INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

An essential aspect of bacterial population genetics that was seldom considered until recently is the large variation in DNA content of different strains in a single species. Plasmids and phages have been known for a long time to be highly variable in their presence in different isolates from the same species but often represent a small fraction of the total genome. The chromosome has generally been considered a well-conserved replicon. The fact that the genetic maps of Escherichia coli K-12 and Salmonella enterica LT2 are colinear (22) and of very similar sizes has weighed heavily in favor of assuming relative conservation of chromosome structure and content over wide phylogenetic distances. However, with the development of rapid chromosome mapping derived from pulsed-field gel electrophoresis techniques it has become more and more clear that chromosome size and organization are far from conserved in many species (27). Even in E. coli, differences of 1 Mb among the chromosomal sizes of different E. coli reference collection (ECOR) strains have been shown previously (2, 3). A similar result was found for two pathogenic strains, in which large DNA insertions were distributed throughout the genome (23). These data indicate that there could be significant differences in gene content among different strains of E. coli. We will refer to this DNA present in some but not all the strains in a bacterial species, regardless of its location (plasmid or chromosome), as accessory DNA. Naturally, an essential question is what is the nature of this accessory DNA. In principle, this extra DNA contained in some strains could be repeated genes or duplicated regions of the chromosome, and therefore, the significance of chromosome size variation in terms of the gene complement carried by different strains is unclear.

In one of the few studies addressing this issue (14), it was shown by subtractive hybridization that two strains of S. enterica belonging to subspecies I and V (considered to be the most widely divergent within the species) had about 20% of their genes not homologous. Some genes or genomic regions are known to have a variable presence in different strains belonging to the same species. The pathogenicity islands described for several pathogens during the last few years (8) are one example. Rhs elements are another well-known example of accessory DNA representing about 0.8% of the E. coli genome (9); they seem to have been acquired through separate transfer events from a GC-rich background (30). In the fully sequenced genome of E. coli K-12, ca. 2% of the genes are related to phages, plasmids, and transposons (4). They are also obvious candidates to represent accessory DNA. However, it is important to underline that, among the E. coli strains whose genome size is known, K-12 has one of the smallest (3). Comparative genomics of three or more strains representing divergent genealogical lineages of the same species would be an ideal way to analyze this problem. However, this kind of information is not yet available.

To contribute to the knowledge of the kind of genes that are present in only some strains of E. coli, we have developed an approach that has been already applied to detect subtractive genetic material in other situations (17). A random amplified polymorphic DNA (RAPD) study was carried out to identify amplified DNA bands present in some but not all the strains included in the ECOR collection. RAPD permits the amplification of genomic DNA with a single primer of arbitrary nucleotide sequence under conditions that favor relatively nonspecific binding of the primer to multiple sites of the template DNA (7). The loss of an RAPD marker can be due either to the mutation of the primer annealing site or to the absence of the whole sequence or genomic region. To differentiate between these two situations, we carried out a hybridization study with RAPD bands as probes. We found two clearly different situations, one in which the band hybridized to all or most of the ECOR strains and another in which only a very limited number of strains showed a clear hybridization signal. The last RAPD fragments were assumed to represent accessory genes or DNA characteristic of some strains.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and culture conditions. The 72 E. coli strains of the ECOR reference collection, which includes isolates from a wide variety of hosts and geographic regions (18), were used. E. coli K-12 strain CECT 102 (Colección Española de Cultivos Tipo, Valencia, Spain) was used for reference purposes. Five strains of E. coli O157 were also included in the study: A8190 and E3406 (O157:H7), provided by T. S. Whittam (Pennsylvania State University, University Park), and E16159 (O157:H45, diffuse adherence), E39233 (O157:H43, heat-labile enterotoxin), and E76561 (O157:H, VT2 and eae genes) provided by H. R. Smith (Central Public Health Laboratory, London, United Kingdom). S. enterica serovar Typhimurium LT2 and Halobacterium sp. DNAs were used as controls. E. coli strains were grown on Luria-Bertani medium at 37°C and stored at 80°C in 15% glycerol.

Isolation of genomic DNA. Bacterial genomic DNA was extracted with InstaGene DNA purification matrix (Bio-Rad Laboratories, Inc.) according to the manufacturer's instructions. The concentration of the DNA samples was spectrophotometrically determined, and aliquots of 100 ng/µl were stored at 20°C.

