An in silico analysis of the genome of M. smegmatis mc²155 revealed that it contains 94 ICDSs 5. The ICDS database was created using a program based on the analysis of physically adjacent genes to predict putative ICDSs in complete genomes. Briefly, pairs of adjacent genes with at least one common homolog are defined as 'coding sequences (CDSs) containing common hits' and may correspond to a pair of adjacent paralogs or ICDSs. We excluded paralogs from the analysis by searching for sequence similarity between the two 'CDSs containing common hits'. The remaining CDSs are considered to be ICDSs, indicating frameshifts or in-frame stop codon insertion, due to sequencing errors or authentic events. These 94 ICDSs account for 1.4% of the total coding capacity of this organism. They may result from mutations acquired during evolution or from errors in genome sequencing.

We resequenced the genome of this strain to determine the status of these ICDSs. We did not resequence 21 ICDSs due to the duplication of some open reading frames (ORFs) or high levels of paralogy. The remaining 73 ICDSs were amplified and sequenced on both strands. We compared the nucleotide sequences obtained with the publicly available genome sequence of M. smegmatis mc²155. We found that 28 of the 73 ICDSs investigated correspond to sequencing errors (Table 1). These 28 genes containing sequencing errors correspond to 4 errors per megabase in the complete genome. In most cases, correction of the error reunified two adjacent ORFs, resulting in a single ORF rather than the two small ORFs of the original sequence (Figure 1).

Three types of error can be distinguished: miscall, overcall and undercall (Table 1) 234. However, no miscalls (incorrect prediction of a specific nucleotide at a given position) were observed within the 28 sequences containing errors, due to the nature of the program used. The predicted amino acid sequences derived from the corrected nucleotide sequences differed greatly from the original predicted sequences and, in all cases, were systematically more similar to their orthologs. In one case (ICDS0089), the ORF containing the frameshift was not even predicted; the frameshift was probably responsible for the non-assignment of this ORF. The genes affected by the sequencing errors encode proteins of several classes, including 'unknown', 'intermediary metabolism', 'regulation' and 'lipid metabolism' (Table 1). The genes containing frameshifts encode proteins of several classes, including all of those cited above (Table 2). No particular pattern of nucleotides was associated with the 28 sequences containing errors or with the 45 sequences containing frameshifts.

As M. smegmatis mc²155 was derived from strain ATCC607, we carried out a comparative analysis of the ICDSs in these two strains. The mc²155 strain was generated from ATCC607 by selection for adaptation to genetic manipulation 13. The mc²155 strain differs phenotypically from its progenitor (ATCC607) in several ways 1314. The frameshifts in mc²155 may well have been acquired recently in the laboratory, due either to counter-selection of pathways of little utility or selection for genetic manipulability. We therefore investigated whether the genes containing frameshifts were acquired before or after the divergence of the two strains. The genome of the ATCC607 strain has not been sequenced, but as both strains belong to the same species (M. smegmatis), the sequencing primers originally designed for the mc²155 strain could also be used for the ATCC607 strain. We resequenced the 45 genes containing a frameshift of mc²155 strain in ATCC607 (Table 2). All these genes but one (ICDS0020) also contain a frameshift in the progenitor (ATCC607), suggesting that these mutations were acquired before the divergence of the two strains. Thus, the selection of the mc²155 strain and its repeated culture in laboratory conditions had no major impact on frameshift acquisition and pseudogene formation.

Our analysis shows that the genome sequence of M. smegmatis mc²155 contains ICDSs, some of which correspond to authentic mutations acquired during evolution, with others resulting entirely from sequencing errors. Our results show that 18 predicted genes do not actually exist in this species (due to fusion of the two ORFs following the correction of the errors) and that one gene was even not predicted in the former sequence, presumably due to these sequencing errors. In all cases, the new predicted genes are actually more similar than previously thought to orthologs in other species.

As ICDSs (corresponding to authentic events or to sequencing errors) accounted for 1.4% of the ORF content of M. smegmatis mc²155, we surveyed a fraction of the proteome to determine the percentage of proteins originating from ORFs not predicted due to misannotations. We carried out two-dimensional electrophoresis of a soluble protein extract. The major spots (120) were excised, digested and analyzed by nano-LC-MS-MS (nanoflow liquid chromatography coupled to tandem mass spectrometry). We were able to identify about 250 proteins unambiguously by comparing the MS-MS data obtained from the tryptic peptides. We compared these MS-MS data directly with public nucleotide sequences, rather than using the classic comparison of MS-MS data with protein sequences 1516 to prevent the introduction of bias. The identification of several proteins for a single spot is not surprising and has been widely reported in proteomic analysis 17. For four spots the tryptic peptides identified by nano-LC-MS-MS analysis matched two contiguous hypothetical ORFs each (Table 3, Figure 2). There are two possible explanations for this finding. Firstly, two different proteins, encoded by two different frames in the same genome region, may be present in the same two-dimensional gel electrophoresis spot. This is unlikely, due to differences in molecular masses (Table 3), but cannot be entirely excluded. Secondly, these peptides may be derived from the same protein. In this case, a bypassed stop codon or a sequencing error could account for such an observation.

For the four proteins concerned, MS-BLAST showed that all the tryptic peptides identified matched the same protein on the basis of sequence similarity with other organisms. We carried out a new search with the MS-MS data obtained for the four two-dimensional gel electrophoresis spots using the corrected sequences obtained after resequencing of all the ICDSs. For all four spots the peptides were found to match in the same frame and new peptides from the proteins were detected (Table 3, Figure 2). We can conclude, therefore, that the four ICDSs detected were due to sequencing errors. These ICDSs are ICDS0019, ICDS0039, ICDS0040 and ICDS0093. We show ICDS0040 as an example in Figure 2.

