The hybridization of the N. lactamica strains (Table 1) to a large proportion of the probes of the pan-Neisseria microarray-v2 is both a validation of the design strategy used for the microarray (see Additional file 3) and highlights the similarity of these species. This data can be easily visualized in its genomic context at http://www.compbio.ox.ac.uk/data, where GBrowse databases detail the N. gonorrhoeae strain FA1090 and N. meningitidis strain MC58 and Z2491 genome sequences, their annotations, microarray probe locations and sequences, and indicate those probes that hybridize with N. lactamica in this study. In general, with N. lactamica neither the intensity of the hybridization signals nor the percentage of the microarray probes hybridized significantly differs from that seen in similar N. meningitidis or N. gonorrhoeae comparative genome hybridizations (Figure 1).

A set of genes identified in the literature 8910 to be involved in the virulence or pathogenesis of the Neisseria spp. (see Additional file 1) were specifically assessed through manual interpretation of the microarray image data. The microarray scan images of each spot feature were individually analyzed to determine hybridization to the probes for this subset of genes. Most of these genes were identified in N. lactamica (Table 2). Of the 127 virulence-associated genes assessed (of which 83 are present in all of the 4 pathogen genome sequences), 85 were present in at least one of the strains of N. lactamica. Many of the 42 remaining pathogen-specific genes are not present in all of the pathogenic Neisseria spp. genome sequences. Based upon microarray hybridizations, 11 of these 'virulence-associated' genes that are universally present in the pathogen sequences did not produce hybridization signals from any of the N. lactamica strains tested (Table 3). Of these, analysis of the currently unpublished N. lactamica genome sequence of another strain of this species (The Wellcome Trust Sanger Institute – unpublished) indicates that divergent homologues of the pilus-associated genes pilC, pilD, pglD, and pglF are present in this sequenced strain of N. lactamica, as is a divergent version of the antigen-encoding ssa1 gene. For the remaining 6 genes, the equivalent genomic region could be found in each case, yet these genes were absent, both here (Figure 2) and in the genome sequence as a whole. In the cases of dca, virG, and the NMB1646 hemolysin gene, alternative genes are present in these locations (Figure 2, panels A, C, and E). These results suggest that of the total number of currently recognized 'virulence genes', excluding the influence of allelic differences, the presence of only six can now be considered to be pathogen-specific. Only two of these six genes, iga and dca, have previously been reported to be pathogen-specific 48911121314.

When the complete microarray dataset was interrogated, 2,137 probes hybridized to at least one of the strains of N. lactamica (see Additional file 2). Of these, 46 are strain specific, with the majority of these (33) being in the strain that was not part of the MLST set. The previously reported study of relatedness 1 of the neisserial species is probably heavily influenced by the genes that are shared between N. lactamica and the pathogenic Neisseria spp.

The N. lactamica strains used in this study possess several genes previously reported as pathogen-specific, including lipA, bioA, secY 9, mafA 15, and dsbA 915. Of the 2,845 probes on the pan-Neisseria microarray-v2, 1,473 will hybridize to orthologues present in all four of the pathogenic sequenced strains, and the remainder addresses strain- or species- specific genes or highly divergent alleles. Of the 1,473 probes to genes that are universally present in the four pathogens' genome sequences, 1,373 hybridize with the N. lactamica strains tested.

Following the design and public availability of the original pan-Neisseria microarray-v1 in 2002 (available from Prof. John Davies, Monash University, Australia) 16, the Bacterial Microarray Group at St. George's (BμG@S) designed a second, independent multi-strain Neisseria microarray that became available in 2004. During the time in which publication of this study was being pursued, the Bacterial Microarray Group published a study of the hybridization of several different Neisseria spp. to their microarray. Readers of other reports of neisserial microarray-based studies should note that the investigators at BμG@S have also chosen to call their microarray the pan-Neisseria microarray, but this is not the same microarray used in our study. In their study, Stabler et al 17 hybridized two N. lactamica strains (L13 and L18) to their microarray. We have previously used these strains for other studies 181920, but they were not included in the current study because they do not currently form part of a collection of known diversity and on their own do not represent a sufficient sample to address our experimental question related to N. lactamica. We had instead chosen to evaluate the gene complements of 12 MLST analyzed strains to assess the genes present in N. lactamica. To this data we added one other N. lactamica strain that we had used previously 1119. In addition, our study sought to identify genes that are present in N. lactamica, not those that are absent from the commensal species, as was the goal of the Stabler et al 17 study.

Several of the genes reported by Stabler et al 17 as being absent from the commensal Neisseria spp., including their two strains of N. lactamica, are present in our N. lactamica hybridizations (Table 4). Whether these are due to differences in strain gene complement, microarray design, or data interpretation cannot be determined. In cases where the gene is present in only a few of our N. lactamica strains (see Additional file 2), the former may be the case, but where the gene probe hybridization is present in the majority of our strains, the later two points must be considered.

