Listeria ivanovii differs from the human pathogen Listeria monocytogenes in that it specifically affects ruminants, causing septicaemia and abortion but not meningo-encephalitis. The genetic characterization of spontaneous L. ivanovii mutants lacking the virulence factor SmcL (sphingomyelinase) led us to identify LIPI-2, the first species-specific pathogenicity island from Listeria. Besides SmcL, this 22 kb chromosomal locus encodes 10 internalin (Inl) proteins: i-InlB1 and -B2 are large/surface-associated Inls similar to L. monocytogenes InlB; i-InlE to –L are small/excreted (SE)-Inls, i-InlG being a tandem fusion of two SE-Inls. Except i-inlB1, all LIPI-2 inl genes are controlled by the virulence regulator, PrfA. LIPI-2 is inserted into a tRNA locus and is unstable – half of it deleting at ≈10−4 frequency with a portion of contiguous DNA. The spontaneous mutants were attenuated in vivo in mice and lambs and showed impaired intracellular growth and apoptosis induction in bovine MDBK cells. Targeted knock-out mutations associated the virulence defect with LIPI-2 genes. The region between the core genome loci ysnB-tRNAarg and ydeI flanking LIPI-2 contained different gene complements in the different Listeria spp. and even serovars of L. monocytogenes, including remnants of the PSA bacteriophage int gene in serovar 4b, indicating it is a hot spot for horizontal genome diversification. LIPI-2 is conserved in L. ivanovii ssp. ivanovii and londoniensis, suggesting an early acquisition during the species’ evolution. LIPI-2 is likely to play an important role in the pathogenic and host tropism of L. ivanovii.
The Gram positive genus Listeria comprises six species, two of which –Listeria monocytogenes and Listeria ivanovii– are pathogenic. Both bacteria are facultative intracellular parasites able to infect macrophages and non-phagocytic cells, such as epithelial cells. After internalization, they undergo a characteristic intracellular infection cycle involving early escape from the phagocytic vacuole, rapid cytosolic replication, actin-based motility, and direct (cell-to-cell) spread to neighbouring cells, where the cycle begins again. Several virulence genes involved in key steps of this cycle are clustered together in a 9 kb locus that is located at the same chromosomal position in L. monocytogenes and L. ivanovii. This central virulence gene cluster or ‘Listeria pathogenicity island 1 (LIPI-1)’ is absent – or present in a non-functional form – in the non-pathogenic Listeria spp. (Vázquez-Boland et al., 2001a). LIPI-1 encodes a pore-forming toxin (listeriolysin O, LLO) and two phospholipases C (PlcA and PlcB) that cooperate to lyse the phagocytic vacuole membrane; an actin-polymerizing surface protein (ActA), responsible for intracellular bacterial motility and cell-to-cell spread; a metalloprotease (Mpl) involved in the maturation of proPlcB; and a transcriptional activator (PrfA) that controls the expression of LIPI-1 genes and of other virulence determinants located elsewhere on the listerial chromosome (Portnoy et al., 2002; Dussurget et al., 2004). The latter include hpt, encoding a hexose phosphate transporter (Hpt) required for rapid cytosolic replication (Chico-Calero et al., 2002), present in both L. monocytogenes and L. ivanovii; and the inlAB operon, encoding two surface proteins (InlA and InlB) that mediate host cell invasion (Cossart et al., 2003), only found to date in L. monocytogenes.
InlA and InlB belong to a family of listerial proteins known as ‘internalins’ characterized by the presence of 22-residue leucine-rich repeats (LRRs), structures that provide a versatile framework for the formation of protein–protein interactions. Via their LRR domains, InlA and InlB interact with their host cell receptors, human E-cadherin (Mengaud et al., 1996) and the hepatocyte growth factor receptor Met (Shen et al., 2000) respectively, leading to internalization. There are two types of internalins: (i) large (> 50 kDa)/surface-associated internalins (LA-Inls), such as InlA and InlB, which possess a C-terminal cell wall attachment (Cwa) domain that ties the protein to the bacterial cell wall (Cabanes et al., 2002); and (ii) small (25–35 kDa)/excreted internalins (SE-Inls), exemplified by L. monocytogenes InlC (Engelbrecht et al., 1996), which lack a membrane anchor and are therefore released into the culture supernatant. While a role in pathogenesis has been clearly established for internalins of the former group, the way SE-Inls contribute to virulence remains unknown. The genome of L. monocytogenes contains between 16 and 18 LA-inl and seven to nine SE-inl genes depending on the strains, all arranged in small chromosomal islets (Glaser et al., 2001; Nelson et al., 2004). Although no LA-Inls have been found to date in L. ivanovii, four SE-Inls encoded in two different loci, i-inlDC and i-inlFE, have been identified in this species (Engelbrecht et al., 1998a,b).
Despite their similar intracellular lifestyles, L. monocytogenes and L. ivanovii differ in pathogenicity. In contrast to L. monocytogenes, which affects a wide range of animal species including humans and birds, L. ivanovii is almost exclusively associated with infections in ruminants, particularly sheep. Furthermore, L. ivanovii causes abortion, enteritis and neonatal septicaemias, but not meningo-encephalitis, which is the hallmark of L. monocytogenes infection in these animals (> 90% of cases) (Vázquez-Boland et al., 2001b). In the mouse model, L. ivanovii appears to be less virulent than L. monocytogenes and proliferates predominantly in the liver, in contrast to L. monocytogenes, which efficiently colonizes both the liver and spleen (Rocourt et al., 1983; Hof and Hefner, 1988). Cell type-related differences in susceptibility to L. monocytogenes and L. ivanovii are also observed in tissue culture models (unpubl. obs. from our laboratory). L. ivanovii therefore provides a unique model to study the molecular mechanisms of bacterial host, tissue and cell tropism.
In our search for species-specific factors that could explain the differences in virulence between the two pathogenic Listeria spp., we previously identified in L. ivanovii the smcL gene, encoding a neutral sphingomyelinase (González-Zorn et al., 1999; Openshaw et al., 2005). We showed that SmcL is responsible for the distinctive phenotype of L. ivanovii on sheep blood agar (SBA), characterized by a strong bizonal haemolysis and a typical synergistic haemolytic effect in the presence of Rhodococcus equi cholesterol oxidase (the so-called ‘CAMP-like’ reaction). We also determined that SmcL is a phagosome-disrupting factor and that its loss attenuates L. ivanovii virulence in the mouse and impairs intracellular proliferation in bovine MDBK epithelial cells. Finally, we showed that the 5′ end of the smcL gene lies contiguous to i-inlFE (González-Zorn et al., 1999), a locus also required for full virulence in mice (Engelbrecht et al., 1998b) (see below).