RAPD PCR amplification. RAPD primers, previously demonstrated to generate polymorphic and reproducible patterns (7), were A1 (5'-TGCGGCTTAC-3'), A7 (5'-TCACGGTGCA-3'), and A10 (5'-GTAGACGAGC-3'). All three primers were synthesized in an Applied Biosystems 396 DNA-RNA synthesizer and diluted to 10 µM-concentration aliquots which were stored at 20°C.

DNA-DNA dot blot hybridization. A 5-µg aliquot of each genomic DNA was transferred to a Hybond N⁺ nylon membrane (Amersham) with a vacuum-blotting apparatus (Bio-Dot apparatus; Bio-Rad Laboratories, Inc.). DNA was fixed to filters by baking it for 2 h at 80°C. Labeling of probes and blot development were performed by using the enhanced chemiluminescence (ECL) system (Amersham). Probes were prepared from selected RAPD PCR bands excised from the gel and purified with the Sephaglas BrandPrep kit (Pharmacia). Following a 1-h prehybridization at 42°C in ECL gold hybridization buffer supplemented with 0.5 M NaCl and 5% blocking agent, 200 ng of labeled probe was added and incubation was carried out overnight at 42°C. Hybridized filters were washed for 5 min at 42°C in 5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), followed by two 20-min stringent washes at 42°C in 6 M urea-0.4% sodium dodecyl sulfate-0.1× SSC. A final 5-min wash at room temperature with 2× SSC was carried out prior to exposing the membranes to the detection reagents. Chemiluminescence was detected on autoradiography films (Hyperfilm ECL; Amersham) after signal generation for 30 min.

Cloning and sequencing of RAPD fragments. Purified RAPD fragments were inserted into pCR 2.1 vector by ligation at 14°C overnight and transformed into INVF' One Shot competent cells (original TA cloning kit; Invitrogen Co., Carlsbad, Calif.) according to the manufacturer's instructions. The presence of insert was confirmed by PCR amplification with the M13 forward (20) primer and the M13 reverse primer.

Computer analysis of nucleotide sequences. Putative open reading frames (ORFs) and G+C moles percent composition analyses were performed with the DNAStar package. Similarities between determined RAPD sequences and those from the GenBank database were calculated by BLAST searches (BLASTN and BLASTX) carried out at the network server of the National Center for Biotechnology Information. BLASTN similarity searches screen the nucleotide query sequence against nucleotide databases; BLASTX similarity searches compare a nucleotide query sequence translated in all reading frames against a protein sequence database. Alignments were performed with ClustalX. Dinucleotide GC relative abundance profiles (*_GC) were assessed through the odds ratio functional *_GC = _GC/_GC, where _G denotes the frequency of nucleotide G and _GC is the frequency of the dinucleotide GC in the sequence under study. Values of *_GC in the range of 1.20 to 1.35 are considered a genome signature for E. coli (10). The codon adaptation index (CAI) was calculated for each ORF according to the method of Sharp and Li (25) at the Virtual Genome Center website (27a). Values of CAI below 0.25 indicate unusual codon usage in E. coli and may signify recent immigration by horizontal transfer (4).

Nucleotide sequence accession numbers. The sequences reported in this paper were deposited in GenBank with accession numbers from AF 127003 to AF 127017.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Analysis of RAPD products. Three 10-mer oligonucleotides previously demonstrated to generate polymorphic and reproducible patterns (7) were selected as RAPD primers. Genomic DNA from all 72 strains of the ECOR collection as well as five representatives of the E. coli serotype O157 (verotoxigenic strains) and the E. coli reference strain K-12 were amplified as described above (Materials and Methods). As expected (29), a high degree of polymorphism was found with the three primers. Figure 1 shows an example of the patterns obtained. Twenty-five different bands were included in the patterns generated with primer A1; among these, four bands were present in over 70% of the strains. In the case of primer A7, 20 different bands were amplified, and four of them were present in over 50% of the strains. A higher degree of polymorphism was found with primer A10, which generated 30 different bands, and among them, only three were present in more than 50% of strains. In total, 75 different bands were generated with the three primers, and only 12 of them were present in more than 50% of the strains. On the other hand, 10 RAPD fragments were present in less than 10% of the strains (see Table 1). These 10 plus three RAPD fragments with widespread distribution (H1, H2, and H4) were selected for further study. Two of them (H4 and H13) were cloned and sequenced from two different strains.

