This is the first large scale survey of intragenomic variation of paralogous 23S rRNA genes. We analyzed 23S rRNA genes from genomes representing 184 prokaryotic species to examine for evidence of ribosomal constraint of rRNA structures at the intragenomic level. Our findings support the hypothesis that individual 23S rRNA genes within a genome are conserved due to such structural constraints.
One level of ribosomal constraint on 23S rRNA genes was observed at the level of primary structure. In the large majority of the genomes (183/184) analyzed, variation of primary structure among paralogous 23S rRNA genes was small (<3%, which is the 16S rrn
-based species boundary) 
. In our dataset, 38% of the 184 genomes containing multiple paralogous 23S rRNA genes were invariable, similar to the finding in 16S rRNA genes 
. Our findings of 0.4% average diversity for intragenomic variation of rRNA sequences among the 113 diversified genomes is higher than the reported 0.17% diversity among 16S rRNA genes 
, but will be more similar to the 16S rRNA genes diversity (0.25%) if all 184 genomes are included in the calculation (Table S2
). Excluding T. tengcongensis
, which has unusually high diversity among 23S rRNA genes, the maximal diversity among paralogous 23S rRNA genes is 1.96%, higher than the maximal variation of 1.26% among 16S rRNA genes 
. In organisms containing multiple rRNA genes, the homogeneity of primary sequences is believed to be maintained through gene conversion by homologous recombination 
. The evolution of these paralogous rRNA genes is concerted in such a fashion that within a species the rRNA genes are nearly identical, whereas orthologous rRNA genes between species can vary considerably 
. However, our study suggests that homologous recombination is not always efficient in keeping the primary structure in check. In T. tengcongensis
, the 4% difference (without considering the IVS) in primary structure between rrnB23S
is in agreement with the high % divergence reported between 16S rRNA paralogues 
. It is tempting to attribute the unusual variation to the presence of IVS in the T. tengcongensis
23S rRNA genes, but variation amongst 23S rRNA genes in six other genomes containing IVS is much lower (). It appears that gene conversion or concerted evolution is one, but not the only, way to meet the requirement of ribosomal constraint.
Another level of ribosomal constraint was observed at the level of 2° structures. For example, the detailed analysis of the 23S rRNA genes of T. tengcongensis
showed only in uncommon instance (10 of 115 point substitutions) a point substitution altered the 2° structure. This observation is in line with the current concept derived from comparing consensus rRNA sequences from closely related species 
in that selection pressure from ribosomal constraints is to preserve 2° structures that are essential for the assembly of functional ribosomes 
. In organisms with a single rRNA operon, ribosomal constraints could follow a simple birth-and-death model by linking the fate of a mutation to the viability of the organisms. However, in an organism with multiple rRNA operons, the nature of the selection pressure is unclear. In Escherichia coli
, which has seven rRNA operons, after deletion of the other 6 operons only a single operon is sufficient to produce >50% of wild-type rRNA levels and support >50% of the wild-type growth rate 
. This experiment suggests that in an organism with multiple rRNA operons, it may not be necessary for survival to keep all rRNA genes under ribosomal constraint. The observed tight constraints on individual 23S rRNA in T. tengcongensis
needs a new explanation. T. tengcongensis
, isolated from a fresh water hot spring in Tengchong, China, is a rod-shaped, gram-negative bacterium that grows anaerobically in extreme environments. It propagates at wide ranges of temperatures from 50° to 80°C and pH values between 5.5 and 9 
. It is possible that all four 23S rRNA genes are essential for survival of T. tengcongensis
in its natural inhabit, though one might be sufficient under laboratory conditions. It also is possible that the divergence of 23S rRNA genes offers competitive advantage in a changing environment in nature and thus are selectively maintained. Ribosomal constraints could be met by maintaining conserved 2° structures through conserved point mutations under sufficient selective pressure to resist the tide of homologous recombination.
Our finding is consistent with the finding by Acinas et al.
of T. tengcongensis 
. Both studies indicate that the 2° structures are conserved despite a high level of change in the primary structures and suggest that the rrnC
operon is functional. Although Acinas et al
. suggested that the rrnC operon could be brought into T. tengcongensis
by horizontal gene transfer, the hypothetic donor has yet to be identified. The ribosomal constraint observed at the level of 2° structures is also evident in 23S rRNA genes in C. hydrogenoformans
, H. marismortui
, S. oneidensis
, and S. pyogenes
Complex rearrangement of the position of nucleotides, the position and size of loops, and the length of stems might be another mechanism to conserve the functional structure of an rRNA gene. The examples shown in N. farcinica and C. perfringens are beyond simple alteration or conservation of the 2° structures since there is no straightforward way to quantify the changes and their effects. Rather, this type of rearrangement seems to aim at conserving the topographic structure in the presence of drastic changes in the primary and secondary structures. However, an experimental approach will be required to address the relationship between the conservation of the topographic structure and function of an rRNA gene.
The concept of rRNA conservation has revolutionized exploration of microbial ecosystems and identifying the causes of diseases of unknown etiologies 
. The most common molecular sequences used for phylogenetic analysis are ribosomal RNAs. The 16S rRNA genes have become the most widely used sequences for phylogenetic analysis as seen by the vast and growing database of 16S rRNA genes 
. The choice of 16S rRNA as the most widely used sequences for phylogenetic analysis has been based on the simple assumption that 16S rRNA is optimal because 5S rRNA is too short while 23S rRNA sequencing requires extra effort without a commensurate increase in phylogenetic signal. However, misidentification has been reported. Strains from different species may have identical 16S rRNA gene sequences, and strains of one species may have 16S rRNA that differ by >4% while 3% is considered as the threshold for within-species dissimilarity 
. On the other hand, the main obstacle limiting 23S rRNA from being used in classification of microorganisms has significantly diminished, as the cost for DNA sequencing is becoming less expensive. Evaluation of intragenomic variation of the rRNA genes has become possible since more and more sequences of whole microbial genomes have recently become available. It would be a logical expectation that the intragenomic diversity of a marker for taxonomic classification of microorganisms is below the level of interspecies diversity. We evaluated the suitability of 23S rRNA genes in this regard. Our study indicates that high-level variation among paralogous 23S rRNA genes might exceed the level of variation among 23S rRNA genes from different strains within a species but will not have a significant effect on the integrity of a 23S rRNA-based phylogenetic system, because its occurrence was rare. The insertions encountered when using 23S rRNA genes in phylogenetic analysis can be easily recognized and eliminated in comparisons of homologous sequences.
Intervening sequences (IVS) were frequently observed in 23S rRNA genes. In the seven species in which IVS were identified, five species contain small IVS that did not contain any ORFs but in D. radiodurans
sp, IVS were large and contained ORFs encoding putative transposases. Experimental studies have demonstrated that IVS in 23S rRNA does not interfere with its functions
. The 23S rRNA of Salmonella typhimurium
LT2 are known to carry IVS at two sites, helix-25 and helix-45, which are excised by RNase III during rRNA maturation, resulting in rRNA which is fragmented, but nevertheless functional 
. Fragmentation of the 23S rRNA also has been observed in diverse Eubacteria 
. 23S rRNAs with insertions have been shown to rescue an E. coli
strain in which all chromosomal rRNA operons were deleted