We report here that smcL and the adjacent i-inlFE locus form part of a 22 kb, species-specific pathogenicity island, LIPI-2, encoding eight additional internalins including two InlB homologues. LIPI-2 is inserted into a tRNA gene and, yet perfectly conserved in all L. ivanovii isolates, is genetically unstable in vitro, a large section of it being spontaneously deleted together with a portion of the flanking core listerial genome.
A spontaneous deletion affecting the smcL–i-inlFE locus of L. ivanovii
The weakly haemolytic derivative 44/2 was isolated by screening a bank of Tn1545 insertion mutants of L. ivanovii ATCC 19119 on SBA (Kreft et al., 1989). 44/2 did not produce sphingomyelinase and the cognate bizonal haemolysis and R. equi CAMP-like reaction (Fig. 1, Table 1), suggesting that the transposon had affected smcL or a regulatory locus required for its expression. However, sequencing of the Tn1545 insertion site of 44/2 did not provide any obvious explanation for the mutant phenotype, suggesting that it was caused by an unrelated mutation. SDS-PAGE analysis of the culture supernatant showed that 44/2 did not secrete a major 27 kDa protein (Kreft et al., 1989) (Table 1). N-terminal sequencing of this protein revealed similarities with internalins. Its gene, i-inlE, was identified by polymerase chain reaction (PCR) using degenerate oligonucleotide primers, and i-inlF by inverse PCR (Engelbrecht et al., 1998b). Various combinations of primers specific for the smcL–i-inlFE locus yielded no PCR products in 44/2, suggesting that the phenotype of the mutant was due to a spontaneous deletion affecting these genes.
Table 1. Differential properties of weakly haemolytic mutants of L. ivanovii (see Fig. 1).
. Activity of the product of the PrfA-regulated plcB gene, determined on egg yolk agar (+, opaque precipitation halo around the colonies).
. Ivanolysin O, the product of the PrfA-dependent hly gene of L. ivanovii; detected by SDS-PAGE Western immunoblotting of culture supernatant proteins with an anti-ILO polyclonal antibody (see Fig. 1C).
. Phenotypic marker of L. ivanovii sphingomyelinase, encoded by smcL, a gene that is not regulated by PrfA. Determined on SBA using R. equi as indicator strain (Ripio et al., 1995) (+, shovel-shaped synergistic haemolysis reaction; –, smaller, onion-shaped reaction).
. Sphingomyelinase activity, mean value of three experiments. The weak background SMase activity in 44/2, GD1 and GD3 is likely due to the wide-substrate-range phospholipase C, PlcB, which is also active on sphingomyelin (Geoffroy et al., 1991).
. The excreted internalin, i-InlE, detected by SDS-PAGE of L. ivanovii culture supernatant proteins (see Fig. 1B).
To determine whether the deletion in 44/2 was due to a repeatable recombination event, we attempted to isolate new mutants with the same characteristics. Appropriate dilutions of stationary phase brain–heart infusion (BHI) cultures of L. ivanovii ATCC 19119 were plated onto SBA and weakly haemolytic variants were checked on egg yolk agar to discard prfA mutations (identified by a negative lecithinase reaction due to the lack of expression of PlcB) (see Table 1). After screening c. 5 × 104 colonies, we identified two further mutants phenotypically identical to 44/2, GD1 and GD3, each derived from an independent experiment (Fig. 1, Table 1). These new mutants were also negative in smcL-, i-inlF- and i-inlE-specific PCRs (not shown). Pulsed field gel electrophoresis of chromosomal DNA from 44/2, GD1 and GD3 digested with SmaI showed in the three mutants a specific pattern characterized by the loss of a band of ≈225 kb and the appearance of a new band of ≈ 260 kb (Fig. 2). Southern blot analysis of these digestions using an smcL-specific probe gave no positive signal in the mutants (not shown). These results indicated that the absence of the smcL, i-inlF and i-inlE genes was associated with a same chromosomal rearrangement involving the loss of a DNA fragment containing at least one SmaI site.
Characterization of the deletable chromosomal region
To identify the boundaries of the deletable chromosomal region encompassing the smcL–i-inlFE locus, the known sequence (Engelbrecht et al., 1998b; González-Zorn et al., 1999) was extended in both directions by genome walking in wild-type (wt) L. ivanovii. PCR reactions with oligonucleotide primers derived from the new sequence were regularly performed on GD3 until a product was obtained. The first positive PCR was obtained with primers PI-22/-17, targeting sequences located ≈ 6 kb away from the 3′ end of smcL(Fig. 3A). The excision point was identified by inverse PCR using GD3 DNA and primers PI-16/-17 and by comparing the sequence of the 3115 bp HincII inverse-PCR amplicon with that of the corresponding wt region. This inverse PCR fragment contained ≈ 2 kb of the chromosomal region flanking the deletion on the right. We subsequently identified the right end of the deletable region by sequencing an inverse-PCR RsaI fragment generated with primers ED-1/-2 using wt DNA, and extended this sequence towards i-inlE by primer walking (Fig. 3A). The complete list of oligonucleotide primers used in this study is available as Supplementary material (Table S1).
PCR mapping using selected primer combinations (forward: PI-26, ED-5; reverse, PI-12, ED-4) on 44/2 and GD1, complemented by sequence analysis of the amplicons bridging the junction regions, yielded the same results as in GD3, confirming that the deletion resulted from an identical recombination event in all three mutants that led to the excision of an 17.8 kb fragment (Fig. 3A). The precise excision points were, on the left, 162 bp upstream of the i-inlI start codon, and on the right, 52 bp upstream the stop codon of surF15/nrt (see below). Both were located in A/T-rich stretches very similar in sequence and which could form hairpin structures individually and a 30-bp-long direct repeat together (Fig. 3B). Thus, the deletion seems to occur through a homologous recombination event favoured by the presence of secondary structures that could facilitate the dissociation and crossing over of the nicked DNA strands.
Comparisons with the genome sequences of L. monocytogenes and Listeria innocua (Glaser et al., 2001) revealed two distinct parts in the deletable chromosomal fragment. A 9.8 kb left section was exclusive to L. ivanovii and comprised seven open reading frames (ORFs) encoding three new SE-Inl proteins (i-InlI, i-InlH, i-InlG) plus SmcL, i-InlF, i-InlE and a protein of unknown function, SurF3 (for sphingomyelinase upstream region ORF 3). The 8 kb right section (RS) contained 11 complete ORFs (surF4/ydeI, surF5, surF6/gpmA, surF7, surF8/dbpA, surF9, surF10/nfh, surF11/stp, surF12/adh, surF13/mfr, surF14) and the 3′ region of surF15/nrt. Nine of these ORFs were present in the other Listeria spp. with the same arrangement, indicating that RS belonged to the core listerial genome (Fig. 3A). The main features of the products encoded by the L. ivanovii genomic region characterized in this study are summarized in Tables S2 and S3.
