Regulation of chemosensory and GABAergic motor neuron development by the C. elegans Aristaless/Arx homolog alr-1

Homeodomain (HD) proteins play crucial roles in multiple aspects of development and differentiation in all metazoans. Although the functions of individual HD proteins and the genetic pathways in which HD proteins act are often remarkably conserved across phyla (e.g. Gehring and Ikeo, 1999; Montalta-He et al., 2002; Reichert, 2002; Zuber et al., 2003), these pathways are often used in different contexts in different species. For example, the Pax-Six-Eya-Dach regulatory network is used in eye development in both vertebrates and invertebrates, but is also required for muscle development in vertebrates (Heanue et al.,1999; Kawakami et al.,2000). Redeployment of a conserved genetic module for the development of different tissue or cell types may underlie key aspects of speciation and the generation of species-specific characteristics. Studying the functions of HD proteins in multiple model systems can lead to the identification of conserved targets and reveal unexpected insights into their regulatory roles.

The functions of the highly conserved paired-type HD protein Aristaless(AL) was first described in Drosophila(Campbell and Tomlinson, 1998; Campbell et al., 1993; Schneitz et al., 1993). In al mutants, the development of wings, legs and aristae, terminal antennal appendages required for auditory functions and hygrosensation(Gopfert and Robert, 2002; Manning, 1967; Sayeed and Benzer, 1996) are affected. AL was shown to act in a network with the Bar HD and LIM1 LIM-HD proteins to regulate development of the pretarsus, the distal-most leg segment(Kojima et al., 2000; Pueyo and Couso, 2004; Pueyo et al., 2000; Tsuji et al., 2000). Vertebrate genomes encode multiple AL homologs that have been classified into three groups based on structural and functional properties, and expression patterns: group I genes are primarily involved in skeletal and craniofacial morphogenesis; group II genes are expressed in the central and peripheral nervous system; and group III genes mediate diverse functions(Meijlink et al., 1999). Recently, the functions of Arx, a group II murine and human homolog of Al, have been described. In mice, knockout of Arx results in animals with thin cerebral cortices, owing to decreased neuroblast proliferation in the neocortical ventricular zone, and defects in the differentiation, proliferation and migration of GABAergic interneurons from the ganglionic eminences to the cortex and olfactory bulbs(Kitamura et al., 2002; Yoshihara et al., 2005). Interestingly, mutations in the human ARX gene have been associated with highly pleiotropic developmental and behavioral anomalies. These include X-linked lissencephaly with abnormal genitalia (XLAG), mental retardation,infantile spasms and epilepsy (Bienvenu et al., 2002; Stromme et al.,2002a; Stromme et al.,2002b). In both mice and humans, Arx is expressed at high levels in the neocortical ventricular zone and in GABAergic interneurons in the ganglionic eminences (Bienvenu et al.,2002; Colombo et al.,2004; Kitamura et al.,2002; Miura et al.,1997; Poirier et al.,2004). These results suggest that ARX plays an important role in regulating both neuroblast proliferation and differentiation and migration of GABAergic interneurons. However, the targets of ARX and its precise roles in neuronal development remain to be elucidated. Because, together with Fragile X syndrome, mutations in Arx may represent the most common cause of mental retardation in males (Sherr,2003), it is important that the functions of ARX-related proteins are further investigated.

C. elegans provides an excellent model system in which the functions of conserved proteins can be genetically explored. In particular,the roles of specific proteins in the development and function of the nervous system can be easily investigated. The adult C. elegans hermaphrodite contains 302 neurons of which 32 are sensory and 26 are GABAergic motoneurons(McIntire et al., 1993b; Ward et al., 1975; White et al., 1986). Molecules and pathways required for the generation and specification of sensory and motoneuron subtypes have been described(Melkman and Sengupta, 2004; Thor and Thomas, 2002). Genes identified to date encode members of well-conserved transcription factor families, including members of several HD protein families. For example,members of the OTX-type and LIM-HD protein families have been shown to play roles in the development and differentiation of several sensory neuron types in C. elegans (Lanjuin and Sengupta, 2004; Melkman and Sengupta, 2004). Vertebrate Otx genes can functionally substitute for C. elegans Otx genes, suggesting conservation of protein function across species (Lanjuin et al.,2003). Similarly, the mouse PITX2 HD protein has been shown to functionally substitute for the C. elegans UNC-30 PITX-type HD protein in the regulation of expression of the glutamic acid decarboxylase gene and differentiation of GABAergic motoneuron subtypes(Eastman et al., 1999; Jin et al., 1994; Jin et al., 1999; McIntire et al., 1993a; Westmoreland et al., 2001). Characterization of additional molecules required for the development of these neuron types not only allows us to understand the principles underlying the generation of distinct neuronal subtypes in an organism, but also provides an opportunity to explore further the roles of conserved proteins in a well-defined system.

Here, we describe characterization of the C. elegans homolog of Arx/Al, alr-1 (Aristaless/Arx-related). alr-1mutants exhibit defects in the specification of the AWA and ASG chemosensory neurons, and we show that similar to its ortholog in Drosophila,ALR-1 acts in a pathway with the LIM1 ortholog LIN-11 to regulate the development of both these neuron types. Intriguingly, we also demonstrate that ALR-1 plays a role in the differentiation of GABAergic motoneurons. In alr-1 mutants, the VD MN type is mis-specified, leading to a partial adoption of DD motoneuron subtype characteristics. These data indicate that some functions of ARX/Aristaless may be conserved across species, and suggest that studying the role of ALR-1 in neuronal development in C. elegansmay provide insights into the functions of this important protein in other organisms.

