ABSTRACT
We present evidence that the agr cell-to-cell communication system is present across firmicutes, including the human pathogen Clostridium perfringens. Although we find that the agr system is evolutionarily conserved and that the general functions which it regulates are similar in different species, the individual regulated genes are not the same. This suggests that the regulatory network controlled by agr is dynamic and evolves rapidly.
The pathogenic firmicute species Staphylococcus aureus has a well-characterized cell-to-cell communication system whose signals are modified peptides (6). This system is encoded in the accessory gene regulator (agr) locus (13). The agr cell-to-cell signaling system is an important regulator of biofilm formation and virulence factor expression (7). The agr locus consists of four genes, which are agrB, agrD, agrC, and agrA (Fig. 1A). The agrD and agrB genes encode the precursor signaling peptide and the enzyme cleaving and modifying the precursor peptide, respectively (Fig. 1B). The resultant signal is called autoinducing peptide (AIP). Systems which are similar to the agr system either by homology or by analogy have been identified in firmicutes outside Staphylococcus. In some cases they have different names. For example, the agr-like system in Enterococcus faecalis is known as the fsr system (12) and the one in Lactobacillus plantarum is called the lam system (3). Here we investigate the distribution of agr-like systems outside Staphylococcaceae in order to understand the conservation and plasticity of their interaction with other cellular components.
In order to define agr homologues in taxa other than Staphylococcus, we built hidden Markov models of protein domains which are considered to be unique to the agr system, namely, the AgrB and AgrD domains. To our knowledge, no function of AgrB apart from processing of AgrD has been reported so far. In a complementary approach, we identified agr homologues with PSI-BLAST searches. In the 384 genomes used in this analysis, 33 instances of the AgrB domain and 18 instances of the AgrD domain were found, all of them in firmicutes. The genomic context of the AgrB homologues was mapped onto a phylogenetic tree built from ribosomal protein sequences (Fig. 1C). All identified cases of the AgrD domain were found in close genomic proximity to a gene encoding an AgrB domain, which means that there are no orphan agrD homologues in the genomes we searched. All 18 loci which contained both the AgrD and the AgrB homologue were considered to be true quorum-sensing loci. Four other loci were also considered to be true quorum-sensing loci because they encoded a histidine kinase and a response regulator linked to the AgrB homologue. In the case of Enterococcus faecalis V583, the agrD homologue fsrD is transcribed together and in frame with fsrB but is translated independently of it (10).
To investigate whether the agr system is an evolutionarily ancient system, we built two phylogenetic trees by use of different methods. The first tree was built from an alignment of AgrB homologues. The second tree was built from an alignment of AgrB, AgrD, AgrC, and AgrA homologues, whose sequences were concatenated (see the supplemental material). The topology of these trees should agree with the topology of the tree built from ribosomal sequences if agr is inherited only vertically. Although these trees generally agree with each other and the ribosomal tree, the agr locus of Clostridium acetobutylicum clusters with Listeriaceae. Clostridium acetobutylicum is, however, a member of the class Clostridia, therefore making its agr system a candidate for horizontal gene transfer. This is also supported by AgrD of Clostridium acetobutylicum showing remarkable similarity to AgrD of Listeriaceae (Fig. 2). To our knowledge, this would be the first reported instance of horizontal gene transfer between Clostridia and Listeriaceae.
In three genomes we were able to identify a putative agr system where, to our knowledge, none had been reported before. These genomes are Moorella thermoacetica ATCC 39070, Desulfitobacterium hafniense Y51, and Thermoanaerobacter tengcongensis. We propose that these genomes might contain a functional peptide quorum-sensing circuit. Although we could not detect genes which take the role of agrD in these species, this could be due to errors in the prediction of small genes or to such genes being nonhomologous and located elsewhere on the genome or fused with other genes, as in the case of Enterococcus faecalis (10). Alternatively, this might represent instances of a system that can recognize peptides secreted by other organisms (cross talk). In the genomes, such as those of Bacillus halodurans, Syntrophomonas wolfei, and the three Clostridium perfringens strains, which contain an AgrB homologue that is not linked to a complete two-component system, agr-mediated quorum sensing might also be functional. This would require the two-component sensing system to be encoded elsewhere in the genome rather than being linked to agrB. However, in the Clostridium perfringens strain ATCC 13124 and strain 13, one of the two paralogues of AgrB cooccurs with a histidine sensor kinase. Peptide-mediated quorum sensing in the Clostridium perfringens species would be of particular interest as this species is the causative agent of the human disease gas gangrene (11). Interestingly, we also identified an uncharacterized gene frequently present in close genomic proximity to the agr locus which encodes a transmembrane protein with conserved residues (Fig. 1C; see also Fig. S1 in the supplemental material).
In the Clostridium perfringens strains as well as in Bacillus halodurans, we were able to identify short open reading frames which might have the function of agrD without having detectable sequence similarity to it (marked with the number 4 in Fig. 1C and highlighted in bold in Fig. 2). Although they lack significant sequence similarity to known AgrD homologues, they possess key sequence features connected to AIP function: a central cysteine residue which is required for the formation of the thiolactone ring (9) and a Pro-X-X-Pro motif which might serve as a recognition site for AgrB (16) (Fig. 2).
