A truncated H-NS-like protein from enteropathogenic Escherichia coli acts as an H-NS antagonist


  • Helen S. Williamson,

    1. Institute of Structural and Molecular Biology, University of Edinburgh, Darwin Building, The King's Buildings, Mayfield Road, Edinburgh EH9 3JR, UK.
    Search for more papers by this author
  • Andrew Free

    Corresponding author
    1. Institute of Structural and Molecular Biology, University of Edinburgh, Darwin Building, The King's Buildings, Mayfield Road, Edinburgh EH9 3JR, UK.
    Search for more papers by this author

E-mail Andrew.Free@ed.ac.uk; Tel. (+44) 131 650 5545 or (+44) 131 650 5338; Fax (+44) 131 650 8650.


The H-NS nucleoid-associated protein of Escherichia coli is the founder member of a widespread family of gene regulatory proteins which have a bipartite structure, consisting of an N-terminal coiled-coil oligomerization domain and a C-terminal DNA-binding domain. Here we characterize a family of naturally occurring truncated H-NS derivatives lacking the DNA-binding domain, which we term the H-NST family. H-NST proteins are found in large genomic islands in pathogenic E. coli strains, which are absent from the corresponding positions in the E. coli K-12 genome. Detailed analysis of the H-NST proteins from enteropathogenic E. coli (EPEC) and uropathogenic E. coli (UPEC) shows that the EPEC protein (H-NSTEPEC) has a potent anti-H-NS function at the classical H-NS-repressed operon proU. This correlates with the ability of H-NSTEPEC to co-purify with H-NS in vitro, and can be abolished by a mutation of leucine 30 to proline which is predicted to prevent the N-terminal region from forming a coiled-coil structure. In contrast, despite being 90% identical to H-NSTEPEC at the protein level, the UPEC homologue (H-NSTUPEC) has only a weak anti-H-NS activity, correlating with a much-reduced ability to interact with H-NS during column chromatography. A single amino acid difference at residue 16 appears to account for these different properties. The hnsTEPEC gene is transcribed monocistronically and expressed throughout the exponential growth phase in DMEM medium. Our data suggest that a truncated derivative of H-NS encoded by an ancestral mobile DNA element can interact with the endogenous H-NS regulatory network of a bacterial pathogen.


The Escherichia coli H-NS protein is a major component of the bacterial nucleoid and has been implicated as a transcriptional repressor of a diverse array of genes, particularly those involved in environmental adaptation or virulence, via a preferential interaction with intrinsically curved DNA (Dorman, 2004). In recent years the family of known H-NS-like proteins in other Gram-negative bacteria has expanded rapidly, largely resulting from the pace of complete bacterial genome sequencing (Tendeng and Bertin, 2003). Laboratory strains of E. coli contain two H-NS-like proteins, H-NS and StpA, and these related proteins can form heteromeric complexes with each other which have the potential to regulate gene expression in novel ways (reviewed in Dorman et al., 1999). Sequencing studies have also revealed that some bacterial species contain multiple hns-related genes, with four such genes being present in Ralstonia metallidurans and Ralstonia solanacearum (Tendeng and Bertin, 2003), and the proteins encoded by these genes have the potential to possess novel specialist functions and/or interact in novel ways. It has also become apparent that certain strains or pathogenic derivatives of E. coli and its close relatives can encode additional H-NS-like proteins on mobile genetic elements or pathogenicity islands, such as Ler of enteropathogenic and enterohaemorrhagic E. coli (Mellies et al., 1999) and the Sfh protein of Shigella flexneri 2a (Deighan et al., 2003). The Ler protein, encoded by the locus of enterocyte effacement (LEE) pathogenicity island, is particularly interesting as it acts as an anti-H-NS factor which activates expression of other genes on the island (Mellies et al., 1999; Bustamante et al., 2001).

In almost all cases, the H-NS-like proteins which have been studied directly or identified by sequencing have a bipartite structure consisting of a C-terminal DNA-binding domain which is attached to an N-terminal domain predicted to form coiled-coil structures via a flexible amino acid linker (Dorman et al., 1999). Across the whole family the C-terminal domain is the best conserved and contains a core motif which can be used as a signature to identify the protein (Tendeng and Bertin, 2003). NMR studies on this domain reveal a mixed α/β structure (Shindo et al., 1995), although DNA binding appears to be via residues situated in flexible loops between the structured regions (Shindo et al., 1999). However, the isolated C-terminal domain binds DNA only weakly (Shindo et al., 1995), suggesting that oligomerization mediated via the N-terminal domain (Ueguchi et al., 1997) is required for high-affinity binding. Structural studies on this domain (Renzoni et al., 2001; Esposito et al., 2002; Bloch et al., 2003) suggest that it is composed of three α-helices, with the longest (α3) forming the core of the coiled-coil interaction while the other two act to stabilize the structure. There is disagreement between these studies over whether the coiled-coil interaction is parallel or anti-parallel, although a recent crystallographic study on the H-NS homologue VicH from Vibrio cholerae (Cerdan et al., 2003) supports the anti-parallel model (reviewed in Rimsky, 2004). However, there is general agreement that the dimer is the core H-NS module, with higher-order oligomers forming upon interaction with preferential sites for DNA binding.

As predicted from this modular protein organization, expression of the N-terminal domain of H-NS can have a dominant-negative effect on H-NS-dependent repression via interaction of the truncated protein with full-length H-NS, thereby disrupting the DNA binding activity of the intact protein (Williams et al., 1996; Free et al., 2001). This effect has largely been studied at the promoter of the proU operon, a classical H-NS-repressed locus which encodes an osmotically inducible glycine betaine uptake system (Higgins et al., 1988). In contrast, the cryptic bgl operon encoding β-glucoside metabolism, which also requires H-NS for silencing, is not derepressed by expression of truncated H-NS derivatives lacking the DNA-binding domain (Free et al., 2001). Instead, such proteins can still repress bgl so long as either they are expressed to a sufficient level or the paralogous StpA protein provides a compensating DNA-binding domain to interact with the truncated H-NS (Dersch et al., 1994; Ueguchi et al., 1996; Free et al., 1998; 2001). At proU, low-level expression of truncated H-NS has no effect on repression by full-length H-NS and has no repressive effect in its absence (Higgins et al., 1988; Ueguchi et al., 1996), pointing at distinct mechanisms of repression acting at the two promoters (Free et al., 2001). However, these studies are largely irrelevant for the natural in vivo situation as they involve laboratory-constructed truncations of H-NS and/or artificial overexpression.

Here we describe a novel family of H-NS-related proteins, encoded by pathogenic derivatives of E. coli, which lack the DNA-binding domain. These truncated (H-NST) proteins cannot repress bgl expression in the same manner as truncated H-NS, but one of them, from enteropathogenic E. coli (EPEC), can act in a dominant-negative manner on H-NS-dependent repression. This activity does not require high level expression, and correlates with the ability of the protein to interact with H-NS and modulate the formation of H-NS–DNA complexes.


Uropathogenic E. coli strain CFT073 and enteropathogenic E. coli strain E2348/69 contain related genomic islands encoding a truncated H-NS-like protein

As part of a survey of genes encoding H-NS-like proteins in the completed and partially completed genomes of bacterial pathogens, we identified via database searches that the uropathogenic E. coli (UPEC) strain CFT073 encodes three chromosomal hns-related genes: the well-characterized hns and stpA genes which are also encoded by E. coli K-12 strains, and a third gene which we term hnsB that is encoded by a region of the chromosome specific to CFT073. The complete genome sequence of strain CFT073 (Welch et al., 2002) indicates that hnsB is contained within a 23.2 kb region of UPEC-specific DNA inserted at the serU tRNA locus (Fig. 1; Table 1). The 134-amino-acid H-NSB protein is 65% identical to H-NS of Erwinia chrysanthemi and appears to function as a classical H-NS-like repressor (Fig. 2; our unpublished data; see below). Our analysis indicates 26 other open reading frames (ORFs) of > 50 amino acids within the hnsB island, although five of these are not annotated in the CFT073 database entry, which also annotates three additional reading frames, two with non-ATG start codons and one overlapping other reading frames that we defined. We designated the ORFs we defined as uORF1–27; their relationship to those annotated in the database entry is shown in Table 1. Among them are genes encoding a putative type IV pilin precursor and tip adhesin, a RepB/MobA-like protein, and a P4-like integrase protein at the serU-distal end of the island, the latter hinting at a previous existence of this chromosomal island as a mobile element. The hnsB gene corresponds to uORF7, and downstream of it in a convergent orientation is a putative operon of six genes (uORF1–6; Fig. 1; Table 1). Of these, uORF5 was of particular interest as it was found to encode an 80-amino-acid protein related to the oligomerization and linker domains of H-NS but lacking the C-terminal DNA-binding domain. Homology of uORF5 to hns is weak overall (29% identity to E. chrysanthemi H-NS over 78 amino acids), but stronger over the first 45 amino acids encoding the core oligomerization motif (Fig. 2). However, the truncated nature of the protein encoded by uORF5 compared with previously identified H-NS-like proteins, and its proximity to the hnsB gene, suggested that it might be a remnant of a full-length hns-like gene whose function had been replaced by hnsB.

Figure 1.

