distal end. The precise stoichiometry of the two molecules remains to be determined.

Use of GFP-tagged S16 LTF enabled identification of OmpC as its primary interaction ligand. Interestingly, an N-terminally truncated GFP–LTF protein featuring only the 183 C-terminal amino acids of gp37 remained functional in binding assays (data not shown). We also show that intact phage adsorption specifically requires S. enterica OmpC, and does not work with E. coli OmpC (Fig. 6B). Further investigations revealed another layer of complexity in S16 LTF attachment. As a single mutant, lpsRe reduced adsorption of phage S16, but not the e.o.p. In contrast to the well-known Salmonella phage Felix-O1, which requires terminal N-acetylglucosamine residue of the outer LPS core for infection (Lindberg, 1967; Lindberg and Holme, 1969), S16 is therefore also able to infect the so-called ‘deep rough’ strains, featuring complete loss of LPS carbohydrate substitution (Re LPS chemotype) in the presence of OmpC. The lpsRe mutant did, however, display an additive effect on reduction of both adsorption (Fig. 6C) and e.o.p. (Fig. 7), if combined with the ΔompC mutation. These findings suggest that OmpC is the preferred ligand for the S16 LTF adhesin. The additive effects observed in the ΔompC lpsRe mutant may be due to interactions of both long and short tail fibre protein (STF, equivalent of the T4 gp12) with conserved portions of the LPS molecule. It cannot be excluded, however, that the S16 LTF may be responsible for this effect by recognizing an alternative receptor besides OmpC, which could also be a component of the LPS. The lpsRe single mutant did not exhibit reduced susceptibility to phage infection, which suggests that a fully intact LPS core structure is not required for irreversible binding of the STF, which is likely represented by S16 gp12 (CDS 93, Table S1). The KDO (2-keto-deoxy-d-octanoate) residues of the inner LPS core are probably sufficient for this purpose, whereas the STF of other T-even phages appear to require heptose residues of the LPS core for binding (Riede, 1987).

The host range of S16 was found to be extremely broad; it can infect all Salmonella species and subspecies and most serovars of the subspecies enterica, while none of the tested Escherichia, Cronobacter, Enterobacter, Citrobacter, Klebsiella, Vibrio, Campylobacter and Pseudomonas strains were susceptible. Therefore, S16 represents an ideal candidate for phage-based intervention, such as pre-slaughter treatment of animals (Wall et al., 2010) or biocontrol in foods (Hagens and Loessner, 2007a; Guenther et al., 2012), which we were able to confirm in a proof-of-concept experiment using spiked chocolate milk. Apart from being virulent and specific, phages for biocontrol should not be able to carry out generalized transduction, the unspecific transfer of pathogen host DNA (Hagens and Loessner, 2010), and our results demonstrate that S16 meets these requirements.

With respect to specific recognition of the host cells, S16 LTF binds to Salmonella with an equilibrium affinity in the nanomolar range, which is comparable to matured polyclonal antibodies (Medina et al., 1997). This high affinity is reminiscent of the properties of cell wall-binding domains of phage-encoded peptidoglycan hydrolases (Loessner et al., 2002; Schmelcher et al., 2010), which are successfully used in bacterial diagnostics for both cell decoration and immobilization (Hagens and Loessner, 2007b; Schmelcher and Loessner, 2008), and renders the S16 LTF component a promising reagent for this purpose.

Experimental procedures

Strains and plasmids

The strains and plasmids used in this study are listed in Table 3; strains used in host range testing only are listed in Table 1. All bacteria were grown in LB media at 37°C under agitation. Concentrations of antibiotics used are as follow: Ampicillin (Amp, AppliChem, Darmstadt, Germany) at 100 μg ml−1; Chloramphenicol (Cm, Sigma-Aldrich, St. Louis, USA) at 25 μg ml−1; Kanamycin (Kan, Sigma-Aldrich) at 200 μg ml−1 for liquid cultures, and 50 μg ml−1 for agar plates; Tetracycline (Tet, Sigma-Aldrich) at 18 μg ml−1.

Phage isolation

S16 was isolated from sewage from a municipal sewage plant in Gelderland, Netherlands. Liquid samples were filtered (0.45 μm), decimal dilutions mixed with different S. enterica indicator strains, and plated on soft-agar overlay plates (Sambrook and Russel, 2001). Phage S16 was selected for further analyses based on its broad lytic activity (host range) against all used indicator strains.

Phage propagation and purification

The virus was propagated using the soft-agar overlay plate-lysate method, PEG precipitated and subsequently CsCl purified following standard procedures (Sambrook and Russel, 2001).