Effect on virulence of the spontaneous deletion and of specific knock-out mutations
The spontaneous GD3 mutant displayed a diminished capacity to survive in the liver of sublethally infected mice (dose: wt = 0.6 × 106, mutant = 6.0 × 106; bacterial loads 24 h post infection: wt = 1.6 ± 1.2 × 105, mutant = 6.5 ± 1.2 × 103; 72 h post infection: wt = 1.1 ± 0.4 × 104, mutant = 3.5 ± 2.9 × 101; data in colony forming units, cfu). GD3 also showed a significantly increased mouse LD50 (5.9 ± 2.2 × 107 versus 7.8 ± 1.5 × 106 for the wt; P = 0.0038). Differences in LD50 were more pronounced using a chick embryo test (1.8 ± 0.4 × 103 for the deletion mutant versus 6.6 ± 4.0 × 101 for wt; P = 0.032; data using 44/2 mutant).
Virulence experiments were also conducted in sheep, the natural host of L. ivanovii (see Experimental procedures). Four of the six lambs infected with wt bacteria (group 1), in contrast to only one for the GD3-infected animals (group 2), developed acute, fatal septicaemia. All infected lambs showed fever and prostration during the first 48 h but these symptoms were more conspicuous and lasted longer in the surviving animals from group 1 (Table 2). L. ivanovii was recovered from all the organ samples from group 1 but from none of the samples from group 2 (except in the case of the lamb that developed terminal septicaemia). Microscopic lesions, especially obvious in the liver parenchyma and myocardium, were consistently found but with differing patterns in the two groups (Table 2). Survivors from group 1 showed massive white cell infiltration and the typical pyogranulomes associated with listerial infection which, in the liver, appeared macroscopically as discrete subcapsular necrotic foci 2–3 mm in diameter (Fig. 4). There was no such multifocal pyogranulomatous reaction in the survivors from group 2 but only a diffuse inflammatory infiltration (Fig. 4), suggesting that the granulomes had resolved earlier and/or that GD3 elicited a weaker cellular response than wt bacteria.
Table 2. Summary of clinical, microbiological and histopathological findings in sheep.
Group 1 (wt)
Group 2 (GD3 mutant)
. Fever, loss of appetite and prostration (+++, present in all animals and observed up to 120 h post infection in survivors; +, less conspicuous and lasting up to 48 h post infection). Mean rectal temperatures between 48 and 120 h post inoculation: non-infected animals = 39.44 ± 0.17°C; group 2 = 39.77 ± 0.15°C (P = 0.144); group 1 = 40.87 ± 0.36°C (P = 0.002 versus non-infected animals; P = 0.005 versus group 2).
. 12 h (n = 1), 24 h (n = 2) and 48 h (n = 1) post inoculation.
. 24 h post inoculation.
. Animal that developed fatal septicaemia.
. As observed in survivor animals. Lesions present in all infected animals included generalized inflammatory infiltration and lymphoid tissue hyperplasia. Significant lesions absent from non-infected animals.
. Macroscopic lesions (subcapsular necrotic foci) only evident in group 1 animals.
. Macroscopic lesions (white infiltration areas in the myocardium) present in group 1 and 2 animals.
The above in vivo effects correlated with a markedly impaired intracellular proliferation capacity of the GD3 mutant in vitro in bovine MDBK cells (Fig. 5A). L. ivanovii infection induces apoptosis in MDBK cells (unpubl. obs. from our laboratory), and this response was also almost completely abolished in GD3 (Fig. 5B). There were, however, no significant differences in internalization between the wt and the GDR mutant (Fig. 5C).
To determine whether the core-genome housekeeping genes from the spontaneously deletable fragment contributed to the virulence defect seen in GD3, we constructed an ATCC 19119 mutant lacking the 8 kb RS region (Fig. 3A; see Experimental procedures). This mutant (ΔRS) was tested, together with two mutants in genes from the L. ivanovii-specific section of the deletable fragment (smcL and i-inlFE; see Table 3), in the MDBK infection model (Fig. 5). The ΔRS mutation did not affect significantly the intracellular proliferation and apoptosis induction capacities. In contrast, the two other mutants were substantially affected in intracellular proliferation (smcL) and apoptogenic activity (smcL and i-inlFE) (Fig. 5A and B).
Table 3. Bacterial strains and plasmids used in this study.
Strain or plasmid
. Special Listeria Culture Collection (H.P.R. Seeliger).
. American Type Culture Collection.
. Collection de l’Institut Pasteur, Paris, France.
. Strain collection of the Bacterial Molecular Pathogenesis group.
Recombinogenic plasmid used in the construction of the ΔRS mutant (pLSV1 inserted with a PCR fragment containing surF3-ydeI and nrt-cfr target sequences)
Thus (i) the spontaneous genomic deletion in GD3 is associated with substantial virulence attenuation, and (ii) this attenuation is likely due to the loss of the L. ivanovii-specific, not the core-genome RS sequences present in the deletable fragment.
LIPI-2, a L. ivanovii-specific pathogenicity island: structure and PrfA regulation
Five more inl genes were found upstream of i-inlI, beyond the left deletion end-point (Fig. 3A): three of these encoded SE-Inls similar to i-InlF/E (i-inlJ, i-inlK and i-inlL), the two other encoded LA-Inls homologous to L. monocytogenes InlB (i-inlB1 and i-inlB2), the first internalins of this type to be identified in L. ivanovii. Upstream of i-inlB2 we found an ORF encoding an orthologue of a conserved hypothetical protein with similarity to phosphodiesterases designated YsnB in Bacillus subtilis, Lmo1240 in L. monocytogenes EGDe and Lin1203 in L. innocua. lmo1240 and lin1203 lie near the orthologue of the first non-L. ivanovii-specific ORF in the RS region of the deletable fragment, surF4, which encodes another conserved hypothetical protein designated YdeI in B. subtilis, Lmo1242 in L. monocytogenes EGDe and Lin1206 in L. innocua (see below and Fig. 8). These findings indicated that the 22 kb chromosomal region encompassed by the ysnB (lmo1240/lin1203)- and ydeI (lmo1241/lin1206)-homologous genes in L. ivanovii constitutes a species-specific chromosomal island (Fig. 3A). As 11 of the 12 genes contained in this island encoded proteins with a known (i.e. SmcL, i-InlF and i-InlE; Engelbrecht et al., 1998b; González-Zorn et al., 1999; Fig. 5A and B) or putative (the rest of the i–Inl proteins) association with virulence, we named it ‘Listeriapathogenicity island 2′ (LIPI-2).
