Salmonella-induced enteritis is associated with the induction of an acute intestinal inflammatory response and net fluid secretion into the lumen of infected mucosa. Proteins secreted by the Inv/Spa type III secretion system of Salmonella play a key role in the induction of these responses. We have demonstrated recently that the Inv/Spa-secreted SopB and SopD effector proteins are translocated into eukaryotic cells via a Sipdependent pathway and act in concert to mediate inflammation and fluid secretion in infected ileal mucosa. Mutations of both sopB and sopD significantly reduced, but did not abrogate, the enteropathogenic phenotype. This indicated that other virulence factors are involved in the induction of enteritis. In this work, we characterize SopA, a secreted protein belonging to the family of Sop effectors of Salmonella dublin. We demonstrate that SopA is translocated into eukaryotic cells and provide evidence suggesting that SopA has a role in the induction of enteritis.
Salmonella cause a variety of infections in humans and animals, ranging from self-limiting gastroenteritis to systemic enteric fever. The pathology associated with enteric Salmonella infections is characterized by a disruption of the normal movement of electrolytes and water ( Rout et al., 1974 ) and a large influx of polymorphonuclear leucocytes (PMNs) into the intestinal mucosa and lumen from the underlying vasculature ( Turnbull and Richmond, 1978; Wallis et al., 1989 ). It is becoming apparent that interactions between Salmonella and intestinal epithelial cells is a crucial step in the induction of host responses leading to the onset of enteritis ( Eckman et al., 1993 ; McCormick et al., 1993 ; 1995a,b;Galyov et al., 1997 ).
The initial interaction between Salmonella and intestinal epithelial cells involves many bacterial proteins and is largely dependent on the function of the Inv/Spa type III secretion system ( McCormick et al., 1995b ; Galyov et al., 1997 ; Watson et al., 1998 ). Type III secretion systems are highly conserved in a variety of Gram-negative pathogenic bacteria (for a recent review, see Hueck, 1998). The main function of these systems is the translocation of effector proteins into target eukaryotic cells (for a review, see Cornelis and Wolf-Watz, 1997 ; Galán, 1999) . Several such effector proteins, SopB, SopD, SopE, AvrA and SptP, have been identified in Salmonella to date ( Kaniga et al., 1996 ; Wood et al., 1996 ; Galyov et al., 1997 ; Hardt and Galán, 1997 ; Jones et al., 1998 ). The Sops are Salmonella-secreted proteins, the expression of which is significantly increased in Salmonella sip mutants ( Wood et al., 1996 ). We have shown recently that at least two Sops, SopB and SopD, play important roles in the induction of an inflammatory response and fluid secretion in the infected ileum ( Galyov et al., 1997 ; Jones et al., 1998 ). Inactivation of sopB and sopD substantially reduced the net fluid secretion and PMN influx into intestinal ligated loops infected with Salmonella dublin ( Jones et al., 1998 ). In addition to this, recent identification of a pathogenicity island involved in enteropathogenicity, SPI-5 ( Wood et al., 1998 ), also suggested that the enteropathogenic responses elicited by Salmonella result from the additive contribution of multiple effector proteins. Here, we report the identification of a gene encoding another Sop effector protein of S. dublin, SopA. We present evidence that SopA is translocated into eukaryotic cells to promote the inflammatory response and the fluid secretion into Salmonella-infected intestines. Moreover, we provide data suggesting that SopA, SopB and SopD may have different roles in recruiting neutrophils to the intestinal mucosa.