Animals were grown using standard methods(Brenner, 1974). A strain carrying integrated copies of the ops-1::dsRed transgene(oyIs47) was generated as previously described(Satterlee et al., 2001).

Expression of the following stably integrated markers was also examined in alr-1 mutants (neuron type examined is indicated in parentheses): gcy-8::gfp (AFD), srh-142::dsRed (ADF), str-1::gfp(AWB), str-2::gfp (AWC), sre-1::gfp (ADL), sra-6::gfp (ASH and ASI), str-3::gfp (ASI), daf-7::gfp (ASI), odr-1::dsRed (AWC and AWB), ttx-3::gfp (AIY) (Hobert et al.,1997; L'Etoile and Bargmann,2000; Peckol et al.,2001; Ren et al.,1996; Troemel et al.,1995; Troemel et al.,1997; Troemel et al.,1999; Yu et al.,1997) (E. R. Troemel, PhD thesis, University of California,1999).

oy42 and oy56 alleles were isolated in EMS mutagenesis screens for altered expression of an integrated odr-7::gfp transgene (kyIs38) or dye-filling defects respectively. The ok545allele was generated by the C. elegans Gene Knockout Consortium. Mutants were outcrossed at least four times prior to further analysis. alr-1(oy42) was mapped with respect to genetic markers, polymorphisms and deficiencies to LG X. Rescue of tested alr-1 mutant phenotypes was obtained with sequences amplified from the cosmid R08B4 (nucleotides 13223-19971). The molecular nature of the lesions in the alr-1alleles was determined by sequencing.

The alr-1p::alr-1 and alr-1p::Al constructs were generated by fusing 2.8 kb of alr-1 upstream promoter sequences to Al or alr-1 cDNAs (gifts of G. Campbell and Y. Kohara,respectively) in a C. elegans expression vector (gift of A. Fire). The unc-30p::alr-1 construct was generated by fusing 2.5 kb of unc-30 promoter sequences upstream of an alr-1 cDNA. unc-30p::unc-55 was a gift from W. Walthall.

Germline transformations were performed using standard techniques(Mello and Fire, 1995). All transgenes were injected at 30 ng/μl along with one of the following co-injection markers: 100 ng/μl pRF4 rol-6(su1006), 50 ng/μl unc-122::gfp (Miyabayashi et al.,1999) or unc-122::dsRed.

A rescuing alr-1 genomic fragment was injected along with osm-6::dsRed (gift of A. Lanjuin) at 15 ng/μl each into an alr-1(oy42) strain containing stably integrated copies of an odr-10::gfp fusion gene (kyIs37). L1-L3 larvae were incubated in 1:500 dilution of DiO for 1 hour. osm-6::dsRed-expressing animals were scored for the presence or absence of odr-10::gfp expression, and the corresponding presence or absence of osm-6::dsRed expression in ASI or ASK neurons on the same side. Presence of the array scored as a result of perdurance of osm-6::dsRed expression is unlikely as we could detect expression in either of the lineally related ASK or ADL neurons on a given side. It is possible that the number of AWA neurons lacking odr-10::gfpexpression is an overestimation, as gfp in the AWA neurons was often difficult to detect in the background of fluorescence because of dsRed expression and dye filling.

Animals with the desired odr-7::gfp-expressing phenotype were selected under 400× magnification, and allowed to recover on food for at least 2 hours prior to analysis. Single animals were assayed essentially as described (Bargmann et al.,1993). The behavior was scored as wild type if the animal entered a region of defined diameter surrounding the odorant without entering a similar region around the control diluent during the course of the assay. Statistical significance was determined using a chi-square test.

Animals were incubated in 1:100 DiI or 1:1000 DiO in M9 for 2 hours, washed twice and let recover on food for at least 30 minutes prior to analysis.

Dauer assays were performed essentially as previously described(Lanjuin and Sengupta,2002).

Worms were placed on one half of a 5 cm plate surrounded by a barrier of 4M fructose with a point source of an attractive odorant on the opposite side. The percentage of worms that stayed within the barrier was scored after 30 minutes.

Sequences encoding the C-terminal 182 amino acids of ALR-1 were cloned into the pGEX4T-1 (Pharmacia) vector. The fusion protein was expressed in E. coli and purified using Glutathione Sepharose beads. Polyclonal antibodies were generated in rats by Cocalico Biologicals. Sera were affinity purified before use. Staining with anti-ALR-1, anti-ODR-7 and anti-GABA antibodies was performed as described(McIntire et al., 1992; Sarafi-Reinach et al., 2001). Animals were viewed under a Zeiss Axioplan microscope equipped with epifluorescence and images were captured using a CCD camera (Hamamatsu). Images were analyzed using Openlab (Improvision) and Adobe Photoshop software(Adobe Systems).