The boundaries of LIPI-2 were identified by aligning the sequenced L. ivanovii region with the lmo1240–1242 region from the L. monocytogenes EGDe genome (Glaser et al., 2001). The sequence homology ended abruptly 263 bp downstream of the stop codon of ysnB and immediately downstream of the stop codon of ydeI. Analyses of the junction regions did not reveal significant direct repeats or signatures of mobile genetic elements. However, we found an arginine tRNA gene (tRNAarg) downstream of ysnB, 79 bp away from the putative LIPI-2 insertion site.
The genetic organization of LIPI-2 is schematized in Fig. 3A. All the genes of the island are transcribed in the same orientation except smcL. Palindromic structures that could function as transcription terminators are present downstream of i-inlK, i-inlB1, i-inlG/smcL and surF3. We had previously shown that the expression of i-inlF and i-inlE is dependent on PrfA (Engelbrecht et al., 1998b), and PrfA boxes (the 14 bp palindromic recognition sequences for PrfA in PrfA-regulated promoters; consensus: TTAACANNTGTTAA; Kreft and Vázquez-Boland, 2001) similar to those present in front of i-inlF, i-inlE and other members of the L. ivanovii PrfA regulon preceded at the appropriate distance all the other LIPI-2 inl genes, with the exception of i-inlB1 (Table S3). Semi-quantitative reverse transcription (RT)-PCR analyses using RNA extracted from wt L. ivanovii and an isogenic ΔprfA derivative (Table 3) confirmed that LIPI-2 genes with PrfA-boxes were regulated by PrfA whereas those lacking the PrfA target sequence (i.e. i-inlB1, smcL and surF3) were not (Fig. 6).
Analysis of the LIPI-2-encoded internalins
The structural features of the 10 members of the internalin family encoded by LIPI-2 are schematically represented in Fig. 7. The LA-Inls, i-InlB1 and i-InlB2, of 1078 and 897 amino acids, respectively, display the typical modular architecture of this type of internalins (Schubert et al., 2001): (i) an N-terminal signal sequence, (ii) a short ‘cap’ domain corresponding to the tip of the mature protein in the crystal structure, (iii) a (large) LRR domain (13 and 18 repeats respectively), (iv) a conserved central inter-repeat region (IR) structurally related to the immunoglobulin (Ig)-like domains present in antibodies and numerous eukaryotic cell-surface proteins, and (v) a C-terminal Cwa domain. i-InlB1 and i-InlB2 do not contain the cell wall-sorting motif (LPXTG) present in most LA-Inls (Dramsi et al., 1997; Raffelsbauer et al., 1998) and many other surface-associated proteins from Gram positive bacteria, which via sortase A mediates the covalent linkage of the protein to the peptidoglycan (Cabanes et al., 2002); instead, like InlB from L. monocytogenes, i-InlB1 and i-InlB2 display a Cwa domain characterized by the presence of ≈ 160-residue repeats comprising each two ≈ 80-amino acids modules starting with the dipeptide sequence GW (Fig. 7). These GW modules mediate a new mechanism of protein attachment to the bacterial surface that uses lipoteichoic acid as ligand (Jonquières et al., 1999). The Cwa domain of i-InlB2 is structurally similar to that of L. monocytogenes InlB, as both have a relatively small number of GW modules (four and three respectively). In contrast, the i-InlB1 Cwa domain is more similar to that of the autolysin/adhesin Ami (Milohanic et al., 2001), both carrying eight GW modules spanning a 645-amino-acid region with an overall sequence identity of 54% (Fig. 7).
The remaining eight LIPI-2-encoded internalins, i-InlE, -F, -G, -H, -I, -J, -K and -L, are SE-Inls. They all exhibit extensive sequence similarity with one another and with other members of this group of Inls previously described in L. ivanovii (i.e. those from the i-inlDC locus; Engelbrecht et al., 1998a) and L. monocytogenes (e.g. the prototype SE-Inl, InlC; Engelbrecht et al., 1996). Structurally, they are very similar to the LA-Inls except that they lack a Cwa domain and have fewer LRR units (two to five), being thus of a significantly smaller size (230–320 amino acids, 22.0–43.6 kDa) (Fig. 7). The only exception is i-InlG, which is substantially larger (615 amino acids, 65.8 kDa) and carries 10 LRRs distributed in two separate regions each with five repeat units (Fig. 7). This unusual SE-Inl is a tandem fusion of two SE-Inls, i-InlG1 and i-InlG2 (313 and 302 amino acids respectively). The fusion occurred between the C-terminus of i-InlG1 and the N-terminus of i-InlG2, with traces of the original signal sequence still being recognizable in the latter, charged or more polar residues replacing hydrophobic or less polar amino acids (Fig. 7). Data on the sequence similarity between L. ivanovii SE-Inls are available as Supplementary material (Table S4).
Conservation of LIPI-2 in L. ivanovii
LIPI-2 encodes very similar proteins that contain internal repeats and hence its DNA sequence could be prone to recombination events. This possibility raised the question of whether LIPI-2 is subject to interstrain variation. LIPI-2 was PCR-mapped in seven L. ivanovii isolates from different geographic origins, including the ssp. londoniensis (Boerlin et al., 1992) (PAM nos. 10, 19, 24, 40, 55, 209 and 709; Table 3). Combinations of two kinds of oligonucleotide primer pairs were used: (i) intragenic, to determine presence or absence of individual LIPI-2 genes; and (ii) ‘bridging’ (with intragenic and intergenic target sequences), to provide information on the size of the chromosomal region and on the gene arrangement. The primers used, the mapping strategy and a summary of the results obtained is available as Supplementary material (Table S1 and Fig. S4).
Amplicons of the expected size were produced with the intragenic primers and the different combinations of bridging primers in all cases (Fig. S4) except ssp. londoniensis, for which the i-inlB2 PCR was negative and the corresponding bridging PCRs gave a smaller product consistent with the absence of this gene. For some strains, in particular ssp. londoniensis, a low yield (see Fig. S4) or even no PCR product was obtained with particular primers, possibly reflecting the existence of strain-specific sequence variability in the target DNA. In these cases the structure of the region could always be resolved using alternative primer combinations.