Identification and mutagenesis of the gene encoding the SopA protein of S. dublin
As a part of our ongoing effort to investigate S. dublin secreted effector proteins further, we isolated the Sop protein filaments produced by a sipB mutant, S. dublin B1, and purified an 80 kDa SopA protein by SDS–PAGE. Our attempt to obtain an N-terminal amino acid sequence of the protein was unsuccessful. To overcome this problem, we hydrolysed the SopA protein with trypsin and purified several peptides. Sequences of several amino-terminal residues were obtained for two such peptides as D-S-P-I-E-F-A-L-P-Q and E-V-L-G-P-V-Q-E-R respectively. To identify the sopA gene, we designed degenerate oligonucleotides corresponding to amino acid sequences from the tryptic peptides (two for each sequence to match coding and non-coding strands of DNA). Two different combinations of these primers were used in polymerase chain reaction (PCR) with chromosomal DNA from S. dublin as a template. No DNA fragments were amplified with A1rev and A2 primers. In contrast, a PCR reaction with A1 and A2rev primers yielded a DNA fragment of approximately 1000 bp. This DNA fragment was cloned into pBluescript, and the resulting plasmid was denoted pP80. The nucleotide sequence of the cloned DNA was then determined and found to contain an open reading frame (ORF). The deduced amino acid sequence of this ORF included sequences identical to those determined from SopA. This suggested that the DNA fragment amplified by PCR with A1 and A2rev and cloned into the pP80 was the fragment of the sopA gene. This DNA fragment was then labelled and used as a probe to detect sopA-specific DNA fragments in different restriction enzyme digests of S. dublin chromosomal DNA by Southern hybridization. Two DNA fragments, an ≈ 3 kb EcoRI–BamHI DNA fragment and an ≈ 2.5 kb PstI–KpnI DNA fragment, that hybridized to the sopA probe were identified and cloned into pBluescript to yield pMW80 and pMW79 respectively ( Fig. 1).
The nucleotide sequence of the relevant parts of the cloned DNA was then obtained. This sequence included a complete ORF capable of encoding a 79.3 kDa protein ( Fig. 2A), which is in good agreement with the molecular mass estimated by SDS–PAGE for SopA. In addition, the deduced amino acid sequence included fragments identical to those obtained by the sequencing of trypsin-digested fragments of SopA. These results suggest that the identified ORF was the sopA gene of S. dublin. The sequence of sopA and its flanking areas has been deposited in GenBank under accession no. AF121227. No significant similarity to the sopA gene product sequence was found among the EMBL entries.
The analysis of the flanking region upstream of sopA revealed the presence of phsA, previously described in Salmonella typhimurium. The phsA gene encodes a component of thiosulphate reductase contributing to anaerobic energy metabolism ( Heinzinger et al., 1995 ). In S. typhimurium, phsA is followed by phsB, phsC and orf4 (a gene homologous to the yeeC gene from Escherichia coli K-12). PCR analysis using phsA–C- and orf4-specific primers confirmed that these genes and their relative order are conserved in S. typhimurium and S. dublin (data not shown; Fig. 1). Downstream of sopA, there is a potential transcriptional terminator composed of an inverted repeat of nine nucleotides followed by a sequence of 48 nucleotides highly homologous to a fragment of the IS1351 from Salmonella enteritidis ( Figs 2B and C). A gene with a high homology to the sbcB gene from E. coli K-12 was found further downstream of sopA ( Figs 1 and 2B). In E. coli K-12, yeeC and sbcB are separated by only a short DNA fragment ( Fig. 1). In addition, E. coli K-12 does not possess genes highly homologous to sopA or phsA–C elsewhere on the chromosome. Thus, sopA and phsA–C are a Salmonella-specific islet. The presence of a truncated transposon sequence immediately downstream of sopA ( Figs 2B and C) indicates that the acquisition of the islet may have occurred via homologous recombination mediated by a transposon. However, it is also possible that the sopA and phsA–C genes were acquired independently, as the G+C content of sopA (44%) is different from that of the phs genes, which is the same as that of the entire Salmonella genome (52%).
To assess the importance of SopA in Salmonella virulence and to investigate its expression, a sopA mutant of S. dublin, S. dublin SA1, was constructed. This mutation did not cause any obvious effect on the growth characteristics of Salmonella in LB medium at 37°C and did not affect the secretion of other proteins (data not shown). We have also constructed a sopA transcomplementing plasmid by cloning a 794-bp-long SmaI–EcoRI DNA fragment from pMW79 into pMW80. This resulted in pMW81, which included the complete sequence of sopA preceded by the DNA fragment containing a sopA upstream area ( Fig. 1).
