BMC Microbiology BioMed Central

Background Coagulase-negative Staphylococcus epidermidis has become a major frequent cause of infections in relation to the use of implanted medical devices. The pathogenicity of S. epidermidis has been attributed to its capacity to form biofilms on surfaces of medical devices, which greatly increases its resistance to many conventional antibiotics and often results in chronic infection. It has an urgent need to design novel antibiotics against staphylococci infections, especially those can kill cells embedded in biofilm. Results In this report, a series of novel inhibitors of the histidine kinase (HK) YycG protein of S. epidermidis were discovered first using structure-based virtual screening (SBVS) from a small molecular lead-compound library, followed by experimental validation. Of the 76 candidates derived by SBVS targeting of the homolog model of the YycG HATPase_c domain of S. epidermidis, seven compounds displayed significant activity in inhibiting S. epidermidis growth. Furthermore, five of them displayed bactericidal effects on both planktonic and biofilm cells of S. epidermidis. Except for one, the compounds were found to bind to the YycG protein and to inhibit its auto-phosphorylation in vitro, indicating that they are potential inhibitors of the YycG/YycF two-component system (TCS), which is essential in S. epidermidis. Importantly, all these compounds did not affect the stability of mammalian cells nor hemolytic activities at the concentrations used in our study. Conclusion These novel inhibitors of YycG histidine kinase thus are of potential value as leads for developing new antibiotics against infecting staphylococci. The structure-based virtual screening (SBVS) technology can be widely used in screening potential inhibitors of other bacterial TCSs, since it is more rapid and efficacious than traditional screening technology.


Background
The genus Bordetella currently comprises nine species, most of which were found to be associated with host organisms [1,2]. Medically the most important species is B. pertussis, the etiological agent of whooping cough for which humans are the only known host. B. parapertussis causes milder forms of whooping cough-like disease in humans. B. bronchiseptica is known to cause respiratory disease in various mammalian species, but only rarely in humans [2]. These "classical" Bordetella species are closely related and the recent determination of their genome sequences confirmed previous suggestions that B. pertussis and B. parapertussis are independent descendents of B. bronchiseptica-like ancestors which during specialization to a single host have sustained a significant erosion of their genetic material [3]. In agreement with their close relationship these organisms produce highly related virulence factors such as several toxins and colonization factors [2].
Among these virulence factors the filamentous hemagglutinin FHA is of particular relevance for pathogenesis. It is an important adhesin and it is required for tracheal colonization in animal models [4]. FHA is a huge protein synthesized as a precursor of 367 kDa which is processed to a mature protein of about 220 kDa by extensive N-terminal and C-terminal modifications involving possibly the Lep signal peptidase and the subtilisin-like autotransporter protein SphB1 [5,6]. FHA is exported to the cell surface and/or secreted via a two-partner export mechanism requiring the FhaC protein located in the outer membrane of the bacteria [7,8]. It has several distinct binding domains involved in adhesion to different substrates. The carbohydrate recognition domain (CRD) probably enables the bacteria to attach to ciliated respiratory epithelial cells and to macrophages [9]. FHA exhibits lectin-like activity for heparin and dextran sulphate possibly involved in the interaction with nonciliated epithelial cells which also contributes to FHA-mediated hemagglutination [10]. Furthermore, there is an Arg-Gly-Asp (RGD) motif which enables FHA to adhere to human monocytes/ macrophages via the leukocyte integrin complement receptor 3 (CR3, alpha M beta 2, CD11b/CD18) [11]. The FHA proteins of the "classical" species are highly related, but not entirely functionally exchangeable. Recently, by use of a B. bronchiseptica strain expressing the B. pertussis FHA protein it was found that the heterologous protein could mediate adherence to various epithelial and macrophage cell lines in vitro. In contrast, in rat infection models significant differences in the host-pathogen interaction were noted for the mutant B. bronchiseptica strain suggesting significantly different activities of the closely related FHA proteins of the classical species and a role of FHA for host adaptation [12]. Two other proteins, FhaL and FhaS, with significant sequence similarity to FHA are present in the members of the B. bronchiseptica cluster, but their functional relevance in virulence is not yet clear [3]. As in case of most other virulence factors, the expression of the fhaB locus is controlled by the BvgAS two-component system which is responsive to environmental stimuli such as temperature or compounds such as MgSO 4 or nicotinic acid [13,14]. The architecture of the fhaB promoter and its activation mechanism by the BvgAS system has been investigated in much detail [15,16]. Finally, FHA is a protective antigen and it is included in acellular pertussis vaccines [2].
