The putative chemoreceptor families of C. elegans ^*

Chemoperception is a central sense of soil nematodes like Caenorhabditis elegans, as it is for most other animals (reviewed in Mori, 1999; Sengupta, 1997; Bargmann, 1997). Following the Nobel-Prize-winning discovery of olfactory receptors (ORs) in mammals (Buck and Axel, 1991) as members of the rhodopsin superfamily of G-protein-coupled receptors (GPCRs) with seven transmembrane domains (7TM), Troemel et al. identified a series of disparate GPCRs as candidate olfactory receptors in C. elegans (Troemel et al., 1995). These identifications were made from some of the earliest sequences reported by the genome project and proved to be the tip of the iceberg. The genes were classified into small gene families and given names starting with sr for serpentine receptor (sra, srb, srd, sre, and sro). Their probable chemosensory function was based on transgene expression patterns in one or more known pairs of chemosensory neurons. Subsequently, Sengupta et al. used a genetic screen to identify odr-10, which they showed encodes a GPCR that mediates olfactory response to diacetyl (Sengupta et al., 1996). With completion of the genome sequence it became possible to flesh out the true extent of these and related gene families encoding GPCRs that are candidate chemoreceptors. Robertson (Robertson, 1998; Robertson, 2001) worked up the relatives of odr-10 as the large str family along with the related sri and srj families, and then also worked up the large related srh family (Robertson, 2000). Annotation and naming of the smaller sra family was completed by HMR as part of the comparison with C. briggsae (Stein et al., 2003), while Chen et al. described the related srab family (Chen et al., 2005), and Thomas et al. described the srz family (Thomas et al., 2005) and the srt family (Thomas, 2006 [in press]). A decade after their first recognition, we describe here all the remaining families and relate them in a series of superfamilies that comprise a remarkable array of roughly 1300 potentially functional candidate chemoreceptor genes, along with about 400 apparent pseudogenes. Together they comprise roughly 7% of the gene complement of this tiny animal, more than any other sequenced genome. This large genetic investment might result from an extreme dependence on chemosensory abilities in the absence of visual and auditory systems. On the basis of genetic complexity, this system probably constitutes the bulk of signal transduction that occurs in C. elegans.

We first discuss these families below in superfamily groupings, loosely defined as consisting of readily recognizable groupings based on protein sequence similarity, as well as sometimes shared intron locations in their genes. We then discuss various aspects of their biology, including new expression pattern results that support their likely function as chemoreceptors. For the rest of this chapter, while fully acknowledging the formal status of all but odr-10 as putative or candidate chemoreceptors, we will refer to them as chemoreceptors. Genetic analysis of sensory-neuron specific G proteins indicates that olfaction, nociception, and pheromone responses are mediated predominantly by GPCRs (e.g., Hilliard et al., 2004; Jansen et al., 1999; Jansen et al., 2002; Roayaie et al., 1998) and it is difficult to see what else these huge families might generally mediate. Nevertheless, it seems likely that some of the genes have been subsequently recruited to other roles such as monitoring of internal chemistry, so the function of any specific gene should be interpreted with caution.

Table 1 lists the named chemoreceptor families by superfamily group with the numbers of apparently intact genes and of apparent pseudogenes in each. Annotation of the genes is not nearly as straightforward as for mammalian olfactory receptors, because the C. elegans genes have complex gene structures with 1–8 introns each, and the intron locations are commonly different between gene families. EST data are of little assistance due to uniformly low expression levels (e.g., of ~160,000 ESTs on WS138, only 81 were from chemoreceptor genes). In the absence of functional information it is not possible to be confident of the functionality of any specific gene, but each has been annotated to encode a seemingly intact full-length protein that aligns well with other family members. It is always possible that particular amino acid changes or small in-frame insertions or deletions might inactivate an otherwise apparently intact receptor. On the other hand, amongst the apparent pseudogenes from N2 is a subset with probable functional alleles in other wild strains of C. elegans (Stewart et al., 2005). In part because of this we have chosen to name both genes and pseudogenes in these families. We have arbitrarily chosen to ignore gene fragments that encode less than 50% of the amino acid length typical for a gene family. The curated annotations and gene names for all families should be correct in WormBase release WS141, though ongoing studies will undoubtedly refine their structure.

An unusual characteristic of the chemoreceptor families is their highly dynamic gene number, reflected within C. elegans by frequent duplications and gene loss. Systematic comparisons with C. briggsae are only beginning, but evidence thus far indicates that the familiar concept of orthology is largely useless for these families due to their dynamic evolution. Figure 4 shows a phylogenetic tree of two of the three families for which a completed C. briggsae annotation is available (Chen et al., 2005). Several large expansions in gene number in C. elegans are apparent and only 7 pairs of genes are convincing one-to-one orthologs. Several additional reciprocal best match pairs (marked "a") are substantially more divergent in protein sequence, which probably indicates that they duplicated well before elegans-briggsae speciation (followed by loss of one copy in both species) and are not orthologs in the strict sense. Similar dynamics are even more extreme in the srz family (Thomas et al., 2005) and the srh families (JHT, unpublished).

Much work remains to be completed, but the broad picture is that the chemoreceptor gene families are in a dynamic state, with fixation of new gene duplicates approximately balancing gene loss (reflected in the large number of defective genes, and in the evidence for ultimate deletion of many genes entirely). Over relatively short periods this dynamic is probably near equilibrium, resulting in a stable number of genes. However, over longer periods unknown factors can change the relative rates of gene gain and gene loss, giving rise to the dramatically different gene numbers in C. elegans and C. briggsae. This pattern is strikingly reminiscent of recent findings in mammalian odorant and vomeronasal gene families, in which gene duplications, gene loss, and lineage specific expansions are common (Del Punta et al., 2000; Glusman et al., 2001; Matsunami et al., 2000). It will be interesting to determine the chemosensory gene distributions for the other Caenorhabditis species currently in the genome sequencing pipeline.

