The lin-11 LIM homeobox gene specifies olfactory and chemosensory neuron fates in C. elegans

In the olfactory and gustatory systems of vertebrates and invertebrates, diversity of sensory neuron function allows animals to fully experience the complexity of their chemical environment, and to make behavioral and developmental decisions accordingly. The distinct sensory functions of each olfactory and gustatory neuron type are defined by the expression of one or more members of the large olfactory or taste receptor gene families, in addition to the differential expression of other signaling genes (Adler et al., 2000; Chess et al., 1994; Clyne et al., 1999; Hatt and Ache, 1994; Meyer et al., 2000; Nef et al., 1992; Ressler et al., 1993; Strotmann et al., 1992; Vassar et al., 1993; Vosshall et al., 1999). The developmental mechanisms that generate this functional complexity are largely unknown.

The nematode C. elegans senses its chemical environment using a small number of relatively well-characterized chemosensory neurons. In total, the functions of 11 pairs of chemosensory neurons have been defined by laser ablation studies, showing that each neuron type responds to a distinct but partly overlapping set of chemicals (Bargmann and Mori, 1997; Troemel, 1999). Volatile odorants are sensed by the bilaterally symmetric AWA, AWB and AWC olfactory neuron types (Bargmann et al., 1993; Troemel et al., 1997). Aqueous attractants are sensed primarily by the ASE neuron type and pheromones by the ADF, ASI, ASG and ASJ neuron types (Bargmann and Horvitz, 1991a; Bargmann and Horvitz, 1991b; Schackwitz et al., 1996). Similar to vertebrates and Drosophila, the distinct response profile of each neuron type is defined by the expression of specific signaling molecules such as olfactory receptors, G proteins and channels (Prasad and Reed, 1999; Troemel, 1999).

The LIM subfamily of homeobox genes has been implicated in generating functional diversity in the nervous system. This class of transcriptional regulators contains two metal-binding LIM domains that mediate protein-protein interactions, and a DNA-binding homeodomain (Hobert and Westphal, 2000). LIM homeobox genes define a ‘combinatorial code’ for the specification of different classes of motorneurons in the central nervous system of vertebrates and Drosophila (Appel et al., 1995; Thor et al., 1999; Tsuchida et al., 1994). In the spinal cord of vertebrates, motorneuron subtypes can be distinguished on the basis of the positions of their soma, their axonal trajectory, their target choice and their patterns of gene expression. Distinct classes of postmitotic motorneurons express unique spatiotemporal combinations of LIM homeodomain proteins, leading to the proposal that these LIM homeodomain proteins, along with other transcription factors, provide these neurons with their unique functional identities (Appel et al., 1995; Thor et al., 1999; Tsuchida et al., 1994). Support for this model has been obtained from knockout and misexpression experiments. In mice, the Lhx3/Lhx4 LIM homeobox genes specify the distinct ventrally extending axonal trajectories of motorneurons (Sharma et al., 1998). In Lhx3/Lhx4 double mutant mice, ventrally projecting motorneurons instead adopt the fate of motorneurons that project dorsally. Conversely, ectopic expression of Lhx3 in the dorsally projecting motorneurons is sufficient to switch their trajectories to those of the ventral motorneurons (Sharma et al., 1998). Similarly, in Drosophila, the lim3 and islet LIM homeobox genes are required to specify the distinctive identities and axonal trajectories of subsets of motorneurons (Thor et al., 1999; Thor and Thomas, 1997).

In C. elegans, members of the LIM homeobox gene family are expressed in largely non-overlapping sets of cell types, indicating that a combinatorial LIM code probably does not function in their specification (Hobert and Westphal, 2000). However, LIM homeobox genes regulate the development and differentiation of several neuron types in C. elegans, suggesting that these genes are important components of the suite of transcription factors that confers a unique identity to each neuron type. Several LIM homeobox genes have been implicated in sensory neuron function. The mec-3 and lim-4 LIM homeobox genes appear to play broad roles in the specification of the mechanosensory and the AWB olfactory neurons. In mec-3 mutants, the six mechanosensory neurons are generated but fail to acquire differentiated features (Chalfie and Sulston, 1981; Way and Chalfie, 1988). In lim-4 mutants, the AWB olfactory neurons fail to acquire any AWB-specific characteristics, but instead adopt an AWC-like default developmental state (Sagasti et al., 1999). The ceh-14 and lim-6 LIM homeobox genes regulate a more restricted subset of the features of the differentiated AFD thermosensory and the ASE chemosensory neurons, respectively (Cassata et al., 2000; Hobert et al., 1999). These findings indicate that LIM homeobox genes play crucial roles in the differentiation of a subset of sensory neuron types in C. elegans, and raise the question of whether members of this gene family are involved in the specification of sensory identity of additional chemosensory neurons.

