A complete re-annotation of the C. jejuni NCTC11168 genome was performed resulting in the reduction of the total number of CDSs from 1654 to 1643. This reduction was due to the merging of adjacent CDSs or the removal of CDSs. Three CDSs originally designated as pseudogenes were removed as a result of merging with adjacent pseudogenes. CDSs designated as pseudogenes were also updated to reflect the complete amino acid sequence for the encoded protein regardless of expression. Phase-variable CDSs that contained an intersecting homopolymeric region between adjacent CDSs on separate frames were merged. This allowed the complete amino acid sequence for appropriate genes to be obtained regardless of phase. Re-interpretation of phase-variable CDSs resulted in removal of seven CDSs. CDS (Cj1520) was removed because of the recently discovered CRISPR structural moieties 20 (See Structural modifications section in Results & Discussion). In total, 11 CDSs were removed from the re-annotated sequence (Table 1). The accurate identification of all CDSs within the genome has implications for downstream applications, such as mutagenesis, microarray design and proteome analysis.

A systematic re-annotation of all CDSs was performed. For the purpose of this re-annotation, all CDSs with additional information have had an 'updated' note qualifier attached. This qualifier contains consistent free-hand descriptions on recently identified motifs, relevant similarity search results and any characterisation work carried out within Campylobacter species/strains or any orthologs in similar microorganisms. Additionally, the 'updated' note qualifier also contains reasoning for including 'putative' or not within the product function. Putative designations infer an accepted product function without definitive evidence. For each CDS, a full literature search was performed. In total, 64.5% of CDSs have had one or more literature qualifier added. Interestingly, from all the literature added (2092), 50.5% have been published after the year 2000. Considering there was no literature qualifier in the original annotation, this data illustrates the depth of research that has been carried out since 2000 and further supports the need to make use of this information in a re-annotation. Detailed statistics on genome modifications are given in Table 2. 18.2% of CDSs have had their product functions updated. 60.5% of CDSs with new product function have been designated with a different functional classification. Additional file 1 gives the outline of functional classification used in this annotation. This description was adopted from the Sanger Institute. A different functional classification may still be within the same field as the previous function, or may be in a completely new area. 97.8% of these CDSs with a new product function and a different functional classification were given a completely new type of functional classification. Additional file's 2, 3 and 4 give in-depth data on the change and distribution of CDSs within these functional categories. Importantly, the number of CDSs in the 'Unknown and other' category has been reduced by 122. Also, the number of CDSs in the 'Miscellaneous' category has risen by 77. This is attributed to the fact that a number of CDSs have new information relating to a product function from uncharacterised motifs and thus the CDSs were not placed into a specific category as yet.

Since the original annotation, significant new information has been derived on the genetic loci encoding the four main carbohydrate surface structures. The C. jejuni N-linked glycosylation pathway (not described in the original annotation), has been fully characterised 9121314. This re-annotation includes the nomenclature for the pglA-K (protein glycosylation) genes and has updated all product functions for genes Cj1119c – Cj1130c. The LOS locus (Cj1131c – Cj1152c) described in the original annotation was updated to include recent product functions and gene names including neuA1, B1, C1 and hldDE 21222324. The O-linked glycosylation loci (Cj1293 – Cj1342) involved in flagellar glycosylation, has been updated to include neu, pse and maf genes 891011. Finally, the capsule locus (Cj1413c – Cj1448c), has now been updated to include kps and hdd genes 567.

Additional genome-wide updates were also carried out, of which a large proportion entailed adding specificity to existing product function. For example, the identification of a new PFAM or PROSITE motif has allowed the product function to become further specified e.g. putative transport protein modified to putative MFS (Major Facilitator System) transport protein. A complete list of changes throughout the C. jejuni NCTC11168 genome is provided in Additional File 5.

Pseudogene identification is a challenging process where discrepancies exist between pseudogene assignment techniques 25. Identifiers include detection of Open Reading Frames (ORF) belonging to a single CDS on multiple frames, the presence of one or more stop codon within a CDS, and extra information from the biology of the microorganism. More recently, comparative genomics has been used as a technique for pseudogene assignment 26. The number of pseudogenes identified in the original annotation of the C. jejuni NCTC11168 genome was 20. We carried out a re-analysis on all pseudogenes in the NCTC11168 genome. The majority of revisions we carried out incorporated multiple features created from different coordinates on more than one frame. This process is often complicated with support needed from FASTA and TBLASTX search results. Completion of this re-analysis resulted in modification of 19 out of 20 pseudogenes (Table 3). The final pseudogene number was 19 due to the merging of two adjacent CDSs designated as pseudogenes (Cj0968/Cj0969).