sns-10(oy42) animals were isolated in a screen for mutants exhibiting altered expression of an odr-7::gfp transgene in the AWA olfactory neurons (see below). Genetic mapping and transformation rescue experiments identified sns-10 as the R08B4.2 gene. Phylogenetic analyses had previously suggested that R08B4.2 encodes the C. elegans ortholog of Drosophila Aristaless, and appears to be most closely related to the vertebrate group II ARX homeoprotein(Galliot et al., 1999)(Fig. 1A). Based on sequence and functional conservation, sns-10 has been renamed alr-1.The homeodomain of ALR-1 is ∼81% and 79% identical to the homeodomains of Drosophila AL and human ARX, respectively, and contains a Q50 residue, characteristic of this subclass of Paired-like homeodomains(Galliot et al., 1999; Meijlink et al., 1999)(Fig. 1B). In addition to a conserved homeodomain, members of the AL/ARX family are distinguished by the presence of a C-terminal 16 amino acids `Aristaless/OAR domain', which is present although not highly conserved in ALR-1(Meijlink et al., 1999). Expression of the Al cDNA driven by the alr-1 promoter rescued alr-1(oy42) mutant phenotypes(Fig. 1D), suggesting conservation of protein function across species.

alr-1(oy42) is a complex deletion-rearrangement that deletes all alr-1 as well as flanking sequences(Fig. 1C) and is therefore a null allele. There are less severe lesions in two additional alr-1alleles (Fig. 1C). alr-1(oy56) is a point mutation in the first splice donor site. RT-PCR and sequence analyses indicated that the transcript arising as a consequence of this mutation would be predicted to encode a prematurely truncated protein. However, wild-type transcripts were also detected at low levels, indicating that alr-1(oy56) is not a null allele. alr-1(ok545) is predicted to encode a protein with a partly deleted homeodomain. All tested defects of alr-1(oy42) mutants could be rescued with a genomic fragment containing only alr-1 sequences(Table 1; Fig. 1D), indicating that these phenotypes were a consequence of loss of alr-1 functions alone.

In alr-1 mutants, expression of the odr-7 nuclear receptor gene was lost in one or infrequently, in both AWA olfactory neurons,as assayed by expression of an odr-7::gfp transgene and by staining with anti-ODR-7 antibodies (Fig. 1D; Fig. 2B-C; Table 1). We also observed ectopic expression of odr-7 in a second sensory neuron at a lower penetrance (Fig. 2D; Table 1). Loss or ectopic expression of odr-7 was not correlated on the left and right sides of individual animals, enabling us to quantitate the mutant phenotype per side(Table 1). ODR-7 regulates the expression of signal transduction genes, including the odr-10olfactory receptor and the gpa-5 Gα protein genes in the AWA neurons (Jansen et al., 1999; Sengupta et al., 1996). As expected, expression of these genes was similarly affected in alr-1mutants (Table 1; data not shown). Gene expression defects were observed as early as the threefold embryonic stage. The AWA neurons exhibited additional defects in alr-1 mutants. In particular, we observed morphological defects in the dendritic structures, such that the dendrites failed to fully elongate to the anterior amphid organ opening (Fig. 2F). Eighteen and 17% of AWA neurons that could be visualized(n>100) exhibited severe dendritic elongation defects in oy42 and oy56 animals, respectively. Mutations in alr-1 did not affect expression of a subset of marker genes in the AWB, AWC, ADL, ASH, ASI, ADF and AFD sensory neurons, and AIY interneurons(data not shown; see Materials and methods). However, 45% of AFD thermosensory neurons and 4% of ADF chemosensory neurons (n>65) also exhibited a range of dendritic defects in alr-1(oy42) animals.

odr-7 mutants are unable to respond to all AWA-sensed attractive odorants including diacetyl (Sengupta et al., 1994). We examined the olfactory behaviors of alr-1(oy42) and alr-1(oy56) mutant animals lacking odr-7 expression (Fig. 3). Although the overall olfactory responses of alr-1(oy42) mutants were diminished, oy42 and oy56mutants expressing odr-7 in both AWA neurons retained attractive responses to diacetyl. However, animals lacking odr-7 expression in both AWA neurons showed a strong defect in this response. This defect was specific to the AWA neurons, as responses to the chemicals benzaldehyde and 2,3-pentanedione, mediated by the AWC olfactory neurons(Bargmann et al., 1993), were unaltered upon loss of odr-7 expression in both AWA neurons(Fig. 3; data not shown).

In odr-7 mutants, the AWA neurons have been shown to misexpress the AWC-specific str-2 olfactory receptor gene(Sagasti et al., 1999). However, in alr-1 mutants, str-2 was not ectopically expressed in the AWA neurons upon loss of odr-7 expression (data not shown). Additional sensory neuron-specific markers examined (see Materials and methods) were also not ectopically expressed in the AWA neurons in alr-1 mutants. To determine whether the AWA neurons in alr-1mutants were generated, and whether they expressed general sensory neuronal features, we examined expression of an osm-6::dsRed fusion gene. osm-6 is expressed in and required for development of the ciliary structures of all sensory neurons in C. elegans(Collet et al., 1998; Perkins et al., 1986). We detected osm-6::dsRed expression in AWA neurons (identified by relative position with respect to additional osm-6-expressing cells),which failed to express the AWA-specific marker odr-10::gfp in alr-1(oy42) animals (n=12 cells). These results suggest that upon loss of ALR-1 function, the AWA neurons are generated and retain general sensory neuronal characteristics, but fail to adopt neuron type-specific fates.

