Background to the Invention Colonization is the first step in the disease process and is often fundamental to bacterial pathogenesis. For many host-adapted bacteria, the site of colonization is the source of infection for further susceptible hosts.

Escherichia coli is a ubiquitous resident of the human gastrointestinal (GI) tract and is transmitted by the feco- oral route. From the GI tract, it can cause enteric, urinary, and systemic disease. Studies on E. coli colonization have identified a variety of bacterial receptors, and their corresponding ligands, that mediate E. coli attachment to, or uptake into, host intestinal cells (see, for example, Gaastra and Svennerholm, Trends Microbiol. 1996; 4: 444-452 and Guerrant et al., J. Infect.

Dis. 1999; 179: 5331-5337).

However, bacterial adhesion and cell entry is only one component of colonization, which is a complex and dynamic sequence of events. To colonize the GI tract, E. coli must survive the low pH conditions of the stomach, the high osmotic pressures in the small bowel, and other innate and acquired defence mechanisms. Once at its site of residency, the bacteria must attach to the mucosal surface and avoid mechanical expulsion by peristalsis. During its transit through, and residency within, the GI tract, the bacterium must also acquire nutrients from the surrounding microenvironment, often competing with the host microbial flora for limited resources.

Strains of E. coli expressing the K1 polysaccharide capsule are the major cause of Gram-negative bacteraemia and meningitis in neonates. Infants often become infected through contact with their mothers or health care workers carrying the bacterium. Furthermore, E. coli K1 isolates

are among the most common cause of ascending urinary tract infection in childhood. In studies of E. coli pathogenesis, host-to-host transmission, GI tract colonisation, systemic disease, and ascending urinary tract infection have all been described. However, colonization is an aspect of E. coli K1 pathogenesis that has not been widely investigated.

Summary of the Invention The present invention is based on the identification of gene sequences within E. coli Kl which may code for proteins that have an important role in colonisation during infection. The genes are therefore important targets for antimicrobial therapy, where disruption of the gene sequences decreases infection of the microbe.

According to the present invention, a peptide (or protein) is encoded by any of the genes defined herein as dgc B, C, D or E from E. coli K1, or a homologue thereof in a Gram negative bacterium, or a functional fragment thereof.

According to a second aspect of the invention, a peptide (or protein) is encoded by any of the genes dgcA, trsE, trsC, fimH, rnr, csgE, fnr, metJ, rpoN, frdA, speA, adh, pgi, treB or emrB, or a homologue thereof in a Gram negative bacterium, or a functional fragment thereof, for therapeutic use, particularly for use in the manufacture of a medicament to treat bacterial infection.

The polynucleotides that encode the peptides (or proteins) may also have therapeutic use.

The peptides (or proteins) may be used as antigens to raise an immune response. Alternatively, they may be used to produce an attenuated microorganism that can be used in a vaccine. For example, an attenuated microorganism may be produced by a deletion in one or more of the genes encoding the peptides (or proteins).

Description of the Invention The present invention is based on discoveries made using signature tagged mutagenesis (STM) (Hensel et al.,

Science, 1995; 269: 400-402) to identify genes required by E. coli K1 for GI colonization.

The gene sequences identified herein are only partial sequences, but have been used to identify the full-length sequence. A skilled person will appreciate how to use the sequences to obtain either the full-length sequence or the encoded product.

A gene as identified herein, or its encoded product, may be used in various ways in antimicrobial therapy. For example, the product may be used as a vaccine, to elicit an immune response. Alternatively, the product may be used to generate a monoclonal or polyclonal antibody, which may have therapeutic or diagnostic use. Techniques for generating antibodies are known to those skilled in the art. The term"antibody"is used herein to refer to whole antibodies, single chain antibodies and antibody fragments, e. g. Fv or Fab fragments.

A microorganism which is attenuated by inactivation of a gene identified herein, may also be used as a vaccine.

The gene may be inactivated by any suitable technique, including a deletion or insertion mutation. It may be desirable to include further attenuating mutations in one or more other genes, to provide a greater degree of certainty that reversion to the wild-type strain cannot occur.

An attenuated microorganism of the invention may also be used as a carrier of another antigen or therapeutic agent. In this way, the microorganisms are useful as delivery vehicles.

A gene or polynucleotide of the invention may also be a useful therapeutic or diagnostic agent. For example, the gene, or functional fragment thereof, may be used in gene therapy. Alternatively, the polynucleotides may be used as probes in a diagnostic test to identify possible pathogenic microorganisms.

A nucleotide sequence identified herein may have homology to sequences in other microorganisms. Homology

refers to either sequence similarity or identity.