We show that the LIM homeobox gene lin-11 is required to specify the sensory identities of the AWA olfactory neurons, and their lineal sisters, the ASG chemosensory neurons. lin-11 is a founding member of the LIM homeobox gene family, and was first identified on the basis of its role in regulating asymmetric cell divisions in vulval cell lineages (Ferguson and Horvitz, 1985; Ferguson and Horvitz, 1987; Freyd et al., 1990). Subsequent work has shown that lin-11 also plays a role in the differentiation of the VC ventral cord motorneurons, and the AIZ interneurons (Hobert et al., 1998) which are a major component of the chemosensory and thermosensory circuits (White et al., 1986). In this study we show that in lin-11 mutants, the AWA neurons lose most AWA-specific characteristics and instead adopt an AWC-like default state. lin-11 functions by regulating expression of the AWA-specific odr-7 nuclear hormone receptor gene, which has been previously implicated in promoting AWA-specific characteristics and repressing the AWC fate (Sagasti et al., 1999; Sengupta et al., 1994). lin-11 also regulates the differentiation of the ASG neurons, such that in lin-11 mutants, ASG-specific genes are not expressed. We show that the distinct functions of lin-11 in the two neuron types are partly determined by the dynamic temporal expression pattern of lin-11. Moreover, we show that misexpression of lin-11 is sufficient to cause ectopic expression of ASG but not AWA-specific genes. Our results provide evidence for hitherto undescribed functions of lin-11, and suggest that a LIM code may be an important contributor to the generation of functional diversity in the sensory neurons of C. elegans.

Animals were grown under standard conditions (Brenner, 1974). Strains were obtained from the Caenorhabditis Genetics Center unless noted otherwise. Strains carrying the following stably integrated fusion genes were provided as indicated: pkIs582 (gpa-5::GFP) (G. Jansen; Jansen et al., 1999); ntIs1 (gcy-5::GFP) (S. Lockery; Yu et al., 1997); oyIs17 (gcy-8::GFP) (J. Satterlee; Yu et al., 1997); kyIs111 (tax-2::GFP), kyIs128 (str-3::GFP), kyIs104 (str-1::GFP), kyIs140 (str-2::GFP), oyIs14 (sra-6::GFP), otIs24 (sre-1::GFP), oyIs34 (T08G3.3::GFP), kyIs37 (odr-10::GFP), kyIs5 (ceh-23::GFP) and kyIs29 (glr-1::GFP) (C. Bargmann; Coburn and Bargmann, 1996; Maricq et al., 1995; Peckol et al., 1999; Sagasti et al., 1999; Sengupta et al., 1996; Troemel et al., 1995; Troemel et al., 1997; Troemel et al., 1999; Zallen et al., 1999); gmIs12 (srb-6::GFP) (G. Garriga; Troemel et al., 1995). The gcy-10/odr-1::RFP plasmid was provided by C. Bargmann (L’Etoile and Bargmann, 2000; Yu et al., 1997).

ops-1::GFP was constructed by amplifying approx. 2.0 kb of T13A10.13 promoter sequences and fusing in frame to GFP. The T13A10.13 predicted ORF shows the highest homology to the vertebrate rhodopsin genes, and is expressed in the ASG and ADL neurons in the head. ops-1::RFP was constructed by replacing GFP sequences with those encoding RFP (Matz et al., 1999). lim-6::GFP is a fusion of a 5.4 kb genomic fragment containing approx. 1.9 kb of promoter and all coding sequences of lim-6 fused in frame to GFP. The expression in AWA and ASG observed with this fusion gene is repressed in a full-length lim-6::GFP rescuing construct (Hobert et al., 1999). These fusion genes were chromosomally integrated to generate oyIs26 (ops-1::GFP) and mgIs20 (lim-6::GFP).

osm-6::lin-11 and odr-3::lin-11 were constructed by first replacing the GFP sequences in the expression vector pPD95.77 with the lin-11 cDNA yk452F7. osm-6 and odr-3 promoter sequences were then inserted upstream (Collet et al., 1998; Roayaie et al., 1998). To make the GFP-tagged full-length lin-11 construct, 3.3 kb of upstream promoter sequences, and all coding sequences of lin-11 up to the STOP codon were amplified in two pieces and cloned in frame upstream of GFP in the expression vector pPD95.77.

Germline transformations were performed using standard techniques (Mello and Fire, 1995). Coinjection markers used were the dominant marker pRF4 rol-6(su1006) at 100 ng/μl, and ofm-1::GFP at 50 ng/μl (Miyabayashi et al., 1999). All other plasmids were injected at 30 ng/μl.

The odr-7(oy9) allele was isolated from the F2 progeny of a mutagenized strain carrying integrated copies of an odr-7::GFP fusion gene. odr-7(oy9) failed to complement the odr-7(ky4) null allele. To identify the molecular lesion in the oy9 allele, odr-7 genomic sequences were amplified from mutant animals in two independent reactions, and the products were directly sequenced. The odr-7(oy9) strain was outcrossed a total of three times prior to characterization.

Sequences coding for residues 1-315 of ODR-7 were fused to GST coding sequences in the vector pGex3X. The purified fusion protein was injected into rabbits to generate polyclonal antibodies (CalTag Laboratories). Antibodies were used without affinity purification to stain adults using the Finney-Ruvkun protocol (Finney and Ruvkun, 1990). Embryos were stained using a modified protocol from Chin-Sang et al. (Chin-Sang et al., 1999). ODR-7 antibodies were used at a 1:20 dilution. Secondary antibodies were used as described previously (Sarafi-Reinach and Sengupta, 2000).

Animals were examined using a Zeiss Axioskop microscope with epifluorescence. Images were captured using a CCD digital camera (Hamamatsu) and PCI digital frame grabber. Images were analyzed and pseudocolored using OpenLab (Improvision) and Adobe Photoshop software.