The intriguing architecture of i-inlG, i.e. an almost perfect tail-head in-frame fusion of two SE-inl genes (see above), led us to ask whether this gene was genuine or was the consequence of a fortuitous rearrangement specific to ATCC 19119. The i-inlG1-G2 junction region from ssp. ivanovii strains PAM 24 and PAM 55 and the ssp. londoniensis isolate PAM 709 was PCR-amplified with primers G-1/G-2 (Table S1). Sequencing of the resulting fragments identified the same in-frame translational fusion, even in the ssp. londoniensis isolate in which significant sequence variability was observed (a multiple sequence alignment of this region is available as Supplementary material; Fig. S5).
These results indicated that, except for the absence of i-inlB2 in the genetically divergent ssp. londoniensis, LIPI-2 is perfectly conserved in L. ivanovii isolates.
Structure of the ysnB-ydeI region in other Listeria spp.
LIPI-2 is absent from the genomes of L. monocytogenes EGDe and L. innocua (Glaser et al., 2001), their corresponding ysnB-ydeI intergenic regions just containing, respectively, one (lmo1241) and two (lin1204 and the lmo1241 orthologue, lin1205) LIPI-2-unrelated genes (Fig. 8). L. ivanovii belongs, however, to a different line of descent of the genus together with the non-pathogenic species Listeria seeligeri and Listeria welshimeri (Collins et al., 1991; Sallen et al., 1996; Schmid et al., 2005). The configuration of the region was therefore studied in the latter two Listeria spp. by PCR mapping. A serovar 4b strain (PAM 14; Table 3), from a phylogenetic subdivision of L. monocytogenes substantially divergent from that of serovar 1/2a to which the sequenced strain EGDe belongs, was also included in the analyses.
Primers that amplified ysnB (YS-3/YS-4) and ydeI (PI-83/−81) from L. ivanovii, L. innocua and L. monocytogenes EGDe generated a PCR product of the expected size in L. seeligeri, L. welshimeri and L. monocytogenes PAM 14. The positions and relative orientations of ysnB and ydeI with respect to adjacent genes of the core genome were checked by PCR (not shown). ysnB-ydeI-bridging PCRs (primers YS-3/PI-81) did not yield products in these bacteria, suggesting that the target DNA region was possibly too large to be amplified. A PCR targeting lmo1241/lin1205 adjacent to ydeI (primers M-1/M-2) yielded the expected 470 bp product in L. monocytogenes PAM 14 but not L. seeligeri or L. welshimeri (Fig. 8, Table S1).
In L. monocytogenes PAM 14, a lmo1241/lin1205-ysnB-bridging PCR (primers YS-3/M-2) yielded an amplicon of 1.3 kb, as with EGDe and L. innocua; in contrast, a ysnB-lmo1241/lin1205-bridging PCR (primers M-1/PI-81) did not yield the 1.8 kb fragment obtained in EGDe but one of ≈ 5 kb as in L. innocua. This PAM 14 DNA fragment encoded a non-internalin LRR-, LPXTG-containing surface protein homologous to Lin1204. Analysis of the recently published genome sequence of L. monocytogenes 4b (Nelson et al., 2004) confirmed that the ysnB-ydeI region of this serovar is similar to that from L. innocua, containing an orthologue of lin1204 (LMOf2365-1254) between ysnB and the lmo1241/lin1205 homologue. However, between ysnB and the lin1204 homologue, the L. monocytogenes 4b genome contains three additional ORFs (LMOf2365-1251, 1252 and 1253) encoding small hypothetical proteins. These ORFs are unique to this serovar except for a small partial duplication of one of them (LMOf2365-1253), located further downstream of the corresponding gene, which is also present at the same relative position in L. innocua (Fig. 8). Interestingly, a 252 bp DNA segment corresponding to the 3′ end of the integrase gene (int) from L. monocytogenes bacteriophage PSA (Zimmer et al., 2003), interrupted with premature stop codons and frameshifts and with the downstream attB integration sequence (AATCCCTCTCAGGACG) overlapping the last-3′ nucleotides of the tRNAarg gene near ysnB, is present in L. monocytogenes 4b (Fig. 8). A careful search, including systematic six-frame protein blast scanning, did not identify prophage sequences in the corresponding regions of L. innocua, L. monocytogenes EGDe or L. ivanovii.
Finally, the adaptor genome walking technique was used to clone fragments downstream of ysnB in L. seeligeri and L. welshimeri. As well as the tRNAarg gene, this approach revealed an ORF encoding a surface protein different from Lmo1241/Lin1205 but similar (63–64% identity) to Lmo0435/Lin0557 in L. seeligeri, and two adjacent ORFs encoding a transcriptional regulator of the TetR-/AcrR-family and a protein similar to the Escherichia coli SugE transmembrane chaperone in L. welshimeri. No LIPI-2-related sequences were detected in L. seeligeri or L. welshimeri by PCR using different combinations of the mapping primers (see above).
Thus, the genomic region flanked by ysnB-tRNAarg and ydeI has a different genetic structure in different Listeria spp. and even different serovars of L. monocytogenes.
Differences in pathogenicity between genetically closely related bacteria often rely on the presence (or absence) of specific genomic islands, known as pathogenicity islands (PAIs), that encode specific virulence functions (Hacker and Kaper, 2000; Ochman et al., 2000). Here we describe a new Listeria PAI, LIPI-2, which is specific to L. ivanovii, a second pathogenic species of the genus that differs from L. monocytogenes in host and target organ tropism. This 22 kb PAI was discovered while characterizing a spontaneous chromosomal deletion affecting smcL and i-inlFE, two contiguous virulence loci previously identified in our laboratories (Engelbrecht et al., 1998b; González-Zorn et al., 1999). LIPI-2 contains eight new inl genes and a further gene encoding a protein of unknown function, SurF3. The deletion, which occurs at an estimated frequency of ≈10−4, eliminates the right half of LIPI-2 plus 8 kb of the flanking chromosomal region (Fig. 3A).
