The AHR-1 aryl hydrocarbon receptor and its co-factor the AHA-1 aryl hydrocarbon receptor nuclear translocator specify GABAergic neuron cell fate in C. elegans

The aryl hydrocarbon receptors (AHR) were first identified by their high affinity binding to TCDD or dioxin, the persistent environmental contaminant(Poland et al., 1976). The protein sequences of AHRs are evolutionarily conserved, and contain a basic Helix-Loop-Helix (bHLH) domain at the N-terminal and a PAS (Per/ARNT/Sim)domain (the PAS domain can be further divided into PAS A and PAS B subdomains)in the middle of the protein (reviewed by Hahn, 2002). The bHLH domain binds DNA, and the PAS domain mediates protein-protein interactions and toxin binding (reviewed by Schmidt and Bradfield, 1996). In mammals, the AHRs mediate many of the diverse biochemical, biological and toxicological responses following exposure to TCDD and related compounds (Fernandez-Salguero et al., 1995; Schmidt et al.,1996) (for a review, see Gonzalez and Fernandez-Salguero,1998). They normally reside in the cytoplasm in a complex with the chaperonins HSP90 (Heat Shock Protein 90) and XAP2/AIP (X-associated protein 2/AhR-interacting protein) (Hahn,2002). Upon toxin binding, the cytoplasmic AHRs translocate to the nucleus, where they dissociate from HSP90 and XAP2, and form a heterodimer complex with a bHLH-PAS protein, the AHR nuclear translocator (ARNT)(Reyes et al., 1992; Probst et al., 1993; Hahn, 2002). The AHR and ARNT heterodimers bind to specific DNA sequences known as Xenobiotic Response Elements (XREs), and can recruit multiple co-activators to regulate gene expression (Kobayashi et al.,1997; Kumar and Perdew,1999; Nguyen et al.,1999; Tohkin et al.,2000).

Invertebrates generally do not have a dioxin-induced toxic response, and none of the invertebrate AHRs tested bind dioxin, leading to the hypothesis that the ancestral role of the AHR family is not in toxin response(Hahn et al., 1997; Hahn, 2002). Studies in Drosophila have supported this idea. Mutations in the Drosophila AHR spineless cause reduced sensory bristles and transformation of distal antenna into distal leg(Duncan et al., 1998). This function of Spineless requires Tango, the Drosophila ARNT(Emmons et al., 1999). Spineless and Tango bind directly and exhibit genetic interactions. Both Spineless and Tango are normally localized in the nucleus, and, interestingly,the nuclear localization of Tango depends on Spineless.

ahr-1 and aha-1 are the C. elegans homologs of AHR and ARNT, respectively (Powell-Coffman et al., 1998). AHR-1 and AHA-1 bind each other in vitro and can bind XREs in a sequence specific manner in vitro. AHR-1 can also bind rabbit HSP90, but not mammalian XAP2(Powell-Coffman et al., 1998; Bell and Poland, 2000). A single ortholog of HSP90 in C. elegans is encoded by the daf-21 gene (Birnby et al.,2000). The null mutation of daf-21 is early larval lethal, and an unusual mutation of daf-21(p673) causes constitutive formation of dauer larvae.

The C. elegans nervous system is composed of 302 neurons(White et al., 1986), 26 of which use γ-amino butyric acid (GABA) as the neurotransmitter and regulate body movement, defecation and foraging behaviors(McIntire et al., 1993). These GABAergic neurons fall into at least five types based on morphology and function, including the 19 type D ventral cord motor neurons, RIS, AVL, DVB and four RME neurons. The UNC-30 homeodomain protein controls the specification of the type D ventral cord neurons(Jin et al., 1994). The LIM-6 homeodomain protein regulates subsets of the differentiated aspects of RIS,AVL and DVB neurons (Hobert et al.,1999). The four RME neurons innervate head muscles to control foraging behavior (White et al.,1986; McIntire et al.,1993). Although sharing similar neurotransmitter specificity and the same synaptic targets, the RME neurons can be further divided into two subgroups, based on cell lineage and gene expression. For example, RMEL and RMER are lineally related, and express lim-6 and the AMPA-type glutamate receptor glr-1 (Hart et al., 1995). RMED and RMEV do not express these markers; instead they express the ivermectin receptor avr-15(Dent et al., 1997). We found that loss of function in ahr-1 transforms RMEL/RMER neurons into RMED/RMEV-like cells. Ectopic expression of AHR-1 in RMED/RMEV can transform them to an RMEL/RMER-like fate. We provide additional evidence that supports an evolutionarily conserved partnership of AHR and ARNT. Our findings are consistent with the notion that an ancestral role of AHR proteins is in regulating cellular development.

