INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Agrobacterium tumefaciens is a gram-negative soil bacterium and the causative agent of crown gall disease, a condition affecting primarily dicotyledonous plant species (reviewed in references 18 and 62). The pathogen incites production of the characteristic tumor through the transfer of a piece of DNA (T-DNA) from the Ti (tumor-inducing) plasmid into susceptible plant cells, with subsequent integration into the host genome. The T-DNA contains genes that direct the biosynthesis of auxin and cytokinin in infected cells (1, 57), resulting in uncontrolled cell division leading to production of the characteristic tumor. The T-DNA also contains genes for the biosynthesis of unique compounds called opines, which the bacterium can utilize as carbon and nitrogen sources (39).

Successful transfer of the T-DNA is dependent on the coordinated expression of virulence (vir) genes located on the Ti plasmid but separate from the T-DNA. Expression of vir genes occurs in response to certain phenolic compounds released from wounded plants (54). This expression is augmented by monosaccharides (5, 52) and an acidic pH (38), which are characteristic of plant wound sites. Expression of vir genes requires virA and virG, which encode members of the family of two-component regulatory systems (60). VirA is a inner membrane-associated histidine protein kinase which autophosphorylates in response to the environmental signals (19, 29). The phosphate moiety is subsequently transferred to the aspartate residue of VirG, which in turn activates transcription from promoters containing a specific 12-bp sequence called the vir box, present in the promoters of all vir genes (30, 44). In addition to virA and virG, other chromosomally encoded genes that have been shown to modulate virulence gene expression either directly or indirectly have been identified in A. tumefaciens (12, 15, 20, 61).

The use of Escherichia coli as a heterologous host in which to study the regulation of A. tumefaciens virulence genes and the mechanism of T-DNA transfer would constitute an ideal model system given the degree of characterization at both biochemical and genetic levels. However, all previous attempts to reconstitute vir gene expression in E. coli have been unsuccessful, possibly because of the presence of unidentified regulatory genes in A. tumefaciens required for vir induction and/or the possibility that E. coli contains a specific repressor(s) of vir gene induction.

A characteristic of vir gene promoters is the absence of a strong 35 sequence (10). DNase I footprinting studies have shown that VirG protects a region extending into where the 35 consensus sequence should be (30, 44). It has been suggested that binding of VirG may functionally replace the 35 consensus sequence, allowing transcription to occur (10, 30). This situation is similar to that for class II catabolite gene activator protein (CAP)-dependent promoters, in which the CAP binding site overlaps the 35 sequence. Studies have demonstrated that transcription at class II CAP-dependent promoters requires interaction between CAP and the  subunit (RpoA) of RNA polymerase (RNAP) (42, 49, 64). Many transcriptional factors are known to require interaction with the  subunit of RNAP, including FNR (59), GalR (9), MarA (24), MerR (7), MetR (23), OxyR (56), OmpR (21), Rob (26), SoxS (25), and TyrR (34) in E. coli and the 29 P4 protein in Bacillus subtilis (38a). A recent study demonstrated that RpoA from E. coli can interact with BvgA from Bordetella pertussis, but conversely, rpoA from B. pertussis cannot interact with CAP from E. coli (54a). Analysis of the  subunit from E. coli indicates the presence of two independent domains, the N-terminal domain and the C-terminal domain (21, 27, 63). The N-terminal domain is involved in the assembly of the core polymerase, while the C-terminal domain is involved in interaction with certain transcriptional regulators (7, 9, 21, 23, 24, 25, 26, 32, 34, 47, 56, 59). Recently, interaction between the N-terminal domain and CAP at class II CAP-dependent promoters has been demonstrated (42, 49).