The deletion was associated with a significant loss of virulence both in in vitro and in vivo infection models, including experimental infections in sheep. In MDBK cells, internalization was not affected, suggesting that the virulence defect is not due to loss of invasiveness – consistent with the deletion not affecting the LIPI-2 LA-Inls, i-InlB1 and –B2, which are highly similar to the L. monocytogenes invasin InlB. The deletion was not associated with any in vitro growth defect (not shown) but the intracellular proliferation capacity in MDBK cells was significantly impaired, indicating that factors specifically contributing to intracellular survival and/or cytosolic replication were absent from mutant bacteria. The capacity to induce apoptosis was also completely lost in the spontaneous deletion mutants. An smcL knock-out mutant showed very similar behaviour, suggesting that much of the virulence defect seen in the MDBK infection model is due to the absence of the sphingomyelinase. Interestingly, an in-frame i-inlFE deletion, although not affecting intracellular survival/proliferation, also resulted in substantially reduced apoptosis. Thus, several LIPI-2 products appear to contribute to the apoptogenic activity displayed by L. ivanovii in MDBK cells. This is the first time that internalin family members are associated with apoptosis, and also the first clue as to the possible role of SE-Inls in virulence – modulation of host cell responses. An involvement of genes from the flanking DNA also affected by the deletion is unlikely as the ΔRS mutant lacking this 8 kb chromosomal region did not show any significant virulence defect in MDBK cells. This is consistent with this region being present in the non-pathogenic L. innocua (thus obviously belonging to the core listerial genome) and encoding housekeeping/metabolic functions not known to be involved in virulence. These housekeeping genes are clearly nonessential as the ΔRS mutation did not affect the growth of L. ivanovii in rich medium (not shown).
LIPI-2 bears a number of the hallmarks of PAIs typically found in Gram negative bacteria, i.e. large size, genetic instability and a clear association with virulence (Hacker and Kaper, 2000). Interestingly, as for most Gram negative PAIs (Hou, 1999; Ochman et al., 2000), but uniquely among the group of low-G + C Gram positive bacteria, in which a few examples of large PAIs have recently been described also (Brown et al., 2001; Novick et al., 2001; Shankar et al., 2002;), the LIPI-2 insertion point is within a tRNA locus. tRNA genes are typical insertion sites for bacteriophages or integrative plasmids, and PAIs often carry mobility genes (or remnants thereof) and are flanked by direct repeats reminiscent of those generated upon integration of mobile genetic elements (Cheetham and Katz, 1995; Hacker et al., 1997; Williams, 2002). In fact, some PAIs seem to retain the ability to be mobilized by their own specific recombinases (Hacker and Kaper, 2000) or are even transferable via helper bacteriophages (Novick, 2003). Most frequently, however, PAIs become stabilized in the host genome and progressively loose their mobilization machinery. No flanking direct repeats or remnants of mobility genes are present in LIPI-2, suggesting a long evolution in L. ivanovii.
Deletion frequencies similar to LIPI-2 have been reported for other PAIs, such as the 17 kb excision involving the cytolysin and esp loci from the 150 kb PAI of Enterococcus faecalis (Shankar et al., 2002) or the deletions affecting the HPI locus from Yersinia (Buchrieser et al., 1998), PAIs I and II of uropathogenic E. coli (Hacker et al., 1997), and the she PAI of Shigella flexneri (Rajakumar et al., 1997). Such rearrangements, which often involve the direct repeats located at both ends of the island (Buchrieser et al., 1998; Middendorf et al., 2001), have been seen as a general mechanism of bacterial virulence modulation (Blum et al., 1994; Shankar et al., 2002). This notion, however, does not seem to fit well with the irreversible character of the deletions. Alternatively, they may simply reflect the instability inherent to any evolutionary ‘new’ DNA combination carrying direct repeats in the absence of in vivo selective pressures for conservation. Our data support this view as the deletion that affects LIPI-2 is apparently due to a spontaneous homologous recombination process involving what appears to be a fortuitous crossover point formed by two 30-bp-long, AT-rich directly repeated sequences, one provided by the core genome and the other by the island (Fig. 3B).
Interestingly, the different Listeria spp. and even serotypes within the same species all carry a different gene complement between the ysnB-tRNAarg and ydeI loci that flank LIPI-2 in L. ivanovii. This suggests that the ysnB-ydeI intergenic region is a ‘hot spot’ for genome diversification via integration of horizontally mobilized DNA. Indeed, a degenerate fragment from the L. monocytogenes bacteriophage PSA was identified at this site in L. monocytogenes serovar 4b, overlapping the last 16 3′ nucleotides of the tRNAarg locus as typically found in other bacteriophage and PAI integration sites (Hou, 1999; Williams, 2002). It is therefore likely that a bacteriophage originally mediated the integration of LIPI-2 into the listerial core genome. A tRNA gene was also found adjacent to another SE-inl locus previously identified in L. ivanovii, i-inlCD (Engelbrecht et al., 1998b), suggesting that inl genes spread in Listeria primarily by transducing bacteriophages.
The smcL gene is transcribed in the opposite direction to the rest of LIPI-2 genes, differs from the surrounding inl genes in that it is not regulated by PrfA (González-Zorn et al., 1999 and Fig. 6), and is flanked by sequences that form a 42-bp-stem hairpin structure and a semi-overlapping perfect 14 bp direct repeat, which could be the target for a specific recombinase (Fig. 3C). In contrast to inl genes, which belong to a multigene family that has specifically evolved and is widespread in Listeria (Glaser et al., 2001), smcL is unique to L. ivanovii within the genus and has close homologues in phylogenetially related bacteria (56% amino acid sequence identity with Bacillus cereus/anthracis and Staphylococcus aureus neutral sphingomyelinases; González-Zorn et al., 1999; Openshaw et al., 2005). These features suggest that LIPI-2 evolved in a stepwise manner by insertion of a horizontally acquired smcL gene into a pre-existing inl locus.
The expressional subordination of most LIPI-2 genes to the LIPI-1-encoded virulence regulator, PrfA (Kreft and Vázquez-Boland, 2001), suggests that LIPI-2 is a more recent development than the ‘primordial intracellular life cassette’ LIPI-1 in the evolutionary history of Listeria. Indeed, the current model postulates that all Listeria spp. derived from a common ancestor bearing LIPI-1 and that the non-pathogenic species arose by subsequent excision (L. innocua and L. welshimeri) or functional corruption (L. seeligeri) of this ancestral PAI (Vázquez-Boland et al., 2001a; Schmid et al., 2005). The ssp. londoniensis carries, at the same chromosomal position, a divergent copy of LIPI-2 (Fig. S5), indicating that LIPI-2 evolved in the primitive L. ivanovii before subspeciation.