The LIM homeobox genes lim-4, lim-6 and ceh-14 regulate different aspects of fate and function of the AWB olfactory, the ASE chemosensory and the AFD thermosensory neurons respectively (Cassata et al., 2000; Hobert et al., 1999; Sagasti et al., 1999). To determine if members of the LIM homeobox family also play a role in the functional specification of the AWA olfactory neurons, we examined expression of the AWA-specific odr-7 nuclear hormone receptor gene in animals carrying null mutations in the LIM homeobox genes mec-3, ttx-3, ceh-14, lim-4, lim-6 and lin-11. odr-7 has previously been shown to be required for all known AWA functions (Sengupta et al., 1994). As shown in Table 1, we find that expression of odr-7 is altered in lin-11 mutants, whereas expression in other LIM homeobox gene mutants is unaffected (data not shown). Approximately 80% of lin-11(n389) null mutants show loss of expression of an odr-7::GFP fusion gene in at least one AWA neuron. To confirm the observed loss of odr-7 expression, we stained wild-type and lin-11 mutant adult animals with polyclonal antibodies against ODR-7 (see Materials and Methods). This antibody is specific for ODR-7, since no staining is observed in odr-7(ky4) putative null mutants (data not shown). Antibody staining of lin-11(n389) mutants confirms the loss of ODR-7 expression observed with the odr-7::GFP marker (Fig. 1; Table 1).

Loss of odr-7 expression could result either from the inability of lin-11 mutants to maintain odr-7 expression into adulthood, or from a defect in initiation of odr-7 expression. To distinguish between these possibilities, we examined odr-7::GFP expression in young lin-11(n389) L1 larvae. As shown in Table 1, odr-7 expression is also lost in larvae, suggesting that lin-11 plays a role in initiating odr-7 expression.

To determine the extent to which AWA-specific gene expression is affected in lin-11 mutants, we examined the expression of two additional AWA-specific genes, the olfactory receptor gene odr-10 and the G-protein α subunit gene gpa-5 (Jansen et al., 1999; Sengupta et al., 1996). Expression of both genes is also reduced in lin-11 mutants (Table 1), suggesting that lin-11 regulates the expression of several, if not all, AWA-specific genes. We were unable to determine whether loss of AWA-specific gene expression in lin-11 mutants is correlated with a loss of AWA-mediated olfactory functions, owing to both the uncoordinated backward locomotion exhibited by lin-11 mutants (Freyd, 1991), and defects in the differentiation of the AIZ postsynaptic partners of the AWA neurons (Hobert et al., 1998). It is interesting to note, however, that even in lin-11(n389) null mutants, the loss of expression of AWA markers is incompletely penetrant, suggesting that lin-11 acts together with another unidentified gene(s) to regulate AWA cell fate.

The major postsynaptic partners of the AWA neurons are the AIY and AIZ interneurons (White et al., 1986). lin-11 is expressed in the AIZ interneurons and has been shown to be required for their function (Hobert et al., 1998). However, expression of lin-11 in the AWA neurons had not previously been described. Therefore, it is possible that the defect in AWA gene expression results from the loss of lin-11 function in the AIZ interneurons, with lin-11 acting non cell-autonomously to regulate AWA cell fate. To address this issue, we examined ODR-7 expression in mutants in which the fate and function of the AIY and AIZ interneurons are affected. In unc-86 mutants, the AIZ neurons are not generated because of a lineage defect (Chalfie et al., 1981), while in ttx-3 mutants, the AIY neurons fail to differentiate correctly (Altun-Gultekin et al, 2001; Hobert et al., 1997). However, we find that ODR-7 expression is unaffected in both mutants (Table 1). ODR-7 expression is also unaffected in animals mutant in components of general synaptic transmission such as snt-1 and unc-13 (data not shown; Ahmed et al., 1992; Nguyen et al., 1995; Nonet et al., 1993). These results indicate that lin-11 does not regulate AWA cell fate via a retrograde signaling mechanism from its AIZ and AIY postsynaptic targets.

The expression patterns of lin-11::reporter fusion genes have been described (Freyd, 1991; Hobert et al., 1998; Newman et al., 1999). lin-11 expression was observed in uterine and vulval cells and in several neuron types, including the ADL and ADF sensory neurons and the AIZ interneurons (Freyd, 1991; Hobert et al., 1998; Newman et al., 1999). However, the fusion genes used did not contain all the genomic sequences of lin-11, leaving open the possibility that sequences contained within the remaining introns direct expression in additional cell types. Moreover, since the minimal genomic sequences required for lin-11 function were not defined by rescue of lin-11 mutant phenotypes, the complete promoter sequences required for lin-11 activity remained undetermined. To examine the expression of lin-11 in more detail, we generated a full length GFP-tagged lin-11 fusion gene containing all identified exons and introns, in addition to 3.3 kb of upstream promoter sequences (Fig. 2A). This fusion gene was stably integrated into the genome, and was used to examine rescue of lin-11 mutant phenotypes. As shown in Table 2, both the odr-7 loss-of-expression phenotype and egg-laying defect of lin-11 mutants are rescued by this transgene, indicating that its expression pattern probably reflects the expression pattern of the endogenous gene.

Expression of lin-11::GFP is first observed after the time when most postmitotic neurons are born (at approx. 300 minutes postfertilization), suggesting that lin-11 plays a role in regulating differentiation, but not generation, of these neurons. In larvae, the LIN-11::GFP fusion protein is localized to the nuclei of multiple neurons in the head and lumbar ganglia (Fig. 2B-E). In addition, lin-11 expression is observed in uterine and vulval cells, as well as in the VC motor neurons as previously described (data not shown; Freyd, 1991; Hobert et al., 1998).