Comparative analysis of secreted proteins produced by wild-type S. dublin 2229 and S. dublin SA1 strains ( Fig. 3) revealed that the 80 kDa SopA protein secreted by the wild-type S. dublin 2229 was not expressed by the S. dublin SA1 sopA mutant strain. No apparent effect of the mutation on the expression and secretion of other proteins was observed (data not shown). The transcomplementation of the mutation by introducing pMW81 plasmid into S. dublin SA1 resulted in the production and secretion of SopA in elevated amounts ( Fig. 3).
Distribution of sopA in Salmonella serotypes
To investigate the distribution of sopA in different Salmonella serotypes, we performed Southern blot analysis on ClaI restriction digests of chromosomal DNA isolated from a variety of Salmonella serotypes using a sopA-specific DNA probe ( Fig. 1). The sopA-specific sequence is present in most enteropathogenic and host-specific Salmonella serotypes, but it is absent in Salmonella arizonae( Fig. 4). Interestingly, it has been found previously that S. arizonae is not able to elicit transepithelial signalling to PMNs in vitro ( McCormick et al., 1995b ). DNA from Shigella sonnei, Yersinia pseudotuberculosis, enteropathogenic E. coli (EPEC) and enterohaemorrhagic E. coli (EHEC) (one strain of each) did not hybridize with the sopA probe (data not shown), indicating that these bacteria do not have a gene highly homologous to sopA.
Analysis of the role of sopA in invasion of eukaryotic cells
To investigate the involvement of SopA in cell invasion, we assessed the sopA mutant strain, S. dublin SA1, in a number of cell invasion assays. In both T84 and HeLa cells, S. dublin SA1 was recovered in only slightly lower numbers than the wild-type strain. However, the reduction in invasiveness was statistically insignificant and only minor compared with the sipB mutant S. dublin B1. Similar results were obtained in an in vivo intestinal invasion assay using bovine ileal ligated loops (data not shown). These data indicate that SopA, similarly to SopB and SopD ( Galyov et al., 1997 ; Jones et al., 1998 ), has no, or only a minor, role in the invasion process.
The SopA protein is translocated into the target cell via a Sip-dependent pathway
To investigate SopA translocation, we constructed a hybrid plasmid pMJ800, in which a DNA fragment encoding the sopA promoter area and the first 253 codons of SopA are followed by the cya gene. pMJ800 was introduced into wild-type S. dublin 2229, into the sipB mutant, S. dublin B1, and into the spaS mutant, S. dublin SpS, to yield S. dublin 2229(pMJ800), S. dublin B1(pMJ800) and S. dublin SpS(pMJ800) respectively. All these strains were found to produce an active fusion protein, and S. dublin 2229(pMJ800) and S. dublin B1(pMJ800) were shown to secrete a fusion protein of the expected size in vitro (data not shown).
The different S. dublin strains were incubated at 25°C overnight. The cultures were then diluted 10-fold in LB, grown for 1 h at 37°C and used to infect cultured J774.2 macrophage-like cells or HeLa cells in the presence or absence of cytochalasin D, which inhibits bacterial invasion. At appropriate time points after the onset of infection, intracellular adenylate cyclase activity was assessed by measuring cyclic AMP. No significant increase in the levels of intracellular cAMP were detected over a period of 2 h when the wild-type S. dublin 2229 or sipB mutant, S. dublin B1, were used to infect HeLa cells (data not shown). This indicated that infection with either of these strains did not cause an increase in intracellular cAMP levels as a result of endogenous adenylate cyclase activities. In a preliminary experiment, we measured the uptake of different S. dublin strains by the J774.2 cells and found that these strains were all taken up in comparable levels (data not shown). When J774.2 cells were infected with different S. dublin strains carrying the pMJ800 plasmid, only infection with S. dublin 2229(pMJ800) resulted in a steady increase in intracellular cAMP, whereas no increase was observed when S. dublin B1(pMJ800) or S. dublin SpS(pMJ800) were used ( Fig. 5A). Infection of HeLa cells with S. dublin 2229(pMJ800) resulted in a steady increase in intracellular cAMP. A similar increase was also observed in the presence of cytochalasin D ( Fig. 5B). In contrast, no change in intracellular cAMP levels was observed when the sipB mutant, S. dublin B1(pMJ800), was used to infect HeLa cells (data not shown). Together, these data suggest that (i) the SopA–Cya fusion protein is translocated into eukaryotic cells via a Sip-dependent mechanism; and (ii) the translocation can be mediated by extracellular bacteria.