More recently, additional species were included in the genus Bordetella. In 1984, B avium, a respiratory pathogen of birds, was first described. The genome sequence of B. avium was recently established and revealed the presence of several factors orthologous to those of the "classical" Bordetellae including FHA and the BvgAS two-component system [17,18]. Among other "novel" Bordetella species occasionally isolated from patients with underlying disease such as B. hinzii and B. trematum, B. holmesii has gained most attention in the past years, since its 16S rDNA sequence suggested this organism to be most closely related to B. pertussis [2,19]. Moreover, B. holmesii is known to contain several copies of the insertion elements IS481 and IS1001, otherwise found only in B. pertussis and B. parapertussis, respectively [20,21]. In addition to its isolation from immunocompromised patients [22][23][24], B. holmesii was found to be able to cause whooping-cough like symptoms in otherwise healthy persons [25,26]. Very little was known about the virulence properties of this bacterium and attempts to identify virulence factors related to those of B. pertussis failed. Recently, we succeeded to identify the BvgAS two-component system of B. holmesii by PCR amplification with degenerate oligonucleotides [27]. Interestingly, in contrast to the close relationship of the 16S rDNA sequences of B. holmesii and B. pertussis, the B. holmesii BvgAS system was found to be more closely related to the orthologous BvgAS system of B. avium than to that of B. pertussis [27].
Based again on a PCR approach with degenerate oligonucleotides we attempted to identify additional putative virulence factors of B. holmesii related to those of B. pertussis. Here we describe the identification and initial characterization of an FHA orthologue of B. holmesii. We show that the FHA protein of B. holmesii is more closely related to that of B. avium than to the FHA proteins of the "classical" Bordetellae. Furthermore, we show that similar to FHA of all other species it is transcriptionally regulated by the BvgAS two-component system suggesting similar virulence strategies in the different Bordetella species.

Results and Discussion
Identification and sequence analysis of the fhaB gene of B. holmesii Based on short regions of highly conserved nucleotide sequences of the 5'-and 3'-ends of the fhaB genes of B. avium and B. pertussis oligonucleotide pairs Fha1F/Fha1R and Fha2F/Fha2R were designed and used for PCR amplification with chromosomal DNA of B. holmesii. For both primer combinations PCR products could be amplified. The sequence analysis of the amplified DNA fragments revealed significant similarity in particular in the deduced amino acid sequence to the respective fhaB sequences of the other Bordetellae. By PCR amplification using the primer pair Fha1F/Fha1R, we found that all B. holmesii strains tested (B. holmesii G7702, B. holmesii G8341, B. holmesii ATCC51541 and B. holmesii No1) harbour orthologues of this fhaB gene (data not shown). Using a genome walking strategy the DNA sequence of the fhaB gene of strain B. holmesii G7702 was completed. The fhaB coding region of B. holmesii starts with a GTG codon and comprises 8,793 bp with a coding capacity for a 304 kDa protein, which is smaller than FHA of B. pertussis (367 kDa) but larger than the respective protein of B. avium (273 kDa). The genome organisation of the fhaB locus of B. holmesii differs from that of the other Bordetella species. In B. holmesii, upstream of the fhaB gene an open reading frame (orfMP) is located with significant similarity to a conserved integral membrane protein of B. bronchiseptica (BB3004) and B. parapertussis (BPP3041) which is transcribed divergently. Directly downstream of the fhaB gene a copy of the insertion element IS1001 is located which is known to occur in B. holmesii and in B. parapertussis. In contrast, in the species belonging to the B. bronchiseptica cluster the fhaB gene is arranged between the divergently transcribed bvgAS locus and the fimbrial operon fimABCD which precedes the fhaC gene whose gene product is involved in the export of FHA. In B. avium an ORF encoding an ABC transporter is present upstream of fhaB, while the downstream gene organization is similar to the B. bronchiseptica cluster (Fig. 1) [3,17,18].