A. Schematic map of the hnsT-containing genomic islands inserted at the serU and asnW tRNA loci of UPEC strain CFT073 and EPEC strain E2348/69 respectively. Open reading frames of greater than 50 amino acids (left to right, uORF1–27 in CFT073 and eORF1–31 in E2348/69) are indicated (see Tables 1 and 2 for details).
B. Detailed map of the uORF1–8 and eORF1–8 regions of the UPEC and EPEC islands, showing the hnsT genes from either organism and the UPEC hnsB gene.

Table 1.  Accession numbers, locations and database matches of open reading frames (ORFs) in the UPEC CFT073serU island.
ORFNameAccession No.SizeLocationSignificant matchesSource%RangeAccession No.
  1. NA, not annotated in the CFT073 genome sequence entry (Accession No. AE014075). Sizes in amino acids are indicated from the most upstream ATG codon – ORFs annotated from upstream non-ATG codons in the CFT073 genome are indicated by asterisks. Location is shown as the start and stop nucleotide of the ORF numbered from the serU-proximal end of the island. The most significant database match is shown, with percentage identity, range in amino acids and accession No. indicated.

 1NP_754308278*  278, 1114Hypothetical protein t0869S. typhi35273NP_804702
 2alpANA 72 1331, 1549Phage CP4 AlpA-like protein (regulation)E. coliO15751 62NP_289200
 3NP_754307100 1619, 1921  –
 4NP_754306 98 1974, 2270  –
 5hnsTNP_754305 80 2343, 2585H-NS N-terminal domainE. chrysanthemi29 78CAC44357
 6ybaANP_754304126 2664, 3044YbaA/ORF2 (type Ia colicin plasmid)E. coli52102G25035
 7hnsBNP_754303134 3512, 3108H-NSE. chrysanthemi65134CAC44357
 8NP_754302159 4150, 3671  –
 9NP_754301243 4324, 5055329αα hypothetical proteinSulfolobus tokodaii26120NP_376744
10repANP_754300193 5721, 5140C-terminus: hypothetical proteinE. coliCFT07393 47NP_753185
Full-length: replication proteinBartonella grahamii 33169NP_696962
11NP_754298320 7074, 8036Hypothetical proteinPhotorhabdus luminescens44320NP_928087
12comENP_754297393 8088, 9269C-terminus: DNA uptake protein (ComE)Clostridium thermocellum37 74ZP_00060722
13lppNA 95 9648, 9361C-terminus of lipoproteinE. coliO15733 63NP_311533
14NP_75429516210247, 9759Hypothetical phage Sfi 11 proteinStreptococcus thermophilus44 36NP_056687
15NP_754293180*10653, 11195  –
16NA 7011226, 11438  –
17NP_75429227812377, 11541Hypothetical transmembrane proteinR. solanacearum66278NP_520733
18NP_75429116912514. 13023Conserved hypothetical proteinBdellovibrio bacteriovorus38169NP_968352
19NP_754290287*14018, 13155C-terminus: hypothetical proteinBrucella melitensis55129NP_540591
20mobANP_75428950015637, 14135RepB/MobA-like conjugal transfer proteinP. luminescens64489NP_929105
21NA 5616160, 15990  –
22NP_754288101*16283, 16588Hypothetical proteinP. luminescens55100NP_929108
23NA 5316806, 16967  – 
24pilSNP_75428718516987, 17544Type IV pilin protein precursorP. luminescens29131NP_929013
25pilVNP_75428648717597, 19060PilV-like protein (plasmid R721)E. coli26403NP_065346
26NP_75428560620995, 19175Putative inner membrane proteinS. typhimurium31591NP_461688
27intNP_75428440622329, 21109P4-family integraseXanthomonas axonopodis54344NP_641995
Figure 2.

Alignment of the H-NS proteins of Erwinia chrysanthemi (CAC44357) and Escherichia coli (P08936) with the UPEC CFT073 H-NSB protein (NP_754303) and with H-NST from CFT073 (NP_754305), EPEC strain E2348/69 and EAEC strain 042. The untranslated N-terminal extension included in database entry CAC44357 is omitted. Identities to this sequence are shown by dark shading, and conservative substitutions by light shading. The helical structure of the N-terminal 50 amino acids defined by structural studies (Esposito et al., 2002; Bloch et al., 2003; Cerdan et al., 2003) is indicated. A numbering scheme which defines the N-terminal methionine of the H-NS family as amino acid 1 is used here and throughout this work.

To test this hypothesis, we searched databases of completed and incomplete bacterial genomes for relatives of the uORF5-encoded protein. This analysis revealed a similar ORF in the incomplete genome of EPEC strain E2348/69 encoding a protein differing from that encoded by uORF5 at eight amino acid positions (Fig. 2; Table 2). We therefore termed this truncated hns-like gene hnsTEPEC, with uORF5 being retrospectively named hnsTUPEC. Analysis of the 51 376 bp contig containing hnsTEPEC showed that it is encoded within a 25 791 bp EPEC-specific island inserted at the asnW tRNA locus (Fig. 1). This island contains 31 ORFs (eORF1–31) of > 50 amino acids, with hnsTEPEC corresponding to eORF5 (Table 2). Moreover, the E2348/69 asnW island is clearly related to the CFT073 serU island, both from comparisons of the ORFs and from direct DNA–DNA comparisons (Fig. 1). This relatedness is mosaic in nature, with large segments of highly similar DNA being interspersed with unrelated sequences. The type IV pilin genes and MobA-like protein are conserved between the two islands, although the latter appears truncated in E2348/69, and although both islands contain a homologous P4-like integrase gene, the EPEC island has also acquired a second P4-like integrase in an extension to the island sequence (eORF29–31). The region upstream of hnsT is well conserved, with uORFs 1, 3 and 4 being similar to their EPEC counterparts, although uORF2 has been replaced by an unrelated DNA segment in the EPEC island. However, this region of DNA similarity ends adjacent to the stop codon of hnsT, and hnsB is therefore absent from the EPEC island (Fig. 1). This, together with the conservation of the reading frame between UPEC and EPEC, suggests that hnsT might have an independent function rather than simply being a gene remnant. We also identified a less closely related hnsT gene in the incomplete genome of enteroaggregative E. coli (EAEC) strain 042 (Fig. 2). hnsTEAEC is surrounded by EAEC-specific sequences which share some features with the CFT073 serU and EPEC asnW islands but are much less closely related, and two similar hnsT genes are also found in the recently completed genome of Erwinia carotovora SCRI1043 (data not shown).

Table 2.  Locations, database matches and predicted functions of open reading frames (ORFs) in the EPEC E2348/69 asnW island.
ORFNameSizeLocationSignificant matchesSource%RangeAccession No.
  1. Sizes in amino acids are indicated from the most upstream ATG codon. Location is shown as the start and stop nucleotide of the ORF numbered from the asnW-distal end of the island. The most significant database match is shown, with percentage identity, range in amino acids and accession No. indicated. Matches to CFT073serU island ORFs are indicated in square brackets.

 1291  381, 1256Hypothetical protein t0869 [uORF1]S. typhi66286NP_804702
 2 60 1675, 1493  –
 3100 1695, 1997Hypothetical protein [uORF3]E. coli CFT07390100NP_754307
 4 98 2042, 2338Hypothetical protein [uORF4]E. coli CFT07397 98NP_754306
 5hnsT 80 2412, 2654Hypothetical protein H-NST [uORF5]E. coli CFT07392 78NP_754305
 6 84 2692, 2946  –
 7190 3749, 3177Hypothetical proteinNeurospora crassa43 57XP_328227
 8 98 4158, 3862  –
 9repA244 4920, 4186RepA (plasmid pRAO1) central regionRuminobacter amylophilus33 88BAA74510
10140 6570, 6992C-terminus: folic acid synthesis proteinPneumocystis carinii23 91AAN38834
11340 7237, 8259  –
12lpp139 8763, 8344N-terminus: hypothetical proteinE. coli CFT07397 68NP_754296
C-terminus: lipoprotein [uORF13]E. coli O15741115NP_311533
13162 9241, 8753Hypothetical protein [uORF14]E. coli CFT07396162NP_754295
14180 9647, 10189Hypothetical protein [uORF15]E. coli CFT07393180NP_754293
1531011221, 10289Hypothetical protein (central region)Mesorhizobium loti31 71CAD31491
1650612754, 11234C-terminus: ATP-binding protein SugRS. typhimurium29110AAD16951
17mobA′37113978, 12863MobA-like protein (truncated) [uORF20]E. coli CFT07375352NP_754289
18 5114180, 14335  –
19 6614856, 14656Unannotated serU island protein [uORF21]E. coli CFT07377 54
2010114979, 15284Hypothetical protein [uORF22]E. coli CFT07391101NP_754288
21pilS18515682, 16239Type IV pilin protein precursor [uORF24]E. coli CFT07398185NP_754287
22pilV48716292, 17755PilV-like protein [uORF25]E. coli CFT07395487NP_754286
2311418214, 17870  –
2428419080, 18226Methyl-accepting chemotaxis proteinWolinella succinogenes25168NP_906977
2530619990, 19070Hypothetical proteinShewanella oneidensis45296NP_716132
26 8320342, 20593Hypothetical proteinHaemophilus ducreyi50 77NP_873834
27int′ 9021002, 20730C-terminus of phage-like integraseE. coli O15780 75NP_289539
28intA42522245, 20968P4-family integrase [uORF27]E. coli CFT07397404NP_754284
2937122447, 23562Hypothetical proteinMethanosarcina mazei34342NP_634583
3023823562, 24278  –
31intB44025667, 24345P4-family integraseS. flexneri56422NP_708738