Host range analysis

Agar plates were flooded with 4 ml of log-phase host cell culture, excess liquid removed, and the plates dried for 30 min at 30°C. From the CsCl purified stock (>1011 PFU ml−1), phage suspensions were prepared by dilution (10−2 to 10−7) in SM buffer, and small amounts (3 μl) of each dilution were spotted onto the bacterial lawns, followed by overnight incubation. To screen larger numbers of strains, spot on the lawn method was employed. Phage is spotted onto the positive control strain (S. Tm. DT7155) in serial dilutions, to determine at which phage concentration the spots are semi-confluently lysed, i.e. single plaques become visible within the macroplaque of the application zone. The concentration used for the single spot assay was the 100-fold concentration of this dilution (Loessner and Busse, 1990). The appearance of single plaques up to complete lysis within the application zones indicates productive phage infection.

Efficiency of plating (e.o.p.) assay

The titre of phage S16 on S. Tm. DT7155 mutants was standardized against the titre obtained on the wild-type strain. This resulted in the efficiency of plating (e.o.p.).

DNA preparation and restriction enzyme analysis

Purified phage particles were dialysed against a 1000-fold excess of SM buffer, and treated with RNase (10 μg ml−1) and DNase (20 μg ml−1) for 20 min at 37°C. After addition of 20 mM EDTA (pH 8) and proteinase K (50 μg ml−1, Fermentas) and incubation at 55°C for 1 h, the DNA was phenol/chloroform-extracted, ethanol-precipitated (Sambrook and Russel, 2001), dissolved in TE buffer, and stored at −20°C. Small aliquots of purified phage DNA (500 ng) were then digested with restriction enzymes (Table S2), and fragments separated electrophoretically alongside a suitable size marker.

Genome sequencing and in silico analyses

Genome sequencing of phage S16 was carried out using pyrosequencing (Roche 454 FLX Titanium reagents; GATC Biotech, Konstanz, Germany). Sequences were assembled using the GS De Novo assembler software (Newbler, Version 2.3, Roche, Switzerland). Further analyses were performed using CLC Main Workbench (Version 6.5, CLC Bio). Regions displaying uncertain consensus sequence were verified by PCR amplification followed by Sanger sequencing. Annotation of the S16 genome was carried out employing the ‘Genome Annotation Transfer Utility’ (GATU; available at http://www.virology.ca/gatu) and the bacteriophage T4 complete genome (NC_000866) as reference (Tcherepanov et al., 2006), and manually refinement. Putative tRNAs were identified using tRNAscan-SE v.1.21 (available at http://lowelab.ucsc.edu/tRNAscan-SE/; Lowe and Eddy, 1997). Homology detection and secondary structure prediction of proteins were carried out using HHpred (http://toolkit.tuebingen.mpg.de/hhpred).

Database accession

The annotated genome sequence of phage S16 has been deposited in GenBank under the Accession No. HQ331142. The complete sequence of phage T4 (NC_000866) was retrieved from NCBI (http://www.ncbi.nlm.nih.gov/genomes), and the genome of T2 was acquired from the T4-like genomes website (http://phage.ggc.edu/).

Transduction assay

Generalized transduction was tested using mutants of Salmonella Tm. DT7155, carrying different selectable antibiotic resistance markers: Δ1493::Cmr (Cm: chloramphenicol) and ΔphoN::Kanr (Kan: kanamycin), which were constructed using site-directed mutagenesis as described further below. Phage lysate was prepared on the Cmr strain (donor), and used to infect the Kanr strain (recipient). Cultures were tested for growth of colonies on plates containing both antibiotics.

Protein synthesis and purification

The long tail fibre gene gp37 of phage S16 was cloned into pHGFP-Ampr (Loessner et al., 2002). This plasmid allows IPTG-inducible gene expression and offers genetic fusion to an N-terminal 6xHis-tag and the GFP-coding sequence. The S16 gene encoding the adhesin gp38 was inserted downstream of the long tail fibre gene to yield a bicistronic transcript (using 5′-AGGAGG-3′ as Shine–Dalgarno sequence, resulting in plasmid pRM1). Finally, the gene for gp57A, a general trimerization chaperone, was placed on a separate plasmid (pRM4), with expression under arabinose control (Guzman et al., 1995). Protein synthesis in E. coli XL1 Blue MRF′ (Stratagene, Basel, Switzerland) was induced with 0.5 mM IPTG (Axon Lab, Baden, Switzerland) and carried out overnight at 20°C. Purification was done via immobilized metal affinity chromatography (IMAC) using a low-density Ni-NTA resin (Chemie Brunschwig, Basel, Switzerland).