We characterized the neuronal expression of lin-11 in further detail. In the lumbar ganglia, lin-11 expression is observed in the PVPL/R and PVQ neurons, and in an additional unpaired neuron that we tentatively identified as DVC or DVA (Fig. 2B,C). In contrast to shorter lin-11::GFP fusion genes (Hobert et al., 1998), the rescuing lin-11::GFP fusion gene is not expressed in the PHA sensory neurons. In the head, as previously reported using shorter reporter fusions, we observed lin-11 expression in the sensory neurons ADF and ADL, in the interneurons AIZ, RIC and AVG, and in another neuron type tentatively identified as either AVH or AVJ (Fig. 2D,E) (Freyd, 1991; Hobert et al., 1998). However, we also observed expression in a number of additional neurons using the full-length lin-11::GFP fusion gene. lin-11 expression was observed in the AVA and AVE interneurons (Fig. 2D,E); the AVA interneurons are the primary command interneurons for the circuit that drives backward locomotion (Chalfie et al., 1985). Expression in AVA might provide a rationale for the uncoordinated backward locomotion of lin-11 mutants (see below). We also detected faint and occasional expression in the ASH polymodal sensory neurons.

Interestingly, we observed lin-11 expression in the AWA neurons in embryos and young larvae, but not in later stages. As shown in Fig. 3A-C, expression of lin-11 in the AWA neurons is observed consistently in three-fold embryos, where expression of lin-11 overlaps with that of ODR-7 (32/32 AWA neurons examined). Expression decreases by hatching such that in L1 larvae, expression in the AWA neurons is fainter and observed only occasionally, and is absent in later larvae and adults (Fig. 3D). We also observed lin-11 expression strongly and consistently in the ASG chemosensory neurons (Fig. 3E), which are the lineal sisters of the AWA neurons (Sulston et al., 1983). Expression in the ASG neurons is observed throughout postembryonic development. Thus, lin-11 expression is regulated differentially and in a dynamic manner in two neuron types (AWA and ASG) that result from the asymmetric division of a precursor neuroblast. We also detected lin-11 expression in a few other non-sensory neurons in the head and tail that we did not identify definitively.

In odr-7 mutants, all known aspects of differentiated AWA fate are lost (Sengupta et al., 1994). odr-7 is expressed throughout postembryonic development in the AWA neurons (Sengupta et al., 1994), suggesting that continuous odr-7 expression is required to maintain AWA cell fate. However, since lin-11 is expressed only during early development in the AWA neurons, lin-11 is likely required to initiate but not maintain odr-7 expression. To determine whether ODR-7 maintains its own expression by autoregulation, we examined the expression of GFP driven by the odr-7 promoter in an odr-7 null mutant. The odr-7(oy9) allele is likely to be null, based on both the nature of its molecular lesion (Q231 to STOP in exon 2) resulting in a predicted truncated protein product, and a lack of detectable ODR-7 protein upon staining with anti-ODR-7 antibodies (data not shown). We find that expression of odr-7::GFP is detected in approx. 96% of young L1 larvae in odr-7 null mutants (n>100), but is abolished in later larvae and adult animals (only 1% express odr-7::GFP; n>100), indicating that ODR-7 autoregulates to maintain its expression in the AWA neurons.

In odr-7 null mutants, approx. 9% of animals retain expression of an odr-10::GFP fusion gene in at least one AWA neuron (Table 3; Sengupta et al., 1996), raising the possibility that ODR-7 acts together with another unidentified gene product(s) to specify AWA fate. LIN-11 could specify AWA fate by regulating the expression of only odr-7, or could regulate additional genes. To address this issue, we examined the expression of an odr-10::GFP fusion gene in a lin-11; odr-7 double mutant. If LIN-11 acts by regulating genes in addition to odr-7 in the AWA neurons, we might expect to observe a stronger odr-10::GFP expression defect in lin-11; odr-7 double mutant animals than in odr-7 single mutants. However, as shown in Table 3, the odr-10::GFP expression defect is not enhanced in lin-11; odr-7 double mutants. Thus, LIN-11 appears to promote AWA cell fate primarily by regulating odr-7 expression.

The AWA and AWB olfactory neurons have been shown to share a common AWC-like default developmental potential (Sagasti et al., 1999). This potential is modified by the expression of odr-7 and lim-4 respectively to result in the generation of three distinct olfactory neuron types. Thus, in odr-7 mutants, the AWA neurons adopt partial AWC-like fate and express the AWC-specific olfactory receptor gene str-2 (Sagasti et al., 1999). Since LIN-11 regulates odr-7 expression, we might expect that the AWA neurons would also adopt AWC fate in lin-11 mutants. We find that approximately 33% of lin-11 mutants express str-2::GFP ectopically in at least one AWA neuron (Fig. 4; Table 4). We also examined the expression of additional AWC markers, including the guanylyl cyclase gene gcy-10/odr-1 and the cyclic nucleotide channel subunit gene tax-2 (Coburn and Bargmann, 1996; L’Etoile and Bargmann, 2000; Yu et al., 1997). Although we failed to observe expression of gcy-10/odr-1 in the AWA neurons (data not shown), approx. 12% of lin-11(n389) mutants ectopically express tax-2::GFP in at least one AWA neuron (n=65). Thus, the AWA neurons appear to adopt partial AWC-like characteristics in lin-11 mutants. However, the extent of fate conversion is clearly incomplete, based on both incomplete expression of AWC-specific markers, as well as incomplete transformation of AWA ciliary morphology. The AWA, AWB and AWC olfactory neurons can be morphologically distinguished by their distinctive specialized ciliary endings (Perkins et al., 1986; Ward et al., 1975). In lim-4 mutants, the cilia of the AWB neurons resemble those of the AWC neurons (Sagasti et al., 1999). However, in lin-11 mutants, the cilia of the AWA neurons acquire a morphology intermediate between those of AWA and AWC (data not shown).