In this report, we show that both rpoA and virG from A. tumefaciens are required for transcriptional activation of a vir promoter (virp) in E. coli. Evidence that VirG interacts with RpoA from A. tumefaciens but not with RpoA from E. coli is also presented. This observation suggests that in order for successful transcription of vir genes to occur, specific interactions between the A. tumefaciens  subunit of RNAP and VirG are required.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, plasmids and media. All strains and plasmids used or constructed in this study are listed in Table 1. Bacterial strains were grown in either LB medium (40), mannitol glutamate Luria salts (MG/L) medium (58), or induction medium containing 1% glucose (61) at 28°C. Induction medium was used for attempts to reconstitute wild-type vir gene induction, while MG/L and LB media were used for strains containing virG(Con) (i.e., pSY215 and pLG2). When appropriate, the medium was supplemented with ampicillin (100 µg/ml), gentamicin (20 µg/ml), kanamycin (50 µg/ml), and tetracycline (20 µg/ml) for E. coli and carbenicillin (100 µg/ml), gentamicin (100 µg/ml), kanamycin (100 µg/ml), and tetracycline (5 µg/ml) for A. tumefaciens. For determinations of -galactosidase activity, 5-bromo-4-chloro-3-indolyl--D-galactopyranoside (X-Gal) and isopropyl--D-thiogalactoside (IPTG) were included at final concentrations of 75 µg/ml and 1 mM, respectively. Acetosyringone (Sigma) and glucose were included, when necessary, at 200 µM and 1%, respectively. For induction assays, E. coli MC4100 containing constructs indicated in Table 1 were grown to stationary phase in 5 ml of the appropriate medium, containing antibiotics as required. These overnight cultures were used to inoculate 125-ml flasks containing 30 ml of the same medium, using 0.5 ml as an inoculum. The cultures were incubated at 28°C with shaking for 16 h and assayed for -galactosidase activity according to the method of Miller (40). For reconstitution of virulence gene expression in E. coli, plasmid constructs pSY204, pLG2, pGP159, and pSL107 were introduced into MC4100 by electroporation. Construct pHO98 containing lac-driven A. tumefaciens rpoA was then introduced into these strains and initially screened on LB medium (pSY204 and pLG2) or induction medium (pGP159 and pSL107) containing appropriate antibiotics, 1 mM IPTG, 75 mg of X-Gal per ml, and 200 µM acetosyringone. Induction assays for these strains were carried out as described above.

Isolation and subcloning of the rpoA locus. A cosmid clone library of A. tumefaciens A136, a derivative of strain C58 lacking the Ti plasmid, in pVK102 was described previously (8). The cosmid clones were transformed into E. coli MC4100 containing plasmid pSY215. Transformants were initially screened for the development of a blue color on induction medium containing X-Gal, indicating expression of the virBp::lacZ fusion. A cosmid clone, designated pBK2, was isolated by this procedure and used for subsequent subcloning attempts, as outlined in Fig. 1.

Overexpression and purification of proteins. For overproduction of the RpoA proteins, rpoA genes from E. coli and A. tumefaciens were PCR amplified from the MC4100 chromosome and plasmid pBX4.1 (Fig. 1), respectively, using primers designed to contain a BamHI restriction site (E. coli, 5'-CCA AAG AGA GGA TCC AAT GCA GGG-3' and 5'-CCT TAA CCT GGG ATC CGG TTA CTC G-3'; A. tumefaciens, 5'-GGA AGG ATC CAA GAT GAT TCA GAA GA-3' and 5'-CCT GGAA TCC TGC AGA TGA CTT ATC TG-3'). The PCR products were initially cloned into PCR2.1TOPO (Invitrogen, Inc.) and then subcloned into pQE vectors (Qiagen) to generate pECH4 and pZL-2, containing N-terminally His-tagged fusions to RpoA of E. coli and A. tumefaciens, respectively.

View larger version (10K): [in this window] [in a new window]   FIG. 1.   Physical and genetic map of recombinant plasmids carrying the rpoA gene from A. tumefaciens. Abbreviations: B, BamHI; E, EcoRI; P, PstI; Sa, Sau3A; St, StuI, X, XhoI. Locations and directions of the lacp for constructs pPS1.3R and pPS1.3 are indicated by arrows.