H-NSTEPEC interferes with H-NS repression of the proU promoter, but cannot re-repress the bgl operon in the absence of H-NS

Having established that the hnsT gene is well conserved between the CFT073 serU and E2348/69 asnW chromosomal islands, we sought to determine whether the protein it encodes (H-NSTUPEC or H-NSTEPEC in CFT073 and E2348/69 respectively) can function in an H-NS-like manner or, alternatively, possesses a novel function. We amplified the hnsT gene plus flanking DNA extending from the stop codon of uORF4 or eORF4 to the start codon of uORF6 or eORF6 from the CFT073 or E2348/69 chromosome, and cloned this region into the low-copy vector pHSG576 to generate pHSGHNSTE and pHSGHNSTU respectively (Table 3). We then tested the effect of these low-copy clones on the H-NS-dependent repression of two well-characterized systems, the E. coli proU and bgl operons. The chromosomal proU–lacZ fusion in strain GM37 (Table 3) produced ≈400 U of β-galactosidase activity in LB medium, and was derepressed two- to 2.5-fold by the hns-205::Tn10 mutation [encoding truncated H-NS lacking the DNA-binding domain (H-NS 1–93)] or the hns-206::Apr null mutation (Fig. 3A). Plasmid pHSGHNSTE encoding H-NSTEPEC did not restore repression of proU in an hns mutant, but was seen to derepress the proU–lacZ fusion in the wild-type strain, raising the level of β-galactosidase expression to that seen in the hns-205 mutant (Fig. 3A). In contrast, the H-NSTUPEC-encoding plasmid pHSGHNSTU had no detectable effect on proU expression in the wild-type or hns mutant strains compared with the vector control (data not shown). Therefore the H-NST proteins cannot act as H-NS-like repressors in this system, but H-NSTEPEC, unlike its UPEC counterpart, has an anti-H-NS effect and disrupts H-NS-dependent repression of the locus.

Table 3. Escherichia coli strains and plasmids used in this study.
Strain/plasmidRelevant characteristicsaReference/source
  • a

    . Transductional crosses are represented as A × B, where A is the recipient and B is the donor.

AF22BGL1 hns-205::Tn10 (BGL1 × GM230)This work
AF24BGL1 hns-206::Apr (BGL1 × PD32)This work
AF27CSH26 φ(proV–lacZ) hns-206::Apr (HM60 × PD32)This work
AF38GM37 hns-206::Apr (GM37 × PD32)This work
AF39TP8503 φ(proU–lacZ) (TP8503 × GM37)This work
BGL1CSH26 φ(bgl–lacZ)Ueguchi et al. (1996)
CFT073Wild-type UPEC O6:H1Mobley et al. (1990)
E2348/69Wild-type EPEC O127:H6Levine et al. (1978)
GM230GM37 hns-205::Tn10Higgins et al. (1988)
GM37MC4100 φ(proU–lacZ)Higgins et al. (1988)
HM60CSH26 φ(proV–lacZ) hns60 zci-506::Tn10Ueguchi et al. (1996)
MC4100araDΔ(arg–lac) wild typeCasadaban (1976)
PD32MC4100 hns-206::AprDersch et al. (1993)
TP8503Δ(lac–pro) wild typeMasters et al. (1989)
XL-1 BluerecA cloning strainOur stocks
pBAD33Arabinose-inducible expression plasmid, CmrGuzman et al. (1995)
pBADH6SHNSTEArabinose-inducible His6-S-tagged H-NSTEPEC in pBAD33This work
pBADH6SHNSTUArabinose-inducible His6-S-tagged H-NSTUPEC in pBAD33This work
pBADHNSTEArabinose-inducible H-NSTEPEC in pBAD33This work
pBADHNSTEA16VArabinose-inducible H-NSTEPEC A16V in pBAD33This work
pBADHNSTEL23IArabinose-inducible H-NSTEPEC L23I in pBAD33This work
pBADHNSTEL30PArabinose-inducible H-NSTEPEC L30P in pBAD33This work
pBADHNSTUArabinose-inducible H-NSTUPEC in pBAD33This work
pBADHNSTUV16AArabinose-inducible H-NSTUPEC V16A in pBAD33This work
pET30bIPTG-inducible His6-S-tag expression vector, KmrNovagen
pET30HHAIPTG-inducible His6-S-tagged Hha in pET30bThis work
pET30HNSTEIPTG-inducible His6-S-tagged H-NSTEPEC in pET30bThis work
pET30HNSTEL30PIPTG-inducible His6-S-tagged H-NSTEPEC L30P in pET30bThis work
pET30HNSTUIPTG-inducible His6-S-tagged H-NSTUPEC in pET30bThis work
pET30LERIPTG-inducible His6-S-tagged Ler in pET30bThis work
pET6HIPTG-inducible His6-tag expression vector, AprCross et al. (1994)
pET6HHNSIPTG-inducible His6-tagged H-NS in pET6HThis work
pET6HHNSBIPTG-inducible His6-tagged H-NSB in pET6HThis work
pHSG576Low-copy cloning vector, pSC101 replicon, CmrHashimoto-Gotoh et al. (2000)
pHSGHNSTEhnsTEPEC in pHSG576This work
pHSGHNSTUhnsTUPEC in pHSG576This work
pJK289Single-copy cloning vector, KmrKato and Ikeda (1996)
pJKHNSBhnsB in pJK289This work
pOSEX2proU promoter vector, Apr, TcrHerbst et al. (1994)
pSPT18SP6/T7 in vitro transcription vector, AprRoche Molecular
pSPTHNSTEInternal fragment of hnsTEPEC in pSPT18This work
Figure 3.

A. Effect of the hnsT low-copy clone pHSGHNSTE or the cloning vector pHSG576 on proU–lacZ expression in strains GM37 (WT), GM230 (hns-205::Tn10) and AF38 (hns-206::Apr). Strains were grown overnight in LB medium.
B. Effect of pHSGHNSTE and pHSG576 on bgl–lacZ expression in strains BGL1 (WT), AF22 (hns-205::Tn10) and AF24 (hns-206::Apr). The lacZ activity is plotted on a logarithmic scale to enable both repressed and fully derepressed levels of transcription to be seen clearly. Strains were grown overnight in LB medium containing 5 mM β-methyl- d-glucoside.

Expression of the bgl operon was assayed using the chromosomal bgl–lacZ fusion in strain BGL1 (Table 3). This fusion was fully silent in the wild-type strain and exhibited only very low (10 U) activity in its hns-205::Tn10 derivative, consistent with previous data (Dersch et al., 1994; Free et al., 1998; 2001), but was derepressed > 100-fold in the presence of the hns-206::Apr null mutation (Fig. 3B). Consistent with the data obtained with the proU–lacZ fusion, the pHSGHNSTU plasmid had little effect on bgl–lacZ expression in wild-type or hns mutant strains (data not shown), suggesting that the H-NSTUPEC protein has little biological activity in H-NS-regulated systems. However, while the data in the presence of the hns-206 mutation suggested that H-NSTEPEC might have a very weak repressive effect at bgl in the absence of H-NS, this was insignificant compared with the highly efficient repression achieved by the H-NS (1–93) protein encoded by the hns-205 strain (Fig. 3B). In the wild-type strain, pHSGHNSTE again elevated bgl–lacZ expresssion to the same level as in the hns-205::Tn10 strain. Therefore, instead of acting as a repressor of bgl but being inactive at proU, as truncated derivatives of H-NS itself are, H-NSTEPEC seems instead to be dominant negative against H-NS at proU and (very weakly) at bgl. Dominant-negative effects of the H-NS N-terminal domain itself at proU are only seen when this domain is overexpressed (Williams et al., 1996; Free et al., 2001), and H-NSTEPEC therefore has a novel anti-H-NS activity functional without high-level overexpression. We also examined the effect of the H-NSTEPEC and H-NSTUPEC proteins on a third H-NS-dependent phenotype, motility. Strains lacking H-NS are non-motile resulting from a failure in flagellar expression (Bertin et al., 1994), but neither of the H-NST proteins was able to restore motility to an hns mutant or significantly to affect motility in the presence of wild-type H-NS (data not shown). The positive H-NS-dependent control of flagellar expression is poorly understood, and may involve both repression of a negative regulator of the flhDC operon and a direct interaction with the flagellar motor protein FliG (Donato and Kawula, 1998; Ko and Park, 2000), but it appears that the H-NST proteins do not affect either of these processes. Therefore, the anti-H-NS effect defined for H-NSTEPEC at proU may be restricted to a subset of H-NS-regulated systems.