Thus, proteome analysis identified errors in sequences that were not predicted to correspond to an ORF. All four cases detected in this way were found to correspond to sequencing errors (Table 1). There is, therefore, strong congruence between in silico data and nucleotide and proteomic analyses.

M. smegmatis mc²155 (ATCC700084) and M. smegmatis NRRL B-692 (Trevisan) Lehman and Neumann (ATCC607) were purchased from the American Type Culture Collection (Manassas, Virginia, USA).

The genome sequence of M. smegmatis mc²155 was taken from the TIGR website 12. The ICDSs were detected using the method developed by Perrodou et al. 5.

The primers used to sequence frameshifts were designed as previously described 5 using an optimized version of the CADO4MI program (Computed Assisted Design of Oligonucleotides for Microarray). It is a freeware (GNU General Public License) accessible online 26. For each genome, sequencing primers are available online 27. The chromosomal DNA of the mc²155 and ATCC607 strains of M. smegmatis used for PCR amplification was purified as previously described 28. Pairs of primers were used for amplification with Pfu Turbo DNA polymerase (Stratagene, La Jolla, CA, USA). PCR samples were run on a 0.8% agarose gel and the fragments were excised from the gel and purified using the QIAquick Gel purification kit (Qiagen Chatsworth, CA, USA). The PCR fragments had a mean length of 300 base-pairs. Purified PCR fragments were used as templates in sequencing reactions with each primer used for PCR amplification. The nucleotide and inferred aminoacid sequences were analyzed with DNA Strider 29. Three independent amplicons were sequenced for each ICDS.

M. smegmatis strain mc²155 (1 liter) was grown in M9 minimal medium (Difco, Detroit, USA) for 5 days and then centrifuged. Bacterial pellets were used for two-dimensional electrophoresis. Unless otherwise specified, all chemicals were obtained from Sigma (St Louis, MO, USA). Dithiothreitol (DTT) and iodoacetamide were obtained from Fluka (Buchs, Switzerland). The pellet fraction was incubated with extraction buffer (50 mM Tris, pH7.5, 1 mM phenylmethylsulfonyl fluoride, 1 mM EDTA, 1 mM DTT, protease inhibitor mixture (complete from Roche, Basel, Switzerland)) for 45 minutes at 4°C. The mixture was sonicated for a few seconds and its protein concentration determined by Bradford assay. The solvent of the protein extract was evaporated off and the protein residue was suspended in rehydration buffer (8 M urea, 2 M thiourea, 4% 3- [(3-cholamidopropyl)dimethylammonio]-1-propanesulfonic acid, 0.5% Triton X-100, 1% DTT, 20 mM spermine, 2% Pharmalyte (Amersham Pharmacia Biotech, Piscataway, NJ, USA)). The sample was incubated for 30 minutes at 20°C and centrifuged at 15,000 rpm at 20°C.

Protein extract was run on a strip of gel of pH range 3 to 10 (Bio-Rad Laboratories, Hercules, CA, USA) for 15 h at 20°C under 50 V in a PROTEAN isoelectric focusing cell (Bio-Rad). Isoelectric focusing was carried out with several voltage steps: 1 h at 200 V, then 4 h at 1,000 V followed by 16 h at 5,000 V and finally 7 h at 500 V at 20°C. The strips were incubated for 30 minutes at 20°C in electrophoresis buffer (50 mM Tris-HCl, pH 8.8, 6 M urea, 30% (v/v) glycerol, 2% (w/v) SDS, and 1% DTT), followed by 30 minutes in the same buffer supplemented with 2.5% iodoacetamide. Electrophoresis in a gradient gel (5% to 20% acrylamide) on a PROTEAN II (Bio-Rad) apparatus at 5 mA for 1 h and 10 mA overnight was used as the second dimension. The gel was stained with Colloidal blue (G260, Sigma); 120 spots were selected by visual inspection and gel slices were excised with a Proteineer SP automated spot picker (Bruker Daltonics, Bremen, Germany) according to the manufacturer's instructions.

The two-dimensional gel spots were excised, washed, destained, reduced, alkylated and dehydrated for in-gel digestion of the proteins with an automated protein digestion system, MassPREP Station (Waters, Milford, MA, USA). The proteins were digested overnight at room temperature with trypsin. They were then extracted with 60% (v/v) acetonitrile in 5% (v/v) formic acid and then with 100% acetonitrile. The resulting peptide extracts were analyzed directly by nano-LC-MS-MS on an Agilent 1100 Series capillary LC system (Agilent Technologies, Palo Alto, USA) coupled to an HCT Ultra ion trap (Bruker Daltonics). This instrument was equipped with a nanospray ion source and chromatographic separation was carried out on reverse phase (RP) capillary columns (C18, 75 μm id, 15 cm length, Agilent Technologies) with a flow rate of 200 nl/minute. The voltage applied to the capillary cap was optimized to -2,000 V. MS-MS scanning mode was performed in the Ultra Scan resolution mode at a scan rate of 26,000 m/z per second. Eight scans were averaged to obtain an MS-MS mass spectrum. The complete system was fully controlled by Agilent ChemStation and EsuireControl (Bruker Daltonics) software. The generated peak-lists of fragments were used for public M. smegmatis genome database searches.