The high level of conservation of LIPI-2 suggests that this element fulfils an important role in L. ivanovii. It is therefore likely that its gain (or loss by some primitive Listeria lineages) had a significant impact on the organism's infectious ecology and, conceivably, in the configuration of the contemporary L. ivanovii. The panoply of internalins carried by LIPI-2 may well be responsible, at least in part, for the specific pathogenic properties of L. ivanovii. Indeed, specific sequence features in the exposed face of the LRR domain of internalins may mediate the recognition of host-, tissue- or cell-specific receptors, as has been shown for InlA and InlB (Marino et al., 1999; Schubert et al., 2002; Cossart et al., 2003). The specific interaction of InlA with E-cadherin appears to be crucial for the tropism of L. monocytogenes for human epithelial cells (Lecuit et al., 1999), and it has been suggested that the diversity of internalins in the different Listeria spp. and subtypes within L. monocytogenes could be at the basis of their differences in virulence and pathogenicity (Cabanes et al., 2002; Nelson et al., 2004). SmcL may also play a significant role in the pathogenic tropism of L. ivanovii, as suggested by its selective lytic activity towards sphingomyelin-rich membranes (González-Zorn et al., 1999; 2000). Finally, our results indicate that LIPI-2 products are involved in apoptosis induction, a property that may contribute to limiting the spread of infection by favouring antigen presentation and the development of a cellular immune response (Weinrauch and Zychlinsky, 1999; Gao and Kwaik, 2000). It is tempting to speculate that the apoptosis caused by L. ivanovii may account for the inability of this species to cause meningo-encephalitis, a process that, as shown for L. monocytogenes, requires a prolonged exposure of the blood–brain barrier to bacteria continually released into the bloodstream from active infectious foci (reviewed in Vázquez-Boland et al., 2001b). Work is in progress in our laboratory to elucidate the role of LIPI-2 and its products in L. ivanovii virulence.
Bacterial strains, plasmids, media and culture conditions
The bacteria and plasmids used in this work are listed in Table 3. Listeria and E. coli strains were cultured in BHI (Difco) and Luria–Bertani (LB) media respectively, with appropriate antibiotics as required.
General DNA techniques
Restriction enzymes were used according to the manufacturer's instructions (Amersham). Plasmid DNA was extracted from E. coli using the Plasmid Purification kit from Qiagen. Chromosomal DNA from Listeria was extracted and purified as described elsewhere (Ripio et al., 1997). PCR was carried out using Appligene, Ecogen or Biotools DNA polymerases for gene detection/mapping purposes, the Expand High Fidelity system (Roche) when the DNA product had to be sequenced (including inverse-PCR), and the Expand Long Template PCR system (Roche) for generating long fragments. The reaction mixes contained 100 ng of DNA template, 200 µM dNTPs, 0.25 µg oligonucleotide primers, 2.5 mM MgCl2, the suitable amount of polymerase buffer and 1 U of polymerase per kb in a 25 µl volume. The standard amplification programme was: 3 min at 94°C; 30 cycles of 15 s at 95°C, 30–60 s at 48–58°C, 1–3 min at 72°C; and 3 min at 72°C. PCR products were purified with the PCR purification kit from Qiagen. Southern blot hybridizations were carried out as previously described (González-Zorn et al., 1999) using as probe digoxigenin (DIG-11-dUTP, Roche)-labelled PCR fragments.
Pulsed-field gel electrophoresis
Genomic DNA was prepared in agarose plugs as previously described (Brosch et al., 1991). After digestion with low-frequency cutting endonucleases, the macrorestriction fragments were resolved with a CHEF-DRIII apparatus (Bio-Rad) using an electric field of 6 V cm−1 and an angle of 12. Migration of the DNA fragments was performed at 15°C in 0.9% agarose gels and 0.5 × Tris-borate-EDTA buffer. Pulse times varied according to the size of the fragments to be resolved.
Genome walking, DNA sequencing and sequence analysis
Genome walking was performed by successive inverse PCRs. Chromosomal DNA samples (10 µg) were digested in 50 µl with a range of restriction enzymes (high-frequency cutting: Taql, RsaI, DraI; medium- to low-frequency cutting: HindIII, HhaI, XbaI, HphI, HincII) at 37°C overnight and precipitated with Colour paint co-precipitant (Novagen) according to the manufacturer's instructions. The precipitate was resuspended and ligated with 8 U of T4 ligase at 16°C overnight in 100 µl. The ligation mixture was precipitated as above and resuspended in 30 µl H2O. Three microlitres samples were used for PCRs, which were performed as described above except that the elongation step was 3–4 min at 68°C. Alternatively, genome walking was carried out using the Universal GenomeWalker Kit (Clontech) following the procedures recommended by the manufacturer. DNA sequencing was performed at the Unidad de Secuenciación Automatizada de DNA, Universidad Complutense de Madrid, using an Applied Biosystems 377 apparatus. All sequences were determined on both strands and confirmed using small PCR products amplified directly from genomic DNA preparations. DNA sequences were assembled using DNA Strider 1.2.1 and homology searches were performed with blast at the National Center for Biotechnology Information, Bethesda, USA (http://www.ncbi.nlm.nih.gov). Multiple alignments were generated with the pileup programme of GCG Wisconsin package 9.0 and displayed graphycally using boxshade 3.2.1 except otherwise stated.
Total RNA was extracted from 10 ml L. ivanovii liquid cultures grown at 37°C to an OD600 = 1.0 according to the protocol of Dietrich et al. (2000) using the EZNA Bacterial RNA Kit (PeqLab). RNA was treated with DNase kit (Ambion) and complete removal of DNA was checked by PCR. This DNA-free RNA was used for reverse transcription (RT) with Moloney Murine Leukaemia Virus reverse transcriptase (M-MulV-RT, Stratagene). Semi-quantitative RT-PCR was carried out as previously described (Ermolaeva et al., 2004) using the First-Strand cDNA synthesis kit (Amersham), serial dilutions of the RNA preparation, and the constitutively expressed sod gene as internal control for linearity. Equal volumes of the PCR mixtures were separated by electrophoresis in 1% agarose gels in TAE containing ethidium bromide and those giving the best discrimination within the linear range were considered for analysis.