Controlled induction of H-NSTEPEC results in strong derepression of proU and weak derepression of bgl in a wild-type strain

The above experiments indicate that H-NSTEPEC mediates a significant anti-H-NS effect at the proU promoter when cloned in low copy in its native context. To study this effect further, we cloned the hnsTEPEC ORF downstream of the arabinose-inducible araBAD promoter in the vector pBAD33 and used controlled, stepwise induction of the promoter with increasing concentrations of arabinose (Guzman et al., 1995) to monitor the effect on H-NS-dependent regulation. An arabinose-inducible H-NSTUPEC was used for comparison. The proU–lacZ strain GM37 is a derivative of MC4100, which contains an araD mutation and is therefore sensitive to the accumulation of toxic intermediates in the presence of arabinose (Guzman et al., 1995). We therefore transduced the proU–lacZ fusion from GM37 into the araD+ strain TP8503 to avoid this complicating factor, generating strain AF39 (Table 3). Strong derepression of proU expression was seen in AF39 when H-NSTEPEC expression was induced with arabinose concentrations of 0.002% or greater, reaching sixfold at 0.2% arabinose (Fig. 4A). The vector control pBAD33 showed no such effect, while the inducible H-NSTUPEC construct exhibited a very slight derepressive effect at higher arabinose concentrations only. This suggests that H-NSTUPEC may not be completely inactive as an anti-H-NS factor, but rather that it has a much weaker anti-H-NS function requiring very high protein concentrations to become apparent. When the arabinose-inducible constructs were tested against the repressed bgl promoter in strain BGL1, a slight derepressive effect was again seen for H-NSTEPEC, although H-NSTUPEC behaved similarly to the vector control (Fig. 4B). This effect is at the limits of detectability for the β-galactosidase assay because of the very low activity of the repressed bgl–lacZ fusion, but taken together with the similar effect seen in Fig. 3B does suggest that H-NSTEPEC can have some anti-H-NS effect at this promoter too, although it is much weaker than at proU. A non-specific repression of the bgl promoter in the presence of 0.2% arabinose, which may be related to catabolite repression, was also observed in this experiment. In contrast, induction of H-NSTEPEC or H-NSTUPEC in the presence of the hns-206 null mutation had no effect on the high, derepressed level of bgl–lacZ expression in this strain (Fig. 4C), indicating that these truncated H-NS-like proteins have no repressive function at this promoter even when highly expressed, in contrast to H-NS itself (Free et al., 2001). Again, non-specific repression of the bgl promoter in the presence of 0.2% arabinose was observed.

Figure 4.

A. Effect of arabinose-dependent induction of H-NSTEPEC (pBADHNSTE), H-NSTUPEC (pBADHNSTU) or the vector pBAD33 on proU–lacZ expression in strain AF39. Cultures were grown overnight in LB medium plus arabinose as indicated.
B. Effect of pBADHNSTE, pBADHNSTU or pBAD33 on bgl–lacZ expression in strain BGL1 (WT). Cultures were grown overnight in LB medium plus 5 mM β-methyl- d-glucoside plus arabinose as indicated.
C. Effect of pBADHNSTE, pBADHNSTU or pBAD33 on bgl–lacZ expression in strain AF24 (hns-206::Apr). Cultures were grown overnight in LB medium plus 5 mM β-methyl- d-glucoside plus arabinose as indicated.

H-NSTUPEC has no significant antagonistic effect against UPEC H-NSB function

Given that H-NSTUPEC seemed to have only a weak anti-H-NS effect unlike its counterpart from EPEC, we wondered whether it might instead act antagonistically against a different H-NS-like protein. The obvious candidate is the UPEC H-NSB protein, which is encoded just downstream of H-NSTUPEC and may have co-evolved in the same mobile DNA element. We therefore cloned the hnsB gene and its flanking sequences extending from the stop codon of uORF6 to the stop codon of uORF8 (Fig. 1) into the single-copy vector pJK289 and introduced the resulting clone pJKHNSB into strain AF27, which contains a chromosomal lacZ fusion to the first gene of the proU operon, proV, and an hns-206::Apr null mutation (Table 3). This alternative fusion to the proU promoter region was used because the φ(proU–lacZ)hyb2 fusion in GM37 and AF39 encodes kanamycin resistance which is incompatible with the pJK289 vector. The proV–lacZ fusion is more sensitive to osmotic induction at low salt concentrations than the proU–lacZ fusion (our unpublished data), and was therefore assayed at the lower NaCl concentration of 50 mM, compared with 170 mM in standard LB medium. Presence of pJKHNSB in AF27 restored proV–lacZ expression to the repressed level of ≈20 U compared with ≈240 U when no H-NS-like protein was present (Fig. 5A), indicating that H-NSB acts in an H-NS-like manner as an efficient repressor of proU. Consistent with this activity, purified His6-tagged H-NSB bound preferentially to the proU promoter fragment from digested pOSEX2 DNA (Fig. 5B) and at slightly higher concentrations to the fragment containing the bla promoter, which is also a known H-NS target (Bertin et al., 1999). However, introduction of the low-copy H-NSTUPEC clone pHSGHNSTU caused no relief of the H-NSB-dependent repression of proV–lacZ (Fig. 5A). Therefore, despite their common origins, H-NSTUPEC has no anatgonistic activity against H-NSB, similar to its lack of effect on repression by E. coli K-12 H-NS.

Figure 5.

A. Repression of proV–lacZ expression in strain AF27 (hns-206::Apr) by single-copy hnsB (pJKHNSB) and lack of effect of low-copy hnsTUPEC (pHSGHNSTU). Vector controls pJK289 and pHSG576, respectively, are also shown. Strains were grown overnight in LB medium containing a final NaCl concentration of 50 mM.
B. Binding of purified H-NSB to fragments of plasmid pOSEX2 containing the proU and bla promoters (indicated by arrows), but not to other vector fragments, in a competitive agarose bandshift. Concentrations of H-NSB in nM are shown.

H-NSTEPEC co-purifies with H-NS in an approximately 1:1 ratio, while only a fraction of H-NSTUPEC co-purifies with H-NS

In order to rationalize the anti-H-NS effect of H-NSTEPEC compared with the lack of effect of H-NSTUPEC, we sought to purify the two proteins to study their in vitro properties. Initial studies with derivatives of the H-NST proteins tagged at the N-terminus with a hexahistidine peptide revealed that the tagged proteins were almost entirely insoluble (data not shown). We therefore employed a longer tag consisting of six histidines followed by an S peptide, which was added to the N-terminus of H-NSTEPEC and H-NSTUPEC by cloning them into the tag-encoding vector pET30b (Table 3). When the tagged proteins were cloned downstream of the araBAD promoter in pBAD33, they exhibited somewhat reduced effects on proU–lacZ expression compared with their untagged counterparts (Fig. 6A), but the strong derepressive effect of H-NSTEPEC was retained, indicating that the His6-S-tag did not eliminate anti-H-NS function. Expression of the pET30b-encoded tagged proteins in E. coli strain BL21 (λDE3) resulted in a significant yield of soluble tagged H-NSTEPEC, but significantly less soluble H-NSTUPEC, although much more H-NSTUPEC was present in the insoluble fraction (data not shown). Significantly, both proteins seemed to co-purify with a contaminant which ran slightly faster than the tagged H-NST protein (Fig. 6B). This contaminant band was not recognized by a labelled S-protein preparation which detects the S-peptide tag of the fusion protein, but did react strongly with a specific monoclonal anti-H-NS anti-serum (Fig. 6B). Therefore, the H-NST proteins co-purify with cellular H-NS when overexpressed. The H-NS/H-NSTEPEC ratio in our preparation appears close to one, but is significantly lower in the H-NSTUPEC preparation, and this correlates with the greater solubility of H-NSTEPEC, which may be mediated by its co-purification with H-NS. As controls, we purified His6-S-tagged derivatives of the EPEC Ler protein, which is related to H-NS and has anti-H-NS function but is thought not to interact with H-NS (Elliott et al., 2000), and the E. coli K-12 Hha protein, which has been found to co-purify and interact with H-NS in previous studies (Nieto et al., 2000; 2002). However, neither Ler nor Hha purified by the method employed here displayed any co-purification with cellular H-NS (Fig. 6B), indicating that the H-NST–H-NS interaction, particularly for the EPEC protein, is much tighter than that previously defined for Hha and H-NS.

Figure 6.

A. Comparison of derepression of proU–lacZ in strain AF39 by untagged (pBADHNSTE, pBADHNSTU) and His6-S-tagged (pBADH6SHNSTE, pBADH6SHNSTU) H-NST proteins induced with arabinose, or the vector pBAD33. Strains were grown overnight in LB medium plus arabinose as indicated.
B. Co-purification of H-NST proteins with H-NS. Silver stain (top), anti-S-tag blot (middle) and anti-H-NS blot (bottom) of preparations of His6-S-tagged H-NSTEPEC and H-NSTUPEC purified by nickel column chromatography. Positions of H-NST (thick arrow; predicted Mr = 14.4 kDa but retarded in gel electrophoresis), H-NS (thin arrow; 15.5 kDa) and molecular weight markers (kDa) are shown on the top panel. His6-S-tagged Ler (19.8 kDa) and Hha (13.3 kDa) proteins purified by identical procedures provide size and blotting controls. Breakdown products of the Ler and H-NST proteins seen on the silver stain and the anti-S-tag blot are indicated by asterisks.