Mass spectrometry

Peptide mass fingerprinting for confirmation of protein identity was performed by ESI-MS/MS on a LTQ-Orbitrap Velos device (Thermo Scientific, Wohlen, Switzerland) at Functional Genomics Center Zurich as previously described (Klumpp et al., 2008; Born et al., 2011).

Cell decoration with GFP-tagged LTF

Protein binding to host cells (binding assay) was tested using log-phase cultures. Bacteria from 0.5 ml aliquots were pelleted, and resuspended in 200 μl of SM buffer. Purified GFP–LTF was centrifuged to remove aggregates (31 000 g, 4°C, 30 min), and approximately 1 μg of protein was added to the cells. After 10 min incubation at room temperature, cells were again pelleted and washed once in 1 ml SM buffer. Microscopy (Axioplan, Carl Zeiss AG, Germany) was performed at 100-fold magnification, with appropriate filter settings: excitation BP 450–490 nm, FT 510 nm, emission wavelength LP 520 nm. Images were recorded using a Leica LP-420 digital microscope camera, and minimally processed using Photoshop software (Adobe).

Surface plasmon resonance (SPR) measurements with GFP-tagged LTF

Surface plasmon resonance (SPR) measurements were carried out on a BIAcore X Analyser (Uppsala, Sweden), essentially as described previously (Schmelcher et al., 2010). An amine coupling procedure was used to immobilize polyclonal goat anti-Salmonella IgG (OAMA02979, Aviva Systems Biology, San Diego, USA) in the two flow cells (Fc1 and Fc2) of the sensor chip (2D carboxymethyldextran hydrogel, SCB CMDP, Xantec bioanalytics, Duesseldorf, Germany). Fixed S. Tm. DT7155 wild-type cells (3% formaldehyde, 10 min) were applied and immobilized in Fc2. Real-time interaction analysis was performed by injecting S16 GFP–LTF in both flow cells, and measuring the Fc2–Fc1 response. Curves were fitted using the predefined ‘1:1 binding with mass transfer’ model (BIAcore evaluation software).

Site-directed mutagenesis

Insertion mutants in E. coli and S. enterica ssp. enterica were created as previously described (Datsenko and Wanner, 2000). Homologous sequences were chosen such that the first 18 and last 36 nucleotides of the gene of interest remained unaltered. The rest of the gene was replaced by a drug-resistance cassette (i.e. ΔompC::Kanr, or just ΔompC for short; see Table 3). Deletion mutants were trans-complemented by using ompC from either S. Tm. DT7155 (ompC(DT)) or E. coli K-12 (ompC(K-12)) cloned on pRM2 and pRM3 respectively (Table 3), which are arabinose-inducible (Guzman et al., 1995).

Phage pull-down assay

One millilitre of samples of bacterial overnight cultures were adjusted to OD600 = 1.0 ± 0.05 (approximately 5 × 108 cfu ml−1), and 10 μl of phage solution (containing 109 PFU ml−1) were added. Samples were incubated in an overhead rotator for 10 min at room temperature, centrifuged (5 min, 20 000 g), and the PFU ml−1 remaining in the supernatant determined in triplicate. The adsorption ratio was calculated from comparison to a negative control in LB medium without cells. Inhibition of adsorption by purified LTF was determined by pre-incubation of cells with ∼20 μg ml−1 GFP–LTF for 10 min before phage was added.

Biocontrol of Salmonella in chocolate milk

Chocolate milk samples were purchased at a local retailer, and monitored for pre-existing contamination with Salmonella (ISO 6579:2002). Samples of 40 ml each were inoculated with 103 cfu ml−1 each of S. Tm. DT7155 (Cmr, Kanr), S. Newport (Kanr) and S. Enteritidis (Cmr). Phage S16, phage Felix-O1-E2 (Guenther et al., 2012), or both together were each added at a concentration of 3 × 108 PFU ml−1. Chocolate milk with and without bacteria served as positive and negative controls respectively. Samples were stored over a period of 6 days at 6.5°C. Following plating on agar plates containing either Cm and/or Kan, colonies could be differentiated and enumerated based on their drug resistance(s).

Statistical analyses

Data values from multiple measurements were averaged, and standard deviations calculated. P-values of Student's t-test (one-tailed, two samples of unequal variance, significance level α = 0.05) were determined.

Acknowledgements

We thank Rudi Lurz (Max-Planck Institute for Molecular Genetics, Berlin, Germany) for his help with electron microscopy, and are grateful to Stefan Miller (Lisando GmbH, Regensburg, Germany) for helpful discussions and advice concerning cloning and expression of the LTF, and to Thilo Fuchs (TU Munich, Germany) for providing the pBAD18 plasmid.

Ancillary