The AWA olfactory and the ASG chemosensory neurons arise from the terminal cell division of the ABpl/raapapaa precursors(Sulston et al., 1983). To determine whether mutations in ALR-1 also affect differentiation of the sibling ASG neurons, we examined expression of the markers ops-1::dsRed and unc-30::gfp in alr-1 mutants. Expression of both markers in the ASG neurons was affected in alr-1mutants (Table 1). Similar to the defects observed in the AWA neurons, marker expression was lost in one or at a lower penetrance, in both ASG neurons. However, we did not detect ectopic expression of either marker. To determine whether gene expression defects in the AWA and ASG sibling neurons were correlated, we examined the expression of both odr-7::gfp and ops-1::dsRed in individual alr-1 mutant animals. As shown in Table 2, loss of marker expression in an AWA or ASG neuron was not correlated with loss of expression in its sibling.

Approximately 8% of alr-1 mutants exhibited ectopic ODR-7 expression in another cell type (Table 1). Although we could not definitively identify the ectopic ODR-7-expressing cells due to some disorganization in the heads of alr-1 mutants, we investigated the possibility that the sibling ASG neurons were adopting AWA fate. Adoption of AWA fate is expected to be concomitant with loss of ASG-specific gene expression. We noted that although only 10-18% of sides in alr-1 mutants failed to express ops-1::dsRed (Tables 1and 2), >99% of sides(n=56) exhibiting ectopic odr-7::gfp expression lacked ops-1::dsRed expression (Table 2). Rarely, we also detected colocalization of ectopic odr-7::gfp with ops-1::dsRed expression. Taken together,these results suggest that a subset of ASG neurons adopt AWA-like characteristics in alr-1 mutants. However, it is possible that additional cell types may also infrequently adopt an AWA fate. Thus, ALR-1 appears to regulate the differentiation of both the AWA and ASG lineal sibling neurons, and moreover, ALR-1 acts to repress the AWA fate in a subset of ASG neurons.

The forkhead domain-containing protein UNC-130 has previously been shown to regulate asymmetric division of the AWA/ASG precursors(Sarafi-Reinach and Sengupta,2000). Similar to alr-1 mutants, the ASG neurons fail to express ASG-like characteristics, and instead adopt an AWA-like fate in unc-130 mutants. Expression of AWA-specific markers is also lost at a low penetrance in unc-130 mutants. To determine whether ALR-1 and UNC-130 act in a linear or parallel pathway to regulate AWA and ASG development, we examined AWA- and ASG-specific gene expression in unc-130;alr-1 double mutants. We observed an almost complete loss of odr-7 expression in unc-130(ev505); alr-1(oy42) and unc-130(ev505); alr-1(oy56) double mutants(Table 3). We also observed a significant increase in the penetrance of loss of ops-1::dsRedexpression in unc-130(ev505); alr-1(oy42) double mutants when compared with either single mutant alone(Table 3). These results suggest that ALR-1 and UNC-130 act in parallel to specify AWA and ASG fates.

Previously, we have shown that the C. elegans ortholog of the LIM1 LIM homeodomain transcription factor LIN-11 specifies the fates of the AWA and ASG neurons (Sarafi-Reinach et al.,2001). In the AWA neurons, transient expression of lin-11during late embryonic/early larval stages is necessary to initiate odr-7 expression, whereas in the ASG neurons, lin-11expression is maintained through all postembryonic stages. This differential temporal regulation of lin-11 expression is important for correct fate specification, as forced maintenance of lin-11 expression in the AWA neurons has been shown to result in defects in AWA differentiation(Sarafi-Reinach et al.,2001).

To investigate possible regulatory relationships between ALR-1 and LIN-11,we examined AWA- and ASG-specific gene expression in double mutants. We found that odr-7 and ops-1::dsRed expression was lost to a similar extent in lin-11(n389); alr-1(oy42) double mutants as in lin-11(n389) mutants alone (Table 3), suggesting that ALR-1 and LIN-11 act in a linear pathway to regulate gene expression in the AWA and ASG neurons. In both neuron types, a higher percentage of lin-11(n389) than alr-1(oy42) mutants exhibited loss of gene expression, indicating that genes in addition to alr-1 may play a role in regulating lin-11. Alternatively, lin-11 may function upstream of alr-1 and additional genes to regulate AWA and ASG fate. To investigate whether ALR-1 acts upstream or downstream of lin-11, we examined the expression of a stably integrated lin-11::gfp fusion gene in alr-1 mutants. We noted that ∼15% of alr-1(oy42) (n=87) when compared with 0% of wild-type AWA neurons (n=26) exhibited persistent expression of lin-11::gfp through late larval stages. We also observed loss of lin-11::gfp expression in 40% of ASG neurons of alr-1(oy42)mutants (n=87). Taken together with the observation that alr-1 expression was not detected in postmitotic neurons (see below),these results suggest that, ALR-1 acts upstream of lin-11 in part to downregulate lin-11 expression in later larval stages in the AWA neurons, whereas ALR-1 promotes lin-11 expression in the ASG neurons. Interestingly, Lim1, the Drosophila ortholog of lin-11 has been shown to be regulated by AL and acts together with AL to regulate the development of the distal-most compartment of the leg(Pueyo and Couso, 2004; Pueyo et al., 2000; Tsuji et al., 2000).