SopA has a role in the induction of enteritis
To study the role of SopA in Salmonella-induced enteritis, we assessed the ability of different S. dublin strains to induce fluid secretion and PMN influxes in ligated ileal loops in two calves. The sopA mutant S. dublin SA1 induced a significantly lower fluid secretory response and PMN influx compared with the wild-type strain in both calves ( Fig. 6), indicating that the sopA mutation reduces the induction of enteritis. We attempted to transcomplement the phenotype of S. dublin SA1 by introducing the pMW81 plasmid into this strain. Although the presence of this plasmid was sufficient to restore the production of SopA ( Fig. 3), transcomplementation of the enteropathogenic response was not observed (data not shown). In order to determine whether the reduced enteropathogenicity of S. dublin SA1 and the absence of transcomplementation was the result of a polar effect caused by the insertion of a suicide plasmid into the sopA gene, we constructed a non-polar deletion mutant of sopA, S. dublin SA2. In addition, we subcloned the sopA-containing DNA fragment from pMW81 (pBluescript-based high-copy-number plasmid) into pBR322 (medium copy number) and into pACYC177 (low copy number) to yield pMW82 and pMW83 respectively. These plasmids were introduced into S. dublin SA2. The analysis of secreted proteins by SDS–PAGE and Western blotting with anti-SopA antibodies revealed that the SopA protein band was not present in S. dublin SA2, but protein production and secretion was restored in the transcomplemented strains (data not shown). The wild-type S. dublin 2229, S. dublin SA2 and the two transcomplemented strains were then assessed for enteropathogenicity in ligated ileal loops. S. dublin SA2 was affected in its ability to induce enteric responses. However, no transcomplementation was observed when pMW82 or pMW83 were used (data not shown). Taken together, these data suggest that the reduced enteropathogenicity observed for S. dublin SA1 and S. dublin SA2 is most probably caused by inactivation of sopA, as both mutants are affected in a similar way. However, we could not confirm this unequivocally as complementation was not achieved by providing sopA in trans.
In order to investigate the activities of SopA further, we compared the effects of mutations in sopA with mutations in other genes encoding secreted proteins in an in vitro PMN transmigration assay. The different S. dublin strains were used to infect the apical surface of T84 cell monolayers, and PMN transmigration was quantified by assaying for PMN myeloperoxidase. Infection of monolayers with the wild-type strain, S. dublin 2229, resulted in a large transmigration of PMNs ( Fig. 7A). In contrast, a non-invasive sipB mutant strain, S. dublin B1, was not able to mediate PMN transmigration. An insertional mutation in sopA abrogated the ability of the mutant strain S. dublin SA1 to induce PMN influx, almost to the same extent as the mutation in sipB ( Fig. 7A). As it has been shown earlier that the SopB and SopD proteins act in a concerted manner to promote the inflammatory responses in Salmonella-infected intestines ( Jones et al., 1998 ), we also included the double sopB/sopD mutant strain, S. dublin SB2SD1, in the in vitro PMN transmigration assay. To our surprise, S. dublin SB2SD1 was as efficient as the wild-type strain in this in vitro model system ( Fig. 7A), suggesting that SopB and SopD are not needed for inducing PMN transmigration in this model system.