A mutation disrupting the putative coiled-coil of H-NSTEPEC abolishes its effect on proU expression in vivo and its co-purification with H-NS

H-NS-like proteins contain predicted or proven coiled-coil structures in their N-terminal domains which can mediate homomeric or heteromeric interactions between these proteins (Williams et al., 1996; Dorman et al., 1999; Rimsky, 2004). A mutation of the leucine at position 30 in E. coli H-NS to proline abolishes the predicted coiled-coil structure, the ability to dimerize and transcriptional repression  (Ueguchi et al.,  1997).  Both  H-NSTEPEC and  H-NSTUPEC also contain leucine at position 30 (Fig. 2), and an L30P mutation is similarly predicted to disrupt putative coiled-coil structures in the N-terminal 40 amino acids of these proteins (see Fig. 7A for H-NSTEPEC). We therefore mutagenized this residue in H-NSTEPEC to proline and studied the ability of the mutant protein to affect proU–lacZ expression and to co-purify with H-NS. H-NSTEPEC L30P was completely inactive in antagonizing H-NS repression of proU–lacZ (Fig. 7B), and a preparation of tagged H-NSTEPEC L30P contained no detectable H-NS protein (Fig. 7C). Like H-NSTUPEC, H-NSTEPEC L30P was also significantly less soluble than wild-type H-NSTEPEC (data not shown), further suggesting that solubility is connected with the ability to co-purify with H-NS. These data suggest that H-NSTEPEC interacts with H-NS via a heteromeric coiled-coil interaction, leading to the formation of H-NS–H-NSTEPEC protein–protein complexes which disrupt normal H-NS repression at the proU promoter and can be co-purified via column chromatography.

Figure 7.

A. Coiled-coil prediction for the H-NSTEPEC wild-type protein (WT) and its L30P mutant derivative. Vertical black bars indicate predicted coiled-coil probability at each amino acid in the protein, with the prediction considered significant if the probability is greater than 0.5 (horizontal bar above graph).
B. Failure of H-NSTEPEC L30P (pBADHNSTEL30P) to repress proU–lacZ in strain AF39, compared with wild-type protein (pBADHNSTE) and the vector pBAD33. Strains were grown overnight in LB medium plus arabinose as indicated.
C. Detection of H-NST proteins via anti-S-tag blot (left) and co-purifying H-NS protein via anti-H-NS blot (right) in preparations of His6-S-tagged H-NSTEPEC, H-NSTUPEC and H-NSTEPEC L30P purified by nickel column chromatography.

One of two amino acid differences between the N-termini of H-NSTEPEC and H-NSTUPEC is key to the ability to derepress proU expression

It is apparent from the results above that, despite their 90% amino acid identity, H-NSTEPEC and H-NSTUPEC differ greatly in their ability to disrupt H-NS-dependent repression and to co-purify with H-NS. Of the eight amino acid differences between the two proteins, six are localized within the last 30 amino acids, which lie outside the core oligomerization region containing the coiled-coil (Fig. 2). The other two changes are within the core region: residue 16 (alanine in the EPEC protein, valine in UPEC) lies in helix 2 of the N-terminal region, while residue 23 (leucine in EPEC, isoleucine in UPEC) lies at the start of the core interaction helix, α3. We therefore mutated these two residues in H-NSTEPEC to their equivalents in the UPEC protein and measured the ability of the mutant proteins to derepress proU expression. While the L23I mutant was essentially unaffected in its ability to derepress proU, the A16V mutant showed only a very weak derepression at high levels of induction similar to the effect of H-NSTUPEC(Fig. 8). This suggests that the amino acid at position 16 is key to the ability of H-NSTEPEC to act as an H-NS antagonist, and points to an important role for the α2 helix in this function. Consistent with this interpretation, mutating residue 16 of the H-NSTUPEC protein to alanine bestows on this protein an anti-H-NS function similar to that of H-NSTEPEC (Fig. 8).

Figure 8.

Effect of H-NSTEPEC A16V and L23I mutants (pBADHNSTEA16V and pBADHNSTEL23I respectively) and H-NSTUPEC V16A (pBADHNSTUV16A) on proU–lacZ expression in strain AF39 compared with the wild-type proteins (pBADHNSTE and pBADHNSTU) and the vector pBAD33. Cultures were grown overnight in LB medium plus arabinose as indicated.

H-NSTEPEC has minimal effects on the efficiency of H-NS binding to the proU promoter

H-NS is known to repress proU expression by binding to a curved downstream regulatory element (DRE) which acts as a nucleation site for the protein (Owen-Hughes et al., 1992). Lower-affinity binding to regions upstream of the promoter can also occur (Lucht et al., 1994) and it is possible that interaction between the H-NS complexes at the two regions might create a repressive, RNA polymerase-trapping complex as suggested for other promoters (Dame et al., 2002; Haack et al., 2003). We used purified His6-tagged E. coli H-NS and our preparation of His6-S-tagged H-NSTEPEC (which is ≈1:1 H-NS:H-NSTEPEC) to determine whether the presence of H-NSTEPEC affected the H-NS complexes on these DNA regions. The prepartions were normalized for H-NS concentration by Western blotting with the anti-H-NS anti-serum, added to labelled proU upstream and proU downstream DNA fragments, and the resulting complexes were analysed by mobility shift assay (Fig. 9). H-NS binding to the downstream fragment occurred efficiently at H-NS concentrations as low as 10 nM and was unaffected by the presence of  H-NSTEPEC. In contrast, H-NS bound the upstream fragment more weakly, with significant binding not being observed at concentrations below 30 nM, consistent with previous observations (Lucht et al., 1994). The formation of this complex appeared to be somewhat disrupted in the presence of H-NSTEPEC, as it did not appear at concentrations below 100 nM H-NS (Fig. 9). This suggests that  H-NSTEPEC has at most a relatively subtle effect on weaker H-NS complexes upstream of the proU promoter despite its effects on transcriptional repression. Disruption of repression by H-NSTEPEC may depend more on a disruption of interactions between the upstream and downstream H-NS complexes than on a major destabilizing effect on either complex individually.

Figure 9.

Electrophoretic mobility shift assay of H-NS binding to proU 5′ (−176 to +5 with respect to the ATG) or DRE (−16 to +174 with respect to the ATG) DNA fragments in the presence or absence of ≈1:1 H-NSTEPEC. Concentrations of H-NS in the range 10–100 nM are shown. Unbound DNA fragments and protein–DNA complexes are indicated by arrows, and the position of a non-specific band in the proU DRE fragment gel is indicated by an asterisk. A weak complex of higher mobility seen in the presence of H-NSTEPEC on the 5′ fragment is indicated by the small arrow.

H-NSTEPEC is expressed during logarithmic growth in E2348/69 and transcribed monocistronically

To study the expression of the hnsTEPEC gene, we performed Northern blotting with an hnsT-specific riboprobe against total RNA isolated from E2348/69 grown in DMEM medium. hnsT mRNA was undetectable in an overnight culture but detectable at low levels throughout the logarithmic growth phase and into early stationary phase (Fig. 10A). Despite the operon-like organization of the hnsT gene (Fig. 1), the RNA was of a size consistent with the gene being transcribed monocistronically (data not shown). To determine whether this corresponded to transcriptional initiation at a promoter upstream of hnsT, we performed 5′ RACE analysis with hnsT-specific downstream primers (see Experimental procedures) on the RNA sample taken from the culture at OD600 of 0.6, in which the hnsT mRNA was most abundant. Analysis of the 5′ RACE products from second-round polymerase chain reaction (PCR) using the TERACE3 primer (Table 4), which binds 161 bp downstream of the ATG, by gel electrophoresis (Fig. 10B), revealed specific products of ≈200 bp and < 100 bp. The latter corresponds to an RNA 5′ end within the hnsT ORF, which would be non-functional for protein expression, but the larger, more abundant product indicates a transcriptional start site within the eORF4-hnsTEPEC intergenic region. Direct sequencing of this RACE product revealed that it actually corresponds to RNAs initiating at three adjacent nucleotides, A2400, G2401 and G2402, within the asnW chromosomal island (Fig. 10B). The second of these initiating nucleotides, 11 bp upstream of the hnsT start codon, was most abundant based on the relative size of the DNA sequencing peaks. Weak potential −10 and −35 regions were identified upstream of this start site, as was a potential stem loop structure followed by a run of A residues which may act as a Rho-independent transcriptional terminator for eORF4 transcription (Fig. 10B).

Figure 10.

A. Growth curve (top) and Northern blot (bottom) of the hnsT transcript in RNA samples from E2348/69 grown overnight (O/N) or to the indicated OD600 in DMEM medium. RNA sample points are indicated on the growth curve by arrows and the corresponding OD600. The hnsT transcript is indicated on the blot by an arrow.
B. Agarose gel electrophoresis of specific 5′ RACE products (R) obtained from the OD600 = 0.6 RNA sample in (A) using hnsT- specific downstream primers. Arrows indicate the major and minor RACE products; M indicates molecular size markers (sizes in bp shown). The lower half of the panel shows the DNA sequence between the eORF4 stop codon and the hnsTEPEC start codon (both in bold) with transcriptional start sites identified by sequencing of the larger RACE product indicated by angled arrows. Heights of these arrows indicate qualitative relative abundance. Potential −10 and −35 sequences are boxed, and the stem loop of a putative Rho-independent terminator is indicated by half arrows above the DNA sequence.