A higher percentage of odr-7 mutants than lin-11 mutants exhibit ectopic expression of str-2::GFP in at least one AWA neuron (78% versus 33%; Table 4). This difference may be partly attributed to the incomplete loss of odr-7 expression in lin-11 mutants (Table 1). However, this fails to account fully for the difference observed. Instead, this difference may arise partly from the fact that str-2 is a marker for only one of two alternative AWC cell fates. str-2 is expressed stochastically in only one AWC neuron in wild-type animals, and this expression is regulated by interneuronal calcium signaling between the two AWC neurons (Sagasti et al., 2001; Troemel et al., 1999). In odr-7 mutants, it is possible that the AWA neurons have acquired an intermediate AWC fate that does not always carry out this signaling correctly, such that 35% of odr-7 null mutants express str-2 in both AWA neurons (Table 4). Since approx. 37% of lin-11 mutants lack odr-7 expression in both AWA neurons similar to odr-7 null mutants (Table 1), we might expect that 35% of these animals (or a total of 13% of lin-11 mutants) would also express str-2 in both AWA neurons. However, as shown in Table 4, only approx. 3.5% of lin-11 mutants misexpress str-2 in both AWA neurons. It is possible that in lin-11 mutants, bilateral str-2 expression may be regulated similarly to its regulation in the AWC neurons. This ability to correctly signal the str-2 ‘ON/OFF’ decision may represent a more complete transformation towards the AWC fate in lin-11 mutants as compared to odr-7 mutants. Consistent with this, we find that the number of animals expressing str-2 in both AWA neurons in a lin-11; odr-7 double mutant is similar to that in lin-11 single mutants (Table 4). These results suggest that although lin-11 appears to act primarily through odr-7 to promote AWA fate, lin-11 may regulate genes in addition to odr-7 to modulate the default AWC fate in the AWA neurons.

Since lin-11 is also expressed in the ASG lineal sisters of the AWA neurons, we examined whether ASG cell fate specification requires lin-11 function, by examining expression of ASG markers in lin-11 mutants. Markers examined include ops-1::GFP (see Materials and Methods), unc-30::GFP (expressed in the ASG neurons in the head) (Jin et al., 1994), and a truncated lim-6::GFP fusion gene that is expressed in the AWA and ASG neurons (see Materials and Methods). In all cases, we observed loss of ASG-specific gene expression in lin-11 mutants (Table 5). The loss of ASG marker expression is also incompletely penetrant in lin-11 null mutants, suggesting that lin-11 functions with other gene(s) in the ASG neurons to regulate ASG fate.

We have previously shown that in unc-130 mutants, the ASG neurons adopt AWA fate, and that these neurons express the underlying default str-2 ‘ON’ AWC fate only upon loss of both unc-130 and odr-7 gene function (Sarafi-Reinach and Sengupta, 2000). Consistent with this, we find that few lin-11(n389) mutant animals express str-2::GFP in ASG (<6%; n=123), indicating that loss of LIN-11 function alone is not sufficient for adoption of the str-2-expressing AWC fate in the ASG neurons.

Since LIN-11 regulates gene expression in both the AWA and ASG neurons, we wished to determine whether loss of gene expression in either an AWA or ASG neuron is correlated with a loss of gene expression in its lineal sister. We examined loss of ops-1::GFP expression in animals costained with antibodies against ODR-7. As shown in Fig. 5, we find that loss of ASG-specific gene expression does not strictly correlate with loss of expression of AWA-specific markers. This result suggests that a defect in fate specification of an AWA or ASG neuron does not result in a concomitant defect in the differentiation of its lineal sister.

In unc-130 mutants, the ASG neurons ectopically express odr-7 and fail to express ASG-specific markers (Sarafi-Reinach and Sengupta, 2000). unc-130 is expressed in the precursors to the AWA and ASG neurons, and encodes a predicted forkhead domain transcription factor (Nash et al., 2000; Sarafi-Reinach and Sengupta, 2000). We have previously proposed that unc-130 is required for the correct expression and/or segregation of cell fate determinants in the AWA/ASG precursors. Since lin-11 is initially expressed in both the AWA and ASG neurons, asymmetric division of the precursors likely results in the differential segregation of other components that specify the distinct AWA and ASG neuron types. Although we have shown that expression of odr-7 is regulated by LIN-11 in the AWA neurons, it is possible that in unc-130 mutants, odr-7 expression in the ASG neurons does not require LIN-11. To address this, we examined the expression of odr-7 in lin-11; unc-130 double mutants. If expression of odr-7 is independent of lin-11 in the ASG neurons, we might expect to observe a small number of animals expressing odr-7 in both the AWA and ASG neurons in lin-11; unc-130 double mutants. However, as shown in Table 6, lin-11; unc-130 double mutants exhibit a defect at least as severe as lin-11 single mutants, with no animals expressing odr-7 ectopically. This result demonstrates that misexpression of odr-7 in the ASG neurons in unc-130 mutants requires lin-11 function.