Typically, the degree of homology will be greater than 20%, preferably greater than 50%, more preferably greater than 80% and most preferably greater than 90%. Determining the level of homology may be carried out using conventional sequence comparison programmes, for example, the BASTN and BLASTX programmes. The homology is determined over at least 20 nucleotides, preferably 50 nucleotides and most preferably the full sequence identified herein. Similar homology in respect of an amino acid sequence may also be achieved.

The present invention also relates to functional fragments of the peptides and genes. The term"functional fragment"is used to refer to a subunit of a sequence identified herein, which retains either functional activity, i. e. its biological role, or therapeutic activity, i. e. its ability to elicit an immune response.

The invention includes, of course, mutated versions of the genes or encoded products, which have reduced or no biological activity. The mutants are intended to produce an attenuated microorgansm. A nucleotide fragment may be 20 nucleotides, preferably 50 nucleotides and more preferably at least 100 nucleotides, in length. An amino acid fragment may be from 10 amino acids, preferably 20 amino acids, more preferably 40 amino acids and most preferably at least 50 amino acids, in length.

Methods for formulating the proteins (peptides), polynucleotides and microbial cells into suitable vaccines and therapeutic compositions will be apparent to the skilled person, based on conventional formulation technology. Suitable diluents, adjuvants or other excipients may also be added.

The following Example illustrates the invention.

Example Construction and characterization of the mutant library In using STM, it is first necessary to construct a mutant library. The E. coli isolate used in this study

expresses the 018 serotype lipopolysaccharide. This serotype is responsible for a significant proportion of all cases of E. coli bacteraemia, and 018 strains have a propensity to cause invasive infections in experimental models.

E. coli CC118-ypir was transformed with pUTmini-Tn5Km2 containing signature tags (de Lorenzo et al, J. Bacteriol.

1990; 172: 6568-72; Hensel et al, supra). A total of 288 individual transformants were screened to identify a subset of 96 plasmids harboring tags that gave consistent and specific signals. Colony blots of the transformants were probed, with all tags labelled with (32P]-dCTP.

Transformants that failed to give a hybridization signal were discarded. The remainder were then analyzed for the presence of cross-hybridizing tags by probing the blots with tags amplified from subsets of the transformants.

96 uniquely tagged pUTmini-Tn5Km2 derivatives were selected for the construction of the mutant library. These plasmids were transformed separately into the donor strain, S117-Xpir, and used in independent matings with the E. coli K1 recipient, RS228nalR. A total of 2,304 mutants were arrayed into 24 pools of 96 mutants each. To determine the diversity of insertion sites of mini-Tn5 in RS228nalR, DNA from 30 ex-conjugates from three different matings was subjected to Southern analysis. DNA from ex-conjugates was digested with Pstl or EcoRl and Kpnl, and the blots hybridized with the kanamycin resistance cassette in mini- Tn5Km2. The results demonstrated that each mutant arose from the single integration of the transposon at a distinct site with no siblings.

To evaluate the proportion of auxotrophic mutants in the library, 480 mutants were replica plated onto complete and minimal media. Five mutants (approximately 1 % failed to grow on minimal media alone, and these were randomly distributed in the STM pools.

Screening the library for colonization-defective mutants

In preliminary experiments, the total number of bacteria and the number of different mutants in the inoculum were adjusted so that consistent hybridization results were obtained from two animals given the same inoculum. Similar results were seen when dot blots were probed with tags from bacteria recovered from the large bowel two days after inoculation with 5 x 108 cfu, containing 45 to 95 mutants. E. coli K1 was detected in the jejunum of animals within one hour of inoculation, and was found in the large bowel 24 hr later where it persisted for at least five days. From day one onwards, the bacterial load in the large bowel exceeded that in the small bowel by at least an order of magnitude. Animals receiving up to 109 cfu bacteria remained well with no signs of gastrointestinal upset, and gained weight at the same rate as non-inoculated animals; bacteraemia was not detected. There was evidence of extensive intra-litter spread of E. coli K1. The small and large bowel of non- inoculated mothers and siblings housed in the same cage as infants given E. coli K1 become colonized with the bacterium. However, animals tested prior to inoculation, or those kept in separate cages did not carry E. coli K1.