C. elegans were grown on NGM plates at 22.5°C under standard conditions, as described by Brenner(Brenner, 1974). Genetic nomenclature follows the standard C. elegans system(Horvitz et al., 1979). The GFP reporter strains used are listed below: juIs76(P_unc-25GFP) (Huang et al., 2002), juIs1 (P_unc-25SNB-1-GFP)(Zhen and Jin, 1999), oxIs12 (P_unc-47flGFP)(McIntire et al., 1997), otIs37 (P_unc-47sGFP)(Aurelio et al., 2003), nuIs25 (P_glr-1GFP)(Hart et al., 1995), otEx406 (P_lim-6rGFP)(Hobert et al., 1999), juEx517 (P_avr-15GFP)(Dent et al., 1997) (this study), vbEx1 (P_clh-6GFP)(Bianchi et al., 2001), iaIs2 (AHR-1::GFP) (Qin, H and J.A.P.-C., unpublished). ZG69 aha-1(ia1); iaEx28 was used as the starting strain to examine the role of aha-1 in RME neurons. aha-1(ia1) is a deletion mutation and the mutant animals die as first stage larvae. This larval lethality is due to the functional requirement of aha-1 in pharynx. The iaEx28 transgene expresses a construct, pHJ32, that has the complete aha-1 genomic coding sequences driven by a 2.5 kb promoter of the C15C8.2 gene, which is active in non-neuronal cells of the pharynx, and the iaEx28 transgene expression restores viability to aha-1(ia1) homozygous animals. Immunostaining with an AHA-1-specific monoclonal antibody (Jiang et al.,2001) confirms that aha-1; iaEx28 animals express AHA-1 in the pharynx, but not in neurons in the head or in other cells in the body(H. Jiang, S. Wu and J.A.P.-C., unpublished).

juIs76(P_unc-25GFP) animals were mutagenized with ethyl methanesulfonate (EMS) following the standard procedure as described (Sulston and Hodgkin,1988). The cell morphology of RME neurons of mutagenized F2 progeny was examined under a Nomarski fluorescence microscope, and mutant animals were recovered to produce progeny. A total of ∼6,200 mutagenized haploid genomes were screened. The ju145 mutation was outcrossed multiple times before further analysis. Homozygous ju145 mutant animals exhibited no discernable abnormalities in locomotion, egg laying, or male mating. The foraging behaviors of ahr-1 animals were observed in parallel with N2, juIs76, glr-1(n2461), unc-25(e156) and unc-47(e542) animals on matched L4 and young adult animals, and the genotypes were blinded. ju145 was mapped to chromosome I because of its linkage to dpy-5. It was further mapped to the unc-29 dpy-24 interval, near the hP6 marker, based on the following data: from ju145/unc-29 hP6 dpy-24 heterozygous animals, 8 out of 37 Unc non-Dpy recombinants, and 5 out of 8 Dpy non-Unc animals, segregated ju145; 0 out of 8 Unc ju145 non-Dpy, and 0 out of 5 Dpy ju145 non-Unc, segregated hP6. ju145 was uncovered by the deficiencies qDf7, qDf9 and mnDf111, but not by dxDf1,dxDf2, eDf3, nDf23, nDf29 or qDf5.

Cosmids were obtained from the Sanger Centre, Hinxton, UK. DNA preparation and subcloning followed standard procedures(Sambrook et al., 1989). The ahr-1 minimal rescuing construct (pCZ466) was generated by cloning a 10 kb ApaI-SpeI genomic fragment into pSL1190. To make the pHT101 Pahr-1GFP plasmid, a 5377 bp HindIII-BamHI genomic fragment, which includes over 3 kb of sequence 5′ to the translational start codon, exon1, intron 1 and part of exon 2, was ligated into the corresponding sites of the pPD95.67 GFP expression vector (a gift of A. Fire). To generate P_unc-25AHR-1, the ahr-1promoter was replaced by the unc-25 promoter at the BstBI site, which is located 14 bp 5′ to the start ATG codon of C41G7.5/AHR-1. Germline transformation was performed following standard procedures(Mello et al., 1991). The co-injection marker was the dominant pRF4 rol-6(su1006) injected at 20 ng/μl. Over 10 transgenic lines expressing P_unc-25AHR-1 were obtained, and all showed similar effects. juEx467(P_unc-25AHR-1) was chosen for further analysis.

