Analysis of nearest neighbor genes in the gamma subdivision of proteobacteria reveals a conserved gene arrangement surrounding the porin protein gene ompF in species closely related to E. coli (Figure 1). OmpF is flanked on one side by aspC, the aspartate aminotransferase gene and by asnS, the asparaginyl-tRNA synthetase gene on the other side. In addition, pncB (nicotinate phosphoribosyltransferase) and pepN (aminopeptidase N) are situated to the right of the ompF gene locus in the schematic shown. Similar gene arrangements for the ompF locus are also in more distantly related organisms, e.g., Buchnera aphidicola str. Sg. The loci in Yersinia species (not shown) have the same gene arrangements, but ompF is referred to as a general porin gene, e.g., see yptb1435, Yersinia pseudotuberculosis IP 32953, complete genome annotation, 17. However an ompF ortholog has been located in this locus in Yersinia species 18. Xenorhabdus nematophilia, which has an ompF -like porin gene (opmP), displays the similar nearest neighbor gene arrangement as that in E. coli 19. X. nematophilia has a similar life cycle as that of P. luminescens.

In P. luminescens, an identical nearest neighbor arrangement is present around the gene annotated as ompN (Figure 1). ompN is flanked by aspC and asnS with pncB and pepN situated adjacent to asnS as is found in E. coli and the other species shown. However, Photorhabdus also has a hypothetical protein gene (plu1752) situated between ompN and asnS (Figure 1). Thus the gene annotated as ompN in P. luminescens has the enterobacterial ompF chromosomal locus signature with the exception of an inserted protein gene. Analysis of nearest neighbor genes and their conservation in related enterobacterial species has been reported before, e.g., for non- coding RNA genes 20. In addition Notebaart et al 21 have presented a detail analysis of nearest neighbor genes, protein amino acid sequence and gene function predictions.

A comparison of amino acid sequence identities of the P. luminescens OmpN with representative gamma-proteobacteria outer membrane proteins is shown in Table 1. The highest identity is with OpmP, the OmpF-like porin in X. nematophilia. However identities with other porin proteins in other bacteria are also high and a clear segregation of OmpF protein sequences is not apparent (Table 1). Due to evolutionary relatedness between many outer membrane proteins, there is high sequence identity between these proteins, but they are not necessarily orthologs of each other and identities between porin protein amino acid sequences is of limited use in gene annotation 18.

ompN is not uniformly found in the gamma-proteobacteria. For example, there is no annotation for the ompN gene in Yersinia pestis and a protein blast search using the E. coli OmpN protein sequence does not yield an orthologous protein in the Y. pestis genomic sequence. Additional characteristics of ompN in other species are discussed below.

mRNA 5' UTR sequences can be important markers in identifying protein coding genes 18. In the study here, a comparative sequence analysis of 5' UTRs and a comparison of mRNA 5' UTR secondary structures were performed. All parameters investigated suggest that ompN is the P. luminescens homolog of ompF.

The P. luminescens ompN mRNA 5' UTR was deduced from alignment of sequences upstream of the ATG coding start site with 5'UTR sequences of Yersinia species (Figure 2). A comparison with Yersinia sequences is pertinent since Yersinia and Photorhabdus species are closely related evolutionarily 1112. The P. luminescens ompN mRNA 5' UTR has a very high nucleotide sequence identity to the ompF mRNA 5'UTR of Yersinia pestis (89.5%). Interestingly, there is 100% sequence conservation at the 3' half of the 5'UTR from positions 61 – 111 (Figure 2). The Photorhabdus ompN 5' UTR also displays a high sequence identity with other gamma-proteobacteria ompF 5' UTRs (data not shown).