We have shown that lin-11 is necessary for the expression of both AWA and ASG neuronal fates. To determine whether lin-11 is also sufficient for either fate, we first misexpressed lin-11 using the odr-3 promoter. odr-3 drives expression strongly in the AWC neurons, and weakly in four other neuron types (Roayaie et al., 1998). Misexpression of lim-4 in the AWC neurons using the odr-3 promoter is sufficient for fate conversion of the AWC neurons to AWB fate (Sagasti et al., 1999). However, we find that misexpression of lin-11 in the AWC neurons does not result in the ectopic expression of either odr-7 or ops-1::GFP in the AWC neurons (data not shown), indicating that in the AWC neurons, lin-11 expression from the odr-3 promoter is not sufficient to drive either AWA or ASG cell fate.

To determine whether lin-11 misexpression in neurons other than AWC is sufficient for expression of AWA or ASG fate, we misexpressed lin-11 using the osm-6 promoter. osm-6 drives expression in most ciliated neurons, including all chemosensory neurons (Collet et al., 1998). We first examined expression of the ops-1::GFP gene in transgenic animals carrying the osm-6::lin-11 construct. As shown in Fig. 6, we observed misexpression of ops-1::GFP in a small subset of ciliated neurons in the head ganglia. The majority of these neurons are in the anterior ganglia, which contain sensory neurons of largely unknown function (White et al., 1986). In addition, we occasionally identified faint expression of ops-1::GFP in the AWA and ASJ chemosensory neurons. Expression of lin-11 from the osm-6 promoter rescues the ops-1::GFP expression defect in the ASG neurons in lin-11 mutants (data not shown). ops-1::GFP is expressed in both the ASG and ADL neurons. Since lin-11 is also expressed in the ADL neurons, it is possible that misexpression of lin-11 drives ectopic expression of ADL and not ASG fate. To address this, we examined the expression of additional ASG markers (unc-30::GFP and lim-6::GFP), and an ADL marker (sre-1::GFP; Troemel et al., 1995) in transgenic animals carrying the osm-6::lin-11 construct. Both ASG markers, but not the ADL marker, were expressed ectopically (data not shown). However, clearly not all neurons in which osm-6 is expressed adopt the ASG fate. This result indicates that lin-11 is sufficient for ASG fate expression in some, but not all ciliated neurons. Interestingly, the ectopic expression of ASG fate does not appear to be dependent on unc-130 function in most cell types, since we continued to observe misexpression of ops-1::GFP in unc-130 mutants (not shown).

We also determined whether osm-6::lin-11 is sufficient to promote AWA cell fate by examining the expression of odr-7 in animals carrying this transgene. We find that odr-7 is not expressed ectopically upon misexpression of lin-11. Instead, misexpression of lin-11 resulted in the occasional loss of odr-7 expression in the AWA neurons (4 independent lines; range of loss in at least one AWA neuron 14-63%; n>50 for each line). Expression driven by the osm-6 promoter appears to peak at L1 larval stages but is maintained into adulthood (Collet et al., 1998). We suggest that the prolonged expression of lin-11 in the AWA neurons from the osm-6 promoter results in the occasional alteration of AWA fate and loss of AWA-specific gene expression. Since misexpression of osm-6::lin-11 results in only occasional weak expression of ASG-specific markers in the AWA neurons (see above), this suggests that temporal misregulation of lin-11 expression in the AWA neurons is only partly sufficient to promote ASG fate.

In addition to the AWA and ASG neurons, lin-11 is also expressed in additional sensory neurons including the ADF, ADL and ASH sensory neurons. To determine whether lin-11 is necessary for the specification of these neurons, we examined the expression of markers for each neuron type, including genes encoding putative olfactory receptors, guanylyl cyclases and transcription factors. We also examined the ability of lin-11 animals to correctly fill with the lipophilic dye DiI. We find that gene expression in the neurons ADL, ADF, ASH, AWB and ASE is largely unaffected in lin-11 mutants, as is the overall pattern of DiI filling (Table 7).

We also examined gene expression in the AVA interneurons. Ablation of the AVA interneurons results in animals that are unable to move backwards normally (Chalfie et al., 1985; Zheng et al., 1999). The paired-type homeodomain gene unc-42 is expressed in the AVA interneurons, and regulates expression of a subset of genes required for AVA function, including the ionotropic glutamate receptor genes glr-1, glr-4 and glr-5 (Baran et al, 1999; Brockie et al., 2001; Hart et al., 1995; Maricq et al., 1995). To determine the role of lin-11 in the AVA interneurons, we examined expression of a glr-1::GFP fusion gene in the AVA interneurons in lin-11 mutants. We find that glr-1 expression in these neurons is unaffected (Table 7). Since UNC-42 directs glr-1 expression in the AVA interneurons, this result suggests that LIN-11 does not regulate unc-42 in these neurons, and that lin-11 may function downstream or in a parallel pathway. We examined lin-11::GFP expression in unc-42(e419) mutants and found that lin-11 expression in the AVA interneurons is unaltered in unc-42 mutants (17/17 animals). Several other NMDA and non-NMDA glutamate receptors are also expressed in the AVA interneurons and may be required for AVA interneuron function (Brockie et al., 2001; Zheng et al., 1999). It is possible that lin-11 regulates a subset of these genes to modulate AVA function, and hence locomotion.