A total of 2,140 mutants were screened for mutants defective in GI colonization. Pairs of animals received 5 x 108 cfu, containing 47 to 95 mutants intragastrically, and bacteria were recovered from the descending colon 48 hr later. In the initial screen, 37 mutants (1. 7%) were present in the inoculum but failed to be recovered from the large bowel of infected rats. All these mutants were re- assessed in animals. Therefore, three pools were assembled containing the mutants identified in the first screen, and the pools re-tested in the infant rat model. Of the 37 original mutants, 19 were consistently attenuated in subsequent examination; all these mutants were prototrophic. This method of evaluation was validated by comparing directly the colonization potential of a selection of mutants against the wild-type bacterium. In

mixed inoculum experiments, nine out of 10 mutants that gave consistently negative hybridization signals were colonization-defective when compared with the wild-type strain. Conversely, two mutants that were not consistently colonization defective when examined in pools, colonized the large bowel at the same level as the wild-type strain.

To determine whether the colonization defective mutants had general growth defects, their growth rates in vitro were compared with the wild-type. Only one of the mutants (rnr) showed reduced growth in shaken liquid media, while two further strains (pgi, fnr) exhibited reduced growth rates in static media.

Characterization of colonization defective mutants The insertion site of mini-Tn5 in all 19 colonization defective mutants was identified. Genomic DNA flanking the insertion sites was amplified, the PCR products cloned, and the nucleotide sequence of the insert determined. To confirm that the DNA sequence flanking the insertion site had been correctly amplified by arbitrary PCR, Southern analysis was performed on eight mutants. Blots wore prepared containing genomic DNA from the mutants and RS228nalR digested with Clal or Pstl, and probed with the corresponding arbitrary PCR product. In each instance the hybridization pattern obtained for the mutant differed from that of the wild-type strain, confirming that sequences flanking the transposon insertion site of mini-Tn5 had been isolated. DNA and protein database searches were undertaken using flanking sequences.

Of the 19 colonization defective mutants, 12 have transposon insertion sites in genes with homologues in the published whole genome sequence of E. coli K12 (Blattner et al., Science, 1997; 277: 1453-1474), and no gene or protein with significant similarity (<101°) was found in database searches for four insertion sites. The mutants fall into five categories that contain transposon insertions in genes encoding: i) cell surface structures, ii) transcriptional regulators, iii) enzymes in metabolic pathways, iv)

proteins with membrane transport functions, and v) proteins of unknown function.

Two mutants (SEQ ID NO. 1 and 2) had disruptions in genes related to trsE and trsC which are involved in lipopolysaccharide (LPS) biosynthesis in the enteric pathogen Yersinia enterocolitica serotype 0: 3 (Skurnik et al. Mol. Microbiol. 1995; 17: 575-594). Neither gene is found in E. coli K12, but both are present in a single locus in Y. enterocolitica that is required for the synthesis of the outer core moiety of LPS. The predicted protein encoded by trsC has similarity to many glycosyl transferases, with the highest level of homology over two predicted functional domains. The derived protein sequence of trsE contains a motif shared by galactosyl and mannosyl transferases in a number of bacteria.

Transposon insertions were also identified in genes involved in the synthesis of type-1 pili and curli, and the sequences are identified as SEQ ID NO. 3 and 4, respectively. Type-1 pili are filamentous surface organelles that bear an adhesin, encoded by fimH, that mediates attachment of bacteria to mannosylated host receptors. FimH is required for adhesion to the bladder mucosal surface via uroplakins (Wu et al., PNAS, 1996; 93: 9630-35 and Mulvey et al., Science, 1998; 282: 1494- 1497), and for the pathogenesis of lower urinary tract infections. Curli are cell surface structures that are encoded by two divergently transcribed operons, csgBA and csgDEFG (Hammar et al., Mol. Microbiol., 1995; 18: 661-670).

The genes in the latter operon are required for the expression of curli, though their specific functions are not known except for csgD, a transcriptional activator for the csgBA operon. The gene mutation was found to be within csgE.

Three mutants had insertions in transcriptional activators. The first is a mutant with an insertion in fnr (fumarate nitrate reduction), was colonization defective and had a markedly reduced growth rate in static culture.

The fnr product modulates the expression of over 70 genes whose products are involved in anaerobic metabolism (Stewart, J. Bacteriol., 1982; 151: 1320-1325 and Jones and Gunsalus, J. Bacteriol., 1987; 169: 3340-49). The identified gene sequence is shown as SEQ ID NO. 5. The second was mutated in metJ, which is a repressor of the methionine synthesis pathway. The third mutant, rpoN, encodes the alternative sigma factor O59 that pleiotropic functions in the cell. The identified gene sequences are shown as SEQ ID NOS. 6 and 7, respectively.

Transposon insertions were also found in genes with general metabolic functions, with several expressed under anaerobic conditions. FrdA (SEQ ID NO. 8) encodes a subunit of fumarate reductase (Lohmeier et al., Can. J.