For purposes of comparing their findings with ours, the reported pilE and associated silent pilS cassettes were not evaluated, since the possession of type II pili by N. lactamica 21 and pathogenic Neisseria strains such as N. meningitidis strain FAM18 22 are well known and cannot be considered a differentiating factor between the pathogenic and commensal Neisseria spp. In addition, the results presented in Table 3 of absent genes in Stabler et al 17 includes separately listed copies of identical genes: NMA1167 and NMA1999 (100% DNA ID; NMB0486/NMB0970/NMB1741); and 9 reported copies of IS30. Discounting the additional copies of genes and the pilE/pilS sequences, 38 genes are reported to be absent in the commensal Neisseria spp. by Stabler et al 17; in contrast, we find that 6 of these are present in our hybridizations with N. lactamica, and a further 14 produced positive hybridizations to our pan-Neisseria microarray-v2 with other commensal species (Table 4). This highlights the problematic nature of reporting an absence based on a lack of hybridization in microarray and similar hybridization-based studies, which is why our study was focused in the genes present in N. lactamica.

Of the remaining 18 genes reported by Stabler et al. that also did not give positive hybridization results in our study, four are within Minimal Mobile Elements (MMEs; NMB0415, NMB0468, NMB0470, and NMB1621), one (NMA0368) is lost upon the acquisition of XNG1985 to XNG1898, eight (NMB0230, NMB0304, NMB1737, NMB1738, NMA0510, NMA1218, NMA1219, and NMA1221) are associated with IS elements, and one is IS150. This is consistent with these being genuinely strain and species differentiating genes based upon other evidence, reflecting their strain-specific and mobile nature. In summary, of the 38 genes reported as absent in the commensals by Stabler et al., 20 are present in our hybridizations and 14 are probably genuine differences, but are associated with strain-specific mobile elements.

It is interesting to note that of the 6 pathogen-specific genes identified in our study (Table 3), iga, NMB1880, NMB1882, virG, and NMB1646 are not reported as absent by Stabler et al 17, however their absence in our hybridizations is consistent with their absence in the N. lactamica genome sequence (Figure 2). The apparent presence of IgA protease in their commensal Neisseria collection is particularly of note, however this is not discussed in their manuscript. In light of the previously established absence of IgA protease from N. lactamica 13, this observation must be assumed to be false positive data, either due to cross-hybridization or data interpretation issues.

Thirteen strains of N. lactamica (Table 1) were investigated for positive hybridization to gene probes on the pan-Neisseria microarray-v2. Twelve of the strains were provided by Julia Bennett and Martin Maiden from their MLST study 35 and were chosen by them as representative strains of the population structure of this species. Additional information on these strains, or representative strains from these STs, can be obtained from the Neisseria MLST Home Page [41]. A thirteenth strain of N. lactamica was obtained previously from David Stephens. This strain has been used by us in previous studies 1920 and was added to this study to determine if additional data could be obtained from this strain. Eight commensal Neisseria spp. were investigated for positive hybridization to probes for genes previously reported by Stabler et al 17 as producing negative hybridizations in their study using the BμG@S neisserial microarray. Neisseria cinerea, Neisseria elongata, Neisseria flavescens, Neisseria macacae, Neisseria mucosa, Neisseria polysaccharea, Neisseria sicca, and Neisseria subflava were obtained from Prof. David Stephens (Emory University), which have been used in previous studies 1119.

The expanded pan-Neisseria microarray-v2 6 was designed and generated using the same strategy as described previously for the pan-Neisseria microarray-v1 16. The microarray includes probes to the genome sequences of N. gonorrhoeae strain FA1090 (AE004969), N. meningitidis strains MC58 10, Z2491 36, and FAM18 (The Wellcome Trust Sanger Institute – unpublished), plus the N. gonorrhoeae strain MS11 Gonococcal Genetic Island 37, neisserial genes from GenBank/EMBL that are not in any of the genome sequences currently available (see Additional file 2 for GenBank Accessions), genes identified through our on-going investigations of MMEs 1619 (see Additional file 2 for GenBank Accessions), and antibiotic resistance cassette markers commonly used in mutagenesis. Each probe of 150–450 bp targets the most conserved (≥ 90% identical over at least 150 bp) non-repetitive region within each coding sequence, maximizing hybridization with common probes to orthologous genes in different strains and species. Any significant cross-hybridization was eliminated, where possible, or noted, thereby minimizing the potential for cross-hybridization with other genes, which would generate false positive results. At the time this microarray was designed and constructed there was no publicly available annotation for the N. gonorrhoeae strain FA1090 genome sequence, therefore the microarray probes were designed using our own annotation of this sequence. This is available to view against the genome sequence, with comparisons to the other annotations, showing the probe sequences, locations, and those hybridized by N. lactamica at http://www.compbio.ox.ac.uk/data. For a more detailed description of the probe design and availability of the pan-Neisseria microarray, see Additional file 3. The PCR product probes generated were suspended in Genetix Spotting Solution for Amine slides and printed with non-adjacent replicates in triplicate onto Genetix Amine microarray slides using a Genetix QArray Mini microarray printer with 150 μm aQu solid pins.