The putative chemosensory genes are strongly clustered in the genome (Figure 3). Most families are heavily enriched on chromosome V, and significant enrichment is also found on chromosomes II and IV. Furthermore, the positions of genes on all of these chromosomes are strongly biased toward the chromosome arms (as defined by increased recombination frequency; Barnes et al., 1995). In addition to these broad patterns of gene clustering, there is a strong tendency for closely related genes to be physically close in the genome (Chen et al., 2005; Robertson, 2000; Robertson, 2001). These patterns are likely attributable to patterns of gene duplication in C. elegans. New duplications are almost always local, often tandem or inverted repeats with little or no intervening sequence between the duplicates (Katju and Lynch, 2003; Semple and Wolfe, 1999). The current gene positions are consistent with a history of frequent local gene duplication and rare large chromosomal rearrangements, which occasionally break up clusters of local related genes. These large rearrangements are rarely if ever interchromosomal translocations; we suspect that most are inversions.

Detailed analysis of the distribution of srh and str genes not on chromosome V showed that small interchromosomal transposition events occur occasionally (Robertson, 1998; Robertson, 2000; Robertson, 2001). These events are followed in some cases by a duplication and diversification process in the new genomic location. As with chromosome V, residence on other chromosomal arms seems to favor the duplication process. The mechanisms underlying this pattern are not known, but are likely to involve the higher frequency of complex DNA repeats found on almost every autosomal arm.

Analysis of the function of C. elegans chemoreceptors has been remarkably resistant to genetic methods. Only odr-10 has been identified in classical forward genetic screens (Sengupta et al., 1996) and the fraction of genes assigned gross phenotypes in the initial global RNAi screen is close to zero (Kamath et al., 2003). This resistance may result from a combination of the expected narrow phenotype range for the genes (and a relative paucity of genetic screens directed toward this) and genetic redundancy among closely related receptors. A few additional chemoreceptor genes have more limited phenotypic information. Deletion or overexpression of sra-13 affects olfactory response to isoamyl alcohol and also affects vulval induction, probably as part of a sensory response to starvation (Battu et al., 2003). One sign of hope is that about 25 chemoreceptor genes were identified as affecting fat accumulation in a global RNAi screen testing specifically for this phenotype (Ashrafi et al., 2003). In addition, RNAi knockdown of str-2 is reported to extend lifespan, though the effect is small (Alcedo and Kenyon, 2004). These RNAi results are uncorroborated by other approaches, but it is possible that environmental chemoperception or internal physiological monitoring by chemoreceptors plays a prominent role in energy resource allocation.

The cellular expression patterns of the putative chemosensory receptor genes are of interest for two reasons. First, their classification as chemosensory receptors is currently best assessed by testing whether expression is predominantly in chemosensory neurons. Second, because of the large set of genes, their patterns of expression provide a remarkable opportunity to study the relationship between cell fate, transcription factors, and cis regulatory sequences. Only one gene in the entire SR class has been rigorously shown to function as a chemosensory receptor (Sengupta et al., 1996; Zhang et al., 1997). Are the other ~1,300 genes really chemosensory receptors, or are they a mixture of chemoreceptors and other types of receptors? Transgenic GFP-fusion expression data bearing on this question are summarized in Table 2. Though the coverage of the superfamily is insufficient to draw detailed inferences, a broad picture can be painted. Many genes are expressed only in chemosensory neurons, consistent with a strictly chemosensory function. There are also exceptions to this rule, which might result from real function in other types of neurons and non-neuronal tissue or from transgenic misexpression in ectopic tissues, possibilities that it is currently not possible to distinguish. In addition to GFP fusion data, one study used microarrays to examine gene expression from cell-sorted AWB neurons (Colosimo et al., 2004), a method that is likely to gain importance in the future. SR genes were prominent among the genes identified, including str-220, srd-16, srd-23, sru-38, and srsx-5. Analysis of SR expression by microarray studies using total worm mRNA is problematic because of their generally low abundance transcript levels; though some genes were assigned to expression mountains (Kim et al., 2001), these assignments should be viewed with some skepticism because absolute expression levels are typically at or close to noise levels.

A final method that shows promise in assigning expression patterns to SR genes is identification and analysis of specific promoter sequence motifs that may be binding sites for transcription factors. Using this approach, one study identified a well-defined E-box motif in the promoters of about 200 SR genes, mostly from the srh and sri families (McCarroll et al., 2005). It was further shown that the E-box is correlated with expression in the ADL chemosensory neuron, and that it was sufficient to direct expression of odr-10 in ADL.

HMR thanks Kim Walden (sru, srv, srx families), Kim Ly (srxa and srz families), Colin Stoetzner (srb and srbc families), Lindsey Schmidt (srw family), and Greg New (srx family) for assistance with annotation of some of the gene families, all the WormBase annotators and Jonathan Hodgkin for their assistance over the past decade, and Cori Bargmann for her enthusiastic support. JHT annotated the srt family and thanks Mary Stewart, Emily Rocke, Evan Galloway, and Joanna Kelley for assistance. HMR was supported by NSF grant IBN-96-04095 and NIH grant 1AI056081; JHT was supported by NIH grant RO1GM48700.