Cis-regulatory mechanisms of left/right asymmetric neuron-subtype specification in C. elegans

The classification of individual neuron types has a long history dating back to the work of Ramon y Cajal (Ramon y Cajal, 1911). Initially characterized by morphological criteria,more modern functional, physiological and molecular approaches have revealed an astounding diversity of cell types in the nervous system, more so than in any other organ (Masland,2004; Nelson et al.,2006). In order to understand the development of an organ composed of such vastly diverse cell types, it is useful to study the development of the simple units that make up the organ. One approach to breaking down the nervous system into simple units is the hierarchical clustering of neurons into groups of neurons that share common features such as morphology or function (`neuron classes' or `neuron types') and further clustering these groups into subtypes or subclasses. These subtypes share fundamental common functions or morphologies, but they differ in a restricted subset of features that define individual subtypes.

How are the regulatory mechanisms responsible for neuron type and neuron subtype specification integrated? Specifically, are the regulatory factors that control neuron type specification completely separate from those that control ensuing subtype specification or are they mechanistically linked? We address this here by exploring the specification of the bilaterally symmetric ASE neuron class in the nematode C. elegans. This neuron class is composed of two neurons, each defining its own subtype: the left ASE neuron(ASEL) and the right ASE neuron (ASER). Both neuron subtypes are morphologically and synaptically indistinguishable and share a large battery of co-expressed genes, but display differential asymmetric expression of a class of putative chemoreceptor genes (reviewed by Hobert et al., 2002).

ASE neuron class specification is controlled by the terminal selector gene che-1 (Chang et al.,2003; Etchberger et al.,2007; Uchida et al.,2003). che-1 encodes a zinc-finger transcription factor that directly controls the expression of its target genes by binding to its cognate binding sequence: the ASE motif. It is the expression of these target genes, including ion channels, neurotransmitter receptors and neuropeptide-encoding genes, that characterizes the terminally differentiated bilateral ASE neurons. While adopting their bilaterally symmetric terminal fate, the ASE neurons execute a further subtype diversification program in which the left and right ASE neurons diversify by differentially expressing distinct members of a family of putative chemoreceptors, encoded by the gcy genes (Fig. 1A)(Ortiz et al., 2006; Yu et al., 1997). This diversification program involves a gene regulatory loop, composed of transcription factors and miRNAs (Fig. 1A) (Johnston et al.,2005), and results in the asymmetric expression of the gcy genes. The expression of components of this regulatory loop, as well as the gcy genes themselves, is lost in che-1 mutants and each of these genes contains ASE motifs in their promoters that are required for their expression in ASE(Etchberger et al., 2007). Therefore, CHE-1 not only determines terminal differentiation features of ASE,but also directly induces downstream regulatory events required for subtype specification.

The necessity of CHE-1 for the specification of both neuronal subtypes,ASEL and ASER, poses the question of specificity. How does CHE-1 induce solely ASEL-specific genes in ASEL and ASER-specific genes in ASER, even though CHE-1 is expressed in both ASEL and ASER, and is able to induce bilaterally expressed genes in both neurons? Genetic evidence has shown that differential expression of genes in the regulatory loop shown in Fig. 1A is required for ASE to adopt either ASEL or ASER fate, but exactly how this mechanism functions at the level of cis-regulatory motifs is unknown. We envision several distinct models that are addressed here in this paper and are schematically presented in Fig. 1B. Model 1 is based on the observation that initially after the birth of ASEL and ASER, the two ASE subtypes appear to share the expression of terminal differentiation markers that eventually become restricted to ASEL or ASER during late embryogenesis and early larval development (Johnston et al., 2005). Model 1 posits that CHE-1 is required only for the induction of the bilateral precursor state, inducing the expression of both terminal markers (such as gcy-7, which is initially expressed in both ASEL and ASER) and regulatory factors that directly control ASEL or ASER-specific genes. These regulatory factors may then be sufficient to induce the L/R diversification, thereby making CHE-1 activity superfluous after this initial specification event (Fig. 1B). In Model 2, CHE-1 is required throughout the lives of the ASE neurons and actively participates in controlling the L/R asymmetry of the gene expression profiles in these cells. As CHE-1 acts in a target gene-specific manner in both ASEL and ASER, the L/R specificity cannot be brought about simply by directly modulating intrinsic CHE-1 activity (e.g. by post-transcriptional mechanisms such as protein modifications or stability). Rather, L/R specificity would need to be brought about by the repression of CHE-1 activity in a target gene-dependent manner through what we term here`co-repressor' molecules (Fig. 1B, Model 2a). Alternatively, CHE-1 may not be sufficient to activate L/R-restricted target genes and may require what we term here`co-activator' molecules (Fig. 1B, Model 2b).

We test these models here in a variety of ways, including temporally controlled gene knock-down, dissection of the cis-regulatory architecture of L/R asymmetrically expressed genes and various genetic manipulations. We find that L/R asymmetric gene expression programs are controlled by complex and diverse cis-regulatory control mechanisms and we show that the affinity of CHE-1 to its binding site plays a role in controlling L/R asymmetric gene expression programs.

All gfp reporter constructs are based on the pPD95.75 vector and were generated either by subcloning or PCR fusion(Hobert, 2002). As indicated in the figure legends, constructs were injected either as circular subcloned DNA or as a linear PCR fragment, generated either by PCR fusion or PCR amplification of a cloned product. DNA constructs were injected into an otIs151 background, which expresses a ceh-36::DsRed2transgene to facilitate the identification of expression of gfp-based reporter genes in ASEL/R. The gut-specific elt-2::gfp reporter (a gift from J. McGhee) was used as an injection marker. Mutagenesis reactions were performed using the QuickChange II XL Site Directed Mutagenesis Kit(Stratagene). A list of DNA sequences can be provided on request.