We have shown that the LIM homeobox gene lin-11 is necessary for the functional specification of the AWA olfactory and ASG chemosensory neurons. In the AWA neurons, LIN-11 functions by regulating expression of the transcription factor odr-7. In addition, lin-11 is sufficient in certain cellular contexts to promote ASG but not AWA fate. lin-11 is expressed postmitotically and dynamically in both cell types. Our findings suggest that the temporal dynamic regulation of lin-11 in these two neuron types, coupled with its interaction with other cell type-specific factors, result in the specification of the distinct AWA and ASG chemosensory neurons.

Temporally regulated transcription factor cascades are critical mediators of neuronal cell fate specification. During Drosophila neurogenesis, sequential expression of the transcription factors Hunchback (Hb), POU-homeodomain proteins 1 and 2 (Pdm; also known as nubbin; nub), and Castor (Cas) in the CNS neuroblasts regulates neuroblast identity and sublineage generation (Brody and Odenwald, 2000; Kambadur et al., 1998). This sequential Hb→Pdm→Cas transition appears to be regulated in a neuroblast-intrinsic manner, in the absence of cell-cell communication. In the AWA neurons, a similar temporally regulated cascade of transcription factors is required for fate specification. In C. elegans, expression of most LIM homeobox genes appears to be maintained throughout the postmitotic life of neurons (Hobert and Westphal, 2000). However, results from this work show that lin-11 is expressed only during a brief temporal window, shortly after the birth of the AWA neurons. This expression is sufficient to activate odr-7, which then autoregulates to maintain its expression and the expression of AWA-specific characteristics. Our data suggest that forced prolonged expression of lin-11 in the AWA neurons may be partly sufficient to alter AWA fate, indicating that strict temporal control of lin-11 expression is critical for correct AWA fate specification. In vertebrates, the expression pattern of LIM homeobox genes is also regulated temporally, and is an important component in determining cell type identity. Isl1 is first expressed early in all motorneurons, while expression of other LIM homeobox genes follows in a temporally stereotyped manner (Appel et al., 1995; Ericson et al., 1992; Pfaff et al., 1996; Tsuchida et al., 1994; Varela-Echavarria et al., 1996). Expression of the Lhx3 and Lhx4 genes has also been shown to be dynamic (Sharma et al., 1998). It will be interesting to investigate whether other LIM homeobox genes in C. elegans are also regulated dynamically.

What are the mechanisms by which lin-11 expression is down-regulated in the AWA, but not in the ASG neurons? Maintenance of expression in one, but not the other sibling cell may arise from cell-cell interactions. Notch signaling between the two daughters of the pIIA sensory neuron progenitors in Drosophila has been shown to be required for the autorepressive and autoactivating functions of Suppressor of Hairless (Su(H)) in each of the two daughter cells, resulting in high levels of Su(H) in one, but not the other cell (Barolo et al., 2000). We have shown that a defect in specification of an AWA or ASG neuron does not result in a fate defect in the sibling neuron, suggesting that cell-cell signaling between the AWA and ASG neurons may not be the primary mediator of regulation of lin-11 expression. Instead, we suggest that this difference in the temporal pattern of lin-11 expression arises from the asymmetric segregation of factors to the AWA and ASG neurons. We have previously shown that the forkhead domain transcription factor UNC-130 plays an important role in the asymmetric division giving rise to the AWA and ASG neurons (Sarafi-Reinach and Sengupta, 2000). We propose that UNC-130 function in the AWA/ASG precursors is necessary for correct segregation of factors required for modulation of LIN-11 function in these two neurons. These factors could function to maintain lin-11 expression in the ASG neurons, or to repress lin-11 expression in the AWA neurons later in development. It is also possible that LIN-11 is autoregulatory, and functions along with these factors to maintain or repress its expression.

We also propose that UNC-130 regulates the asymmetric segregation of factors that work together with LIN-11 to regulate either AWA- or ASG-specific gene expression (Fig. 7). In unc-130 mutants, incorrect segregation of factors to the ASG neurons may result both in the downregulation of lin-11 expression and promotion of AWA-specific gene expression, thereby converting the ASG neurons to an AWA fate. These factors are likely to be different for each of the AWA and ASG cell types (X for AWA; Y for ASG in Fig. 7). The functions of LIM homeobox genes have been shown to be modified by interaction with a number of different proteins, including members of the paired-type homeodomain family, POU-homeodomain proteins and other LIM homeobox proteins (Bach et al., 1997; Bach et al., 1995; Feuerstein et al., 1994; Jurata et al., 1998; Schmeichel and Beckerle, 1994; Xue et al., 1993). It is unlikely that lin-11 works with other LIM homeobox genes in the AWA and ASG neurons, since none of the six additional identified LIM homeobox genes in C. elegans is expressed in the AWA or ASG neurons, and mutations in these genes do not affect odr-7 expression (Hobert and Westphal, 2000; this work).