Biochem., 1981; 59: 158-164) which is part of a pathway used as a terminal electron acceptor during anaerobic respiration. The enzyme arginine decarboxylase (encoded by speA) catalyses the first step in the degradation of L- arginine to succinate, and is a component an acid survival system in E. coli that protects the stationary-phase cells in acidic environments (Castanie-Comet et al., J.

Bacteriol., 1999; 181: 3525-3535). This gene sequence is identified as SEQ ID NO. 9. Alcohol dehydrogenase, encoded by adh (SEQ ID NO. 10), is involved in fermentation, reducing acetyl CoA to ethanol, thereby generating two molecules of NAD+ to act as electron acceptors under anaerobic conditions (Gupta and Clark, J. Bacteriol., 1989; 171: 3650-3655). Additionally, an insertion in pgi (SEQ ID NO. 11) was identified, the gene product, glucose phosphate isomerase, catalyses the interconversion of glucose-6- phosphate and fructose-6-phosphate. Loss of pgi affects glycolysis with carbon diverted into the hexose monophosphate shunt, and leads to growth defects in conditions that do not provide glucose-6-phosphate either directly or through catabolism (Vinopal et al., J.

Bacteriol., 1975; 122: 1172-4).

Insertions in genes for two putative membrane- associated transporters were also identified. Under low osmolarity, the di-saccharide trehalose is phosphorylated and taken up by the treB gene (SEQ ID NO. 12) product which is part of the phosphotransferase system (Boos et al., J.

Bacteriol., 1990; 172: 3450-61). EmrB (SEQ ID NO. 13) is a component of the ErmAB efflux pump, one of a family of proteins known as membrane-fusion-proteins that includes AcrAB (Lomovskaya and Lewis, J. Bacteriol., 1992; 177: 2328- 34; and Miller and Sulavik, Mol. Microbiol., 1996 21: 44- 448). This gene was not identified in E. coli K12, but is present in Rickettsia prowazekii.

A mutant was also found to have a disruption in a gene corresponding to the rnr gene (SEQ ID NO. 14) in Shigella (Cheng et al., J. Biol. Chem., 1998; 273: 14077-14080). The mutant has a general growth defect which may be responsible for its attenuation.

Five of the colonization defective mutants have insertions in genes of currently unknown function that were designated dgc (defective in G1 colonisation) A, B, C, D and E, and are shown as SEQ ID NOS 15 to 19, respectively.

For one of these genes, dgcA, a homologue was present in the whole genome sequence of E. coli K12 (Blattner et al., supra), while for the remainder, no related sequences were identified through database searches. The colonization potential of the dgcB strain was reduced compared with the wild-type isolate, but the strain grows equally well in vitro, compared to the wild-type. dgc genes are present in other enteric pathogens To determine whether the four dgc genes absent from E. coli K12 are restricted to E. coli Kl or have related sequences in other enteric pathogens, low stringency Southern analysis was performed. Blots containing DNA from a range of bacterial pathogens (UPEC, EPEC, EHEC, ETEC, enteroaggregative E. coli, Shigella flexneri, Salmonella typhi, Salmonella typhimuiium, Yetsinia enteracolitica and Yersinia pseudotuberculosis) were probed with fragments of

the dgc genes under conditions that allow sequences of 60% identity to be detected. Only a probe generated from dgcD gave a signal from E. coli K1 alone. All the others gave a signal with Shigella flexneri. When dgcB was used as a probe, a strong band was also detected in DNA from uropathogenic E. coli, with weak signals in isolates of EAggEC, EPEC and EHEC. Bands were detected in the same organisms when dgcE was hybridized with blots except that there was no related sequence in EAggEC. Therefore, the dgc genes are not found only in the E. coli isolate used in this study but are widespread among pathogenic strains of E. coli and Shigella. Interestingly, sequences related to the dgc genes are not seen in Salmonella typhi or S. typhimurium indicating that the genes were probably acquired some time after Salmonella diverged from E. coli.

Homologues of trsE and trsC in E. coli K1 are involved in LPS bibsynthesis To determine whether the homologues of trsC and trsE in RS228nalR function in LPS biosynthesis, the mutant strains were subjected to SDS-PAGE analysis. The trsE and trsC strains had altered glycoforms on LPS analysis compared with the wild-type and another colonization defective mutant (pgi). The core portion of the LPS of both the trsC and trsE mutants was affected. The trsC mutant had lower molecular weight core glycoforms, while the trsE mutant structures were of similar size but of altered relative abundance in comparison with the wild- type. The 0 antigen of both the trsC and trsE appeared unaffected by the changes in the core structure. The results indicate that the trs homologues in this isolate of E. coli K1 function in LPS blosynthesis and, as in Y. enteracolitica, the genes are required for the expression of the complete LPS molecule.