RNAi by feeding was performed (see Ahringer, 2005) in the nre-1 lin15b sensitized background(Schmitz et al., 2007). For scoring of F1s, 10-15 staged L3/L4 hermaphrodites were placed onto plates seeded with dsRNA-expressing clones and incubated for 5 days at 22°C before scoring the F1 progeny as adults on a Zeiss axioplan microscope. In the case of P0 scoring, 100-150 staged L1s were placed on the plates and incubated for 3-4 days before scoring as adults.

CHE-1 and CEH-36 proteins were expressed and purified, and binding reactions were performed (see Etchberger et al., 2007). For quantitative cold competition assays, different types of cold competitor (ASE motifs from genes tested) were added at specific ratios relative to the labeled ceh-36 probe to the protein/ceh-36 probe mixture after 10 minutes of the initial binding reaction and allowed to incubate with the protein/probe complex for an additional 10 minutes before loading on the gel. After detection of bound probe using the Molecular Dynamics PhosphorImager and quantification with ImageJ, the fraction of bound labeled ceh-36 probe compared with bound uncompeted probe was plotted against the molar ratio of unlabeled to labeled probe (Glass et al.,1988). The affinities of each probe relative to the ceh-36 probe were calculated by comparing the slopes of a plot of[(1-Y)/Y] versus amount of competitor, where Y is the fraction of bound labeled probe at a given ratio of unlabeled to labeled probe(Glass et al., 1988).

che-1 null mutant animals do not initiate ASE differentiation, as evidenced by a complete lack of expression of ASE differentiation markers at all stages of embryonic development (Chang et al., 2003) (data not shown). To assess whether the function of che-1 function is restricted to the initiation of ASE differentiation or extends to the maintenance of the differentiated state(Fig. 1B, Model 1 versus Model 2), we reduced che-1 activity in a temporally restricted manner by performing RNAi on transgenic animals expressing either an ASEL-specific marker (lim-6::gfp) or an ASER-specific marker (gcy-5::gfp). The ASE neurons undergo their final division at ∼380 minutes post-fertilization (Sulston et al.,1983). che-1 expression can be observed as early as one cell division before the birth of ASEL and ASER, and persists through adulthood (B. Tursun and O.H., unpublished). We delivered che-1 dsRNA by feeding animals at the larval L1 stage (i.e. long after ASE has been generated in mid-embryogenesis). If che-1 activity was required solely for the initiation of gene expression in ASEL/R, then only the F1 progeny of RNAi-treated animals and not the P0 animals should show a loss or reduction of gfp expression. If che-1 activity was continuously required during the lifetime of the animal, the removal of che-1 would result in a loss or reduction of gfp expression in the acutely RNAi-treated P0 animals. After treatment of staged L1 animals with che-1 RNAi, we observed a significant reduction of both ASEL-specific and ASER-specific marker expression in both ASEL and ASER in adult P0 animals (Fig. 2A,B). These results indicate that che-1 is required not only for the initiation but also for the maintenance of gene expression in ASEL/R.

The continuous requirement of che-1 activity for ASEL/R gene expression means that L/R asymmetric transcription of terminal differentiation genes necessitates a mechanism that restricts the activity of bilaterally expressed CHE-1 in a target gene-dependent (i.e. promoter-dependent) manner to either ASEL or ASER. `Promoter-dependent' means that bilaterally expressed genes respond to CHE-1 in both ASEL and ASER, whereas L/R asymmetrically expressed genes respond only to CHE-1 in either ASEL or ASER. As mentioned in the Introduction, this may involve either promoter-dependent co-activator or co-repressor mechanisms (Fig. 1B). To test these models, we analyzed the cis-regulatory architecture of four asymmetrically expressed genes that are direct targets of che-1, as they each contain ASE motifs required for expression in the ASE neurons (Etchberger et al.,2007; Sarin et al.,2007): two putative chemoreceptors of the guanylyl cyclase family,ASEL-expressed gcy-7 and ASER-expressed gcy-5(Yu et al., 1997); the ASEL-expressed LIM-homeodomain transcription factor lim-6(Hobert et al., 1999); and the ASEL-expressed microRNA lsy-6(Johnston and Hobert, 2003). The activation model (Fig. 1B,Model 2b) predicts that the ASE motifs of each of these genes are not alone sufficient to promote expression in both ASE neurons, whereas the repression model (Fig. 1B, Model 2a)predicts that the ASE motif alone can drive expression in both ASE neurons. We find that each isolated ASE motif from the four asymmetrically expressed genes when fused to gfp produces reporter gene expression in both ASE neurons (Fig. 3). As this bilateral expression is lost in the context of larger regulatory elements,these regulatory elements must therefore contain information that somehow restricts CHE-1 and ASE motif activity. To dissect the nature of the additional regulatory information, we undertook a systematic deletion analysis of the L/R asymmetrically expressed regulatory regions of these genes(Fig. 4).

