INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The rRNA genes are the most frequently used markers for the study of molecular phylogeny for reasons that have been repeatedly reviewed (1, 30, 32), although their applicability to establish organismal phylogenies in prokaryotes has been recently contested (19, 26, 33). The rrn operons are among the few genes in prokaryotic genomes that are often repeated. In Escherichia coli seven rrn operons are located throughout the half of the chromosome containing the origin of replication (6). Little is known about the mode of evolution of these genes. A standard way to approach this problem is to study sequence variation within populations of relatively recent divergence, so that the general pattern of molecular evolution can still be discerned (31). Lately, we have studied the variation affecting different regions of the rRNA operons in E. coli and related species (2, 10, 22, 23), specifically, the rrs gene and the internal transcribed spacer (ITS) region between the rrs and rrl genes. As expected, relatively little variation of single nucleotide substitutions for the rrs gene was found, and higher variation for the rrs-rrl ITS was found. There were hypervariable regions in both elements where clustered nucleotide substitutions occurred, generally conserving the rRNA secondary structure (compensatory mutations). In some regions, such as helix V1 of the rrs gene, several variants were found. The different sequence variants differed by single or compensatory mutations, and possible relationships among the different sequences were shown (23). This mode of gradual change fits well with a model of evolution by neutral change and genetic drift. However, the most frequent situation in both the rrs gene and the rrs-rrl ITS (2) was the finding of a restricted number of highly divergent sequences that maintained the same secondary structure. This "jumpy" mode of variation is difficult to explain by neutral-change models. Sometimes the sequences found are similar to those of other species (5), and some researchers have suggested lateral transfer and recombination to explain their presence in E. coli. Regardless of their origin, molecular coevolution with ribosomal proteins or the transcript processing machinery could be an important reason for the preservation of these sequences (8).

The most remarkable difference found between the rrs gene and the rrs-rrl ITS was the presence in the latter of relatively large insertions and deletions, as exemplified by the rsl loop, that alter the secondary structure dramatically. A mode of evolution in which cumulative insertions and deletions would eventually alter the overall sequence of the rrs-rrl ITS was suggested (2, 22). This would explain the great variation found when organisms with increasing phylogenetic distances are compared, as opposed to the rrs gene, in which variation on a larger phylogenetic scale is relatively proportionate.

Here, we have studied another region within rRNA operons that includes the rrl gene and the spacer region downstream (rrl-rrf ITS). The rrl gene is apparently less restricted in size variation (11), and a wide divergence exist throughout the class Proteobacteria (21). In addition, the rrl gene is also assumed to be more variable at the sequence level than the rrs gene, particularly for shorter phylogenetic distances. Representatives from different clusters determined by multilocus enzyme electrophoresis (MLEE) of the E. coli ECOR collection (25), plus one O157 serotype isolate, were selected for sequencing. Hypervariable regions have been sequenced to assess intercistronic heterogeneity. Finally, the determined variants of two helices were examined by PCR analysis throughout the ECOR collection.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and cultivation. E. coli strains of the ECOR collection, comprising isolates from a wide variety of hosts and geographic regions (25), were used. Strain A8190 of E. coli serotype O157:H7 (kindly provided by T. S. Whittam, Pennsylvania State University) was also included. Strains were grown in Luria-Bertani (LB) medium for 18 h at 37°C and stored at 80°C in 15% glycerol.

DNA extraction. Bacterial genomic DNA was extracted by using the InstaGene DNA Purification Matrix (Bio-Rad Laboratories, Inc.) in accordance with the manufacturer's indications. DNA concentrations were spectrophotometrically estimated and diluted to 100 ng/µl.

PCR amplification. Long-distance PCR was used to amplify the complete rrn operon and the upstream flanking sequence of ca. 1,000 bp. Previously described primers (2) designed on the basis of specific sequences located upstream of the rrs gene of E. coli K-12 were combined with primer 5SR (5'-ATGCCTGGCAGTTCCCTACT-3'). Reaction mixtures of 50 µl contained 1 µl of genomic DNA, 60 mM Tris-SO₄ (pH 9.1), 18 mM (NH₄)₂SO₄, 1.8 mM MgSO₄, 0.2 mM each deoxynucleotide, 0.2 µM each primer, and 2 U of Elongase Enzyme Mix (Gibco BRL, Life Technologies Ltd.). PCR was performed with a PTC-100 thermal cycler (MJ Research, Inc.). The mixtures were subjected to 35 cycles of 94°C for 30 s, 55°C for 30 s, and 68°C for 6 min. A final extension step of 68°C for 5 min was allowed. Probes designed on the basis of two variable regions of the rrl gene were H25R I (5'-CACACGCCTAAGCGTGCTCC-3'), H25R II (5'-CACCCCATTAAGAGGCTCCC-3'), and H25R III (5'-GTCACACAGATTGCTCTGTG-3') for helix 25 and H63R I (5'-CAGCTCCATCCGCGAGGGAC-3') and H63R II (5'-CAGCTCCACGAGCAAGTCGC-3') for helix 63. The PCR conditions used with these probes were as previously described (2). PCR products were analyzed in 1% agarose gels, stained with ethidium bromide, and visualized on a UV transilluminator.