An example of the difficulty and complexity associated with pseudogene designation is observed when viewing the CDSs Cj0522, Cj0523 and Cj0524 within C. jejuni NCTC11168. These three CDSs are represented as one whole CDS on a single frame within C. jejuni RM1221 (Cje0628). The three CDSs are large enough to be represented as individual CDSs and in C. jejuni NCTC11168 have been represented on more than one frame. The question can be asked as to whether these CDSs (which are intact in C. jejuni RM1221), represent a pseudogene in C. jejuni NCTC11168. Given the fact that in C. jejuni RM1221 these three CDSs do actually code for a product (Na/Pi-cotransporter, putative), it is more likely that they represent a pseudogene in C. jejuni NCTC11168. In this re-annotation, our intention was to carry out a full mark up of existing pseudogenes, however, the potential for a pseudogene has been noted.

The frequency and importance of pseudogene formation in microorganisms has attained added significance in recent years with the emergence of genome reduction theories and enhanced virulence through pathoadaptive mutations 2728. Recent studies have suggested that ever increasing non-functional genes are being identified within microorganisms and in particular are more common in genomes of recently evolved pathogens, than in their benign or free-living relatives 25. The number and type of predicted pseudogenes within C. jejuni NCTC11168 and C. jejuni RM1221 are compared in Additional File 6. Observing CDS location rather than CDS function was carried out for this comparison. This was to ensure variation in product function naming does not exclude identical pseudogenes, which are represented on the same row. Currently, the C. jejuni 81–176 genome has not been fully annotated so could not be used in this comparison. This is also the case for C. coli RM2228, C. lari RM2100 and C. upsaliensis 3195 which only have an estimation of pseudogene numbers based on a subset of genes 15. In C. jejuni NCTC11168, 63% (12/19) of the pseudogenes are shared with C. jejuni RM1221. In contrast to 19 pseudogenes in C. jejuni NCTC11168, C. jejuni RM1221 contains 47 pseudogenes. Assuming these are genuine pseudogenes this would imply C. jejuni NCTC11168 (1980 human isolate, UK) and C. jejuni RM1221 (2000 chicken isolate, US), share a core set of ancestral pseudogenes. Even with the variation of isolation dates, source and geographical location, there is substantial conservation of pseudogene type. It is speculative to suggest when and how the additional pseudogenes in C. jejuni RM1221 arose, or when and how the C. jejuni NCTC11168 genome lost CDSs as pseudogenes since divergence occurred.

The significance of pseudogenes in early genome annotations were frequently ignored, as these were considered as sequencing artefacts. However, given the recent realisation of the importance of pseudogenes in pathoadaptive mutations, a greater significance is placed on their identification 2629. An example of this is the re-analysis of Escherichia coli K-12, which has predicted an additional 161 from the original single pseudogene identified 28. The same study also indicated pseudogenes are continually generated, with existing pseudogenes being eliminated over a period of time 28. Pseudogenes can accumulate in the genomes of some bacterial species, especially those undergoing processes like niche adaptation, host specialisation or weak selection strength 30. Analysis of further Campylobacter strains and species along with additional epsilon proteobacteria species will aid our understanding on this emerging area of interest. Also, greater understanding of pseudogene dynamics and in particular innovative pseudogene identification techniques will yield more information about the actual number and purpose of these entities within microorganisms.

Phase-variable CDSs containing hypervariable regions were also analysed. The initial annotation gave a number of hypervariable sequences found within the C. jejuni genomic shotgun sequence 4. These hypervariable sequences are scattered throughout the genome, however, there is a large cluster within both the O-linked glycosylation and capsule loci. Further research on these loci have illustrated the impact of phase-variation on microorganism pathogenicity 113132. Table 4 shows phase-variable CDSs that have been modified.