We conducted a scanning mutagenesis of the cis-regulatory region that is required and sufficient for the ASEL-specific expression of gcy-7(gcy-7^prom2, 188 bp) (Etchberger et al., 2007). We first created seven non-overlapping deletion windows of 22-32 bp in length (Fig. 4A). Previous results have shown that the che-1-responsive ASE motif of the gcy-7 promoter lies at the junction of deletion windows 3 and 4, the only deletions that resulted in loss of expression in ASEL (Etchberger et al.,2007). Of the seven deletion windows, three constructs (del2, del5 and del7) show a partially penetrant derepression of gfp expression in ASER (Fig. 4A; see Figs S1 and S2 in the supplementary material). A triple deletion (del2,5,7) does not further increase the penetrance of de-repression, even though the levels of de-repressed gfp appear to be somewhat increased(Fig. 4A; see Fig. S2 in the supplementary material). These findings support the `co-repressor model' in Fig. 1B.

To refine our mapping of the repressive regulatory architecture, we generated smaller deletions within the identified repressive sequences. For all three regions, our analysis revealed discrete repressive sequences 6-8 bp in length that recapitulated the derepression effects of the larger deletions(del2b; del4b; del7a) (Fig. 4A;see Fig. S2 in the supplementary material). We also deleted small phylogenetically conserved motifs within the ASE motif-containing region del3(del3a) (see Fig. S3 in the supplementary material) and deletion window 6(del6a) (see Fig. S3 in the supplementary material). As a result of both deletions, we observed partial derepression of gfp in ASER(Fig. 4A; see Fig. S2 in the supplementary material). None of the four small motifs show notable similarities to one another, but each is phylogenetically conserved in the gcy-7 promoter of three other nematode species (see Fig. S3 in the supplementary material). One of the four motifs, del4b, matches a predicted NK-2 homeodomain binding site (TAAGTT)(Noyes et al., 2008). Other ASEL-specific gcy genes have some, but not all, of these motifs. We deleted two of these motifs (TTCGGGAG and GGAAAAAA) in gcy-14 and gcy-20, but these deletions did not affect ASEL-specific expression of these genes (data not shown).

We noted a predicted K50-homeodomain binding site (TAATCC) directly downstream of the ASE motif. Deletion of this motif (del4a) from the gcy-7 minimal promoter resulted in a loss of gfp expression in ASEL. This is surprising as this deletion construct retains its ASE motif,which, like any other ASE motif tested, drives bilateral expression in complete isolation (Fig. 3). In the context of the gcy-7 promoter, the CHE-1-binding ASE motif,therefore, needs to collaborate with a `co-activator' motif(Fig. 1B, Model 2b) to activate ASEL-specific expression.

The `co-activator' motif also confers a left/right asymmetric activity to the ASE motif as a construct that contains only the ASE motif and the neighboring co-activator motif is asymmetrically expressed, with gfplevels in ASEL being substantially higher than in ASER(Fig. 4A) (compare ASE^gcy-7#1 with ASE^gcy-7#2). In other words, such a minimal motif is derepressed in ASER owing to the loss of the repressor motifs described above, but it still retains L/R asymmetry to some degree.

Intriguingly, the loss of ASE expression observed upon loss of the co-activator motif (del4a) is completely suppressed if the del4a deletion is combined with deletions of the three repressor motifs: del2, del5 and del7. The resulting quadruple mutant construct is expressed in a bilaterally symmetric manner in both ASEL and ASER(Fig. 4A). We interpret these findings to mean that the ASE motif of gcy-7 can, in principle, drive bilateral expression in ASEL and ASER; this activity alone is repressed through repressive cis-regulatory motifs, but the presence of the TAATCC motif can overcome this repression specifically in ASEL(Fig. 4A). In summary, gcy-7 is regulated by a combination of the `co-activator' and`co-repressor' models as shown in Fig. 1B.

We started the analysis of the cis-regulatory region of the gcy-5gene again with a scanning deletion analysis on the fragment of the gcy-5 regulatory control region (gcy-5^prom2, 306 bp),which is required and sufficient to drive ASER-specific expression of a reporter gene (Etchberger et al.,2007). We deleted, in a non-overlapping manner, 24-31 bp windows(Fig. 4B). Three of the 11 deletion windows (del3, del7 and del11) each resulted in a low penetrant de-repression of gfp expression in ASEL, while leaving expression in ASER unaffected (Fig. 4B; see Figs S1 and S2 in the supplementary material). Double (del7,11) and triple deletions (del3,7,11) exhibited an increase in the penetrance of de-repression of gfp in ASEL (see Fig. S2 in the supplementary material). However,neither the double nor triple deletion constructs showed equivalent bilateral expression in ASEL/R. Bilateral expression is observed, however, if all sequences around the ASE motif are chopped away, leaving only a 24 bp minimal element (Fig. 3)(Etchberger et al., 2007). We conclude that repressive elements within region 3, 7 and 11 are, in part,responsible for repression of gcy-5 expression in ASEL. Notably, in contrast to the gcy-7 regulatory region, the gcy-5regulatory region does not appear to contain a `co-activator' motif as no mutation other than the ASE motif deletion results in a loss of reporter gene expression.

To define individual sequence motifs within these repressive regions, we created smaller deletions within each repressive region in an attempt to recapitulate the derepression resulting from the larger individual deletions(Fig. 4B; see Fig. S2 in the supplementary material). For del3 and del7, we were able to identify short sequences (8 bp and 7 bp, respectively), the loss of which resulted in de-repression of gfp expression in ASEL. Neither motif shows unambiguous matches to known transcription factor-binding sites predicted by MatInspector (Cartharius et al.,2005). For del11, we were unable to identify any discrete motif,even though our smaller deletions targeted nearly the entire original deletion window (Fig. 4B; see Fig. S2 in the supplementary material).