In vitro transcription assays. For analysis of abortive initiation, 2 pmol of recombinant RNAP holoenzyme containing either E. coli or A. tumefaciens RpoA was incubated with 10 pmol of template DNA in 10 µl of transcription buffer (40 mM Tris-HCl [pH 7.9], 40 mM KCl, 10 mM MgCl₂) for 10 min at 23°C in the presence or absence of the VirG(Con) protein (2 pmol). As the template, a 130-bp EcoRI-HindIII fragment of plasmid pAA121 containing the galP1 promoter (63 to +44) or a 380-bp PCR product of plasmid pSM243cd (53) containing virBp (248 to +81) was used (4, 10). Abortive initiation reactions were started by the addition of 0.5 mM initiating dinucleotide (CpA for the galP1 promoter and ApU for virBp promoter) and 0.5 µM [-³²P]UTP (galP1) or [-³²P]-ATP (virBp) (3,000 Ci/mmol). After 25 min of incubation at 37°C, reactions were terminated by the addition of an equal volume of urea loading buffer. The reaction products were resolved on urea-polyacrylamide (20% polyacrylamide, 19:1 acrylamide:bisacrylamide) gels and visualized by autoradiography.

Mobility shift assays. Electrophoretic mobility shift DNA binding assays were carried out with a PCR-amplified virBp labeled with [-³²P]dCTP. PCR mixtures included 20 µCi of [-³²P]dCTP, plasmid pSM243cd as the template, primer 1 (5'-TTC CAC GGT GAC GCA TCG AAT G-3'), and primer 2 (5'-CCC CGA TCT CTT AAA CAT ACC TTA TCT CC-3'). Unincorporated nucleotides were removed by using a Wizard PCR Preps DNA purification system kit from Promega. Mobility shift reaction mixtures contained 1,600 cpm of ³²P-labeled virBp, 50 mM KCl, 20 mM Tris-HCl (pH 7.0), 10 mM MgCl₂, 1 mM dithiothreitol, 10% glycerol, and 20 µg of herring sperm DNA per ml. Where indicated, 1.5 µM VirG, which was sufficient to saturate all VirG binding sites on the promoter, and 150 to 600 nM (final concentration) His-RpoA were added. Reaction mixtures were incubated for 30 min at 22°C, loaded onto 6% polyacrylamide-10% glycerol vertical slab gels in 0.5× Tris-borate-EDTA buffer (46), and size fractionated by electrophoresis at 20 V/cm at 4°C for 2 h. Following electrophoresis, the gels were dried and autoradiographed overnight at room temperature.

Nucleotide sequence accession number. The rpoA sequence has been deposited into GenBank (accession no. AF111855).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Identification of an A. tumefaciens gene that is required for vir gene expression in E. coli. To reconstitute vir gene induction in E. coli, plasmid construct pGP159 containing virA and virG under the control of their native promoters and a lacZ reporter gene fused downstream of virBp (virBp::lacZ fusion) was introduced into E. coli MC4100. The resulting strain failed to activate transcription of the fusion in response to acetosyringone. The inability of pGP159 to activate transcription may have been due to a lack of expression of the virA and virG genes, which are under the control of their native promoters. To address this possibility, we used plasmid pSL107, which contains virA and virG under the control of the lac promoter (lacp) and virBp::lacZ. Introduction of pSL107 into MC4100 also failed to activate expression of the vir gene fusion in the presence of acetosyringone and IPTG. The observed lack of expression was not due to a lack of expression of virA and/or virG, as both proteins were present at detectable levels by Western blotting (data not shown). Since VirA and VirG were present, we reasoned that the VirA/VirG signal transduction mechanism may not be functional in E. coli. This possibility was addressed through the use of pSY215, which contains a constitutively active virG(Con) under the control of lacp and virBp::lacZ. The virG(Con) is able to activate expression of the virBp::lacZ fusion in A. tumefaciens strains independently of virA and acetosyringone (17, 31, 45). When pSY215 was introduced into MC4100, once again we observed no expression of the virBp::lacZ fusion, which suggested that additional genes from A. tumefaciens may be required or that E. coli may contain specific repressors of vir gene expression.