The putative P. luminescens ompN mRNA 5'UTR also shares signatures of the generalized ompF 5' UTR secondary structure. These signatures are highly specific to ompF mRNAs and consist of a long stem loop which starts at ~5–10 nt from the 5' end of the 5' UTR 232425 (Figure 3). The stem contains several bulged and looped out positions. In addition, there is also a small stem loop close to the 3' end that encompasses the Shine- Dalgarno (S-D) ribosome binding site (e.g., GAGG in E. coli and Y . pestis). The Photorhabdus ompN mRNA 5' UTR closely conforms to the gamma-proteobacteria ompF mRNA 5' UTR secondary structure motif. Figure 3 shows a comparison with the Y. pestis ompF mRNA 5' UTR secondary structure. There are minor differences in the upper portion of the large stem loop, however base pairing differences are characteristic of ompF mRNA 5' UTR structures 24. Secondary structures of E. coli ompC, ompN and other outer membrane protein mRNA 5' UTRs do not display the ompF 5' UTR signatures.

Another important signature of ompF is the conservation and specificity of its upstream regulatory sequences. OmpR, a transcription factor found in E. coli and related organisms, is part of the two-component signal transduction regulatory locus ompB 27228. OmpR transcriptionally regulates expression of ompF and ompC genes in response to osmolarity change in E. coli and related species. Four OmpR binding sites F1–4 are located upstream of ompF in E. coli 2930 (Figure 4a). Figure 4b shows an alignment of OmpR binding sites F1, F2, and F3 upstream of ompF from several species, including the proposed site upstream of P. luminescens ompN. F 1–3 sites in P. luminescens were deduced by the alignment of sequences upstream of the ATG translational start site of ompN and of ompF from the four enteric bacteria shown. The E. coli sites as described by Bergstrom et al 29 were used as a base line. These sites are involved in transcriptional activation in E. coli. There is a comparable sequence identity between Photorhabdus and the three enterobacteria ompF upstream sites F1–3 (Table 2). In addition, most nucleotide positions crucial for OmpR binding in E. coli 3132 are conserved in P. luminescens sites, e.g., A₄ C₅ and A₁₄C₁₅ in F1, A₃₄, C₃₅ in F2, and C₅₃ in F3. Conservation of these crucial positions suggests that the P. luminescens sites function in OmpR binding. A GNNNC motif found in OmpR binding elements 32 is present in the P. luminescens sequence, G₁₁AAAC₁₅, albeit there is only one copy which is in the F1 site (Figure 4b).

The F4 site differs from F1–3 in that it is involved in repression and not activation of ompF expression 29. An hypothetical F4 site was also located upstream of ompN in Photorhabdus based on alignment of sequences with the nt positions and sequence described for the E. coli F4 29 (Figure 5). However, F4 sequences have diverged and are more complex than sites F1–3, e.g., Y. pestis and Y. enterocolitica species reveal low identity with the other gamma-proteobacteria (e.g., 28.6% and 33.3% identity respectively, compared with the E. coli F4 site) (Table 3). Although this is close to random identity, the Yersinia species have a GNNNC motif. The identity between Photorhabdus and the E. coli F4 sites is much higher, 52.4%, however the putative P. luminescens F4 lacks the GNNNC motif (Figure 5). Thus, it is unlikely that this sequence functions as a repressor of the proposed P. luminescens ompF.

Structurally, OmpF consists of trans-membrane beta strands and eight loops. Loop 3 (L-3) of OmpF is critical in formation of the channel pore and conductance and thus is a major functional determinant of the porin protein 3334. The amino acid sequence of L-3 of the proposed P. luminescens OmpF-like porin was aligned with L-3 sequences from related bacteria (Figure S1, ,Supplement, Additional file 1). The Photorhabdus sequence displays a high identity to the L-3 loop sequences of X. nematophilia OpmP (88% identity) and the Yersinia OmpF L-3 (82%) (Table S1, Supplement, Additional file 1). The sequence identity to the E. coli OmpN L-3 is 68%. Interestingly, there are substantial differences between the other gamma- proteobacterial OmpF L-3, especially between E. coli and Salmonella (49% Identity) (Table S1, Supplement, Additional file 1). This suggests a complex selection pressure in amino acid sequence in related gamma enterobacteria and possible structure-function differences of this critical loop. An L-3 phylogenetic tree is in Figure S2, Supplement, Additional file 1).