ΔRS mutant construction
Targeted excision of the 8 kb RS of the spontaneously deletable L. ivanovii chromosomal fragment was carried out in ATCC 19119 by allelic exchange via homologous recombination. Oligonucleotide primer pairs LIP5/LIP6R and LIP7R/LIP8 (Table S1) were used to PCR-amplify two DNA fragments corresponding to the 3′ end of surF3 plus surF3-ydeI intergenic region (573 bp) and the portion of the nrt gene just preceding the spontaneous deletion point (1534 bp) respectively (see Fig. 3A). Primers LIP6R and LIP7R carried overhanging complementary extensions (underlined in Table S1) that were used to fuse the two fragments by ‘recombinant PCR’ (Pogulis et al., 1996). The resulting chimeric PCR product was inserted into pTOPO, sequenced on both strands and transferred to pLSV1 to generate pLS58 (Table 3). This recombinogenic plasmid was introduced into ATCC 19119 by electroporation and transformants were selected by incubation at 30°C on BHI agar plates containing 5 µg ml−1 erythromycin (BHI-erm). Plasmid integration was selected by subculturing for 2 days at pLSV1 vector non-permissive temperature (41.5°C) in BHI-erm and checked using appropriate oligonucleotide primers derived from vector, insert and chromosomal sequences. The second crossover leading to plasmid excision (loss of erythromycin resistance marker) was identified by replica plating in BHI and BHI-erm after regularly subculturing the single-crossover recombinant at 41.5°C without selective pressure. Erythromycin-sensitive colonies were screened by PCR for the presence of the 8 kb RS deletion using oligonucleotides LIPUP and LIPDO external to the recombinogenic DNA fragment inserted into pLS58 (Table S1), and checked for the absence of the RS sequences using ad hoc primer combinations. The ΔRS mutant was phenotypically indistinguishable from its wt parent strain.
Haemolysin, PC-PLC and sphingomyelinase assays
Haemolysin and PC-PLC (lecithinase) assays were performed on Columbia SBA plates (Biomérieux) and BHI-egg yolk agar plates respectively, as described by Ripio et al. (1995). Sphingomyelinase activity was determined on early stationary phase culture supernatants as described in González-Zorn et al. (1999).
Listeria ivanovii culture supernatant proteins were isolated from 10 ml stationary cultures as described previously (Engelbrecht et al., 1996) except that precipitation was carried overnight on ice with 10% TCA. SDS-PAGE and Western immunoblotting were performed as described in Ermolaeva et al. (2004).
Cell-culture infection assays
MDBK (Madin–Darby bovine kidney) epithelial-like cells, obtained from ATCC, were cultured in 24 well plastic dishes (Costar) at 37°C under 5% CO2 in DMEM medium (Biowittaker) without antibiotics and containing 10% fetal bovine serum (FBS; Biowittaker) until 80% confluence. To prepare the inocula, the bacteria were grown in BHI until OD600 = 1.5, collected by centrifugation, washed three times in PBS, and stored in vials at −80°C in 20% glycerol PBS. Except otherwise stated, bacteria were added to cells at a multiplicity of infection of 50:1. Infected monolayers were centrifuged immediately for 3 min at 172 g at room temperature, incubated for 30 min at 37°C, washed twice with Dulbecco's PBS (Gibco) to remove non-adherent bacteria, and incubated during 1 h in DMEM containing 100 µg ml−1 gentamicin (Sigma) to kill extracellular bacteria. Cells were then lysed with 1% Triton X-100 (Sigma) in PBS and the lysates were plated for bacterial counting. Invasion was determined 1 h after gentamicin addition (t = 0). For bacterial intracellular survival/growth assays, immediately after t = 0 the cell culture medium was changed to contain 25 µg ml−1 gentamicin and at the specified time points the cells were washed, lysed and the number of cfu determined.
Apoptosis was quantified by flow cytofluorimetry using Annexin V-FITC staining (Beckman Coulter). Briefly, infected MDBK monolayers were washed with PBS and treated for 2 min at 37°C with Trypsin-EDTA solution (Biowhittaker). The detached cell suspension was brought up to 2 ml with 1600 µl of DMEM−10% FBS and split into two 1 ml alliquots. The suspension typically contained ≈ 3 × 106 cells ml−1. One of the alliquots was processed for apoptosis determination as follows. Cells were washed twice in 5 ml ice-cold binding buffer and added 1 µl of Annexin V-FITC solution and 5 µl of propidium iodide (to exclude necrotic/late apoptotic cells). After a 10 min incubation on ice, cells were washed and resuspended in 1 ml of cold binding buffer and analysed in an Epics XL-MCL FACS machine (Coulter), recording 10 000 events per sample. Cells treated with 200 nM staurosporine (Sigma) for 4 h were used as positive control. The other alliquot was processed for cfu determination as described above. Results were expressed in normalized apoptosis units, calculated dividing the percentage of apoptotic cells by the no. of cfu per well and multiplying by 10n, n being the minimum order of magnitude necessary to eliminate the negative exponent.
Mouse experiments were carried out as previously described (Chico-Calero et al., 2002) using 22–25 g specific pathogen-free BALB/c females (Harlan Ibérica). Lethality tests in chicken embryos were carried out by chorio-allantoic inoculation as previously described (Terplan and Steinmeyer, 1989). LD50s were calculated using the method of Reed and Muench (1938). For experiments in sheep, 1-month-old conventionally raised lambs of local breedings, obtained from farms with no previous history of listeriosis, were divided into two groups: group 1 (n = 3) was infected with wt L. ivanovii and group 2 (n = 3) with GD3 mutant. A third group was used as uninfected control. Two experiments were carried out, one using animals of ‘Churra’ breed and the other of ‘Talaverana’ breed. The infectious doses ranged between 3.5 × 1011 and 4 × 1011 cfu. Animals were euthanized 6 days after infection or earlier if showing prodromic signs of fatal outcome. For microbiological analyses, samples of liver and spleen tissue were taken aseptically, homogenized using sterile saline solution and plated onto SBA. For histopathological analyses, samples of tissue from different organs were fixed in formaldehyde and processed for haematoxylin/eosin staining at the diagnostic facility of the Departamento de Patología Animal II, Facultad de Veterinaria, Universidad Complutense de Madrid (Spain). All animals were kept in isolators and fed autoclaved food and water ad libitum.
Statistical significance of virulence data was assessed by a two-tailed unpaired Student's t-test or the Wilcoxon/Kruskal–Wallis test using JMP 3.1.5. software (http://www.jmp.com).
We thank J. Rocourt and C. Jacquet for hosting one of us (G.D.-B) in their laboratory and for expert guidance in pulsed-field gel electrophoresis, P. Garrido for excellent technical assistance and help with the sequencing, M. Bailey for help and advice with flow cytofluorimetry, A. Rodríguez for the histopathological analyses, S. Steinmeyer and M. Loessner for the chick embryo tests, and R. Schlesinger for the 44/2 mutant. This work was supported by the Spanish Ministry for Education and Science (Grants BMC2000-0553, SAF2001-1403 and SAF2004-04317), the European Commission (QLG2-CT-1999-00932), and in part by the Wellcome Trust (programme Grant 074020). G.D.-B and S.M.-A. were supported by fellowships from the Madrid Autonomous Government and the Free State of Bavaria respectively.