In a second experiment, we assessed whether the nature of the mutation in sopA was important for the observed phenotype and whether the sopA mutation could be complemented in trans. The wild-type S. dublin 2229, S. dublin SA1, S. dublin SA2 and S. dublin SA2 transcomplemented with pMW83 were assessed for their ability to induce PMN transmigration. S. dublin SA2 was affected in its ability to induce PMN influx to approximately the same extent as S. dublin SA1. The introduction of pMW83 into S. dublin SA2 resulted in a restoration of the virulent phenotype to a level similar to that of the wild-type strain ( Fig. 7B). Taken together, these data suggest that SopA is an effector protein needed for inducing PMN migration across a model epithelial monolayer.
Infections of intestinal mucosa with enteropathogenic Salmonella strains result in an intense inflammatory response, consisting of PMN migration towards and subsequently across the epithelial monolayer into the lumen ( Turnbull and Richmond, 1978; Wallis et al., 1989 ). This transmigration involves movement through several anatomic compartments, each with their own microenvironment: (i) the well-characterized emigration of PMNs from the microvasculature; (ii) subsequent migration of PMNs across the lamina propria; and (iii) transepithelial migration. It is therefore possible that different steps of PMN movement are regulated by different pathogen-elicited signals. Although the precise mechanisms driving all these steps are not well characterized, it is evident that the interaction of bacteria with epithelial cells results in the production of a set of inflammatory regulators by epithelial cells ( Eckman et al., 1993 ; McCormick et al., 1993 , 1998). The interaction between Salmonella and epithelial cells is a complex process involving many bacterial proteins and is largely dependent on the function of the Inv/Spa type III secretion system. It has been shown recently that several Inv/Spa-secreted proteins are translocated into eukaryotic cells. This translocation and the intracellular activities of the translocated effector proteins appear to be essential for virulence. In this work, we characterized another Inv/Spa-secreted and -translocated effector protein, SopA. We have shown that two different mutations within sopA have a profound effect on the ability of S. dublin to elicit enteropathogenic responses. Both insertional and deletion sopA mutants were significantly less enteropathogenic than the wild-type strain in intestinal ligated loops, an in vivo model system for enteropathogenesis ( Wallis et al., 1995 ). Furthermore, using a transwell in vitro model system ( Madara et al., 1992 ; McCormick et al., 1993 ), we demonstrated that the sopA mutant strains of S. dublin were impaired in their ability to promote PMN movement across the model epithelial monolayer. Thus, the sopA mutant strains were affected in their enteropathogenicity both in vivo and in vitro. Interestingly, this was in contrast to the effects elicited by a double sopB/sopD mutant of S. dublin: despite being affected in enteropathogenicity in vivo, this mutant strain was as efficient as the wild type at recruiting PMNs across the model epithelial monolayer. These data suggest that different Sops may be involved in the control of different stages of PMN influx. It is feasible that SopB and SopD may be needed for the release of signals leading to the attraction of PMNs from the peripheral blood into the lamina propria, whereas SopA may be involved in orchestrating the migration of PMNs across the epithelial monolayer. The precise nature of such signals and the specific contributions of different Sops will require further investigation.
The interaction between Salmonella and the gut epithelia is a complex process requiring the co-ordinated actions of many proteins. This interaction results in a number of responses of even greater complexity. Therefore, it is conceivable that even a minor deviation in the timing of effector protein expression, or in its quantity, may be of great significance in the final effect. Moreover, it has been observed earlier that complicated regulation is an intrinsic feature of the phsA–sopA intergenic region ( Heinzinger et al., 1995 ). We believe that a likely explanation for the difficulties we encountered with transcomplementation experiments in an in vivo model system is to be found in the complexity of the virulence phenotype and the regulation of sopA expression. No transcomplementation of the virulent phenotype was achieved in an in vivo model system of enteropathogenesis when either polar or non-polar sopA mutants were complemented by three different sopA-containing plasmids with different copy numbers. However, a sopA deletion mutant could be transcomplemented in an in vitro model system when a low-copy-number complementing plasmid was used. Although the absence of transcomplementation in vivo does not allow us to make an absolutely definitive conclusion that SopA is an effector of enteropathogenicity, we feel that, on the balance of the evidence obtained, this is a most likely conclusion. We have demonstrated earlier that two Inv/Spa-secreted and -translocated effectors, SopB and SopD, play an important role in the induction of inflammation and fluid secretion in the infected ileum ( Jones et al., 1998 ). The SopA protein described here is likely to be another member of the family of Salmonella secreted effectors delivered inside host cells and acting in concert to induce enteritis. Although the specific contribution of SopA in eliciting pathogenic responses will require further clarification, it is highly likely that SopA translocation and its intracellular activity are of importance.