The repressor motifs share little sequence similarity with each other and are not conserved in C. briggsae. The C. briggsae gcy-5promoter, when injected into C. elegans, does not drive detectable gfp expression in ASE neurons (data not shown). This finding is also consistent with our previous observations that, in spite of strong synteny of at least some of the individual loci, the regulation of several other gcy genes and also their genomic organization has significantly diverged in C. briggsae (Ortiz et al., 2006). Last, we also did not find the motifs conserved in other ASER-restricted gcy genes, such as gcy-1 and gcy-22.

Previous promoter analysis has identified a 200 bp minimal cis-regulatory sequence of the lim-6 locus that is required and sufficient to direct expression of a gfp reporter gene in ASEL(Etchberger et al., 2007). We generated eight non-overlapping deletions of 24-28 bp within the minimal lim-6 promoter. Four out of the eight deletions (del4, del6, del7 and del8) resulted in the loss of gfp expression in ASEL(Fig. 4C)(Etchberger et al., 2007). In contrast to our analysis of gcy-5 and gcy-7, none of the deletions resulted in the derepression of gfp in ASER.

The lim-6 minimal promoter contains two ASE motifs, one in deletion window 6 and one in deletion window 7. The targeted deletion of each of these ASE motifs individually resulted in a loss of gfp expression(Etchberger et al., 2007). However, as mentioned above, two additional deletion windows, del4 and del8,also resulted in loss of gfp expression(Etchberger et al., 2007). The del4 and the del8 sequence share no similarity with the ASE motif consensus sequence or with one another. They are also poorly conserved in other nematode species (see Fig. S3 in the supplementary material).

To identify the positive-acting cis-regulatory motif(s) in each region, we generated sub-deletions. No sub-deletion in del8 recapitulated the effect of the complete deletion, but within del4, two adjacent deletions, del4a and del4d, resulted in a loss of gfp expression in ASEL(Fig. 4C). In contrast to the temporally persistent loss of expression observed from the del4 construct,both del4a and del4d displayed normal ASEL-specific expression up to early larvae stages (L1/L2), but not in later staged larvae or adults(Fig. 4C).

Taken together, the ASE motifs of lim-6, even though supporting bilateral expression in isolation, are prevented from doing so in a full promoter context and require additional co-activator motifs to support expression in ASEL.

The 931 bp intergenic region between the lsy-6 miRNA hairpin and its upstream gene contains all the regulatory information required for ASEL-specific expression and function of the ASEL-fate inducer lsy-6(Johnston and Hobert, 2003; Sarin et al., 2007). Successive deletions identified a 95 bp element (prom3) that is sufficient to recapitulate expression only in ASEL (Fig. 4D). As the lsy-6 locus is highly conserved across four nematode species (Fig. 4D), we used sequence conservation as a guide for further deletion analysis. We deleted four conserved sequence patches contained within the minimal 95 bp regulatory element. Deletion of a putative E-box, a predicted bHLH protein-binding site (del5; CANNTG) resulted in a complete loss of gfp expression. Therefore, as in the case of gcy-7 and lim-6, within its normal genomic context, the ASE motif (which remains intact upon deletion of the E-box) is not sufficient to drive ASE expression but requires an additional regulatory motif to be active. Moreover,we found that deletion of another conserved motif, del3, resulted in a completely penetrant de-repression of gfp expression in ASER(Fig. 4D, see Fig. S2 in the supplementary material) (gfp de-repression is only partially expressive and cannot be further enhanced in a del3; del4 double deletion). No obvious transcription factor-binding sites match this sequence, as assessed by MatInspector (Cartharius et al.,2005).

Similar to the gcy-7 case, the loss of gfp expression observed upon deletion of the E-box containing co-activator motif is suppressed if the repressor motif del3 is also mutated. The E-box therefore functions to counteract the co-repressor motif del3. If the co-repressor motif is absent, the co-activator motif becomes superfluous and the ASE motif can function in isolation. Notably, though, the ASE-motif-dependent expression in the del3/del5 double mutant is still biased to ASEL, suggesting additional mechanisms to confer L/R asymmetry.

Taken together, the restriction of ASE motif activity in the lsy-6promoter appears most similar to that of gcy-7: each promoter contains repressive elements and in each case additional co-activator motifs are required for ASE motif expression. However, the identified cis-regulatory elements share no similarity to one another.

In conclusion, the four cases analyzed here demonstrate that the regulation of asymmetric gene expression in ASEL/R requires a remarkably distinct spectrum of cis-regulatory mechanisms that differ in a gene-specific manner. In each case, promoters contain an ASE motif that in complete isolation drives bilaterally symmetric expression in ASEL and ASER. In the context of its normal promoter, ASE motifs do not display this sufficiency. The activity of the ASE motif either becomes restricted by repressive cis-regulatory elements;or it becomes dependent on the presence of additional activating cis-regulatory motifs; or it becomes restricted by a combination of both co-activators and co-repressors. None of the isolated additional cis-regulatory elements displays any similarity to one another.