Chromosomal DNA was extracted using a phenol-based method 38 from the strains listed in Table 1, following growth on GC media (BD) with Kellogg and ferric nitrate supplements 39 at 37°C, 5% CO₂ (v/v) overnight.

20 μg of DNA was fluorescently labeled through direct incorporation of FluoroLink™ Cy3-dCTP or FluoroLink™ Cy5-dCTP (Amersham Pharmacia Biotech) using 5 units of DNA polymerase I, Klenow fragment (Bioline) and 3 μg random hexamer primers (Invitrogen) in a 50 μl reaction for 1.5 to 4 hours at 37°C. Unincorporated nucleotides and random primers were removed using QIAquick Nucleotide Removal columns (QIAGEN) according to the manufacturer's instructions.

Hybridizations were conducted in 4× SSC, 0.3% SDS under LifterSlips (Erie Scientific) at 65°C overnight. The slides were washed in 65°C 1× SSC, 0.05% SDS for 2 minutes, 0.06× SSC for 2 minutes, and fresh 0.06× SSC for 2 minutes before drying using an airbrush. Microarray slides were scanned using a ScanArray Express HT (Perkin Elmer).

Thirteen N. lactamica strains were assessed, twelve of which were selected for their diversity based on previous MLST analysis 40. The thirteenth strain was provided from the collection of Prof. David Stephens (Emory University). Each strain of N. lactamica was hybridized to two slides, was paired with different strains in each hybridization, and was labeled with Cy3 on one slide and Cy5 on the other. This generated triads of strains and slides; for example strain 'A'-Cy3 was hybridized with strain 'B'-Cy5, strain 'B'-Cy3 with strain 'C'-Cy5, and strain 'C'-Cy3 with strain 'A'-Cy5.

Microarray spot features were identified, quantified, and assessed using BlueFuse for Microarrays v2 (BlueGnome). Initially, all of the probes for a set of 127 previously described virulence genes for the pathogenic Neisseria (see Additional file 1) were visually checked for hybridization for each strain on both slides.

The complete dataset was then addressed using the BlueFuse for Microarrays (BlueGnome) pON score. The pON score non-comparatively evaluates the data from each microarray spot and uses this to report a probability of there being a hybridization signal for each spot. Unlike ratio-metric methods, this score is not dependent upon or influenced by the signal in the other channel. A pON score of zero indicates that there is no evidence for spot hybridization, while a score of one indicates that there is strong evidence that the spot has hybridized. The pON score is quantitatively calculated from two features of the data: the intensity of signal above background, and the circularity and uniformity that is characteristic of a microarray spot. First, the spot is evaluated to determine the proportion of pixels within it that are statistically inconsistent with the Bayesian calculated background noise, thus the amount of the spot that is determined to be hybridization signal. Next, since a simple signal score can be influenced by inconsistent background and other features that are not hybridized microarray spots, the characteristics of the spot feature are taken into account, i.e. whether it is round and the signal across it is uniform. This serves to discriminate between signals that are simply large, high intensity noise as opposed to those which are consistent with a hybridized microarray spot. The exact details of the calculation of this score are proprietary to BlueGnome. In CGH applications, it therefore provides an independent assessment of gene presence or absence in each channel. Each probe was printed onto the pan-Neisseria microarray slide in three non-adjacent locations, therefore three pON scores were available for each probe. The images were manually flagged within BlueFuse to remove all spots that were generating clearly artefactual data (such as due to dust and local hybridization effects), and the data removed from subsequent analysis. Regions of the slide were manually assessed to determine the range of pON scores that could consistently be relied upon to reflect a strong positive hybridization. Additionally, the pON scores were determined for control hybridizations to the pan-Neisseria microarray-v2 with the strains against which the probes were designed. For this study, the pON score for a positive probe hybridization was purposely selected to be highly stringent, at pON > 0.67, so that the genes called as present within N. lactamica could be relied upon to be present in this species; a lower pON score might reflect hybridization of probes to divergent alleles and orthologous genes of different function. The control hybridizations suggest that with the pathogenic Neisseria spp. pON scores > 0.49 are reliable, therefore the high threshold for this study should be very robust. When all three probe spot replicates had pON scores > 0.67 then the presence of that gene within the N. lactamica genome was accepted without further review of the slide image. For any other probes with pON scores of > 0.67 for one or two spots, the slide image was reviewed to ascertain whether hybridization was present, indicating gene presence in the strain. The results were then combined for all 13 N. lactamica strains. The data reported here is therefore from a pool of data that minimizes strain-specific influences on the data, either due to the absence of certain sequences in certain strains or the presence of differences within divergent gene families between strains. In all, with 8,535 probe spots on each microarray (2845 probes printed in triplicate), 17,070 probe spots were assessed per strain, for a total of 221,910 features that were evaluated in the whole study. This does not include the data for the other commensal Neisseria spp. hybridizations. The images for these were manually assessed for the presence of positive hybridization in the 32 gene probes that produced negative hybridizations in both our analysis of N. lactamica and that of Stabler et al 17.