To identify the lesions in ju145, the genomic DNA, including all exons and exon-intron junctions, was amplified from mutant and wild-type animals. DNA sequences were determined using ³³P-labeled primers and the fmol sequencing kit (Promega), and were confirmed on both strands and from DNA prepared in independent PCRs.

Anti-GABA staining was performed using the glutaraldehyde-paraformaldehyde fixation procedure described by McIntire et al.(McIntire et al., 1992). GFP reporter expression was directly observed under a 63× objective of a Zeiss Axioskop fluorescence microscope equipped with a HQ-FITC filter(Chroma). Images were captured using an AxioCam camera (Zeiss) and analysed using the Axiovision software (Zeiss).

The four RME motor neurons are located at the nerve ring and innervate the head muscles. The positions of their cell bodies show a fourfold symmetry with respect to the nerve ring (Fig. 1A): dorsal (RMED), ventral (RMEV), left lateral (RMEL) and right lateral (RMER) (White et al.,1986). RMEL and RMER are lineally related, and RMED and RMEV come from two separate lineages (Sulston et al., 1983). Each RME cell sends out two processes circling around the nerve ring, with the distal ends of these processes forming synapses to head muscles. RMED and RMEV have additional processes extending along the dorsal and ventral cords, respectively, and the function of these processes is not known.

We used a GFP transgene driven by the unc-25 promoter to visualize the RME neurons [the genotype of the marker is juIs76[P_unc-25GFP](Huang et al., 2002)](Fig. 1B). unc-25encodes glutamic acid decarboxylase/GAD(Jin et al., 1999). In a visual screen for abnormal expression patterns of juIs76[P_unc-25GFP], we isolated the ju145mutation because the mutant animals showed specific defects in the RMEL and RMER, but not RMED and RMEV, neurons (Table 1; and see below).

We mapped ju145 to chromosome I between unc-29 and dpy-24, close to a polymorphism hP6(Fig. 2A; see Materials and methods), and rescued the RMEL/R cell defects with transgenes containing the cosmid R30. A 10 kb rescuing DNA fragment contains two predicted genes: ahr-1 (C41G7.5), the C. elegans ortholog of the aryl hydrocarbon receptor (Powell-Coffman et al., 1998), and C41G7.6, which resides in the third intron of ahr-1 and is transcribed in the opposite direction(Fig. 2B). The function of C41G7.6 is not known, nor is its expression pattern. We determined the sequences of the genomic DNA from ju145 animals and identified a C to T nucleotide transition that would result in an Ochre stop at Gln302 in the PAS domain of AHR-1 (Fig. 2C). This point mutation is located 2 kb upstream of C41G7.6, and is unlikely to affect the expression of C41G7.6. As shown below, we were able to rescue the RME neuron defects in ju145 animals by expressing ahr-1specifically in RMEL/R neurons. The RMEL/R defects remained the same in ju145/Df animals (see Materials and methods). These results are consistent with the conclusion that the RMEL/R neuron defects of ju145 animals are due to the complete loss of function of ahr-1.