How are the co-activator and co-repressor motifs identified above regulated? Previous genetic analysis has revealed a number of transcription factors involved in determining L/R asymmetry(Fig. 1A). However, little to no information about possible binding sites of these transcription factors exists, with the exception of the K50-homeodomain protein CEH-36, a C. elegans OTX homolog and the only K50-homeodomain protein known to be expressed in ASE (Lanjuin et al.,2003). K50-homeodomain proteins bind to a TAATCC core consensus sequence (Treisman et al.,1989) and such a sequence is required for the L/R asymmetric activation of gcy-7 in ASEL (Fig. 4A) (del4a) (see Fig. S3 in the supplementary material). Notably,the mutant phenotype of ceh-36 is consistent with ceh-36activating gcy-7 expression, as gcy-7 expression is lost in ceh-36 mutants (Chang et al.,2003). We find that bacterially produced CEH-36, but not two other tested homeodomain proteins, can indeed bind to the TAATCC motif in vitro (see Fig. S4 in the supplementary material). CEH-36 binding occurs in parallel and independent of the binding of CHE-1 to the neighboring ASE motif (see Fig. S4 in the supplementary material). Of the other ASEL-specific gcy genes(gcy-6,-14,-20), or any other ASEL-specific regulatory factor(lim-6, lsy-6), only the regulatory region of gcy-14contains a TAATCC motif in its functionally relevant regulatory region(lsy-6 also contains such a motif but it is not required for its correct expression; data not shown), thereby corroborating the overall theme that the mechanisms of L/R asymmetric regulatory control are diverse.

Apart from cis-regulatory motifs that collaborate with the ASE motif, we asked whether intrinsic features of an ASE motif may also bear relevance to its bilateral versus L/R asymmetric activation. Specifically, we wondered whether the ASE motifs of asymmetrically expressed genes have a weaker affinity for CHE-1 compared with the affinity of bilaterally expressed ASE motifs, which would allow co-activating and co-repressing cis-regulatory mechanisms to modulate the activity of the ASE motifs of asymmetrically expressed genes. This hypothesis was prompted by the creation of a metric, the`ASE motif score', which measures the quality of a match of a given ASE motif to the ASE motif consensus sequence(Etchberger et al., 2007). According to this metric, the asymmetrically expressed gcy-5 gene has a lower motif score than the bilaterally expressed ceh-36 gene (0.63 versus 0.72 in a scale from 0.56 to 0.73)(Etchberger et al., 2007). To put the relevance of this metric to an experimental test, we performed electrophoretic mobility shift assays and found that bacterially produced CHE-1 indeed binds the ASE motif from ceh-36 with higher affinity than the ASE motif from the gcy-5 locus(Fig. 5A,D). The basis for the difference in motif score and binding affinity appears to be a differential contact of the respective ASE motifs by CHE-1. Mutating individual zinc fingers and testing mutated proteins in gel shift assays demonstrates that CHE-1 contacts the ceh-36 ASE motif with three out of its four zinc fingers, whereas it contacts the gcy-5 ASE motif only with two out of four zinc fingers (Fig. 5A).

The low-affinity binding site in gcy-5 may be required to enable the negative regulatory elements (`co-repressor motifs') described above to counteract bilateral ASE motif activation. This hypothesis makes the prediction that if the low affinity ASE motif in the gcy-5 locus were switched with a high affinity ASE motif from the ceh-36 locus, the gcy-5 promoter may become active in both ASEL and ASER. We indeed find that a mere substitution of 4 bp, which transforms the gcy-5 ASE motif into that of the ceh-36 ASE motif, causes de-repression in ASEL(Fig. 5B). To test whether this logic also applies to an ASEL-specific gene, we analyzed the gcy-14promoter, which also contains an ASE motif with a motif score lower than that of ceh-36 (0.69 versus 0.72). We experimentally confirmed the lower affinity of the gcy-14 ASE motif compared with that of the ceh-36 ASE motif using competitive gel shift assays(Fig. 5D). As is the case for gcy-5, swapping the gcy-14 ASE motif with that of ceh-36 within the context of the gcy-14 promoter also results in bilateral expression of the reporter(Fig. 5B).

Although the logic of asymmetrically expressed genes harboring low affinity ASE motifs appears to apply to some cases, it does not apply generally. The asymmetrically expressed lsy-6 gene also contains a low-affinity ASE motif, as corroborated by competitive gel shift assays, but the two L/R asymmetrically expressed gcy genes gcy-7 and gcy-20do not (Fig. 5D). Moreover,swapping the ASE motifs of the L/R asymmetric gcy-7 and gcy-22 genes with that of the bilateral, high-affinity ceh-36 ASE motif does not cause de-repression in the contralateral neuron (data not shown). These findings further underscore the diversity of the cis-regulatory mechanisms that control L/R asymmetric gene expression.

Further insights into the importance of CHE-1 DNA-binding affinity with regard to the cis-regulatory logic of left/right asymmetric subtype specification were obtained by the recovery of an unusual class of che-1 alleles. In a previously described screen for mutants that affect ASE specification, we described a mutant locus, lsy-14, that resulted in the conversion of ASEL fate to ASER fate, exemplified by ectopic expression of the ASER fate marker gcy-5 in ASEL and a concomitant loss of the ASEL fate marker lim-6(Sarin et al., 2007)(Fig. 6A). This mutant phenotype is a characteristic phenotype of gene regulatory factors that act within the gene regulatory loop described in Fig. 1A. For example, lsy-6 miRNA mutants cause a similar `2 ASER' phenotype as lsy-14 mutants (Johnston and Hobert, 2003).