To further investigate the proposed regulatory relationships among alr-1, lin-11 and unc-130, we also examined marker expression in lin-11; unc-130 double mutants. In both AWA and ASG neurons, we observed a highly penetrant loss of marker expression in lin-11(n389); unc-130(ev505) double mutants(Table 3), consistent with the hypothesis that ALR-1 functions in a linear pathway with LIN-11 but in a parallel pathway with UNC-130 to regulate cell fate.

To examine the expression pattern of ALR-1, we raised polyclonal antibodies against the less well-conserved C-terminal sequences of ALR-1. Staining was first evident in 1.5-fold embryos (Fig. 4B) and although the spatial expression was dynamic, stained neuronal and non-neuronal cells were also observed at later embryonic stages. In larvae and adults, ALR-1 expression was observed in multiple neuronal and non-neuronal cells (including epidermal cells) in the head, neuronal cells in the tail and in the GABAergic DD and VD motoneurons (MNs) in the ventral nerve cord (Fig. 4F,H). Consistent with ALR-1 being a transcription factor, expression was exclusively nuclear in all observed cell types. No staining was observed in alr-1(oy42)mutants (data not shown). To determine whether ALR-1 was expressed in the AWA and ASG neurons, we examined colocalization of anti-ALR-1 staining with expression of odr-7::gfp and ops-1::dsRed, which are expressed postmitotically. However, we did not observe colocalization with these markers in threefold embryos, larvae or adults(Fig. 4I,J; data not shown),although transient expression cannot be ruled out. As ARX has been shown to be expressed in and required for the differentiation of GABAergic neurons in vertebrates (Colombo et al.,2004; Kitamura et al.,2002; Poirier et al.,2004), and ALR-1 is expressed in the GABAergic ventral cord motoneurons, we further investigated whether ALR-1 is expressed in additional GABAergic cells by examining colocalization of anti-ALR-1 staining with unc-47::gfp expression. unc-47 encodes a vesicular GABA transporter and is expressed in all GABAergic neurons(McIntire et al., 1997). Intriguingly, we observed ALR-1 expression in 24 of 26 GABAergic neurons,including the 13 VD and 6 DD, and the RME L/R, AVL, RIS and DVB neurons throughout postembryonic development (Fig. 4F,H; data not shown).

As ALR-1 is expressed in multiple cell types, it may act cell autonomously or cell nonautonomously to regulate AWA/ASG development. To address this issue, we first attempted to rescue the chemosensory neuronal defects by expressing an alr-1 cDNA under specific promoters. However,expression of alr-1 driven by the unc-130, osm-6 or the unc-30 promoters did not rescue the alr-1(oy42) phenotypes(data not shown).

The ASI chemosensory neurons are lineally related to the AWA and ASG neurons such that these three neuron types arise from the common ABp(l/r)aapap precursors via two (AWA and ASG) or three (ASI) additional cell divisions(Fig. 5A)(Sulston et al., 1983). Additional cells arising from these precursors include the AIB interneurons and a cell that undergoes programmed cell death. The ASI neurons can be relatively easily identified via their characteristic cell body positions and via filling with lipophilic dyes (Perkins et al., 1986; White et al.,1986). An additional chemosensory cell type that also fills with dye and is easily identified is the ASK neuron, which arises from a more lineally distant precursor (Fig. 5A) (Sulston et al.,1983). We reasoned that if alr-1 acts in the AWA/ASG/ASI lineage, presence or absence of odr-10::gfp expression in the AWA neurons should correlate more highly with the presence or absence of alr-1 rescuing sequences in the lineally related ASI neurons than in the ASK neurons. However, if alr-1 acts elsewhere, then we may detect higher correlation with the presence or absence of alr-1-coding sequences in ASK (if alr-1 acts in this lineage) or similar correlation with expression in both cell types (if alr-1 acts in a distinct lineage).

To test this hypothesis, we generated alr-1(oy42) animals carrying rescuing alr-1 genomic sequences together with an osm-6::dsRed marker gene on an extrachromosomal array. osm-6::dsRed is expressed in multiple ciliated sensory neuron types,including the ASK and ASI neurons (Collet et al., 1998) (A. Lanjuin, unpublished). Animals carrying this array are mosaic, as the array is lost randomly at each mitotic division. We first identified transgenic animals which expressed odr-10::gfp in an AWA neuron on a side. We then determined whether the array was present in the ASI or ASK neurons on that side by examining expression of the osm-6::dsRed transgene. Cell identification was further facilitated by dye filling. We found that 82% of odr-10::gfp-expressing sides also expressed dsRed in the ASI neurons, whereas 59% of sides expressed dsRed in the ASK neurons(Fig. 5B). We could not always detect osm-6::dsRed expression in the odr-10::gfp-expressing AWA neuron because of the relatively weak expression of dsRedcompared with gfp. Conversely, 77% of sides that failed to express odr-10::gfp also failed to express dsRed in the ASI neurons,when compared with 48% of sides that failed to express in the ASK neurons(Fig. 5B) (see Materials and methods for additional details). These results imply that alr-1 acts in the AWA/ASG/ASI lineage to regulate development of the AWA and ASG neurons.