View larger version (121K): [in this window] [in a new window]   FIG. 1.   Example of the electrophoresis patterns generated by RAPD amplification of some ECOR collection strains and E. coli K-12 with primers A1 and A7. The molecular size marker used is a 1-kb ladder (Gibco BRL, Life Technologies Ltd.). The strain number is indicated above each lane, and the primer used is shown below. Arrows point to some of the fragments used as hybridization probes in dot blot analysis.

Distribution of the RAPD fragments in the ECOR collection. The selected RAPD fragments were used as hybridization probes to determine the distribution of these sequences among the collection of strains tested. The selected bands (Table 1) were purified from the gel and used as probes against a dot blot containing total DNA from the whole set of strains, as well as DNA from other organisms included as controls (S. enterica LT2 and Halobacterium sp.). Hybridization results are summarized in Table 1 and Fig. 2, and they showed that (i) as expected, each probe hybridized strongly to the genomic DNA of the strain from which the band was isolated; (ii) among the RAPD fragments selected for their low frequency, six showed a very restricted distribution by hybridization, i.e., four fragments (H8, H9, H10, and H12) hybridized solely to the genomic DNA extracted from the strain of origin (indicating that those sequences were present in only a single strain within the set), one fragment (H11) hybridized with one additional strain, and another one (H13) hybridized with three strains; (iii) the remaining seven fragments, including the three selected for their high frequency (H1, H2, and H4), had a widespread distribution, hybridizing with between 40 and 100% of strains (Table 1). Those three plus H7 also hybridized with S. enterica LT2, showing a supraspecific distribution. None hybridized to Halobacterium sp. DNA. Several strains that did not generate the corresponding RAPD marker gave a positive hybridization result. This was probably due to the modification of the primer annealing sequence or the loss of a short DNA stretch including the primer target.

View larger version (33K): [in this window] [in a new window]   FIG. 2.   Schematic representation of the distribution of the RAPD fragments among the E. coli ECOR collection strains, as well as the five E. coli O157 strains and K-12, as determined by dot blot hybridization results (H1 to H13 [see Table 1 for details]). Positive results in dot blot analysis are indicated by black dots; open dots correspond to the positive controls, i.e., the strain of origin of the RAPD fragment. The ECOR collection strains are grouped in a dendrogram based on MLEE results (24) with the different cluster designations on the outside branches.

Sequencing of RAPD fragments. The 13 RAPD fragments used as probes were cloned and sequenced as indicated above (Materials and Methods). Additionally, two of them were sequenced in duplicate, i.e., the same band was cloned and sequenced from a second strain. The 15 RAPD fragments sequenced could be classified into three categories with different putative roles, frequencies of distribution, and putative origins (Table 1).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The approach that we have followed here permits the identification of sequences that are present in only some strains within a single species. The two fragments present in all the strains are part of ORFs with high similarity to E. coli K-12 housekeeping (HK) genes. Obviously, this kind of sequence was not the target of our study, and they were included only as a kind of blank or negative control. In both cases, the primer annealing target is included within the ORFs, which obviously contributes to the stability of the RAPD product. However, none of the bands were present in more than 75% of the RAPD patterns generated from the strains studied. In these strains, the absence of the RAPD marker must be due to point mutations in the annealing target. In fact, the conservation of the ORF coding for DamX in ECOR 43 was only 92.9% at the amino acid level, which is quite low for an HK gene (31). The rest of the RAPD fragments assayed (all the ones selected by their restricted distribution as RAPD markers) were detected by hybridization only in some strains. That supports this strategy for recovering accessory DNA from the whole genome of bacterial strains.