We mapped the lsy-14(ot101) locus to a small interval on chromosome I and obtained rescue of the mutant phenotype with a single fosmid(Table 1), which contains several genes, including che-1. A genomic fragment containing only the che-1 locus also rescues ot101(Table 1). Moreover, ot101 fails to complement the che-1 null allele ot66 (Table 1). We sequenced the che-1 locus in ot101 mutants and found a mutation in the linker region between zinc finger 1 and 2(Fig. 6B). The identification of ot101 as a che-1 allele is an unexpected result as che-1 null alleles cause a `no ASEL/R' phenotype (class III phenotype), rather than a `2 ASER' (class II) phenotype(Chang et al., 2003). We then revisited some previously described che-1 alleles, all of which had previously only been analyzed by expression of the lim-6 ASEL fate marker, which is lost in all available che-1 alleles(Sarin et al., 2007). We find that in three additional che-1 alleles, the loss of lim-6::gfp is not a reflection of overall ASE fate loss as previously assumed, but is rather accompanied by de-repression of ASER fate(Table 1). All che-1alleles that cause the `2 ASER'/class II phenotype cluster around the second zinc finger (Fig. 6B).

Further investigating the class II phenotype of these che-1alleles, we find that the expression of four out of four tested bilaterally expressed markers is unaffected in ot101 animals(Fig. 6A). Analyzing more asymmetric fate markers, we find that besides gcy-5, another ASER fate marker, gcy-22 is de-repressed in ASEL and another ASEL fate marker besides lim-6, gcy-7, is lost, even though at much lower penetrance than the lim-6 ASEL fate marker(Fig. 6A). The ASEL-restricted bistable loop component lsy-6 is also affected, but the ASEL terminal fate marker flp-4 is not (Fig. 6A). The distinct effect on different asymmetric markers suggests that che-1(ot101) does not simply affect a single regulatory component but may act independently on several different, ASE-motif containing, asymmetrically expressed genes. Consistent with this notion, che-1(ot101) displays complex interactions with components of the bistable feedback loop. The `2 ASEL' phenotype observed upon forced bilateral expression of the lsy-6 miRNA is mildly affected by ot101 in ASEL, and strongly affected in ASER (see Table S1 in the supplementary material). Forced bilateral expression of ASEL-fate inducer die-1 is also affected by ot101 in ASER, but not in ASEL (see Table S1 in the supplementary material). Taken together, class II che-1 alleles have the unusual feature of not affecting bilateral fate specification of ASE, but modify the left/right subtype diversification program. The end result is a partial switch of ASEL to ASER fate. What are the molecular features of class II che-1 alleles that might explain this effect?

To understand the phenotype of class II che-1 alleles in more detail, we examined the effect of these mutations on protein function. Notably, in contrast to strong loss of function or null alleles, which are either early nonsense mutations or reside in zinc finger 3 and 4(Sarin et al., 2007), all class II alleles relate somehow to the second zinc finger of CHE-1. In ot124, ot153 and ot223, the structural features of the second zinc finger are affected, including a residue that is directly involved in DNA binding (Fig. 6B). In ot101, the linker between the first and second zinc fingers is affected (Fig. 6B). Linker regions have previously been shown to be essential for DNA binding of adjacent zinc fingers (Choo and Klug,1993). We bacterially produced CHE-1 proteins that harbor the ot124 and ot223 point mutations in the second zinc finger. We find that each protein displays a decreased DNA-binding affinity across five different tested ASE motifs, both of high and low affinity(Fig. 6C). Even though it is unclear how mutations in zinc finger 2 interfere with overall DNA binding of CHE-1, we conclude that a reduction in DNA-binding activity causes a disruption of L/R asymmetrically expressed genes (hence, causing a class II Lsy phenotype), but leaving bilaterally expressed cis-regulatory control regions unaffected. This underscores the importance of DNA-binding affinity for controlling L/R asymmetric gene expression programs in the ASE neurons.

Last, we also approached the che-1-binding site affinity issue from the angle of increasing ASE motif occupancy by raising the levels of wild-type CHE-1 protein. This may overcome the effect of co-repressor sites in asymmetrically expressed promoters and/or alleviate the need for a co-activator and thereby result in an ectopic expression of asymmetrically expressed genes in the contralateral neuron. We generated multicopy arrays of the che-1 locus by injecting a fosmid containing the entire che-1 locus into a wild-type background and then examined expression of two L/R asymmetrically expressed cis-regulatory regions (from the gcy-5 and gcy-7 locus). We find that both reporters become ectopically expressed in the contralateral neuron [29% of animals express gcy-7 ectopically in ASER (n=65); 18% express gcy-5ectopically in ASEL (n=56); no contralateral expression of either gene is observed in wild-type animals]. It is not clear whether this effect is direct or mediated by intermediary regulatory components. This is the same issue as with the interpretation of the results of lowering of che-1activity in class II mutant alleles, in which we cannot definitely determine whether observed effects are direct or indirect. Nevertheless, we can conclude that increases or decreases in CHE-1 activity impact the left/right asymmetric execution of cell fate, while not affecting the ASE bilateral fate.

We can draw the following conclusions from our studies: (1) CHE-1 is not only required to induce ASE fate embryonically, but is also continuously required throughout the life of the neuron to maintain L/R asymmetric ASE cell fate; (2) in contrast to bilaterally expressed target genes of CHE-1, the regulation of asymmetrically expressed target genes involves the modulation of the activity of bilaterally expressed CHE-1 at the cis-regulatory level by additional regulatory motifs which restrict CHE-1 activity in a target gene-dependent manner to either ASEL or ASER; (3) the binding affinity of CHE-1 to its target site plays an important role in restricting the CHE-1 activity to ASEL or ASER; (4) the cis-regulatory architecture that restricts CHE-1 activity in a L/R asymmetric manner is remarkably diverse. Rather than using a simple cis-regulatory motif (as is the case for ASE-motif/che-1-dependent ASEL/R bilateral specification), different L/R asymmetrically expressed genes appear to use distinct cis-regulatory strategies. We discuss those points and their implications individually below.