Bacterial strains and growth conditions
The bacterial strains used in this study are listed in Table 1. Strains were grown in LB broth or on LB agar.
Protein purification and raising of anti-SopA antibodies
Secreted proteins were isolated from insoluble protein filaments deposited by S. dublin FMB1 when grown at 37°C ( Wood et al., 1996 ). The filaments were collected by pipette, washed in phosphate-buffered saline (PBS) and dissolved in SDS–PAGE loading buffer as described previously ( Wood et al., 1996 ). The proteins were separated by SDS–PAGE, blotted onto Immobilon membrane and stained with Coomassie blue. The SopA protein band was excised, and the protein was subjected to sequence analysis.
Polyclonal antibodies against SopA were raised in mice as described previously for the SopB protein ( Galyov et al., 1997 ).
Cloning of the sopA gene of S. dublin and construction of S. dublin mutants
Two pairs of degenerate oligonucleotide primers, A1 (5′-gaYWSNccNatHgaRttYgc-3′) and A1rev (5′-ggNaRNgcRaaYtcDatNgg-3′); A2 (5′-gaRtggYtNggNccNgtNcaRga-3′) and A2rev (5′-cYKtcYtgNacNggNccNaRccaYtc-3′), were designed to match coding and non-coding DNA strands corresponding to amino acid sequences of two internal fragments of SopA obtained by its digestion with trypsin, D-S-P-I-E-F-A-L-P and E-W-L-G-P-V-Q-E-R respectively. A1–A2rev and A1rev–A2 combinations of primers were used in PCR in an attempt to amplify a sopA gene fragment using S. dublin 2229 chromosomal DNA as template. The PCR reaction with A1–A2rev primers resulted in an approximately 1 kb DNA fragment. This DNA fragment was cloned into pBluescript plasmid vector to yield pP80. The cloned DNA fragment was labelled and used as a probe in a Southern blot to detect sopA-specific fragments in different restriction enzyme digests of S. dublin chromosomal DNA. Two DNA fragments, an EcoRI–BamHI DNA fragment (approximately 3 kb) and a PstI–KpnI DNA fragment (approximately 2.5 kb), that hybridized to the probe were identified and cloned into pBluescript to yield pMW80 and pMW79 respectively. Plasmids used for transcomplementation studies were constructed as follows. A 794-bp-long SmaI–EcoRI DNA fragment from pMW79 was cloned into pMW80 to yield pMW81. The DNA fragment cloned into pMW81 was then subcloned into pBR322 and pACYC177 to yield pMW82 and pMW83 respectively.