The five fragments included in category II are present in 42.3 to 97.4% of the assayed strains as shown by hybridization. H5, H6, and H7 are clearly related to E. coli phages and/or plasmids, and therefore, there is little doubt about their nature. The RAPD approach followed would retrieve both chromosomal and extrachromosomal fragments, and we have no information as to their location. However, integration of plasmidic and/or phage material into the chromosome seems to be a common fact (6), and therefore, the location of gene clusters in chromosomal or extrachromosomal units can vary rapidly in evolutionary time. Plasmids and phages are the paradigm of accessory DNA, and, if anything, the relatively low number of fragments of this group found here is surprising. However, some of the fragments included in category III could also represent this kind of genetic element, albeit unknown ones. It is remarkable that two of those phage-plasmid-related fragments are found in most of the ECOR strains, but the ORF related to phage T4 found in H5 has a distribution with a certain phylogenetic consistency, being present in most strains of groups A, B1, and E; scarce in B2; and absent in group D strains. Phage 21 has also been reported to have a distribution correlated with the MLEE groups, although its evolution is not correlated with that of the host (28). The consistency in the compositional parameters and codon usage found for these sequences is remarkable. They are slightly divergent from the average for the E. coli K-12 genome, and the CAI corresponds to that of E. coli K-12 genes expressed at a low level. H3 and H4 also had high similarities to sequences described for the E. coli K-12 genome. H3 contains the carboxyl end of PitA plus a part of the PitA-UspA intergenic region. This region, as in E. coli K-12, contains an ORF coding for a hypothetical protein, YhiO, which was very conserved in ECOR 44, with 99.1% similarity to E. coli K-12 at the amino acid level. Its function might be important because the degree of conservation is consistent with that of HK genes. Something similar applies to H4, which contains an ORF highly similar to a paralogous copy of dnaK found in the E. coli K-12 genome. Paralogous copies of essential genes have been shown to represent significant parts of the relatively large bacterial genomes fully sequenced (E. coli and Bacillus subtilis) (4, 13). In E. coli K-12, nearly a third of its genes have paralogous copies in the genome (4). The function of those supplementary copies of genes is unclear. The presence of this paralogous gene in only a limited group of strains could indicate that, in this case, it is either a nonessential copy with an adaptive function, i.e., required only to survive under certain conditions, or a pseudogene. In fact, the relatively high conservation found for the three strains in which this fragment has been sequenced (K-12, ECOR 2, and ECOR 44) supports the first notion. The absence of the RAPD fragments equivalent to H3 and H4 from many strains in which their presence was detected by hybridization can be justified by the location of one of the primer annealing targets for both fragments in intergenic regions that are possibly subjected to a higher rate of change.

The fragments classified in category III have indeed a very restricted distribution throughout the ECOR collection. Four of them were found in only one strain. This fact indicates a very recent acquisition and/or a high adaptive specificity of this genetic material. On the other hand, the fragments within this category that were found in more than one strain were present in relatively distant strains (H13 in groups B1, D, and E and H11 in groups B2 and E). The ORF within H10 shows a residual but significant similarity (53.4%) to a hypothetical protein described for E. coli K-12; for the rest, the lack of similarity with E. coli K-12 sequences confirms that they represent actual differential DNA without any homology to the DNA present in some E. coli strains. The compositional parameters (percent G+C, *_GC) and CAI are widely divergent from those of the average E. coli K-12 genome, and they are reminiscent of the 10% with atypical CAI values described for the K-12 genome (4). This is normally interpreted as the result of recent immigration by horizontal transfer (3, 15, 16). Low G+C values (around 40%) have also been found in pathogenicity islands (5, 19). The kinds of strain-specific sequences that we have retrieved are very unlikely to represent repeated copies of other parts of the genome. Furthermore, the fact that, among the 10 fragments with restricted distribution studied, four were restricted to the strain of origin stresses the importance of the differential gene complement carried by different strains.

Most population genetic studies of bacteria are concerned with sequence variation for a single gene (allelic variation). However, a major contribution to the intraspecific variation in bacteria could be ascribed to the presence of different pools of accessory genes in different strains rather than different alleles of the same genes (the latter is the case in the populations of higher organisms). If these different sets of adaptative genes carried by different lineages within the species are very large, they would be a major factor to consider in bacterial evolution. As stated elsewhere by Reanney (21), "In an ecosystem where the microbiota frequently exchange these genes, the size of the entity upon which selection acts may exceed by orders of magnitude the size of the gene pool of a clonal population."