Promoter reuse in prokaryotes.

Anecdotal evidence shows promoters being reused separate from their downstream gene, thus providing a mechanism for the efficient and rapid rewiring of a gene's transcriptional regulation. We have identified over 4000 groups of highly similar promoters using a conservative sequence similarity search in all fully sequenced prokaryotic genomes. About 6% of those groups are shared between bacteria from different taxonomic depth, including different genera, families, orders, classes and even phyla. Database searches against known mobile elements and RNA motifs have indicated that regulatory motifs such as riboswitches could be moved around on putative mobile promoters.

Reuse of protein coding DNA sequences through gene duplication and horizontal gene transfer is a well-known and profound innovative force in nature; in sharp contrast to this, the mobility of a gene’s transcription regulatory function encapsulated in its promoter region is much less known. There are a few well studied classes of mobile genetic elements that harbour functional promoters, like Correia elements, ERICs and REPIN. But also examples of duplicated promoters not associated with known mobile elements, suggest that promoter reuse could represent a rampant and rapid mechanism of gene rewiring. In a recent publication Blount et al. identified a promoter capture event as a crucial step in the evolution of aerobic citrate utilization by a population of Escherichia coli in a long-term evolution experiment, and speculate that promoter capture may be an important and little appreciated adaptive force in genome evolution. Similarly, Bongers et al. described the activation of a silent lactate dehydrogenase gene by promoter recruitment in Lactococcus lactis. In both studies insertion sequences (IS) were involved in promoter mobility, though Blount et al. also found cases that were not associated with IS elements. In order to estimate the relevance of promoter recruitment in genome evolution we made a conservative inventory of such events in prokaryotes, which was recently published in Nucleic Acids Research.

Tip of the Iceberg

To assess the extent of promoter reuse in bacteria we looked for groups of bacterial genes per genome that share highly similar sequences upstream of their transcriptional start site, but do not have obvious flanking paralogous coding sequences. More specifically, we extracted in silico the DNA region between positions -150 and -50 relative to the start of translation for all genes in a genome (including plasmids), except when this overlapped with the coding region or promoter region of a flanking gene. In Escherichia coli the majority of the transcriptional start sites were shown to be between 20 and 40 nucleotides upstream of the translational start site, so most of our upstream sequence fragments will not contain the important -10 (Pribnow) box, but should include the -35 element. Using BLAST we then searched for sequence pairs that matched with 80% or more nucleotide identity over at least 50 nucleotides, to select for highly similar regions rather than for short conserved DNA elements. Sequences with more than one hit in the database were clustered into families. Sequence pairs that in addition showed a high nucleotide identity in their adjacent coding sequences were assumed to be paralogs and excluded because for this study we were interested in the independent mobility of promoters, not duplicated regions (for details see the materials and methods in our NAR paper).

We analyzed all available complete prokaryotic genomes (1,362; July 2011) and even with our strict selection criteria found over 4,000 families of highly similar sequences upstream of apparently unrelated coding sequences. The majority of these families actually consist of pairs that on average share 92% nucleotide identity, meaning that at least 46 out of 50 base pairs were conserved, but we also found pairs that were completely identical over 100 base pairs. Whether this level of high identity is the result of a strong selective pressure, or indicative of recent duplication events remains to be investigated. We termed these homologous non-coding sequences Putative Mobile Promoters, PMPs. In fact, some of these sequences likely are not promoters but have a different function that causes their conservation. Looking for known elements in our PMP set we actually found 42 tRNAs, 83 resembled other RNA families like the regulatory riboswitches, and interestingly 210 were known insertion sequences. The > 4000 families that our study uncovered represent only a small sample of a large pool of repeated DNA in promoter regions, a conservative reference of promoter reuse in prokaryotes. We anticipate many relevant examples of the phenomenon remain undetected because of our strict criteria. For example, filtering out paralogous genes also removes mobile promoters that extend into the coding region, like reported cases of Correia elements that overlap with an ORF. In addition, our initial extraction of promoter sequences is sensitive to wrongly annotated translational start sites, which is a known issue with genome annotation pipelines.

Horizontal Promoter Transfer?

More surprising even than the large number of promoter pairs sharing high nucleotide identity within one bacterial genome is that about 6% are shared between distantly related species. Clustering these based on sequence similarity resulted in 62 distinct groups, of which four are present in species that are related only by belonging to the same phylum (Table 1). As expected, inter-taxon transfers seem to decrease with phylogenetic distance and at the domain level, i.e., between Archaea and Bacteria, no transfer events were observed. Some non-coding sequence elements like tRNAs are very well conserved over large evolutionary distances, but if highly similar sequences are found only in small number of distantly related species horizontal gene transfer is a more likely scenario. The large majority of the PMPs are located on a chromosome, but for one group of PMPs all members are in fact on plasmids. These plasmids are associated with multiple-drug resistance in pathogenic Salmonella and are frequently transferred between bacterial species.

Table 1. Number of PMPs shared by bacterial genomes from different genera, families, orders, etc

Branch point	Count
Genera	28
Families	12
Orders	9
Class	9
Phylum	4
Domain	0
	62

Although the genetic code for translating DNA to protein is extremely well conserved between species as distant as Escherichia coli and Homo sapiens, transcriptional cis-regulatory elements are much more variable and their activity can differ even between strains of the same species., It can therefore be expected that the 62 homologous PMPs are not primarily transcription factor binding sites, but rather have other (regulatory) functions causing their high conservation. Indeed, two of the PMPs that are shared between families of bacteria are known S-adenosylmethionine (SAM) binding riboswitches. The other 60 PMPs however did not resemble any of the RNA families included in the Rfam database, so their function at present remains uncovered.

We conclude that we have uncovered a large number of putative mobile promoter families, present in numerous bacterial genomes. These may be involved in rapid adaptive processes via transcriptional rewiring, or include post-transcriptional regulatory functions. The ways these PMPs move within and between genomes is still unknown, but due to the large number of families, this may include diverse mobilization mechanisms.

Finally, although transcription regulation in eukaryotes is more complex than in bacteria, it seems obvious that also in eukaryotes promoter reuse offers a mechanism for rapid adaptation of gene expression. It would therefore be very interesting to extent our analysis to this domain, especially now more genomes and transcriptomes are becoming available that greatly facilitate the mapping of the core promoters.