It has been noted previously that lin-11 is expressed in cell types that make direct physical connections with each other (Hobert et al., 1998; Newman et al., 1999). For instance, lin-11 is expressed in the ADL neurons that synapse onto the AIZ interneurons (this study; Hobert et al, 1998). Similarly, lin-11 is expressed in the VC and PVQ synaptic pairs (this study; Hobert et al, 1998). lin-11 has also been shown to be expressed in components of the developing hermaphrodite egg-laying machinery, including the innervating VC motorneurons, uterine cells and vulval cells that must contact each other correctly in order to assemble a functional egg-laying system (Freyd, 1991; Hobert et al., 1998; Newman et al., 1999). Results in this study extend these observations, since we found that lin-11 is expressed in the AWA-AIZ and ASH-AVA sensory neuron-postsynaptic interneuron pairs. We speculate that lin-11 may regulate the expression of genes required for homophilic adhesion in members of each connected circuit. Intriguingly, in the uterine cells, but not in the VC neurons (O. H., unpublished), lin-11 has been shown to regulate the cdh-3 gene, encoding a member of the cadherin superfamily implicated in cell-cell adhesion (Newman et al., 1999; Pettitt et al., 1996). Similar coordinate expression of transcription factors has been noted in the sensory-motor circuit in the vertebrate spinal cord, where it has been shown that expression of the same ETS gene by different motorneuron pools and their corresponding sensory afferents may be required for functional synaptic connections (Arber et al., 2000; Lin et al., 1998).

The C. elegans genome is predicted to encode a total of seven LIM homeobox genes. Of these, five are expressed in largely non-overlapping sets of chemosensory neurons, raising the possibility that a LIM code (although non-combinatorial), may function in some chemosensory neurons for the specification of their functional identities. lin-11 is expressed in the AWA, ASG, ASH, ADF and ADL chemosensory neurons (this study; Freyd, 1991; Hobert et al., 1998); lim-4 is expressed in the AWB olfactory neurons; ceh-14 is expressed in the AFD thermosensory neurons (Cassata et al., 2000); lim-6 is expressed in the ASE chemosensory neurons (Hobert et al., 1999); and ttx-3 is expressed in the ASI and ADL chemosensory neurons (Altun-Gultekin et al., 2001). The ADL neurons are the only sensory neuron type that expresses more than one LIM homeobox gene. In addition, each of these LIM homeobox genes is also expressed in a number of other neuronal and non-neuronal cell types.

These LIM homeobox genes appear to specify different aspects of neuronal cell fate in different sensory cell types. Sensory functions and gene expression in the ADL neurons are largely unaltered even in lin-11; ttx-3 double mutants (Altun-Gultekin et al., 2001). In ceh-14 mutants, the AFD neurons have defective function, and somewhat defective ciliary morphology, but the overall pattern of differentiation, including gene expression, appears unaltered (Cassata et al., 2000). However, similar to the mec-3 gene in the mechanosensory neurons, the lin-11 and lim-4 genes appear to specify overall aspects of the functional identities of the AWA, ASG and AWB chemosensory neurons. Thus, in the mechanosensory and some chemosensory neurons, LIM homeobox genes act to promote most, if not all, cell type-specific characteristics, and in the case of the olfactory neurons, to repress a default AWC-like developmental state. Misexpression of lim-4, but not lin-11 confers AWB fate to the AWC neurons (Sagasti et al., 1999; T. M. and P. S., unpublished), suggesting that LIN-11 and LIM-4 have distinct protein-specific functions, as opposed to similar functions distinguished only by expression in different cellular contexts (Li and Noll, 1994; O’Keefe et al., 1998). Thus, although LIM homeobox genes appear to play different roles in different sensory cell types, we suggest that a LIM code is an important determinant of final differentiated identities of a subset of sensory neurons.

The distinct functional identity of a cell type is defined by the unique complement of transcription factors expressed in it at defined times. This study shows that LIM homeobox genes are crucial components of these combinatorial transcription factor profiles, and in a subset of sensory neurons are sufficient to confer most, if not all cell type-specific characteristics. In other sensory neurons, LIM homeodomain proteins function together with other transcription factors to confer cell fate. These factors are expected to be expressed in specific subsets of sensory neurons in a lineage-dependent manner. For example, while unc-130 is clearly required to repress the AWA fate and promote ASG fate in the ASG neurons, it is not required for ASG-specific gene expression when lin-11 is ectopically expressed in other neuron types, suggesting that factors that work with lin-11 can be regulated by independent signals in different lineages. The fate of a single cell is determined by combinatorial logic integrating multiple developmental signaling modules, with the output at the level of gene regulation (Flores et al., 2000; Ghazi and VijayRaghavan, 2000; Halfon et al., 2000; Mitani et al., 1993; Wray, 1998; Xu et al., 2000; Yuh et al., 1998). It is the combined and precisely coordinated expression of transcription factors that allows a neuron to adopt an identity that distinguishes it from other neuron types. This study has led to the identification of additional components of the developmental cascades that result in the generation of two specific sensory neurons. In the future, it will be important to identify the complete complement of genes and regulatory modules that work together in specific combinations to generate the many different neuron types that constitute the nervous system.

We are especially grateful to Maura Berkeley and Julia Thompson for excellent technical assistance, and Cori Bargmann for help with cell identification. We thank Cori Bargmann, Susan Birren, Anne Lanjuin, Katie Nolan and John Satterlee for critical comments on the manuscript, Andy Fire and Cori Bargmann for plasmids, Yuji Kohara for the lin-11 cDNA, and Gary Ruvkun in whose lab the ops-1 gene was identified. The anti-ODR-7 antibody was generated by P. S. in the lab of Cori Bargmann. We also acknowledge helpful suggestions and advice provided by members of the Sengupta lab. Several strains used in this work were obtained from the Caenorhabditis Genetics Center. T. S.-R. was a predoctoral fellow of the Howard Hughes Medical Institute. This work was partly funded by NIH grant GM56223, the Packard Foundation and the Whitehall Foundation (P. S.). O. H. was funded by the Human Frontiers in Science Program, the Klingenstein Foundation and the Searle Scholars Program.