Table 4.  Oligonucleotide primers used in this study.
PrimerSequence (5′−3′)aUse
  • a

    . Sequences complementary to the target DNA are shown in upper case, sequences unique to the oligonucleotide primer in lower case. Restriction sites used for cloning are underlined.

HHAEFctagcaccATGgCCGAAAAACCTTTAACGConstruction of Hha expression clone
HHAERctagcaggatccTTAGCGAATAAATTTCCATACConstruction of Hha expression clone
HNSEFctagcaccATGgGCGAAGCACTTAAAATTCTGConstruction of H-NS expression clone
HNSER2ctagcaggatccTTATTGCTTGATCAGGAAATCConstruction of H-NS expression clone
HNSBUEFctagcaccATGgGTGAAGCTCTTAAGGCACTGAACConstruction of H-NSB expression clone
HNSBUERctagcaggatccTTAGATAGCAAAGTGTTCCAGTGTCTTConstruction of H-NSB expression clone
HNSTEUFctagcaggatccGATATGTTCTGGTGTCTATCGGCCCConstruction of H-NST expression clones
HNSTEURctagcaaagcttCCTCCATGTTTTTGATATAGATACGConstruction of UPEC H-NST expression clone
HNSTEUERctagcaggatccTCAGTCAATGAGATCTTCTGGCGAAACConstruction of EPEC H-NST expression clone
HNSTL30P1GAACAGCTTcTAAAAAAGTTCAGGFirst round of EPEC hnsT L30P mutagenesis
HNSTL30P2CCTGAACTTTTTTAgAAGCTGTTCFirst round of EPEC hnsT L30P mutagenesis
HNSTL30P3GAACAGCTTccAAAAAAGTTCAGGSecond round of EPEC hnsT L30P mutagenesis
HNSTL30P9CCTGAACTTTTTTggAAGCTGTTCSecond round of EPEC hnsT L30P mutagenesis
HNSTUFARActagcactgcagGGGGCTGATAACAAATATTATConstruction of arabinose-inducible UPEC H-NST
HNSTEUFARActagcactgcagGGGGCTGAAAACAAATCTTATConstruction of arabinose-inducible EPEC H-NST
HNSTERFctagcactgcagGAAGCCGCAACCACTGACCConstruction of hnsT riboprobe
HNSTERRctagcagaattcGGCGAAACCCGGTCACATGCConstruction of hnsT riboprobe
LEREFctagcaccATGgGGAGATTATTTATTATGConstruction of Ler expression clone
LERERctagcaggatccTTAAATATTTTTCAGCGGConstruction of Ler expression clone
PETBADF2ctagcagctagcGTGAGCGGATAACAATTCCCConstruction of arabinose-inducible tagged H-NST
PETBADRGGCCGCAAGCTTGTCGACConstruction of arabinose-inducible tagged H-NST
PPROU1CGATTTGCTCTCAGCCCproU promoter upstream bandshift probe
PPROU2GCCATGCAATAGAAAGATTCCproU promoter upstream bandshift probe
PPROU3GGAATCTTTCTATTGCATGGCproU promoter downstream bandshift probe
PPROU4GACAAATATCTCGCCTTCTTCproU promoter downstream bandshift probe


We have defined in this study a novel family of proteins which are related to the N-terminal oligomerization domain of H-NS-like proteins, but lack the C-terminal DNA-binding domain thought to be necessary for high affinity DNA binding. The H-NST proteins are most closely related to H-NS throughout the first 50 amino acids of the protein, which in H-NS is composed of three α-helices mediating dimer formation (Esposito et al., 2002; Bloch et al., 2003; Cerdan et al., 2003); the longest stretches of conserved residues are within helix α3, which is the core dimerization interface (Fig. 2). Based on this relatedness, H-NST proteins might be predicted to be able to interact with H-NS and modulate its gene regulation activity, and our detailed studies of H-NST from EPEC show that this is indeed the case for this protein.

H-NSTEPEC displays effects on H-NS-dependent gene regulation which are both similar to and distinct from those of an artificially truncated or mutated derivative of H-NS lacking a functional DNA-binding domain. Like such mutant forms of H-NS, the H-NSTEPEC protein can derepress proU expression in the presence of the wild-type H-NS protein – a classical dominant-negative effect. This indicates that the EPEC protein can interact efficiently with H-NS via their related oligomerization domains, and that the lack of DNA-binding domain in the H-NST partner disrupts the ability of the interacting H-NS monomer to bind DNA. Indeed, this interaction is shown by the co-purification of overexpressed H-NSTEPEC with cellular H-NS in an ≈1:1 ratio, and by the abolition of both this co-purification and the dominant-negative effect on proU expression by an L30P mutation disrupting the α3 helix which is the coiled-coil dimerization interface. However, it seems likely that H-NSTEPEC may exert a more efficient dominant-negative effect than does an artificial truncated derivative of H-NS. Truncation mutations of the hns gene are normally expressed at high levels, as the already active hns promoter is further derepressed because of the relief of negative autoregulation in the presence of the truncation (Falconi et al., 1993; Free et al., 2001). A low-expressing clone of the truncated hns60 allele with a point mutation in the −35 region of the promoter is unable to exert a dominant-negative effect on proU (Free et al., 2001). In contrast, despite the low-level expression of the hnsTEPEC gene shown here by Northern blotting, a low-copy clone of the locus exerts a readily detectable dominant-negative effect. H-NSTEPEC may more readily disrupt H-NS-dependent repression because amino acid differences from H-NS in the region outside the core dimerization domain may prevent an H-NS-H–NSTEPEC heterodimer from forming higher-order oligomers either with similar heterodimers or with H-NS homodimers. Therefore, a lower concentration of truncated protein would be necessary to render an H-NS complex containing it non-functional. Nevertheless, higher-level induction of H-NSTEPEC expression from an arabinose-inducible construct leads to a greater dominant-negative effect, presumably because of the trapping of a greater proportion of cellular H-NS in such dead-end heterodimeric complexes.

The second major difference between H-NSTEPEC and a truncated derivative of H-NS lies in their effects on bgl repression. The bgl locus is unusual in that H-NS proteins lacking a functional DNA-binding domain maintain fully functional repression of the promoter (Higgins et al., 1988; Dersch et al., 1994; Ueguchi et al., 1996). This seems to be a combination of the ability of truncated H-NS to interact with the paralagous StpA protein, which provides the missing DNA binding function (Free et al., 1998; 2001) and, if the H-NS truncate is expressed to sufficiently high levels, a poorly understood StpA-independent mechanism which may involve the alternative sigma factor RpoS (Ohta et al., 1999; Free et al., 2001). These results suggest that repression of the bgl promoter can tolerate significant alteration of the H-NS-dependent repressing nucleoprotein complex without an effect on expression. Even high-level expression of a truncated H-NS derivative has no dominant-negative effect at bgl, mediating instead efficient repression in the absence of wild-type H-NS (Free et al., 2001). In contrast, expression of H-NSTEPEC even to high levels cannot re-repress bgl in an hns null mutant, but instead causes a weak but detectable dominant-negative effect in a wild-type strain. Again, this may be related to an inability of H-NS–H-NSTEPEC heterodimers to form higher-order oligomers, thus disrupting even the tolerant nucleoprotein complex at the bgl promoter. H-NSTEPEC may also be unable to interact efficiently with or form oligomeric complexes with StpA, thus limiting the potential for this protein to co-repress with the H-NST protein in an hns null mutant. As the mechanism by which truncated H-NS performs StpA-independent repression of bgl when present at high levels is unclear, the failure of H-NSTEPEC to mediate a similar effect is difficult to analyse.

Our data show clearly that H-NSTEPEC can exert a significant dominant-negative effect on H-NS-dependent repression, but if such an effect was widespread across the plethora of H-NS-regulated genes in the cell, it would be highly deleterious and rapidly selected against. This is evident from the poor growth of strains containing the arabinose-inducible H-NSTEPEC (but not H-NSTUPEC) construct in the presence of high concentrations of arabinose (>0.1%; data not shown). Widespread H-NST-mediated derepression in EPEC strain E2348/69 is probably not a significant problem because of the low basal expression of the hnsT gene, and the effect of H-NST is probably modulatory rather than major at most H-NS-regulated promoters. However, one instance in which this protein may have an important effect is in modulating H-NS-dependent repression of the LEE pathogenicity island. At the LEE2–5 promoters, H-NS acts as a repressor but its effect is counteracted by the H-NS-like LEE-specific regulator Ler, which binds to the same promoter regions responsible for repression (Bustamante et al., 2001; Sánchez-SanMartín et al., 2001; Haack et al., 2003). As Ler does not exert a general anti-H-NS effect (Elliott et al., 2000), and can bind to the LEE promoter regions with greater specificity than can H-NS (our unpublished data), it most probably acts as an anti-H-NS factor at these promoters by displacing H-NS from its binding sites. Therefore, a slight destabilization of H-NS complexes on the LEE promoters by H-NSTEPEC could preferentially assist Ler to activate these promoters in the absence of a major effect on other H-NS-dependent repression. Indeed, we observe that even in the absence of Ler some derepression of LEE2, LEE3 and LEE5 can be effected by low-copy H-NSTEPEC, and this effect also extends to the LEE1 promoter which transcribes the ler gene itself (our unpublished data). Moreover, in the presence of H-NSTEPEC a low-copy Ler clone becomes a more effective activator of the LEE5 promoter, similar to the effect seen when the trans-activator of Ler, PerC (Mellies et al., 1999; Porter et al., 2004), is provided. Therefore, H-NSTEPEC may increase LEE activation under conditions of low Ler expression, although further work is required to assess the importance of this effect in vivo. Our studies on hnstTEPEC transcription in DMEM medium, which is optimal for the expression of LEE-encoded virulence factors (Kenny et al., 1997), suggest that the gene is transcribed at a low constitutive level during logarithmic growth. However, it is possible that certain environmental conditions may induce higher hnsT expression leading to more general effects on gene expression. We have been unable to detect increased hnsT transcription after exposure of EPEC to environmental shifts encountered in the host: osmotic shock, temperature upshift and low to neutral pH shift (our unpublished data). Searches for other conditions which may induce hnsT, together with studies on the phenotypes of mutations in the hnsT gene, are therefore required to clarify the role of this novel protein in gene expression in EPEC strains.