Sequencing of the rrl gene and the rrl-rrf ITS. PCR products were purified by using a QUIquick PCR Purification Kit (Quiagen) in accordance with the manufacturer's protocol. Nucleotide sequences were determined in an ABI PRISM 377 sequencer (Perkin-Elmer) as recommended by the manufacturer. The sequencing primers for the rrl gene were 23SOF (5'-GTGAGTCTCTCAAATTTTCG-3'), 23S3F (5'-AGTAGCGGCGAGCGAACGGG-3'), 23S6F (5'-GCGTACCTTTTGTATAATGG-3'), 23S8F (5'-ATAGCTGGTTCTCCCCGAAA3'), 23S11F (5'-AAGAAAGCGTAATAGCTCAC-3'), 23S16F (5'-TACCCCAAACCGACACAGG-3'), 23S20F (5'-TAGCGAAATTCCTTGTCGGG-3'), 23S23F (5'-GGGTAGTTGACTGGGGCGG-3'), and 23S26F (5'-CAAGGGTATGGCTGTTCGCC-3'). Primer 5SR was used to sequence the rrl-rrf ITS.

Data analysis. Alignment and phylogenetic analysis of the rrl gene sequences were carried out by using the programs ClustalW and Mega (Molecular Evolutionary Genetics Analysis, Version 1.01), respectively. Comparisons with sequences in the GenBank database were achieved by BLAST searches (BLASTN). The GenBank accession numbers of the rrl genes of strains ECOR 35, ECOR 52, ECOR 40, ECOR 10, ECOR 24, and A8190 are AF053963 to AF053968, respectively.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

The six E. coli strains were subjected to locus-specific PCR using primers located upstream from each rrn operon. The strains were selected considering the diversity found in both the rrs gene and the rrs-rrl ITS (2, 23), as well as their relationships shown by MLEE (16). ECOR 10 is highly related to E. coli K-12 and was used as a reference. ECOR 24 was selected because it contains unusual intercistronic homogeneity and the rsl loop in all of the rrs-rrl ITSs. ECOR 35 is relatively homogeneous regarding differences among the rrn operons at the level of the rrs-rrl ITS. ECOR 40 belongs to an MLEE cluster with a peculiar insertion in the rrs-rrl ITS that is also found in Salmonella spp. and contains a high level of intercistronic heterogeneity. ECOR 52 is a representative of the MLEE B2 group that represents a highly homogeneous phylogenetic cluster that lacks the variants found in other MLEE clusters. Finally, A8190 is a representative of verotoxigenic E. coli serotype O157, which has recently been shown to be widely divergent from other E. coli lineages (27). PCR amplification yielded amplicons of ca. 6 kb. Complete rrl gene sequences of a single operon from each strain were determined in order to examine representatives of the operons containing one (rrnC and rrnB) and two (rrnD) tRNA genes. Figure 1 summarizes the 67 polymorphic sites found from alignment of the six rrl genes fully sequenced. The pattern of variation was highly reminiscent of that found for the rrs gene (23), and the percentage of polymorphic sites is about the same. All of the rrl genes were almost identical in size, 2902 to 2905 bp. Intervening sequences, like those found in Salmonella rrl genes (4), were not detected. They were not found in the rrs genes of E. coli strains (5, 23) either, and if they are present at all in E. coli, they should be much rarer than in Salmonella spp.

View larger version (52K): [in this window] [in a new window]   FIG. 1.   Variable positions in the rrl genes of E. coli K-12 and other strains used in this study. Periods indicate the same nucleotide composition as the rrlA sequence of E. coli K-12 (Escherichia coli Database Collection). The numbers at the top indicate positions in the sequences reported by Brosius et al. (3). Insertion of nucleotides G and C between positions 545 and 546 is indicated by the arrow.