Characterization of the identified gene. The DNA sequence of the entire 1.3-kb StuI-PstI fragment was obtained from both strands. Analysis of the DNA sequence revealed an open reading frame of 1,008 bp, in the predicted transcriptional direction, sufficient to encode a polypeptide of 336 amino acids. Nucleotide and protein searches of GenBank and SwissProt databases indicated a high degree of similarity to rpoA genes, encoding the  subunit of RNAP. Comparison with rpoA from E. coli indicated 62.15% sequence similarity and 51.4% sequence identity at the amino acid level. Three highly conserved regions can be identified from the distribution of homologous amino acid residues between E. coli and A. tumefaciens rpoA homologues (Fig. 2). One region extends from residues 30 to 51 near the N terminus (20 of 22 identical), with the other two present in the C-terminal domain extending from residues 256 to 270 (13 of 15 identical) and from residues 276 to 315 (30 of 42 identical). A notable difference is the presence of an additional eight residues at the C terminus of the A. tumefaciens RpoA compared to RpoA of E. coli. A potential Shine-Dalgarno sequence, GAAGGT, was found extending from 7 to 12 bp upstream of the proposed ATG initiation codon of rpoA. Analysis of partial DNA sequence obtained from pBKS7.0 and pBKE4.8 indicated the presence of regions upstream and downstream of rpoA with a high degree of sequence similarity to rpsK and rplQ, encoding S11 and L17 ribosomal proteins, respectively, possibly forming an operon structure similar to that of E. coli.

View larger version (51K): [in this window] [in a new window]   FIG. 2.   Amino acid sequence alignment of the A. tumefaciens (A. tum.) and E. coli rpoA gene products. Identical amino acid residues at a given position are marked by asterisks, and conserved substitutions are marked by dots. Gaps introduced to allow optimal alignment are designated by dashes. Regions of highly conserved sequence as described in the text are designated in boldface.

View larger version (17K): [in this window] [in a new window]   FIG. 3.   Assembly of His-RpoA into RNAP and in vitro transcription assays. (A) SDS-polyacrylamide gel electrophoresis analysis of reconstituted RNAP containing E. coli , ', and 70 subunits and His-RpoA from E. coli (E.c.) or A. tumefaciens (A.t.) and stained with Coomassie brilliant blue B. (B) In vitro transcription from virB and galP1 promoters in the absence and presence of VirG(Con). Results of a multiround transcription reaction using reconstituted RNAP containing His-RpoA from either E. coli or A. tumefaciens are shown. Where indicated, 2 pmol of VirG(Con) was added to the reaction mixture.

In vitro transcription of virBp. To confirm the results observed for the in vivo assays involving the rpoA subclones and virG(Con), we carried out in vitro transcription assays. Purified individual components of E. coli RNAP, , ', and ⁷⁰, were mixed with His-RpoA of either E. coli or A. tumefaciens and high-molecular-weight RNAP complexes were purified on a sizing column (see Materials and Methods). As shown in Fig. 3A, both RpoA molecules were able to successfully assemble with the E. coli , ', and ⁷⁰ subunits into complete RNAP holoenzymes. When tested for in vitro transcription, the two RNAP holoenzymes were equally efficient in initiating transcription from a ⁷⁰-dependent E. coli galP1 promoter (Fig. 3B), demonstrating that the hybrid RNAP containing RpoA of A. tumefaciens is a functional enzyme. Furthermore, no significant differences were detected in the amount of the transcript produced in the presence or absence of virG(Con). When virBp was used as a template, the E. coli RNAP could activate low-level transcription, but no difference was evident with or without the VirG(Con) protein. In contrast, the hybrid RNAP containing A. tumefaciens RpoA was able to activate transcription from virBp at low levels, and addition of VirG(Con) increased transcription by four- to fivefold, as measured by quantification of the gel (Fig. 3B). These results confirm the in vivo assays demonstrating that only RNAP containing RpoA of A. tumefaciens is able to efficiently initiate transcription from virBp in a VirG(Con)-dependent manner. The inability of E. coli RNAP to efficiently express the vir fusion even in the presence of VirG(Con) suggests that E. coli RpoA cannot make the required contacts with VirG(Con).