We next investigated whether the development of GABAergic neurons was affected in alr-1 mutants. As the development of the 13 VD and 6 DD ventral cord motoneurons has been studied extensively, we focused our attention on these cell types. We observed ectopic expression of the DD MN-specific marker, flp-13::gfp(Kim and Li, 2004) in all alleles of alr-1 mutant animals(Table 4; Fig. 6D). On average, there were 10-13 flp-13::gfp-expressing cells in alr-1 mutant adults when compared with six in wild-type adults. All flp-13::gfp-expressing cells were stained with anti-GABA antibodies indicating that the additional cells were GABAergic(Table 4; Fig. 7). However, the total number of GABAergic MNs in the ventral nerve cord were unaltered in alr-1 mutants as determined by expression of an unc-25glutamic acid decarboxylase fusion gene(Jin et al., 1999), and by staining with anti-GABA antibodies (Table 4). These results indicated the possibility that the 13 GABAergic VD MNs were adopting DD MN characteristics in alr-1 mutants. Unlike the DD MNs, which are generated embryonically, the VD MNs are born postembryonically at the L2 larval stage(Sulston, 1976; Sulston and Horvitz, 1977; Sulston et al., 1983). Ectopic flp-13::gfp-expressing cells in alr-1 mutants were not observed prior to the L2 larval stage(Table 4, Fig. 6H), further suggesting that the VD MNs were expressing DD MN-specific genes in alr-1mutants. Consistent with this hypothesis, markers for other MNs (DA and DB, unc-129::gfp; VA and VB, del-1::gfp; VC, ida-1::gfp) were unaffected in alr-1 mutants (data not shown).

In adult animals, the VD MNs innervate ventral muscles, whereas the DD MNs innervate dorsal muscles (White et al.,1986). The locations of the VD and DD MN synapses can be visualized by examining localization of a SNB-1 synaptobrevin/VAMP::GFP fusion protein expressed under the unc-25 promoter(Hallam and Jin, 1998; Jin et al., 1999; Nonet, 1999). In wild-type animals, GFP puncta are observed in the dorsal (DD synapses) and ventral cord(VD synapses) (Hallam and Jin,1998). We reasoned that if the conversion of VDs into DDs was complete, then in alr-1(oy42) mutants, we would observe a dorsal shift in the localization of GFP puncta. However, SNB-1::GFP localization in alr-1 mutants was indistinguishable from that in wild-type animals,(data not shown), suggesting that VDs may adopt only partial characteristics of the DD MNs. Consistent with this, alr-1 mutants are not uncoordinated, as might be expected for animals with gross alterations in MN functions or connectivity (McIntire et al., 1993a; McIntire et al.,1993b).

The UNC-55 COUP-TF-like nuclear hormone receptor has been shown to be expressed in the VD MNs, where it acts to repress DD MN identity(Walthall and Plunkett, 1995; Zhou and Walthall, 1998). In unc-55 mutants, the VDs adopt a DD wiring pattern, and this change in innervation pattern is believed to underlie the locomotion defects of unc-55 mutants (Walthall and Plunkett, 1995; Zhou and Walthall, 1998). In agreement with a VD-to-DD fate transformation in unc-55 mutants, we observed ectopic flp-13::gfpexpression in unc-55(e402) mutant adults but not in L1 larvae(Fig. 6F,I; Table 4)(Shan et al., 2005). ALR-1 may act downstream or upstream of, or in parallel to UNC-55 to regulate MN differentiation. However, alr-1 expression in the MNs was unaffected in unc-55(e402) mutants (an average of 18±1 ALR-1-expressing cells were observed in unc-55(e402) when compared with 18±2 in wild-type animals; n>22 each). Similarly, unc-55::gfp expression was also unaffected in alr-1(oy42) mutants (an average of 14±6 unc-55::gfp-expressing cells were observed in alr-1(oy42) when compared with 13±5 in wild-type animals; n>100 each), suggesting that UNC-55 and ALR-1 may act in parallel to repress flp-13::gfp expression in the VD MNs. As ALR-1 is expressed in both the VD and DD MNs, whereas UNC-55 is expressed only in the VD MNs, we next determined whether ectopic expression of UNC-55 in the DD MNs was sufficient to repress flp-13::gfp expression. However, the expression pattern of flp-13::gfp was unaltered in transgenic animals expressing unc-30::unc-55 (Table 4). These results suggest that both UNC-55 and ALR-1 are necessary to repress flp-13::gfp expression in the VD MNs, but are not sufficient to repress expression in the DD MNs.

Consistent with alr-1 expression in multiple neuronal and non-neuronal cell types, alr-1 mutants exhibit additional phenotypes. A subset of sensory neurons in the amphid and phasmid sensory organs fill with lipophilic dyes such as DiI (Perkins et al., 1986). Developmental or structural defects in either the supporting sheath and socket cells or in the sensory neurons result in dye-filling defects (dyf phenotype). Ninety-three and 34% of alr-1(oy42) mutant amphids and phasmids, respectively(n=224), exhibited dye-filling defects. Consistent with a defect in the supporting non-neuronal cells, dye filling was affected in an all-or-none manner, such that either all amphid neurons on one side of an alr-1mutant animal failed to dye fill or all neurons dye filled in the wild-type pattern. dyf mutants exhibit additional pleiotropies such as the inability to avoid osmotic shock (Osm phenotype) and failure to enter the alternate dauer developmental stage (Daf-d phenotype)(Starich et al., 1995). alr-1 mutants are both Osm [∼65% of alr-1(oy42) and 61%of alr-1(oy56) mutants (n>100) failed to avoid a high osmolarity solution] and Daf-d [15% alr-1(oy42) and 13% of alr-1(oy56) animals formed dauers under conditions where 75% of wild-type animals form dauers (n>200)]. Taken together with the observation that the differentiation of dye-filling neurons appeared to be unaffected in alr-1 mutants, these results suggest that the Dyf phenotype may be due to defects in the amphid sheath or socket cells. These defects may be structural, as the expression of a subset of sheath and socket cell differentiation markers was unaltered in alr-1 mutants (data not shown). It is unlikely that the AWA and ASG differentiation and/or generation defects are a secondary consequence of the defects in the amphid support cells, as the Dyf phenotype was not correlated with the defects in either gene expression or morphology of the AWA and ASG neurons. In addition, ODR-7 expression was unaltered in daf-6(e1377) and che-14(e1960)mutants, which exhibit defects in the development of the amphid support cells and additional hypodermal cells (Albert et al., 1981; Michaux et al.,2000) (data not shown). Thus, ALR-1 may affect the differentiation of both a subset of sensory neurons, as well as the supporting non-neuronal cells.