As well as CDS updates, novel features were also added to the re-annotation. For example, the incorporation of the recently identified Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) regions within Campylobacter 203334. CRISPR regions are thought to be mobile elements. In conjunction with this, the three CDSs upstream of the CRISPR repeats were identified as CRISPR associated proteins and this concurs with existing CRISPR structures. To date, there has only been one identified CRISPR region within C. jejuni and this has now been incorporated within the genome. As a result, one CDS (Cj1520) has been removed. This CDS was previously annotated as having five repeat regions. Thus, the genome now contains a CRISPR repeat region in place of the removed CDS.

Additional genome searches included RFAM database search to discover any non-coding RNAs. This search identified two new non-coding RNA structures. RFAM RF00169, a bacterial signal recognition particle (SRP) RNA, was identified upstream of Cj0046. The SRP is a universally conserved ribonucleoprotein involved in the co-translational targeting of proteins to membranes 3536. Also, RFAM RF00059 a thiamin pyrophosphate (TPP) riboswitch (THI element) was identified upstream of Cj0453 (thiamin biosynthesis protein ThiC). The RFAM motif is a conserved structure (THI element), involved in thiamin-regulation 37.

The final step of the re-annotation process was the incorporation of Gene Ontology (GO) annotation. GO annotation attempts to link three structured, controlled vocabularies (ontologies) that describe gene products in terms of their associated biological processes, cellular components and molecular functions in a species-independent manner 38. This re-annotation aims to incorporate GO annotation using two automated methods. Upon submission to EBI, the EMBL file will carry a GOA link that lists the GO annotation identified automatically by EBI. A second version of GO was generated by performing a reciprocal FASTA search using RM1221. This was created and submitted to GeneDB. GO annotation is a valuable feature in current annotation techniques that can expedite systems biology approaches to genome analysis.

Manual re-annotation of all previously annotated C. jejuni NCTC11168 CDSs 4 was carried out based on results from BLASTP 39 and FASTA 40 sequence comparisons using non-redundant databases. Re-annotation was based, wherever possible, on characterised proteins or genes 4. Additional functional data was provided by using the PFAM 41 and PROSITE 42 motif databases. New searches carried out in this re-annotation included running RFAM 43 database and also the programs TMHMM 44 and SIGNALP 45.

This re-annotation included a complete literature search of all CDS numbers and gene names using PubMed 46, HighWire Press 47, Scirus 48 and Google Scholar 49. Artemis software release 8 50 was used during re-annotation. The re-annotated sequence was submitted to the EMBL public database and also to GeneDB 51 and CampyDB 52. The EMBL file included an 'original' and 'updated' note qualifier, a 'product' qualifier and each CDS represented with a unique 'locus_tag' qualifier. Appropriate 'gene' qualifiers were also present. The GeneDB submission included all the above and extra qualifiers 'colour' and 'literature'. This re-annotation also included for the first time a Gene Ontology (GO) annotation of the NCTC11168 genome sequence. This was created automatically on submission to EMBL and can be accessed via the GOA link. A separate GO annotation was created within GeneDB by carrying out a reciprocal FASTA comparison with C. jejuni RM1221 and adopting the GO annotation of orthologous CDSs.

Advances in genome annotation techniques that were unavailable during the original annotation have led to updated interpretation of pseudogenes and phase-variable CDSs. Using guidance from TBLASTX search results, we carried out a full re-analysis of all pseudogenes. CDSs designated as pseudogenes have been updated to reflect the complete amino acid sequence for the encoded protein regardless of expression. This has caused differences from the amino acid sequence of the previous annotation. Some pseudogene modifications entailed merging two or more adjacent, in frame CDSs (previously annotated as separate pseudogene CDSs), to create a single pseudogene containing internal stop codons. In other cases, pseudogene features were created with multiple coordinates representing one or more frameshift in the CDS – these had previously only detailed the start and stop coordinates so did not reflect the true position of the non-mutated CDS. In both cases the assignment of coordinates was based on matches to homologues determined through FASTA searches.

CDSs containing an intersecting homopolymeric tract were merged to reflect the complete amino acid sequence for appropriate genes regardless of phase. This is analogous to the scenario described above for frameshifted pseudogenes. This modification was carried out for two CDSs with an intersecting homopolymeric tract. The joining of such CDSs was not undertaken in the original annotation.