In summary, a) high identity with ompF mRNA 5' UTR sequences, b) shared mRNA 5' UTR secondary structure motifs, c) similarities between neighbor genes in the P. luminescens ompN locus and genes of the gamma-proteobacteria ompF locus, d) high sequence identities in upstream transcriptional regulatory sites F1–3, and e) identity of OmpF loop L-3 amino acid sequences all suggest that the annotated ompN gene is an ompF-ortholog in P. luminescens.

ompC, which codes for the smaller porin protein was not detected in P. luminescens. For example, ompC nearest neighbor genes such as ubiG, gyrA, rcsC, and rcsB are grouped together in E. coli and Yersinia pestis and are also conserved as a group in P. luminescens, but the Photorhabdus locus does not show an annotated ompC 15. Blast searches of the P. luminescens genome using the ompC coding sequence, the ompC mRNA 5' UTR sequence, and the ompC amino acid sequence from E. coli and Y. pestis did not yield sequences orthologous to ompC (data not shown). In addition, various other E. coli porin protein sequences were employed to search the P. luminescens genome using the NCBI protein tblastn program, however no other porin protein sequences were detected with the exception of the proposed ompF and ompF duplication plu1752.

In E. coli and related species, the micF gene is an independent transcriptional unit with its own promoter and it does not overlap other genes 35. However, micF and ompC are in the same chromosomal locus and are linked by a regulatory region (253 bp in E. coli) that includes binding sites for the transcription factor OmpR 2. These serve as sites for both micF and ompC transcriptional activation by OmpR 36. This linkage region is conserved in all enterobacteria known to contain a micF gene 18. The region is complex and has binding sites for eight transcription regulators of E. coli micF 31837; also, see 3839. In P. luminescens, both the micF and ompC genes are missing from the chromosomal locus where they normally are found. micF may have been transposed to another region of the chromosome, however blast searches with micF gene sequences from E. coli and Yersinia species, as well as with the conserved 253 nt regulatory region did not reveal homologous sequences in the P. luminescens genome.

Strategies employed to further search the P. luminescens genome for micF were based on conserved portions of the micF sequence and phylogenetically conserved micF RNA/ompF mRNA 5'UTR duplex structural motifs. In all organisms where micF RNA is found (more than 6 species), there is a total conservation of the13 nt sequence at the 5' end 18. This sequence is ^5'GCUAUCAUCAUUA^3' and it forms a major part of the base pairing between micF RNA and ompF mRNA 5' UTR. A hypothetical P. luminescens micF RNA would consist of the same sequence based on the highly conserved portion of ompF mRNA 5'UTR (see Figure 2), with the exception of C₂ to U₂ change (Figure 9).

Intergenic regions (i.e., non protein coding regions) of the P. luminescens genome were scanned for sequences that provides a structure similar to the RNA/RNA duplex structure found in enterobacterial species. Permutations in the hypothetical P. luminescens 5' end micF sequence shown in Figure 9 involved the use of four random substitutions in this sequence (in addition to the two C to U variables shown in Figure 9 left). Also, added structural constraints for the total hypothetical micF RNA structure used consisted of a rho-independent termination motif at 35–85 nt downstream from the 3' end of the 13 nt sequence. The P. luminescens genome was scanned and sequences analyzed. A match with forty- three positives was found for a putative micF RNA with the above mentioned constraints. These positives were further analyzed for ability to form an RNA/RNA duplex structure that is consistent with the conserved generalized micF RNA/ompF mRNA 5' UTR duplex structures of various enterobacterial sequences 18. Five positives were found, however four of these have sequences in the 3' region of the hypothetical RNA chain that partly overlap protein coding genes and thus were discounted. The fifth sequence did not overlap a coding region and provided a similar interaction displayed by the Yersinia RNA/RNA duplex (Figure S3a, S3b, Supplement, Additional file 1). However a) this structure did not display the ompF 3' end/micF 5' end RNA/RNA duplex motif consisting of "blunt ends" (Figure 9, right) 18, and b) the percent identity of this fifth sequence with the Y. pestis micF sequence is low, 35.2%. Therefore this "fifth positive sequence" was also discounted as a micF candidate, albeit we cannot exclude that this may constitute an RNA transcript from promoter search analyses (Supplement, Additional file 1). No other intergenic sequences provided all the prerequisite RNA/RNA duplex signatures. A third type of search was performed with a hypothetical P. luminescens micF sequence that would form a perfect 14 bp complement to the ompF 5'UTR 3' region (see Methods and Materials) and no positives were found in a search of the P. luminescens genome.