Unlike its counterpart from EPEC, the H-NSTUPEC protein is unable to exert a significant dominant-negative effect on H-NS: at best it achieves a very small derepressive effect when expressed maximally from the arabinose-inducible construct. This inactivity correlates with a lesser ability of H-NSTUPEC to interact with H-NS, as witnessed by the fact that only a proportion of overexpressed H-NSTUPEC co-purifies with H-NS. Presumably this weaker interaction is readily out-competed by H-NS–H-NS interactions unless H-NSTUPEC is expressed to very high levels. The difference in activity between the two H-NST proteins seems to be accounted for by a change of residue  16  from  alanine  to  valine  in  the  UPEC  protein, as the phenotype of H-NSTEPEC A16V is like that of H-NSTUPEC while H-NSTUPEC V16A has H-NSTEPEC-like activity. In either of the two proposed H-NS N-terminal structures, residue 16 (which is also alanine in H-NS) lies within the α2 helix outside the core dimerization interface (Esposito et al., 2002; Bloch et al., 2003), and may therefore instead affect higher-order complexes which we propose to be important for the anti-H-NS function of H-NST. The H-NST protein from EAEC strain 042, which we have not studied in detail, also contains alanine at position 16 (Fig. 2), as do the two H-NST proteins found in the recently completed genome sequence of E. carotovora strain SCRI1043 (Accession No. NC_004547). Therefore, the UPEC protein may have acquired an inactivating mutation, perhaps because its anti-H-NS effect was deleterious in the context of strain CFT073 gene regulation. The CFT073serU island which encodes hnsT also encodes the closely linked hnsB gene which, given its potent hns-like effect, may also counteract any H-NST-dependent effects by increasing the hns gene dosage in this strain relative to non-pathogenic E. coli strains. Interestingly, despite its obvious relatedness to the serU island, the E2348/69 asnW island lacks an hnsB gene, which may enhance the effect of hnsT in this strain. Although they appear more distantly related to the serU and asnW islands, the genomic islands encoding hnsT genes in EAEC 042 and E. carotovoraSCRI1043 also contain hnsB-like genes (our unpublished data); E2348/69 is therefore unique in lacking such a gene associated with hnsT. Given the probable origin of these islands as mobile DNA elements, there is also the potential for strain-to-strain variation in the hnsT (and hnsB) coding capacity of a particular pathogen. Indeed, EPEC strain B171-8 seems to lack an hnsT gene (M.E. Porter and A. Free, unpubl. data). This mirrors the situation in Shigella flexneri, where the sfh gene encoding an H-NS-like protein is found in strain 2457T on a plasmid that is also present in Salmonella typhi, but is absent in other Shigella serotypes (Beloin et al., 2003). Such variation has the potential to modify the gene regulation patterns, and hence the virulence capacity, of individual pathogenic strains.

H-NSTEPEC is novel because it is a naturally occurring H-NS-interacting protein with an anti-H-NS phenotype. Previously, the only example of such a protein was the gene 5.5 protein of bacteriophage T7 (Liu and Richardson, 1993). Unlike H-NSTEPEC, gene 5.5 protein is unrelated to H-NS at the sequence level, although it is of similar size (99 amino acids) to the N-terminal region of the protein and also co-purifies with cellular H-NS. Gene 5.5 protein  also  has  a  similar  anti-H-NS  effect  to  H-NSTEPEC on proU expression, and although it is not predicted to form a coiled-coil by the Lupas algorithm, an L30P mutation can also abolish the interaction and the anti-H-NS effect. It is most likely that this protein can form a similar interaction with H-NS to that of H-NSTEPEC which prevents correct H-NS oligomerization and hence transcriptional repression. Gene 5.5 of phage T7 is highly expressed, and the function of its product has been postulated to be sequestration of H-NS to prevent its interference with phage DNA and RNA synthesis (Liu and Richardson, 1993). It is therefore distinct from H-NST proteins, which are presumably integrated into the regulatory circuits of the pathogens which encode them and are likely to have more subtle effects on gene expression as a consequence of lower expression. Future studies on the distribution of H-NST proteins among bacterial species and their contributions to gene regulation in these organisms will be of great interest.

Experimental procedures

Bacterial strains and plasmids

The bacterial strains and plasmids used in this study are listed in Table 3, and the oligonucleotide primers used in their construction are shown in Table 4. Cloning was routinely performed in E. coli K-12 strain XL-1 Blue. The UPEC and EPEC hnsT loci plus flanking DNA were amplified from CFT073 or E2348/69 chromosomal DNA using primers HNSTEUF + HNSTEUR and HNSTUF + HNSTUR, respectively, and cloned into the low-copy vector pHSG576 after digestion with BamHI + HindIII  to  generate  pHSGHNSTE and pHSGHNSTU. To construct arabinose-inducible H-NST clones,  the  hnsTEPEC and  hnsTUPEC ORFs were  amplified from pHSGHNSTE and pHSGHNSTU using the primers HNSTEUFARA + HNSTEUR and HNSTUFARA + HNSTUR, respectively, digested with PstI + HindIII, and cloned downstream of the araBAD promoter in similarly digested pBAD33 to give pBADHNSTE and pBADHNSTU. Tagged H-NST derivatives for expression and purification (pET30HNSTE and pET30HNSTU) were made by amplifying the hnsT ORFs from pHSGHNSTE and pHNSHNSTU using HNSTEUF + HNSTEUER and HNSTEUF + HNSTEUR, respectively, digesting with NcoI + BamHI and cloning into pET30b; the entire tagged ORFs were also amplified with PETBADF + PETBADR and cloned into KpnI–HindIII-digested pBAD33 to give arabinose-inducible constructs (pBADH6SHNSTE and pBADH6SHNSTU) to test functionality of the tagged proteins. Tagged derivatives of the control proteins Hha and Ler were constructed by amplifying the genes from MC4100 or E2348/69 chromosomal DNA using primers HHAEF + HHAER and LEREF + LERER, respectively, then cloning into pET30b as described above. The hns ORF was amplified using primers HNSEF + HNSER2, digested with NcoI + BamHI and ligated into pET6H to generate a His6-tagged expression clone, pET6HHNS. The UPEC hnsB gene was amplified from CFT073 chromosomal DNA using primers HNSBUFD/S + HNSBUR, digested with BamHI + HindIII and cloned into the single-copy vector pJK289 to generate pJKHNSB. A His6-tagged expression clone, pET6HHNSB, was generated from this using primers HNSBUEF + HNSBUER, followed by NcoI–BamHI digestion and cloning into pET6H. All constructs were verified by sequencing the entirety of the inserted DNA.

Growth conditions and enzyme assays

For analysis of proU–lacZ and bgl–lacZ transcription, strains were grown overnight in LB medium at 37°C unless indicated, and assayed for β-galactosidase activity as described by Miller (1992). Cultures were assayed in duplicate and the assays  repeated  at  least  twice.  Standard  deviations  were less than 10%. Antibiotic selection with chloramphenicol (20 µg ml−1), ampicillin (50 µg ml−1) or kanamycin (50 µg ml−1) was as appropriate. For assay of the bgl–lacZ fusion, the gratuitous inducer β-methyl- d-glucoside was added at a final concentration of 5 mM. Arabinose-inducible H-NST derivatives were induced with arabinose concentrations in the range 0.00002–0.2%.

Transduction with bacteriophage P1cml

The hns-206::Apr and hns-205::Tn10 alleles were transduced from the donor strains PD32 and GM230, respectively, into the bgl–lacZ strain BGL1 and the proU/V–lacZ strains GM37 and HM60 (hns-206::Apr only) using bacteriophage P1cml as described by Silhavy et al. (1984). The φ(proU–lacZ)hyb2 fusion from GM37 was similarly transduced into the araD+ strain TP8503 by selecting for the kanamycin resistance encoded by this fusion.