Immunohistochemical analysis of GI colonization To characterize further the colonization defect in the mutants, immunohistochemistry was used to determine the fate of E. coli Kl in the GI tract. This allowed the wild

type and individual mutant strains to be followed during colonization, and to compare their distribution. To visualize bacteria, sections were incubated with a monoclonal antibody (mAb) that recognizes the a2-8 linked neuramininc acid structure of the E. coli Kl polysaccharide capsule. No bacteria were visualized in sections from animals receiving phosphate buffered saline (PIBS) alone indicating that the animals were not colonized with E. Coli K1 prior to the start of experiments. Wild-type bacteria (RS228nalR) are evacuated rapidly from the stomach after inoculation, and appear in the lumen of the jejunum 2 hr later. After 24 hr, bacteria are detected throughout the small and the large bowel. By the following day, the density of infection in the large bowel increases, whereas fewer bacteria are detected in the jejunum. The bacteria are predominantly in the lumen of the large bowel in association with faecal matter, and are also present in small numbers in close proximity to the mucosal surface of the bowel. No staining was seen in sections from inoculated animals when incubation with the primary antibody was omitted.

The behaviour of five mutants (trsE, dgcA, dgcB, dgcC, and dgcD) was examined using immunohistochemistry.

The extent of colonization at each time point in the GI tract was quantified by counting the number of bacteria under a microscope at a magnification of 100x. Examination of tissue sections from animals receiving the dgcA mutant showed that, similar to the wild-type strain, the bacterium was present in the jejunum soon after inoculation. However this mutant was not detected in the small or the large bowel at 24 hrs after inoculation. A similar pattern of infection was seen with the trsE mutant. The remaining mutants have distinct patterns of colonization. The dgcB mutant was detected in a similar distribution to the parental strain but at lower density at all locations.

Although the levels of the dgcC mutant are greatly reduced in the small intestine, significant numbers are found in

the descending colon. The dgcD mutant appears to have a specific defect for lower GI colonization; this strain colonizes the small intestine as efficiently as the wild- type but is absent entirely from the colon. The results are shown in Table 1.

Table 1 Location Time post Infecting strain inoculation/hr RS228 dgcA dgcB dgcC dgcD trsE Jejunum 24 24 30 1 27 4 18 0 (6) (2) (10) (6) (6) (0) Descending 48 796 0 112 208 0 0 colon (165) (0) (68) (28.8) (0) (0) In Table 1, the values in brackets represent the standard error of the mean.

Competitive index Competitive index studies were also carried out. In direct competition experiments, mutant (nalR, kanR) and wild-type bacteria (nalR) were grown to mid log in LB broth, equal amounts of bacteria (107 each in 100 ml PBS) were mixed then administered intragastrically to animals.

After 48 hr, animals were euthanised and bacteria recovered by plating dilutions of homogenized large bowel to media containing nal alone. To determine the proportion of wild- type to mutant bacteria in samples, 200 colonies were replica plated to media with or without kan. The colonization potential of each mutant was analyzed in two or more animals, and the results given as an average (Table 2). The competitive index (CI) was calculated as the proportion of mutant to wild-type bacteria recovered from animals divided by the proportion of mutant to wild-type in

the inoculum. A competitive index of less than 1 indicates that the mutant is avirulent.

Table 2 Disrupted gene or Identity/Function in vivo Cl homologue organism Cell Surface Structure trsC 43%, Ye LPS biosynthesis (YE) 0. 038 trsE 48%, Ye LPS biosynthesis (YE) 0. 37 fimH 100%, Ec fimbrial adhesin 0. 21 Transcriptional Regulators fnr 100%, Ec anaerobic metabolism 0. 02 rpoN 100%, Ec alternative sigma factor metJ 100%, Ec methionine synthesis Metabolic adh 100%, Ec fermentation frdA 100%, Ec electron transport chain 0. 30 pgi 100%, Ec glycolysis 0. 06 speA 100%, Ec arginine decarboxylase 0. 37 Membrane Transport treB 100%, Ec trehalose transport emr8 32%, Rp transporter (RP) Function Unknown dgcA 51%, Ec hypothetical protein 0. 079 dgcB 0. 240 dgcC 0. 04 dgcD 0. 09 In Table 2, Ec refers to E. coli ; Ye to Yersinia enterocolitica ; and Rp to Rickettsia prowazekii.