The sopA mutants of S. dublin were constructed as follows. An internal DNA fragment of sopA was amplified by PCR with the custom oligonucleotides SA1 (5′-tgaagatatctcgaggcgcaattaat-3′) and SA2 (5′-taaggtgtttagatctttcggct-3′) and cloned into the suicide plasmid vector pDM4 ( Milton et al., 1996 ). The resulting plasmid was conjugated from E. coli S17.1 into S. dublin 2229, and Cmr transconjugants were obtained. One of these clones was denoted S. dublin SA1 and chosen for further experiments. The correct insertion of the suicide vector into the sopA gene was confirmed by PCR. A non-polar in-frame deletion mutant of sopA was constructed by a double recombination event using pDM4 suicide plasmid as a vehicle. Four oligonucleotide primers, SPA1 (5′-tgatacaattctcgagacacgtaataatatt-3′), SPA2 (5′-aaggctggactacattagaattccttatc-3′), SPA3 (5′-aattctaatgtagtccagcctcaacc-3′) and SPA4 (5′-cgtaccgaAGATCTcgtaatcgtggaa-3′), were designed to generate a fusion DNA fragment covering the 5′- and 3′-end areas of sopA and carrying a 2343-bp-long deletion of almost the entire gene with the exception of the first and last codons. Two DNA fragments were first amplified by PCR with S. dublin 2229 chromosomal DNA as a template using SPA1–SPA2 and SPA3–SPA4 respectively. These two DNA fragments were then purified and used in a mix as a template in a PCR reaction with SPA1 and SPA4. The resulting DNA fragment was cloned into the suicide plasmid vector pDM4 ( Milton et al., 1996 ) to yield pDELA1. pDELA1 was transferred from E. coli S17.1 into S. dublin 2229 by conjugation, and Cmr transconjugates were obtained. The suicide plasmid was then excised from one of these clones by a second recombination event as described by Milton et al. (1996) . The Cm-sensitive recombinants were obtained and screened by PCR for a mutated allele. Several clones carrying the deletion of expected size were identified. One of these was designated S. dublin SA2 and used in further experiments. S. dublin B1 (sipB), S. dublin SB2 (sopB) and S. dublin SB2SD1 (sopB, sopD) have been described previously ( Wood et al., 1996 ; Galyov et al., 1997 ; Jones et al., 1998 ).
Construction of sopA–cya gene fusion
The plasmid pMJ800 containing sopA–cya was constructed as follows. A DNA fragment including 297 bp upstream of sopA followed by 253 sopA codons was amplified by PCR with SAcyaF (5′-gccgcagatatcccggatcccggtgaa-3′) and SAcyaR (5′-gcagactacatcctgcagtgtcgaataa-3′) primers. This fragment was cloned into the pBScya plasmid ( Jones et al., 1998 ) to form an N-terminal fusion of SopA onto Cya.
Measurement of translocation of SopA–Cya fusion protein in HeLa and J774.2 cells
Bacterial cultures were grown overnight at 25°C with appropriate antibiotics. One hour before the experiment, the bacterial cultures were diluted 10-fold into fresh LB medium without antibiotics and incubated at 37°C. Bacteria were washed and resuspended in 1 ml of cell culture medium (DMEM +10% FCS without antibiotics). A sample of 100 µl of this bacterial suspension (MOI 20:1) was then added to monolayers of HeLa cells or J774.2 cells in 24-well plates. Cytochalasin D (final concentration 1 µg ml−1) was added 20 min before infection. After the appropriate incubation period, the monolayer was washed with ice-cold PBS buffer, and then the cells were lysed using 100 µl of 0.1 M HCl with gentle agitation for 20 min. An aliquot of 500 µl of 0.1 M NaOH/2% Na2CO3 was added to neutralize the pH ( Rozengurt et al., 1981 ). The lysate (50 µl) was assayed directly using the cAMP assay kit (Amersham Biotrak TRK 432) to determine the amount of cAMP present.
Protein concentrations of the cAMP assay samples were determined using the BCA protein assay kit (Pierce) according to the manufacturer's instructions.
Bovine ligated ileal loop assays
The experimental techniques used in the ligated loop assay have been described in detail elsewhere ( Wallis et al., 1995 ; Watson et al., 1995 ). All experiments were performed within the mid-ilea of 28-day-old Friesian bull calves. The inocula were prepared by culturing bacteria overnight in LB broth at 25°C with shaking at 150 r.p.m. The cultures were diluted 1:3 in fresh broth and incubated at 37°C for 2 h. The bacterial cultures were adjusted to approximately 1 × 109 cfu ml−1 by optical density, and 5 ml of culture was injected into each loop.
Approximately 50 ml of blood was removed from the calves 1 h before injection of the loops. PMNs were isolated, labelled with 111In and injected intravenously back into the calves from which they were removed. The secretory response [volume of fluid within a loop/length of loop (ml cm−1)] and the influx of PMNs, as assessed by the magnitude of γ-irradiation emitted from 111In-labelled PMNs within each loop, were recorded 12 h after injection of the loops. The PMN influx ratio was defined as the PMN influx in the test loops/the PMN influx in the negative control loops. The PMN influx ratio in the negative control loops was therefore equal to 1.00.