The overall similarity of the nucleotide sequence of the fhaB gene of B. holmesii with those of B. avium and B. pertussis is low (between 50% and 52%) explaining the failure of previous attempts to detect an fhaB orthologue in this species by Southern hybridization experiments (data not shown). Even in those regions of the B. holmesii fhaB gene (e.g. within the first 600 bp) in which the predicted amino acid sequences of the FHA proteins are quite conserved (see below), the nucleotide sequence similarity is quite limited (59%). On the level of the deduced amino acid sequences, the B. holmesii FHA protein shows an over- avium. Region A comprises approximately amino acids 1-850, region B amino acids 950 to 1,640, region C amino acids 1,900 to 2,600, and region D amino acids 2,830 to 2,930 of the B. holmesii FHA protein; these regions correspond to amino acids 1-820, 840-1,540, 1,600-2310, and 2,520-2621 of the B. avium protein, respectively. all similarity of 55% to the FHA protein of B. avium (Fig.  1). The similarity to the FHA, FhaL and FhaS proteins of the members of the B. bronchiseptica complex is less pronounced, e.g. 39.2% to FHA of B. pertussis as calculated using the EMBOSS pairwise alignment algorithms http:// www.ebi.ac.uk/emboss/align/ (Table 1). With a sequence similarity of 37.3% the most similar protein outside of the genus Bordetella is an adhesin-like protein from Yersinia frederiksenii, an environmental bacterium that can also be associated with a wide variety of host organisms [28]. This analysis shows that FHA of B. holmesii is more similar to the B. avium protein than to the related factors of the mammalian pathogens, in line with previous observations that the BvgA, BvgS and RecA proteins of B. holmesii and B. avium are more closely related to each other than they are to those of the B. bronchiseptica complex [29,30]. This contrasts previous data obtained by comparison of the 16S rDNA sequences, which placed B. holmesii as a new species closer to B. pertussis than to B. avium. On the other hand, recent evidence indicated that the gene encoding the 16S RNA of B. holmesii may have been acquired horizontally from B. pertussis [30].
For the Sec-dependent secretion across the cytoplasmic membrane, the B. holmesii FHA protein has an extended N-terminal signal sequence, which is very similar to that of the B. pertussis protein. In B. pertussis, the signal sequence is cleaved at an alanine at sequence position 71 probably by the Lep signal peptidase [6]. The B. holmesii FHA protein has also an alanine residue at this position suggesting its processing at this site after transport through the cytoplasmic membrane. In addition, the Nterminus of B. pertussis FHA harbours a domain about 250 amino acid residues in length which is essential for secretion through the outer membrane according to the twopartner secretion model, the so-called TPS domain [31]. The exact nature of the transport signal is not known so far, but two sequence motifs (NPNL and NPNGI) are well conserved among two-partner secreted proteins and at least the NPNL motif plays a role in the secretion of B. pertussis FHA [32]. Both sequence motifs are also present in the FHA protein of B. holmesii suggesting that it is also exported via a two-partner secretion mechanism.
Similarities of the B. holmesii protein with domains of the B. pertussis FHA protein possibly involved in adhesion including the heparin binding (aa 442-862) and carbohydrate recognition domains (CRD) (aa 1141-1279) are very limited and the elucidation of relevant binding activities must await the biochemical characterization of the B. holmesii FHA protein. An intriguing feature of the FHA protein of B. pertussis is the presence of an RGD motif (aa 1097-1099) enabling the protein to interact with integrin receptors [2]. The B. holmesii protein does not contain an RGD motif, instead at sequence position 742-744 it harbours a KGD motif (Fig. 2). In some cases, KGD motifs may be involved in integrin binding [33,34], however, ascribing such a function to the KGD motif of the B. holmesii protein must await future experimental analysis.