Specific interaction between VirG(Con) and RpoA of A. tumefaciens. Since the C-terminal domain of the E. coli RpoA was known to interact with the A+T-rich UP element of certain promoters (47), we first tested to see if the E. coli and A. tumefaciens RpoA proteins have different affinities for virBp. As shown in Fig. 4 (lanes 3 to 5), A. tumefaciens RpoA was able to shift the mobility of the labeled virBp at the highest concentration used (600 nM), whereas the same concentration of E. coli RpoA did not, suggesting that the RpoA of A. tumefaciens has a higher affinity for virBp than RpoA of E. coli. Similar mobility shift assays were carried out to determine if His-RpoA of E. coli and A. tumefaciens interact differently with VirG(Con) at virBp. Increasing amounts of RpoA were used in combination with a saturating quantity of VirG(Con) for virBp (Fig. 4, lane 2). The concentration of VirG(Con) used was determined through separate mobility shift assays in which increasing amounts of VirG(Con) resulted in two separate shifts in mobility, corresponding to binding of VirG(Con) at the two vir boxes of virBp (data not shown). At the three concentrations of E. coli RpoA used, there was no additional shift in the mobility of the promoter-VirG^con complex (Fig. 4A, lanes 6 to 8). However, when RpoA from A. tumefaciens was used, two separate shifts were observed (Fig. 4B, lanes 6 to 8), which suggests specific interactions between VirG(Con) and RpoA from A. tumefaciens. These results provide further evidence that RpoA from A. tumefaciens may possess a higher affinity for VirG(Con) compared to RpoA from E. coli.

View larger version (64K): [in this window] [in a new window]   FIG. 4.   Protein-protein and protein-DNA interactions between RpoA, VirG(Con), and virBp, determined by electrophoretic mobility shift assays of 32P-labeled virBp containing purified VirG(Con) and His-RpoA from E. coli (A) or A. tumefaciens (B). Lane 1, promoter only; lane 2, promoter and VirG(Con); lane 3, promoter and 150 nM RpoA; lane 4, promoter and 300 nM RpoA; lane 5, promoter and 600 nM RpoA; lane 6, promoter, VirG(Con), and 150 nM RpoA; lane 7, promoter, VirG(Con), and 300 nM RpoA; lane 8, promoter, VirG(Con) and 600 nM RpoA.

Combination of virA-virG and rpoA of A. tumefaciens is not sufficient to reconstitute acetosyringone-mediated vir gene induction in E. coli. To determine if the signaling mechanism, resulting in vir gene activation, can be reconstituted in E. coli, we used wild-type virA and virG in combination with A. tumefaciens rpoA. MC4100 harboring two plasmid constructs, one containing a virBp::lacZ fusion as well as wild-type virA and virG under their native promoters (pGP159), and the other containing a lac-driven A. tumefaciens rpoA gene (pHO98), did not show any significant increase in -galactosidase activity in the presence of acetosyringone (data not shown). The possibility that lack of expression of virA and/or virG accounts for this result was again addressed through the use of pSL107, which contains lac-driven virA-virG and virBp::lacZ. We were still unable to obtain significant expression of the fusion in MC4100 harboring pSL107 and pHO98. Introduction of pHO98 into MC4100(pLG2), containing lac-driven virG(Con) and virBp::lacZ, however, resulted in a significant increase in -galactosidase activity, demonstrating that pHO98 is able to produce a functional RpoA protein. These results suggest that the signal transduction mechanism of VirA and VirG may not be functional in E. coli. Possible explanations for this observation are discussed below.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The purpose of this study was to determine the feasibility of obtaining expression of virulence genes in a heterologous E. coli background. The ability to use E. coli as a heterologous system would provide a valuable tool for studying these processes. We report here the identification of a chromosomally encoded A. tumefaciens rpoA gene and demonstrate that it constitutes one of the required components for expression of a virBp::lacZ gene fusion in a heterologous E. coli background. The rpoA gene of E. coli has been extensively studied, particularly with regard to interactions with transcriptional regulators. This suggested to us that interaction between VirG and RpoA may be required for efficient transcription of virulence genes.