In vitro PMN transepithelial migration assay
The physiologically directed (basolateral to apical) PMN transepithelial migration assay using T84 epithelial cell monolayers has been described previously ( McCormick et al., 1993 ). Human PMNs were isolated from healthy volunteers, as described elsewhere ( Parkos et al., 1991; McCormick et al., 1993 ). Briefly, PMNs were routinely isolated from anticoagulated (13.2 g of sodium citrate and 11.2 g of dextrose in 500 ml of water, pH 6.5) whole blood (150–500 ml) collected by venepuncture from normal donors of both sexes. The buffy coat is obtained via a 400 g spin at room temperature. Plasma and mononuclear cells are removed by aspiration, and the majority of erythrocytes are removed using a 2% gelatin sedimentation technique, as described previously ( Henson and Oades, 1975). Residual erythrocytes are then removed by gentle lysis in cold NH4Cl lysis buffer. This technique allows for the rapid (90 min) isolation of functionally active PMNs (> 98% as detected by trypan blue exclusion) at greater than 90% purity. PMNs are subsequently suspended in modified HBSS (without Ca2+ and Mg2+, with 10 mM HEPES, pH 7.4) at a concentration of 5 × 107 ml−1.
Before the addition of PMNs to the assay system, inverted cell culture inserts ( Parkos et al., 1991; McCormick et al., 1993 ) were rinsed extensively in HBSS(+) to remove residual serum components. Salmonella strains were prepared by washing twice in HBSS(+) and resuspending at a final concentration of approximately 5 × 109 cfu ml−1. Inverted cell culture inserts were removed from each well and placed in a moist chamber, such that the epithelial apical membrane was oriented upwards. The bacteria (25 µl aliquots or approximately 1.25 × 108 bacteria) were gently distributed onto the apical surface and incubated for 60 min at 37°C. Non-adherent bacteria were removed by washing three times in HBSS(+) buffer. The inverted cell culture inserts were then transferred back into the 24-well tissue tray containing 1.0 ml of HBSS buffer in the lower (apical membrane now oriented upwards and colonized with Salmonella) reservoir and 160 µl in the upper (basolateral interface) reservoir. PMNs (40 µl; 1 × 106) were added to each cell culture in the basolateral bath and incubated for 110 min at 37°C. Positive control transmigration assay was performed by the addition of a chemoattractant [1 nM fMLP (n-formyl methyonyl-leucyl-phenylalanine)] to the opposing apical reservoir. All experiments were performed in a 37°C room to ensure that epithelial monolayers, solutions and plasticware were maintained at a uniform temperature of 37°C.
Transmigration was quantified by assaying for the PMN azurophilic granule marker myeloperoxidase as described previously ( Parkos et al., 1991 ). After each transmigration assay, non-adherent PMNs were extensively washed from the surface of the cell culture inserts, and PMN cell equivalents, estimated from a standard curve, were assessed as the number of PMNs associated with the cell culture inserts and the number that had completely traversed the cell culture insert (i.e. into the basolateral reservoir).
Invasion of cultured eukaryotic cells
The invasiveness of S. dublin strains for HeLa cells was determined by a gentamicin protection assay as described by Wood et al. (1996) .
The S. dublin invasion into T84 intestinal epithelial cell monolayers was assessed as follows. Infection of T84 monolayers was performed by the method described previously ( McCormick et al., 1993 ).
The work at the Institute for Animal Health (IAH) was supported by the Biotechnology and Biological Sciences Research Council (BBSRC) UK, including BBSRC ASD grant 201/A07439 and BFP99326, and the Ministry of Agriculture, Fisheries and Food, UK. B.A.M. was supported by the National Institute of Health grant DK-50989. R.R. was supported by the Swedish Medical Research Council grant (16x-12216).