A transcription factor cascade involving the UNC-130 forkhead domain protein and the LIN-11 LIM-HD protein specifies the fates of both the AWA and the ASG lineal sisters (Sarafi-Reinach et al., 2001; Sarafi-Reinach and Sengupta, 2000). The strong synergistic effect of mutations in both unc-130 and alr-1 on AWA and ASG fate specification suggests that these two proteins act in parallel to regulate AWA and ASG development (Fig. 8A). We have previously suggested that UNC-130 regulates the asymmetric cell division of the AWA/ASG precursors by ensuring segregation of the `AWA potential' to only the AWA daughter cells (Sarafi-Reinach and Sengupta, 2000). Asymmetric cell division has been extensively studied in the Drosophila nervous system (for reviews, see Betschinger and Knoblich, 2004; Jan and Jan, 2001). Asymmetric localization of intrinsic factors including transcription factors such as Prospero to the neuroblast basal cortex and subsequent inheritance by one of two daughter cells has been shown to be essential for the generation of two distinct daughter cell types. Failure to localize to the cortex or incorrect localization results in the loss of daughter cell types, duplication of sister cell fates and/or defects in daughter cell differentiation. Analogous to Drosophila neuroblast cell divisions, we suggest that the defects in AWA and ASG fate specification in alr-1 and unc-130 mutants arise as a consequence of improper asymmetric localization and segregation of downstream effector(s), which act combinatorially to regulate AWA and ASG fate. As UNC-130 and ALR-1 act in parallel pathways, they may regulate different sets of effectors perhaps by regulating the transcription of different molecules required to mediate their localization and/or segregation. Alternatively, these proteins may act in parallel pathways to co-regulate a partly shared set of target genes. Interestingly, the penetrance of defects in AWA fate specification is different in alr-1 and unc-130mutants, such that a higher percentage of unc-130 than alr-1mutants exhibit ectopic AWA cells generated at the expense of ASG cells,whereas more alr-1 than unc-130 mutants exhibit loss of odr-7 expression. These phenotypes may arise as a consequence of differential requirements for UNC-130 and ALR-1 in the specification of the AWA and ASG neuron types.

One of the downstream molecules regulated by ALR-1 is lin-11(Fig. 8A). Although lin-11 is required for the differentiation of both the AWA and ASG neurons, the temporal regulation of lin-11 expression in these neuron types is distinct. In AWA, lin-11 is expressed transiently,disappearing around the L1 larval stage(Sarafi-Reinach et al., 2001). Loss of alr-1 disrupts this temporal regulation, allowing expression of lin-11 to persist to the L4 larval stage in a small number of animals. As the number of animals in which mis-regulation of lin-11expression is observed is too low to fully account for the observed loss of odr-7 expression in alr-1 mutants, alr-1 probably plays an additional role in promoting AWA fate. In ASG, ALR-1 acts to promote lin-11 expression, which is maintained throughout the life of the animal. Thus, a core genetic regulatory network comprising ALR-1, LIN-11 and UNC-130 is used in both the AWA and ASG lineal siblings, but this network functions differently in each of these cells to specify their distinct fates. We note, however, that although the simplest model proposes that ALR-1 acts cell autonomously in the AWA/ASG lineage, it remains possible that ALR-1 acts cell nonautonomously to regulate development of these neuron types.

alr-1 is expressed in 24 out of 26 GABAergic neurons in C. elegans, suggesting a role for this gene in GABAergic neuron development and/or function. Although we did not detect gross abnormalities in the development of additional ALR-1 expressing GABAergic neurons in alr-1mutants, we have shown that alr-1 may play a role in the differentiation of the VD motoneurons. The inhibitory VD (DD) MNs innervate the ventral (dorsal) body muscles and are in turn innervated by excitatory cholinergic MNs which also innervate dorsal (ventral) body wall muscles(White et al., 1986). Thus,alternate contraction of ventral or dorsal muscles is accompanied by relaxation of dorsal or ventral muscles respectively, resulting in the characteristic sinusoidal locomotory motion(McIntire et al., 1993a; McIntire et al., 1993b; White et al., 1986). The D-type MNs may also regulate wave amplitude(McIntire et al., 1993b).