Regulation of expression of the B. holmesii fhaB gene
Upstream of the fhaB gene a divergently transcribed gene is present suggesting that the fhaB gene has a promoter of its own. To analyse the regulatory region of the fhaB gene (P fhaB ) it was cloned upstream of a promoterless gfp gene in the broad host range vector pMMB208 and the resulting plasmid was introduced into B. holmesii. By primer extension analysis with a gfp-specific primer three transcripts (P1 -P3) were identified initiating at sequence positions -58 (P1), -71 (P2) and -88 (P3) with respect to the translational start codon of the fhaB gene (Fig. 3). No obvious -10 and -35 regions could be observed upstream of any of these putative transcriptional start sites. To investigate whether, similar to the other Bordetellae, the BvgAS system is involved in the control of fhaB expression, primer extension analysis was performed on RNA extracted from a B. holmesii bvgAS mutant [27] harbouring the plasmid-borne P fhaB -gfp fusion. Interestingly, the longest of the three transcripts (P3) was not detected in the bvgAS mutant indicating that the synthesis of this transcript is controlled by the BvgAS system, while the other transcripts are constitutively produced under our experimental conditions (Fig. 3). Similarly, the fhaB gene of B. pertussis is controlled by a BvgAS regulated promoter, but some constitutive expression was noted [35].
Although the fhaB gene is transcribed in B. holmesii grown at standard culture conditions, various attempts to detect . It is therefore possible, that the translation efficiency of the fhaB gene is low, which may be in line with the fact that the open reading frame starts with a GTG codon, or that the protein is processed to smaller fragments than the related proteins of the other Bordetellae.
To further investigate the transcriptional regulation of the fhaB gene by the BvgAS system we performed DNA binding experiments in vitro with purified recombinant BvgA BH of B. holmesii [27]. In fact, in band shift experiments binding of the phosphorylated but not of the unphosphorylated BvgA BH protein to the fhaB upstream region could be detected (Fig. 4). Binding was specific since addition of unspecific competitor DNA did not interfere with binding of BvgA BH -P even in the presence of a 1,000 fold excess of competitor DNA (data not shown). To further characterize BvgA binding to the promoter region, DNase I footprint analysis with BvgA BH of B. holmesii was performed. Footprint experiments were carried out on a 312 bp DNA segment ranging from nucleotide position +29 to -283 as numbered with respect to the translational start site of the fhaB gene. The addition of BvgA BH -P to the reaction mixture resulted in a large region protected from DNase I digestion ranging from position -40 to -243 with respect to the start codon of fhaB. Within the protected region the appearance of a regular pattern of hypersensitive sites every 10 to 11 nucleotides could be observed (Fig. 5), a phenomenon which was noted previously in the case of the promoter of the bvgAS operon of B. holmesii [27]. Surprisingly, the protected area covers all three transcriptional start sites mapped by primer extension analysis and, accordingly, includes the corresponding core promoter elements. It is not clear whether this observation has in vivo relevance.
A search for putative BvgA binding sites within the fhaB promoter region revealed the presence of four sequence motifs termed BS1 to BS4, respectively, with similarities to the well defined BvgA binding sites in B. pertussis (Fig. 6 affinity for BvgA BP binding centered at position -88.5 relative to the transcriptional start, as well as two additional low affinity binding sites centered at position -67.5 and directly adjacent to the -35 region [14,15]. Cooperative binding of BvgA BP to the secondary binding sites which show only limited similarity to the high-affinity inverted repeat motif is required for full transcriptional activation of the fhaB promoter of B. pertussis [15,35]. Remarkably, the positions of the low-affinity BvgA BP binding sites in the fhaB promoter of B. pertussis and of BS3 and BS4 in the upstream region of fhaB from B. holmesii are almost identical. However, while the high-affinity binding site in the fhaB promoter of B. pertussis is located immediately 5' adjacent to the low-affinity sites, the centers of the inverted repeat sequences BS2 and BS3 are located in a distance of 51 bp. The most prominent sites showing hypersensitivity to DNase I cleavage in footprint experiments map to the region flanked by BS2 and BS3 (Fig. 5).
To investigate a functional role of these putative BvgA binding site(s) we performed a deletion analysis of the fhaB promoter region and carried out band shift assays with progressively 5'-truncated DNA fragments lacking BS1 to BS4. As shown in Fig. 7, BvgA BH -P bound equally well to DNA fragments comprising the four putative BvgA binding sites BS1 to BS4 and to a DNA fragment lacking BS1 suggesting a negligible role of BS1 for the activation of the BvgA-P dependent promoter of fhaB. In agreement with this assumption, BS1 was only partially protected by BvgA BH -P binding in DNase I footprint experiments (Fig.  6). When BvgA BH -P was incubated with a DNA fragment lacking both BS1 and BS2 still a distinct DNA-protein complex was formed, however, binding was significantly weaker since a much higher amount of BvgA BH -P was required to achieve a band shift. No significant binding of BvgA BH -P was detectable to DNA fragments containing only BS4 or no BS box at all. These data suggest a functional role of the highly similar BS2 and BS3 sites for binding of BvgA BH -P to the fhaB upstream region.