Thus, a) no micF-conserved 5' end 13 nt sequence (nt sequence in Figure 9, right), which has a rho -independent termination motif at less than 85 nt downstream of the 13 nt sequence, matched in the intergenic regions of P. luminescens, b) no P. luminescens intergenic sequences provide all the prerequisite signatures of the micF RNA/ompF mRNA 5' UTR duplex. In addition, the P. luminescens chromosomal locus where one would expect to find micF is partly rearranged and does not contain micF (or ompC).

If a putative P. luminescens micF RNA/ompF mRNA 5' UTR duplex structure diverged significantly from the phylogenetically conserved structures used for comparison, a Photorhabdus micF sequence would have been missed. Since the P. luminescens ompF mRNA 5'UTR has a secondary structure that closely conforms to the generalized 5' UTR structure and the 3' end of this UTR sequence that would interact with a hypothetical micF RNA is, with the exception of one nt change, completely conserved, it is less likely that a Photorhabdus RNA/RNA duplex structure would exist that diverges significantly from the consensus. The E. coli and Y. pestis intergenic regions of the chromosome were also scanned using the above parameters and the respective micF sequences were detected in these controls.

The regulatory micC RNA was isolated and characterized in E. coli and homologous micC sequences were detected in close relatives such as Klebsiella pneumonie, but not in Yersinia species 4. This RNA is 109 nt in chain length and regulates ompC expression post-transcriptionally in response to stress conditions 4. Since the target ompC gene does not appear to be present in P. luminescens, it appears unlikely fhat P. luminescens would have micC. However we considered the question of whether "fossil" micC sequences may still be present. The P. luminescens genome, including the genomes of Y. pestis and Y. psedutuberculosis (which were not previously shown to have micC), were scanned for sequences that match E. coli micC. Blast searches with the NCBI Discontinuous Blast Program and utilizing the E. coli micC sequence as the query did not yield a match.

micC nearest neighbor genes were analyzed as well. In E. coli K12, ydpK (pyruvate-flavodoxin oxidoreductase gene) is situated 30 bp downstream from micC and ompN is 227 bp upstream of micC 4. ydpK has not been annotated in P. luminescens and a blast search of the P. luminescens genome using the Y.pestis pyruvate-flavodoxin oxidoreductase amino acid sequence yielded only minor similarities to unrelated proteins. Thus we could not locate a locus encompassing the micC nearest neighbor genes in the P. luminescens genome.

Since Yersinia species have an ompC, theirgenomes were further searched for a micC homolog by analyzing sequences in a putative micC locus. ydpK is present in Y. pestis and Y. psedutuberculosis but annotated as nifJ. A hypothetical protein (with conserved domains similar to adenine nucleotide alpha hydrolases) is the nearest downstream neighbor (the location of micC and ompN in E. coli K12). An analysis of the sequence immediately downstream of nifJ was made to search for a potential sequence that may represent a micC homolog in Y. pestis and Y. psedutuberculosis. The 275 bp sequence (positions 2662970-2663245) downstream of the end of 3' end of nifJ and upstream of the start of YPT 2254 (hypothetical protein) contains a probable rho- independent termination motif, however it did not yield significant sequence identity when aligned with the E. coli micC sequence (unpublished).