The GABAergic DD motoneuron marker flp-13 is ectopically expressed in the VD MNs in alr-1 mutants, whereas expression of additional GABAergic markers, common to both VD and DD motoneurons is unaffected. In addition, the synaptic connectivities of the VD MNs were also unaltered. We could not further investigate the extent to which the differentiation of the VD MNs was affected in alr-1 mutants, owing to the lack of additional VD- or DD-specific markers. The flp-13 gene is predicted to encode at least six FMRFamide-related neuropeptides(Li et al., 1999), two of which have been biochemically isolated from C. elegans(Li et al., 1999; Marks et al., 2001). Peptides encoded by flp-13 have been shown to cause a dramatic inhibition of locomotory behavior and paralysis when injected into Ascaris suum,and inhibit pharyngeal activity in C. elegans(Marks et al., 2001; Rogers et al., 2001),suggesting that these neuropeptides may act to modulate the inhibitory functions of GABA at the neuromuscular junction. Restriction of flp-13 expression to the DD MNs in wild-type animals may be important for precise modulation of locomotory behaviors under specific conditions.

The COUP transcription factor UNC-55 has been previously shown to prevent the expression of the DD synaptic pattern in the VD MNs(Walthall and Plunkett, 1995; Zhou and Walthall, 1998). We have shown that similar to ALR-1, UNC-55 also represses expression of flp-13::gfp in the VD MNs, although, unlike UNC-55, ALR-1 does not affect the synaptic pattern of VD motoneurons. ALR-1 is not sufficient to repress flp-13 expression in the absence of UNC-55 function and vice versa, suggesting that functions of both proteins are necessary for repression of flp-13 expression in the VD MNs. Thus, ALR-1 acts together with a member of the well-conserved COUP transcription factor family to regulate the differentiation of a specific GABAergic MN subtype(Fig. 8B). However, mutations in unc-55 do not affect AWA development, and mutations in unc-130 and lin-11 do not alter flp-13::gfpexpression (T.M. and P.S., unpublished), suggesting that ALR-1 functions in different pathways to regulate chemosensory and motoneuron development.

Our results indicate that ALR-1 acts in distinct transcriptional cascades to regulate asymmetric cell division of a neuronal precursor and to specify the characteristics of a GABAergic MN subtype in C. elegans(Fig. 8). These processes have parallels to the processes regulated by ARX in vertebrates. In arxmutant mice, neuroblast proliferation in the cerebral cortex is decreased(Kitamura et al., 2002). Neuroblast proliferation in the ventricular zone occurs via temporally regulated symmetric and asymmetric cell divisions that generate additional neuronal precursors and postmitotic neurons(McConnell, 1995). We speculate that ARX may regulate these cell divisions perhaps by regulating the localization or segregation of determinants such as Numb or Notch(Petersen et al., 2002; Shen et al., 2002; Wakamatsu et al., 1999; Zhong et al., 1996; Zhong et al., 2000; Zhong et al., 1997). ALR-1 acts in part by temporally restricting expression of lin-11 in the AWA neurons, and by promoting lin-11 expression in the ASG neurons. Interestingly, expression of the LIM homeobox genes Lhx6 and Lhx9 is abolished in the neocortex and thalamic eminence,respectively, in Arx mutant mice, whereas the domain of Lhx6expression in the ganglionic eminences is enlarged(Kitamura et al., 2002). Taken together with the observation that lim1 and al function in a network to regulate Drosophila leg development(Pueyo and Couso, 2004; Pueyo et al., 2000; Tsuji et al., 2000), these findings suggest that regulatory mechanisms between ARX proteins and LIM-HD proteins may be conserved across species.

ALR-1 acts together with the UNC-55 COUP transcription factor to regulate the differentiation of a GABAergic MN type in C. elegans. A COUP-TF protein and the PRDL-B Aristaless/ARX homolog have been shown to act in a network to regulate neurogenesis in Hydra(Gauchat et al., 2004). In vertebrates, COUP transcription factors have been implicated in neurogenesis,neuronal differentiation, migration and axonal guidance(Qiu et al., 1997; Tripodi et al., 2004; Zhou et al., 1999; Zhou et al., 2001). Interestingly, COUP-TFI and COUP-TFII exhibit overlapping spatiotemporal expression patterns with ARX in the developing neocortex, as well as in the lateral and medial ganglionic eminences, which give rise to GABAergic interneurons (Jonk et al.,1994; Liu et al.,2000; Qiu et al.,1994). Moreover, COUP-TFI is co-expressed with the GABAergic neuron marker calbindin in the cortex(Tripodi et al., 2004). These findings suggest the intriguing possibility that COUP and ARX function together to regulate neuronal, and in particular GABAergic, neuronal development. Our results suggest that ARX proteins function in partly conserved genetic networks to regulate the development of different tissue and cell types in different species, and raise the possibility that identification of potential interactors and targets of ALR-1 in C. elegans may aid in elucidating ARX function in brain development in vertebrates.

We thank M. Shibuya for technical assistance; W. Walthall, M. Tucker and M. Han for communication of results prior to publication; W. Walthall, G. Campbell and Y. Kohara for reagents; A. Lanjuin, Y. Jin, O. Hobert and members of the Sengupta laboratory for discussion and critical comments on the manuscript; the Caenorhabditis Genetics Center for strains; and the C. elegans Gene Knockout Consortium for the ok545 allele. This work was supported by the NIH (GM56223 to P.S.).