To test whether these in vitro data have also relevance in vivo, DNA fragments containing various pieces of the fhaB upstream region were cloned in a promoterless gfp expres-Binding of BvgA BH to the fhaB upstream region of B. holmesii sion vector and transferred to the B. holmesii wild type and the bvgAS mutant strain. GFP expression directed from the different constructs was used as a measure for promoter activity. GFP expression was strong in the B. holmesii wild type strain harbouring construct pMMB208-fhaP-gfp0 containing the entire fhaB upstream region, while very low amounts of GFP were detected in the bvgAS mutant harbouring the same plasmid (Fig. 8, compare lanes 1 and  2). These data confirm the BvgAS mediated regulation of fhaB expression which was already observed on the transcriptional level (Fig. 3). Moreover, corroborating the in vitro data, GFP expression was virtually absent in strains B. holmesii G7702 (pMMB208-fhaP-gfp4) and B. holmesii G7702 (pMMB208-fhaP-gfp6) whose gfp expression plasmids contained only site BS4 or did not contain a BS site at all (Fig. 8, lanes 5 and 6). In agreement with the results of the DNA binding experiments a dramatic increase in GFP expression could be noted when the fhaB upstream region in the gfp expression plasmids comprised BS3 or BS2 and BS3 in addition to BS4 (Fig. 8, lanes 4 and 3). The virtually identical GFP expression directed from the pMMB208-fhaP-gfp3 (containing BS3 and BS4) and pMMB208-fhaP-gfp2 (containing BS2 to BS4) plasmids was surprising since the EMSA studies reported above revealed a relatively weak binding of BvgA BH -P to a promoter fragment containing BS3 and BS4, while binding to a fragment comprising BS2 to BS4 was very efficient suggesting a prominent role of BS2 for transcriptional activation. Since BS3 is fairly similar to the consensus BvgA BP binding motif and is located at the appropriate position to allow the interaction between BvgA BH and the C-terminal domain of the α subunit of RNA polymerase, BvgA BH binding to BS3 facilitated by DNA topology effects might be sufficient to fully activate the plasmid-borne fhaB pro-moter in pMMB208-fhaP-gfp3. In the presence of the full length fhaB promoter cooperative protein interactions are likely be involved in the binding of BvgA BH to the BS2/BS3 region.
To investigate whether the fhaB promoter of B. holmesii is also recognized by the BvgA protein of B. pertussis, the pMMB208-fhaP-gfp0 plasmid containing the entire promoter region of the B. holmesii fhaB gene fused to GFP was introduced into the B. pertussis strains Tohama I (TI) and BP359. Strong GFP expression was observed by immunoblot analysis in the wildtype strain TI (pMMB208-fhaP-gfp0), while expression of the reporter gene was hardly detectable in the bvgAS mutant BP359 (pMMB208-fhaP-gfp0) (data not shown). Interestingly, primer extension analysis performed with RNA extracted from TI (pMMB208-fhaP-gfp0) and BP359 (pMMB208-fhaP-gfp0) using a gfp-specific oligonucleotide demonstrated that in the wild type strain transcription of gfp starts at two sites, which, however, are identical to the start sites of transcripts P1 and P3 synthesized from constitutive (P1) and bvg-dependent (P3) dependent promoters in B. holmesii. Moreover, as observed in B. holmesii, in B. pertussis the promoter directing the synthesis of transcript P3 is not transcribed anymore when the BvgAS system is inactivated (data not shown). These data suggest that the activation mechanism of the fhaB promoter of B. holmesii by the BvgA proteins of B. holmesii and B. pertussis is remarkably similar. This is surprising since it was recently shown that the BvgA protein of B. holmesii does not bind to and cannot activate the fhaB promoter of B. pertussis, although in particular in its C-terminal output domain it is highly related to the BvgA protein of B. pertussis [27,36].
Immunoblot analysis of protein lysates of B. holmesii strains with a polyclonal anti-GFP antiserum