The amenability of C. elegans to RNAi and the possibility to time its delivery allowed us to ask the fundamental question whether gene regulatory factors that turn on a specific terminal neuronal fate also are continuously required to maintain this fate throughout the life of the neuron. We find this to indeed be the case for the che-1 transcription factor. Using a temperature-sensitive allele, such sustained function could also be demonstrated for the unc-4 homeobox gene, which acts to determine specific synaptic inputs (Miller et al., 1992), but to our knowledge this issue has not been addressed for other regulatory factors that control neuronal differentiation in any system. Consistent with a maintenance function, che-1 is expressed in the ASE neurons throughout the life of the animal (data not shown). Sustained che-1 activity is ensured by che-1autoregulating its own expression via an ASE motif(Etchberger et al., 2007).

The continuous expression and requirement for che-1 suggests that CHE-1 does not only have a role in initiating the terminal fate of ASE neurons, which is, immediately after the birth of the ASE neurons, initially bilaterally symmetric (Johnston et al.,2005). Rather, che-1 also appears to be required for the progression of the hybrid precursor state to the L/R asymmetric terminal state. We infer this not only from the continuous expression and requirement of che-1, but also from the presence of functionally required ASE motifs in the cis-regulatory regions of L/R asymmetrically expressed genes. We can exclude the possibility that CHE-1 acts first to initiate L/R asymmetric genes via the ASE motif, and then to only indirectly restrict expression of L/R asymmetric genes in either ASEL or ASER, via the activation of intermediary transcription factors. If this were the case, our promoter analysis would have identified cis-regulatory motifs that can instruct ASE expression independently of the ASE motif. Even though we have found motifs that are required for activation of L/R asymmetric genes, these motifs only act in conjunction with the ASE motif. Moreover, the elimination of repressor motifs `bilateralizes' the expression of normally L/R asymmetric genes, again arguing that bilateral CHE-1 is continually able to drive gene expression in both ASEL and ASER, but is prevented from doing so in a target gene-dependent manner.

CHE-1 acts as a terminal selector gene that determines overall ASE fate by directly activating the expression of a large battery of bilaterally expressed genes (Etchberger et al., 2007; Hobert, 2008). In addition to this neuron class specification function of CHE-1, we have uncovered several cis-regulatory mechanisms that restrict CHE-1 activity on several target gene promoters to either the left or right ASE neuron, thereby driving a subtype specification program that diversifies these two cellular subtypes from one another (Fig. 7). The cis-regulatory mechanisms that restrict CHE-1 activity are promoter dependent and remarkably diverse. In three promoters (lim-6, gcy-7, lsy-6), we have found evidence for the existence of three distinct unrelated activator elements with which CHE-1 cooperates to promote L/R asymmetric expression. Notably, these activator elements are only required in the context of the complete regulatory module, as the ASE motif alone can drive bilateral ASE expression when in complete isolation, but is apparently not able to do so if in the context of the whole regulatory element. In such context, a cooperating activator motif is required for ASE expression (K50 binding site in gcy-7 and bHLH binding site in lsy-6). It is possible that bilaterally expressed ASE motifs also require such positively cooperating factors.

Regulation of the gcy-7 and lsy-6 promoters, even though relying on distinct motifs, appears to share a similar logic. CHE-1 appears to be engaged in a tug-of-war with repressor elements in the promoter, which it can only overcome by the presence of an additional activator motif that cooperates with the ASE motif. The activator motif is only required if the repressor elements are present. If both are removed, CHE-1 exerts bilateral control over the promoter in both ASEL and ASER.

By contrast, no discrete repressor motifs were found in lim-6 and no discrete co-activator motifs were found in gcy-5. It appears striking that none of the four promoters analyzed here uses similar strategies, as neither of the identified activator or repressor motifs show any similarity to one another. Moreover, even though some repressor motifs found to be required in one promoter (gcy-7) are present in other promoters (gcy-14, gcy-20), these motifs are not required in these other cases (data not shown). These findings may indicate an independent evolutionary recruitment of the many L/R asymmetrically expressed gcygenes into the regulatory network that controls L/R asymmetry of these neurons. This diversity is consistent with differences in the expression of gcy genes in different nematode species(Ortiz et al., 2006). This plasticity in the composition of the L/R asymmetric terminal features may relate to the sensory function of ASE neurons that may need to adapt to distinct environmental cues in a species-specific manner.

Previous genetic analysis has revealed several candidates for activator and repressor factors that may act through the cis-regulatory elements that we have described here to restrict CHE-1 activity(Fig. 1A). These include the zinc-finger transcription factor DIE-1 [genetically required to activate ASEL-expressed genes and repress ASER-expressed genes(Chang et al., 2004)], the LIM homeodomain protein LIM-6 [required to repress gcy-5 in ASEL(Hobert et al., 1999), Fig. 1A], the zinc-finger protein FOZI-1 [required to repress gcy-7 in ASER(Johnston et al., 2006), Fig. 1A], and two other, as yet uncloned lsy genes with similar phenotypes to lim-6[lsy-20 and lsy-26(Sarin et al., 2007)]. DNA-binding sites are not known for DIE-1, LIM-6 or FOZI-1, and in vitro gelshift assays have not detected binding of these factors to ASEL or ASER-specific promoters (data not shown). However, the putative K50-homeodomain binding site identified in the gcy-7 promoter is a likely binding site for the CEH-36/Otx homeodomain protein. CEH-36 binds to the motif in vitro and gcy-7 fails to be activated in ceh-36mutants. The asymmetric, ASEL-fate inducing activity of CEH-36 is also evident at the cis-regulatory level as the CEH-36 binding site will convert a bilaterally expressed ASE motif into a regulatory motif that is more strongly expressed in ASEL than in ASER (ASE-gcy-7#2 construct in Fig. 4). Therefore, CEH-36 activity must be somehow lateralized, even though CEH-36 is expressed in both ASEL and ASER.