Most polymorphic sites clustered in five secondary-structure loops, viz., helices 25, 45, 63, 79, and 98 (helix numbering is that of Larsen [20]; Fig. 1 and 2). In helices 25, 63, and 98, although the secondary structure is well preserved, the sequence is often remarkably divergent. A moderate degree of polymorphism was observed in helices 45 and 79, and some intermediate forms could be related by single and compensatory mutational events. In helix 25, the three forms detected had a distribution clearly associated with the MLEE clusters. Partial sequences of the most variable regions were obtained from a larger number of rrn operons of the six strains previously amplified by PCR using operon-specific primers. The types of primary structures detected are summarized in Table 1. These sequencing results confirmed the polymorphism already found in the fully sequenced rrl genes; no new forms were detected in the hypervariable helices under study (Fig. 2 and Table 1). Except for helices H25 II, H25 III, and H98 II, all of the structures were previously found in E. coli K-12 (data from the Escherichia coli Database Collection). A remarkable aspect of the results presented in Table 1 is the high degree of intercistronic homogeneity found within each strain. This homogeneity has been confirmed by a PCR test using specific probes for helix 25 and, to a lesser degree, helix 63 (Fig. 3). E. coli K-12 and strain ECOR 10 (which are highly related) were the most heterogeneous, in agreement with previous analysis of the rrs gene (23) and the rrs-rrl ITS (2).

View larger version (26K): [in this window] [in a new window]   FIG. 2.   Secondary structures of variable regions in the rrl genes from E. coli strains. The numbers indicate positions reported by Brosius et al. (3). Conserved nucleotides are in boldface. In parentheses are the nucleotides that have been determined in the rrl genes of the following strains: helix 25, (U)1 in rrlH of ECOR 40, (U)2 in rrlB, rrlC, rrlE, and rrlG of ECOR 35, and a (U)3 insertion in rrlB, rrlC, and rrlD of ECOR 24 and rrlG of A8190; helix 45, (G)1 in rrlB, rrlG, and rrlH of ECOR 10 and rrlB, rrlC, and rrlD of ECOR 24, (U)2 in rrlC of ECOR 24; and (G)3 in rrlA and rrlC of A8190; helix 63, (A)1 in rrlC of ECOR 40 and (U)2 in rrlB and rrlC of ECOR 10 and rrlB, rrlC, and rrlD of ECOR 24; helix 79, (G)1 in rrlA, rrlC, rrlD, and rrlE of ECOR 52 and (G)2 and (G)3 in rrlB and rrlC of ECOR 10 and rrlA, rrlC, and rrlG of A8190; helix 98, (A)1 in rrlC of A8190 and (G)2 in rrlA and rrlE of ECOR 52.

View larger version (21K): [in this window] [in a new window]   FIG. 3.   Distribution of structures I, II, and III of helix 25 and structures I and II of helix 63 among the ECOR strains as revealed by a PCR test using sequence-specific probes. The MLEE relationships shown in the tree are those described by Herzer et al. (16) modified by the addition of E. coli K-12 and A8190.

For helices showing widely divergent sequences (helices 25 and 63), PCR primers were designed and used as probes to screen the ECOR collection in order to evaluate their distribution throughout the phylogenetic expanse of the species (Fig. 3). The sequence of helix 25 I, found in E. coli K-12, was restricted to MLEE cluster A plus two strains of B1. The probe for ECOR 25 II hybridized with most of the strains of groups B2, D, and E. Altogether, the distribution of these different forms seemed to correlate quite well with the clusters defined by MLEE. Finally, the probe for the sequence of helix 25 III, found originally in strain A8190, was restricted to a group of strains within cluster B1 of MLEE and only two strains outside this cluster, ECOR 12 of cluster A and ECOR 37 of cluster E. All of the other representatives of group E were negative with this probe. In a previous study of mdh sequences using representatives of E. coli, only ECOR 37 fell within the cluster that includes representatives of the O157:H7 serotype (27). Only four strains in the ECOR collection were positive by more than one H25 probe, showing remarkable intercistronic homogeneity at this level. A different result was obtained with probes for the two sequences found in helix 63. Twenty-five strains were positive with both probes, 25 were positive only by the probe for helix 63 I (half of them belonging to group B2), and the remaining 22 strains were positive only by the helix 63 II probe. In this case, intercistronic heterogeneity must be as high as that of E. coli K-12 (Table 1).

The sequences found in the hypervariable loops were compared to those in databases to assess similarity to the equivalent loops in other bacterial species. No significant match was found, with the single exception of helix 98 II, which was most similar (only two nucleotides were different) to the equivalent helix of Klebsiella pneumoniae. However, the tetranucleotide (AACG) at the end of helix 25 III was found in many members of the class Proteobacteria, including Thiobacillus cuprinus, Nannocystis exedens, and Stigmatella aurantica.