Udekwu et 10 characterized a small regulatory RNA termed micA that post- transcriptionally controls E. coli OmpA synthesis. The micA sequence was also identified in other enterobacteria, including Yesinia pestis 10 but not in more distantly related organisms.

OmpA, the target of micA, has been annotated in the P. luminescens genome at positions 2117794-2118900 40. The putative P. luminescens ompA mRNA 5' UTR sequence was deduced from an alignment of sequences upstream of the P. luminescens ompA ATG start with E.coli and Y. pestis ompA mRNA 5' UTR sequences (Figure S4, Supplement, Additional file 1). The 5' UTRs of the 3 species show a divergence in sequence, particularly that of P. luminescens, but the region encompassing ~30 nt at the 3' end of the mRNA 5' UTR, the site that interacts with micA RNA in E. coli 10, is highly conserved and shows an identity of ~80% between P. luminescens and E.coli and Y. pestis ompA mRNA 5' UTR sequences.

The P. luminescens genome was searched for sequences similar to micA. A blast search using the E. coli and Y. pestis micA nucleotide sequences did not yield a match. In E. coli and Y. pestis genomes, micA is found between luxS and gshA 10. To further search for the RNA gene, the nucleotide sequence between luxS and gshA in P. luminescens was scanned for homology to micA sequences. This region in P. luminescens was aligned with E. coli and Y. pestis micA sequences and a putative micA homologous sequence was identified (Figure 10). Position 52 shown in Figure 10 is the 5' start of the micA gene and positions 1–51 encompass the upstream promoter region. micA sequence identities between the four organisms shown are in the range of 58.8 – 61.8%. The 23 bp segment at the 5' end region of the putative P. luminescens micA, that includes sequences that would interact with the P. luminescens ompA 5' UTR, shows a slightly higher identity of 65.2% compared with both E. coli and Y. pestis micA 5' end regions. In addition, the upstream -10 and -35 promoter sequences are nearly identical between P. luminescens and Y. pestis (Figure 10).

The P. luminescens sequence displays the major signatures of micA RNA, i.e., a rho-independent termination motif at the 3' end and the potential of 5' end sequences to base pair with region of the mRNA 5'UTR that covers the S-D sequence. Secondary structure models of the putative P. luminescens micA RNA together with E. coli and Y. pestis micA RNA structural models are in Figure S5, Supplement, Additional file 1. Figure 11 shows proposed micA RNA/ompA mRNA 5' UTR duplex interaction for P. luminescens, together with duplexes from related enterobacteria. All duplex models were obtained by the mfold program 26. Although there are variations in duplex structures, the P. luminescens duplex shows close similarities to the other structures, e.g., shielding of the mRNA S-D ribosome binding site by the 5' end region of micA RNA, similarity in the length of the RNA/RNA duplex interaction and a similarity in 5' nucleotide position of micA RNA that participates in base pairing. Thus we propose that the sequence at positions 1451569 – 1451500 of the P. luminescens genome represents the micA gene homolog in P. luminescens.

By bioinformatics methods, Griffiths-Jones et al 42 and Hershberg et al 20 predicted and cataloged the small non-coding RNAs/RNA genes present in E. coli and related organisms. Seventeen are regulatory RNAs in E. coli (Table S2, Supplement, Additional file 1). Of the total small RNAs reported for P. luminescens on the Rfam website 43 (also D. Schones, Cold Spring Harbor Laboratory and N. Delihas, unpublished data), five genes encode regulatory RNAs (Tables S2 and S3, Supplement, Additional file 1). The work reported here adds micA to this list. Twenty-five E. coli ncRNAs are of undetermined function (Table S4, Supplement, Additional file 1) and P. luminescens shares four of these ncRNA genes.