Next, we asked whether the genes which are transcriptionally regulated by agr are conserved between different species with a known agr quorum-sensing system. Differences in gene expression levels between wild-type and agr knockout strains have previously been obtained for Lactobacillus plantarum WCFS1 (14), Enterococcus faecalis OG1RF (1), Staphylococcus aureus NCTC 8325 (4), Staphylococcus aureus UAMS-1 (2), and Staphylococcus epidermidis (15). We used these data to compare genes which are differentially expressed in these agr mutants (see Fig. S3 in the supplemental material). According to our criteria (see the supplemental material), 175 genes were differentially expressed in Staphylococcus aureus, 54 in Lactobacillus plantarum, 54 in Enterococcus faecalis, and 238 in Staphylococcus epidermidis. After defining clusters of orthologous genes, we asked if there were any orthologues which were differentially expressed in all three genomes. We did not find any cluster which was differentially regulated in all three genomes (Fig. 1D). Therefore, it appears that there is no significant propensity for orthologous genes to be regulated by the agr system in the three different species. Using the Infernal program (5) to detect small noncoding RNAs, we found that regulation of genes by agr via the global regulator RNAIII (Fig. 1B) occurred only in the Staphylococcus species. This suggests that RNAIII is an evolutionary innovation unique to Staphylococcaceae and that the downstream regulators under the control of agr can evolve rapidly.
In agreement with previous observations, genes which are involved in biofilm formation, such as those encoding exopolysaccharide biosynthesis proteins, are downregulated by agr in Staphylococcus aureus and Enterococcus faecalis (8). Some exopolysaccharide biosynthesis genes, such as those of the cps2 cluster, are upregulated in Lactobacillus plantarum. This seems to contradict the observation that deletion of lamA, the agrA homologue, has the phenotypic effect of reduced adherence to a glass substratum compared to results for the wild type. The production of virulence proteins is upregulated by the agr-like system in Enterococcus faecalis and Staphylococcus aureus N315, which are both virulent. A clear conclusion is that the agr system is not a universal regulator of virulence factors only and is also active in nonpathogenic firmicutes. Furthermore, we conclude that although the functions which are regulated, including biofilm formation, are similar, the individual genes are not. This suggests that the regulatory network controlled by agr is flexible and can adapt rapidly to changing environments. Therefore, the agr system might be of use in the emerging field of synthetic biology.
(A) Organization of the agr locus in S. aureus, with Pfam domains highlighted in color. nt, nucleotides. (B) Cartoon of the basic interactions of the agr system in S. aureus. The interactions of the proteins encoded by the agrBDCA operon lead to a positive-feedback loop and further agrBDCA expression. Blue arrows, phosphorylation; green arrows, transcriptional interactions. RNAIII and hld, which encodes δ-lysin, are controlled by the same promoter and share the same transcript. The second promoter controls agrBDCA. The agrC and agrA genes encode the transmembrane kinase which senses the AIP (cyan circles) and a response regulator (red hexagon) which is phosphorylated by AgrC and affects changes in gene expression, respectively. Phosphorylation of the AgrA receiver domain by AgrC leads to AgrA activation and DNA binding. AgrA positively regulates transcription of the agr operon. This results in the production of a high level of AIP, leading to even higher agr expression. Apart from the agrBDCA promoter, another promoter positively regulated by AgrA is the one controlling expression of a small noncoding RNA, RNAIII. The RNAIII promoter is located next to the agrBDCA promoter but is transcribed in the opposite direction. RNAIII rather than AgrA is the immediate regulator of most of the genes regulated by the agr operon. (C) Tree of firmicutes based on ribosomal sequences with bootstrap values (1,000 iterations). Where applicable, the genomic organization of the agr locus homologue is depicted to the right of the operational taxonomic unit. One instantly obvious finding is that homologues of the agr system are spread across firmicutes and are not limited to a single phylogenetic class. Furthermore, the genomic organization of the agr system is conserved outside the family Staphylococcaceae, that is to say, in most cases agrB is linked to a two-component signal transduction system. While each genome contained only a single copy of the AgrD domain, some contained more than one copy of the AgrB domain: M. thermoacetica (three copies), D. hafniense Y51 (three copies), C. perfringens (two copies), and C. perfringens ATCC 13124 (two copies). (D) Venn diagram showing numbers of differentially expressed genes and how many orthologous clusters are differentially expressed in more than one genome.
Alignment of potential AgrD homologues. GenBank identifiers are given in the column to the right. Red, experimentally determined signaling peptides; yellow background, residues strongly conserved across the homologues; green background, residues conserved between Listeria and Clostridium acetobutylicum. Boldface type shows potential AgrD homologues which were identified by genomic context (linkage to AgrB) rather than by homology. The conserved cysteine, which is required for the formation of the thiolactone ring, and the Pro-X-X-Pro motif are highlighted at the top of the alignment (see text for details).
ACKNOWLEDGMENTS
A.W. thanks the Medical Research Council for funding.
A.W. and M.M.B. thank Ingmar Schäfer, Aswin Seshasayee, and Subhajyoti De for critically reading the manuscript.
FOOTNOTES
- Received 18 July 2007.
- Accepted 28 September 2007.
- Copyright © 2008 American Society for Microbiology