The inability of pGP159 or pSL107 to activate transcription of the virBp::lacZ fusion suggested either that signal transduction between VirA and VirG was not functional or that additional genes were required from A. tumefaciens for activation. The use of pSY215 containing a virG(Con) mutant allowed us to evaluate vir gene expression in a virA-independent manner, eliminating the need for signal transduction. The lack of expression obtained with pSY215 in E. coli, combined with its ability to function in a Ti plasmidless A. tumefaciens strain, suggested that additional A. tumefaciens genes were required for expression. The introduction of pPS1.3 containing lac-driven rpoA into MC4100(pSY215) resulted in a significant (40-fold) increase in transcription of the virBp::lacZ fusion compared to the control vector. Verification that rpoA is required was obtained through the use of subclone pPS1.3R, which did not activate expression of the fusion. This construct is identical to pPS1.3, but the direction of transcription of rpoA is opposite that of the lac promoter. The observation that A. tumefaciens rpoA was able to complement a temperature-sensitive rpoA mutant in E. coli demonstrates an ability to function at essential E. coli promoters. This is evident from the in vitro transcription assay, where the hybrid RNAP was as effective as E. coli RNAP in transcribing a ⁷⁰-dependent galP1 promoter (Fig. 3B). While we were able to significantly increase expression of the virBp::lacZ fusion (Table 1), the level of expression was lower than that in A. tumefaciens (31). The relatively low expression of the virBp::lacZ fusion may have been a consequence of the presence of RNAP containing RpoA of E. coli. To remove possible interference from E. coli RpoA, in vitro transcription assays using reconstituted RNAP holoenzymes containing His-RpoA from either E. coli or A. tumefaciens were carried out. Using purified E. coli , ', and ⁷⁰ subunits, we were able to demonstrate that both of the His-RpoA proteins were able to assemble into multisubunit RNAP holoenzymes (Fig. 3A). The results of the in vitro transcription assays demonstrated that VirG(Con)-dependent transcription of the virBp::lacZ fusion requires RNAP containing A. tumefaciens RpoA, although the two reconstituted holoenzymes exhibited essentially identical activity in transcription from the galP1 promoter, with no significant difference in the presence or absence of VirG(Con) (Fig. 3B). Another possible explanation for the relatively low induction in E. coli may due to the presence of E. coli sigma factors in the RNAP holoenzymes. It is conceivable that E. coli sigma factors have a lower affinity for virBp than sigma factors from A. tumefaciens. Although the vegetative sigma factor from A. tumefaciens has been identified (50), it is unclear whether this or an alternative sigma factor is involved in vir gene transcription.

Previous reports have identified the presence of an UP (upstream) element in certain E. coli promoters which is required for optimal transcription (14, 41, 47). This element extends from 40 to 60 bp upstream of the transcription start site and is highly A+T rich. Interestingly, virBp contains an A+T-rich sequence from 40 to 60 that overlaps the VirG binding sites. Whether this region of the promoter constitutes a true UP element is unknown. From our gel shift assays, the A. tumefaciens RpoA appears to have a higher affinity for virBp than E. coli RpoA (Fig. 4), although the importance of this observation is unclear at this time. The results of the mobility shift assay suggest that E. coli RpoA is unable to bind to VirG(Con) at the virBp. In contrast, A. tumefaciens RpoA appears to exhibit cooperative binding with two distinct shifts in the mobility of the VirG(Con)-virBp complex. Taken together, these results indicate that RNAP containing E. coli RpoA may be unable to interact effectively with VirG(Con) and therefore cannot activate transcription from virBp. Since virBp contains two binding sites for VirG, the presence of two shifts obtained with increasing amounts of A. tumefaciens RpoA may be a consequence of RpoA interacting with VirG at each vir box.

In examining vir gene expression in E. coli, our attempts to reconstitute wild-type virulence gene expression in E. coli were not successful. The use of virA and virG under lacp control means that sufficient levels of VirA and VirG should be present for signal transduction to take place. One possible explanation may be that E. coli is unable to correctly insert VirA into the inner membrane. Alternatively, even though VirA may be inserted into the inner membrane correctly, dimerization of VirA which is required for activity in A. tumefaciens (43) may be defective. A more likely explanation may be that additional genes from A. tumefaciens are required for efficient signal transduction. An unresolved question is the exact mechanism of sensing of phenolic inducers by the VirA/VirG system. The two possible mechanisms involve direct binding of the inducer by VirA or binding by a second receptor which then interacts with VirA. Although genetic evidence supporting direct binding of inducers by VirA has been reported (36, 37), all attempts to demonstrate direct binding by VirA have been unsuccessful. Conversely, there are reports of studies in which binding of phenolic compounds by proteins other than VirA have been detected (13, 35), although there is no evidence to link these proteins with vir gene induction. The search for additional A. tumefaciens genes involved in the signal transduction should be simplified by our finding that VirG(Con)-mediated expression of virulence genes requires RpoA from A. tumefaciens. This work provides the basis for future studies designed to examine vir gene expression as well as the T-DNA transfer process in E. coli.