23S-5S rRNA gene ITS region (rrl-rrf ITS). This spacer region is highly conserved among the rrn operons of E. coli K-12. In this strain, 57 of the 94 nucleotides of this ITS are presumed to form a secondary structure by pairing with the corresponding stretch downstream of the 5S rRNA gene (rrf) (3). However, strains in the present study showed significant variations affecting mostly the region that does not contribute to the secondary structure of the transcript (Fig. 4A). Particularly, one of the operons of ECOR 40 (group D) and the three operons of ECOR 52 (group B2) are widely divergent from E. coli K-12 in their rrl-rrf ITS sequences. A remarkable intercistronic heterogeneity was found affecting the determined rrl-rrf ITS of ECOR 40. 

View larger version (18K): [in this window] [in a new window]   FIG. 4.   Alignment of rrl-rrf ITS sequences determined for ECOR strains and A8190 and for rrnA of E. coli K-12. (B) Alignment of rrl-rrf ITSs of E. coli K-12, ECOR 40, and S. typhi. Mosaic similarities are underlined. Periods indicate the same nucleotide composition.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

23S rRNA gene. Overall, the sequence variation found in the rrl gene was smaller than expected. The 23S rRNA gene is larger and shows a higher rate of variation than does the rRNA of the small subunit (11, 18). When different genera of the class Proteobacteria are compared, the similarity of the 16S rRNA genes is about 4% higher than that of the 23S rRNA genes (21). In a previous study (23), five complete rrs gene sequences from three strains in the ECOR collection showed 34 polymorphic sites (unpublished data). In the present study, 67 polymorphic sites were detected in the six fully sequenced rrl genes. Therefore, at this level, variation per nucleotide seems to be equivalent for both genes. Similarly, an unexpected conservation was found for the size of the rrl gene. Even disregarding the contribution of frequent intervening-sequence insertions, the size of the 23S rRNA molecule in members of the class Proteobacteria varies in a range of ca. 200 nucleotides (21). The size of the rrl genes within the ECOR collection varied by only three nucleotides; apparently, as happens with the spacer regions (2, 22), variability at the interspecific level cannot be used to infer intraspecific variability.

23S-5S rRNA gene ITS region (rrl-rrf ITS). Sequencing of the rrl-rrf ITS has revealed the existence of an undescribed hypervariable region affecting ca. 40 nucleotides before the 5' end of the rrf gene (Fig. 4A). This region seems to be the most variable found throughout the rrn operon. We have found again a strong heterogeneity affecting ECOR 40, a strain that is highly heterogeneous in the rrs gene and the rrs-rrl ITS (2, 23). The mosaic relationships found in the spacer regions of E. coli K-12, ECOR 40, and S. enterica serovar typhi are remarkable (Fig. 4B). This situation is often found in comparisons of homologous genes of bacteria (9, 24, 28), and it has generally been attributed to recombination. However, recombination acting over such short stretches of nucleotides appears unlikely. In spite of the wide variation at the nucleotide sequence level, the size of the spacer was remarkably conserved for the strains sequenced. This is in contrast to the rrs-rrl ITS, where size varies widely, even among orthologous (containing the same tRNA gene) operons (3).

Molecular evolution of the rrn operons in E. coli. Only restricted information regarding the sequence diversity in roughly half of the ribosomal operons studied here is available, because only the hypervariable regions were sequenced in more than one operon per strain. However, from the partial sequences and the PCR probe screening, it is quite apparent that this section of the ribosomal operon behaves similarly to the other half. The information accumulated regarding the sequence variation in different strains of the ECOR collection allows us to draw a general picture of how the evolution of the operon has proceeded. As expected, variation is concentrated in specific regions, and the sequence variation preserves the secondary structure (i.e., covariation) (13). This much can be predicted from what is known about the variation at the interspecific level. However, there are two markedly different modes of variation. One is exemplified by helix V1 of the rrs genes or helices 79 and 45 of the rrl genes that exhibit a gradual variation that could be explained by a lack of functional restriction and genetic drift. Other regions, such as V6 of rrs genes or helices 25, 63, and 98 of rrl genes only showed a few different, widely divergent sequence types, but no gradual variation was present. In the latter, a high level of similarity to the equivalent region in other bacterial species is sometimes found. These widely divergent variations could have originated by mutation or by horizontal transfer. In any case, the lack of gradual variation among the ECOR representatives studied hints of a more restricted freedom of variation by functional restriction and/or coevolution with ribosomal proteins. The spacer regions vary in a similar pattern, with relatively few isolated nucleotide substitutions and most of the polymorphisms clustered in areas that sometimes also correspond to conserved secondary-structure loops. The rrs-rrl ITS contains important insertion and deletion events that are not found in the structural genes and could be the main force behind the wide divergence found when spacers from distantly related species are compared.