RNA extraction, probe synthesis and Northern blotting

Total cellular RNA was extracted from cultures grown to the indicated OD600 in high-glucose DMEM medium containing Hepes and lacking phenol red (Gibco-Invitrogen Corporation, Cat. No. 21063-029) by cell lysis in boiling RNA Extraction Buffer [20 mM sodium acetate (pH 5.2), 2% SDS, 0.3 M sucrose] followed by phenol extraction and DNase I treatment as described previously (Free and Dorman, 1994). An internal fragment (175 bp) of the hnsTEPEC gene was amplified with the primer pair HNSTERF + HNSTERR, digested with  PstI  and  EcoRI  and  cloned  into  the  in  vitro  transcription vector pSPT18. The resulting plasmid construct (pSPTHNSTE) was linearized with HindIII and transcribed in vitro with T7 RNA polymerase to generate a digoxigenin-UTP-labelled hnsT-specific RNA probe using a DIG-RNA labelling kit supplied by Roche Molecular according to the manufacturer's instructions. Samples of total cellular RNA (20 µg) were electrophoresed on MOPS-formaldehyde-agarose gels, transferred to Hybond-N+ positively charged membranes (Amersham), and hybridized with the labelled RNA probes overnight. After stringency washes, bound probes were detected with anti-digoxigenin-alkaline phosphatase conjugate and the chemiluminescent substrate CSPD (Roche Molecular) as described previously (Free and Dorman, 1994).

5′ RACE analysis of hnsT transcripts

Total cellular RNA from EPEC strain E2348/69 grown to mid-logarithmic phase (OD600 = 0.6) in DMEM (see above) was used as a template for first-strand cDNA synthesis employing a SMARTTM RACE cDNA amplification kit (BD Biosciences) as described by the manufacturer, except that the oligo (dT) primer supplied with the kit was replaced with the hnsT-specific primer TERACE1, which primes from 224 bp downstream of the hnsT initiation codon. RACE PCR was then performed with the kit-specific Universal Primer Mix and the hnsT-specific TERACE2 primer, and nested PCR on these products was carried out using Nested Universal Primer A from the kit and the TERACE3 primer, which primes from 161 bp downstream of the hnsT initiation codon. These steps were performed as described by the kit manufacturer, except that an annealing temperature of 55°C was used. Products were analysed on a 1% agarose gel, and the ≈200 bp band was purified from the gel and sequenced directly using the TERACE3 primer.

Protein purification and mobility shift assays

Five hundred millilitre cultures of BL21 (λDE3)/pLysS transformed with pET30HNSTE, pET30HNSTU, pET30HHA, pET30LER, pET30HNSTEL30P, pET6HHNS or pET6HHNSB were grown to mid-log phase (OD600 ≈ 0.6) in LB under kanamycin or ampicillin selection as appropriate, then induced with IPTG at a concentration of 1 mM for 3 h. Cells were pelleted by centrifugation, washed in 40 ml of 50 mM Hepes, 0.1 M NaCl, and resuspended in 15 ml of His-Tag binding buffer (5 mM imidazole, 0.25 M NaCl, 20 mM Tris pH 7.9, 10% glycerol, 0.1% Triton X-100, 10 mM β-mercaptoethanol). After cell lysis by sonication (MSE Soniprep 150; three to six pulses of 1 min at 5 µm amplitude interspersed with cooling on ice), cell debris was pelleted by centrifugation at 16 000 r.p.m. for 20 min, and the cleared supernatants were loaded onto columns of Ni2+-NTA agarose (Qiagen) equilibrated with His-Tag binding buffer. The columns were washed twice with 15 ml of His-Tag binding buffer supplemented with 30 mM imidazole to remove weakly bound proteins, and the specifically bound proteins were then eluted with His-Tag binding buffer supplemented with 0.5 M imidazole. Eluted proteins were dialysed against 20 mM Tris pH 7.9, 50 mM KCl, 5 mM NaCl, 1 mM EDTA, 1 mM DTT, 20% glycerol, 0.1% NP-40 overnight, and then stored in aliquots at −80°C. Fragments for use in mobility shift assays were amplified with primer pairs PPROU1 + PPROU2 (amplifying between −176 and +5 with respect to the proU start codon) and PPROU3 + PPROU4 (amplifying between −16 and +174 with respect to the proU start codon) and end-labelled with digoxigenin-ddUTP using an end-labelling kit (Roche Molecular) according to the manufacturer's instructions. Approximately 1.2 ng of labelled probe was mixed with purified protein in binding buffer (10 mM Tris pH 7.5, 1 mM EDTA, 5 mM NaCl, 50 mM KCl, 8% glycerol, 0.05 mg ml−1 BSA, 1 mM DTT) in the absence of non-specific competitor DNA (which effectively competes even strong H-NS binding sites in vitro) and incubated at room temperature for 20 min. Protein–DNA complexes were electrophoresed on native 6% polyacrylamide gels at 100 V, 4°C for 5 h and then transferred to Hybond-N+ membranes (Amersham). The labelled DNA probes were then detected using the alkaline phosphatase/CSPD-based chemiluminescence system (Roche Molecular) as described for Northern blots. H-NSB binding to the proU promoter was assayed by competitive agarose bandshift using plasmid pOSEX2 (Table 3) digested with EcoRI + PstI + PvuII to yield a ≈700 bp proU promoter fragment, a ≈1.1 kb bla promoter fragment, plus vector fragments of ≈1.5 kb and ≈2 kb. These fragments (≈100 ng of proU promoter) were incubated with purified H-NSB in 40 mM Hepes pH 8, 100 mM potassium glutamate, 10 mM magnesium aspartate, 10% glycerol, 0.022% NP-40, 0.1 mg ml−1 BSA at room temperature for 20 min, then electrophoresed on a 3% Molecular Screening Agarose gel (Roche Molecular) in TBE buffer at 4°C overnight. After electrophoresis, gels were stained with ethidium bromide.

Site-directed mutagenesis of H-NSTEPEC

Site-directed mutagenesis of the hnsTEPEC and hnsTUPEC genes was performed using the pBADHNSTE template, oligonucleotide primers as listed in Table 4 and a Quikchange site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions. The L30P mutation required two base pair substitutions, which were introduced sequentially in two rounds of mutagenesis. To generate a tagged derivative of the L30P mutant protein, the pBADHNSTEL30P template was amplified with the HNSTEUF + HNSTEUER primer pair and cloned into pET30b as described for the wild-type hnsT gene.

Protein analysis and Western blotting

Purified protein samples (≈10 ng) were electrophoresed on 12% SDS-polyacrylamide gels and either analysed with a Silver Stain Plus kit (Bio-Rad) according to the manufacturer's instructions or electroblotted onto Hybond-ECL membranes (Amersham). Membranes were blocked with 5% protein-grade blotting reagent (Bio-Rad) in TBS-T (0.14 M NaCl, 20 mM Tris pH 7.6, 0.1% Tween 20), then incubated for 30 min with S-protein-HRP conjugate (Novagen) diluted 1:5000 in TBS-T to detect S-tagged proteins, or for 60 min with a monoclonal anti-H-NS antibody raised against amino acids 1–64 of Salmonella typhimurium H-NS (Sonnenfield et al., 2001) diluted 1:100 in TBS-T to detect H-NS. After stringency washes of 15 min and 3 × 5 min in TBS-T, bound S-protein conjugate was detected with the ECL Western Blotting Analysis System (Amersham) as directed by the manufacturer. The bound anti-H-NS antibody was detected after similar stringency washing by incubating for 60 min with anti-mouse IgG-HRP conjugate (Amersham) diluted 1:5000 in TBS-T, followed by a second round of stringency washes and ECL detection as for S-protein-HRP.

Measurement of bacterial motility

The swarming diameter of bacterial strains spotted onto tryptone swarm plates (Bertin et al., 1994) was measured after growth overnight (15 h) at 30°C.

Computational analysis

Protein and DNA sequences were aligned by the blast program (Altschul et al., 1990) at http://www.ncbi.nlm.nih.gov/BLAST/. Homologues of the UPEC H-NST protein in completed and incomplete microbial genomes were identified by tblastn searching at http://www.ncbi.nlm.nih.gov/sutils/genom_table.cgi, and the contigs containing the EPEC and EAEC hnsT genes (51 376 bp and 55 389 bp, respectively, as provided by the Wellcome Trust Sanger Institute) were downloaded. The sequence of the EPEC E2348/69 hnsT gene and its immediate flanks as determined by us has been submitted to GenBank (Accession No. AY684972). Coiled-coil predictions were performed via the Lupas algorithm (Lupas et al., 1991) using MacStripe 2.0b1 (http://www.york.ac.uk/depts/biol/units/coils/mstr2.html), with the MTK matrix, a window length of 28 and weighting of the hydrophobic positions.


We would like to thank the staff of the ICMB sequencing service for DNA sequencing, Padraig Deighan for providing strain PD32, Robyn Emmins and Garry Blakely for strain TP8503, and Seiichi Yasuda of the Cloning Vector Collection, National Institute of Genetics, Japan for the pBAD33, pHSG576 and pJK289 vectors. We are especially grateful to Megan Porter for many invaluable discussions, and to Julian Parkhill for discussions on chromosomal islands. This work was supported by a Wellcome Trust Research Career Development Fellowship (Grant No. 065574/Z/01/Z) to A.F.