Another factor that contributes to the restriction of CHE-1 activity is the affinity of the CHE-1/ASE motif interaction, which we again find to be important for some, but not all, promoters. The affinity argument stems largely from three observations. First, swapping a high-affinity ASE motif from a bilateral promoter into that of two different asymmetric promoters results in the partial bilateralization of promoter activity. We interpret this to mean that, at least in some cases, CHE-1 binding to the ASE motif is weak so as to make it susceptible to the effect of repressor motifs. Increases in the affinity of the CHE-1-binding site counteract the repressive effect. Second, increases in CHE-1 expression disrupt L/R asymmetric promoter activity. Third, we have identified an unusual class of alleles of che-1, which separates the activity of che-1 on bilaterally expressed versus L/R asymmetrically expressed promoters. These alleles cause a general decrease in ASE motif affinity and lead to a disruption of L/R asymmetric gene expression, while leaving bilateral expression intact. We note, however, that the importance of affinity may not be a general theme as not all motif swaps yielded the same results and the effect of the unusual che-1 alleles does not extend to every single L/R asymmetrically expressed gene.

Although CHE-1 activity must somehow be integrated with other transcription factors such as CEH-36, which are genetically required for L/R asymmetry, we also note another, as yet completely unexplored, layer of regulatory control that may play a role in the system. All cis-regulatory elements described here(including the ASE motifs) are located in remarkable vicinity to predicted translational start sites and hence may be components of the RNA-Pol II-binding core promoter. Unfortunately, owing to the phenomenon of trans-splicing, transcriptional start sites and hence core promoter sites are difficult to map in C. elegans. As previous work in other systems has shown shifts in core promoter selectivity during development(Juven-Gershon et al., 2008),it will be intriguing to investigate how the regulatory elements defined here relate to Pol II function.

The findings described here provide insights into how cellular fates become progressively restricted during development. Transcriptional regulators often define broad domains of gene expression that become further restricted,refined and diversified through added layers of regulatory control. In the context of terminal neuronal differentiation, an important class of regulatory proteins are what we term `terminal selector genes'(Hobert, 2008). These encode for transcription factors that determine the terminal identity of individual neuron types by directly controlling the expression of terminal gene batteries. CHE-1 is such a terminal selector gene, which directly controls the expression of a large battery of cell fate markers that are shared by ASEL and ASER (Etchberger et al., 2007). Other examples for terminal selector genes can not only be found in C. elegans but also in vertebrates, particularly in the brain(Hobert, 2008). Although terminal selector genes define the properties of individual neuron classes,ensuing subtype specification events further diversify neuron classes, as is the case in the diversification of the ASEL and ASER subtypes.

Together with our previous analysis of gene regulatory factors in ASE, the data presented here demonstrates that terminal selector genes participate directly in the subtype diversification that follows neuron type specification. The way that CHE-1 achieves this feat may reveal a common theme in gene regulatory networks that serve to diversify gene expression programs. CHE-1 interacts with other regulatory modules in a feed-forward loop (FFL)motif configuration. A conventional FFL consists of a transcription factor A,controlling factor B, and factor A and B controlling together a target C(Alon, 2007). Such simple FFL motifs can have specific properties such as persistence detection or response acceleration. CHE-1 acts in a more complicated version of the FFL. Besides activating a single transcription factor (CEH-36) with which it collaborates to regulate the expression of a target gene (gcy-7), CHE-1 also activates multiple components of the bistable regulatory loop shown in Fig. 1A; the regulatory loop provides a net activity output that then cooperates with CHE-1 in a promoter-specific manner on a given target gene(Fig. 7). For some genes, the loop provides a positive output with which CHE-1 needs to interact to be able to turn on a target gene (e.g. lim-6 in ASEL). For other target genes, the loop provides a negative output (e.g. in the form of the LIM-6)that restricts the ability of CHE-1 to turn on a target gene (e.g. prevents the activation of gcy-5 in ASEL).

Apart from whatever precise kinetic properties such network motif configuration may convey, one may view such network architecture as being reflective of the evolution of gene regulatory networks. The two ASE neurons may have initially been identical, with CHE-1 controlling the exact same set of genes in ASEL and ASER. This ancestral state may still be reflected ontogenetically in the hybrid precursor state through which CHE-1 passes after the ASE neurons are born. As a segregation of certain features (such as chemoreceptors) into distinct cells (i.e. ASEL and ASER) can convey beneficiary selective advantages to an animal [increases in discriminatory properties (Pierce-Shimomura et al.,2001; Suzuki et al.,2008)], additional regulatory mechanisms may have been implemented downstream of CHE-1 to restrict CHE-1 activity to a subset of target genes(Fig. 7).

Even though not dissected to the same extent as the ASE system, FFL-loop dependent subtype specification mechanisms also occur in other systems, such as the vertebrate retina (Hsiau et al.,2007) or in the fly ventral nerve cord(Baumgardt et al., 2007), and may provide a commonly used regulatory logic for subtype specification.