In ahr-1(ju145) animals the expression level of P_unc-25GFP was normal in all four RME neurons. RMED and RMEV had normal cell morphology (Fig. 1B). However, RMEL and RMER exhibited abnormal cell morphology such that, in addition to their circular nerve processes, each extended an ectopic process of variable length growing along the lateral side of the body(Fig. 1B, Table 1). To confirm that this morphological defect is not an artefact of the expression of P_unc-25GFP transgene, we first examined the expression of a unc-47 GFP reporter gene P_unc-47flGFP. unc-47 encodes a GABA transporter and is normally co-expressed with unc-25/GAD in the GABAergic neurons(McIntire et al., 1997; Eastman et al., 1999). We found that in ahr-1(ju145) animals the expression of a transgene P_unc-47flGFP that contains the full-length unc-47promoter was normal in four RME neurons, and RMEL/R exhibited abnormal cell morphology as seen with P_unc-25GFP (n>100,data not shown). We then performed immunocytochemical staining using anti-GABA antibodies to assess the expression of GABA and the cell morphology of RME neurons. We found that in all four RME neurons in ahr-1(ju145)animals GABA expression was indistinguishable from wild type, and the RMEL/R neurons showed morphological defects identical to those seen with P_unc-47flGFP and P_unc-25GFP(Fig. 1C; n>20). To further explore how other differentiated aspects of the RME neurons are affected, we examined the synaptic connections of these neurons using a synaptic GFP marker juIs1[P_unc-25SNB-1-GFP](Zhen and Jin, 1999). In wild-type animals, RMEL/R and RMED/V neurons make numerous synapses onto head muscles at the distal ends of their ring processes(Fig. 3A). We found that in ahr-1(ju145), fewer synapses were made between RMEL/R and head muscles (Fig. 3B), whereas the synapses between RMED/V and head muscles appeared indistinguishable from wild type (data not shown). These observations indicate that the ahr-1mutation specifically affects cellular differentiation of RMEL and RMER neurons. Despite these defects in RMEL/R neurons, the foraging behavior of ahr-1(ju145) animals is largely normal by double-blind tests,suggesting that RMEL/R neurons in ahr-1 mutants provide sufficient synaptic outputs for head muscle movement.

The altered cell morphology of RMEL/R in ahr-1(ju145) resembles that of RMED/V cells, suggesting that RMEL/R neurons may be transformed into RMED/V. To further address this interpretation, we examined the expression of several GFP markers that are differentially expressed in RMEL/R and RMED/V neurons. GLR-1, an AMPA-type glutamate receptor, is expressed in RMEL, RMER and other neurons, but not in RMED and RMEV(Hart et al., 1995)(Fig. 3C). In ju145animals, RMEL/R cells failed to express P_glr-1GFP, whereas P_glr-1GFP expression in other neurons was normal(Fig. 3D). LIM-6, a homeodomain protein, is normally expressed in RMEL/R but not in RMED/V cells(Hobert et al., 1999). We found that the expression of a transgene, P_lim-6rGFP, in RMEL/R was abolished in ju145 animals(Fig. 3E,F). It has been reported that in lim-6 mutants P_glr-1GFP expression is normal and RMEL/R neurons are normal(Hobert et al., 1999),indicating that the effect on P_glr-1GFP in ahr-1(ju145) animals is independent of lim-6. In addition,we found that the expression of a transgene P_unc-47sGFP that contains a shortened unc-47 promoter, and which is expressed specifically in RMEL/R but not in RMED/V(Aurelio et al., 2003), was abolished in RMEL/R neurons of ahr-1(ju145) animals(Fig. 3G,H). Similarly, the expression of a CIC chloride-channel gene, P_clh-6GFP, in RMEL/R neurons was also abolished in ahr-1(ju145) animals (see Fig. S1). By contrast, when we examined the expression of P_avr-15GFP, which is normally expressed in RMED/V, but not RMEL/R, neurons (Dent et al.,1997) (Fig. 3I), we found that the P_avr-15GFP marker was now expressed in RMEL/R cells in ju145 animals(Fig. 3J-L; Table 2). Taken together, our results support the conclusion that the ahr-1(ju145) mutation alters multiple differentiated features of RMEL/R and transforms them into RMED/V-like neurons.

To understand how ahr-1 functions, we examined its expression pattern using a GFP reporter gene (Chalfie et al., 1994). A 5.3 kb ahr-1 genomic fragment that includes over 3 kb of upstream sequences and the first two exons was used to drive GFP expression in transgenic animals. This transgene was expressed in many neurons and the expression pattern was largely identical to a functional AHR-1::GFP [the ahr-1 expression pattern is currently being investigated in more detail (H. Qin and J.A.P.-C., unpublished)]. We found that the Pahr-1 GFP was expressed in RMEL and RMER, but not in RMED and RMEV, cells (Fig. 4).

The expression of ahr-1 in RMEL/R and the mutant phenotypes of ahr-1 suggest that ahr-1 may act as a cell-type-specific determinant to control RMEL/R fate. To further test this hypothesis, we investigated the effect of ectopic ahr-1 expression in RMED/V neurons. We used the unc-25 promoter to express ahr-1 in the four RMEs (see Materials and methods). The expression of ahr-1 in RMEL/R from this transgene fully rescued the cell morphology defects of ahr-1(ju145) (Table 1), consistent with the conclusion that ahr-1 functions in RMEL/R cells. Moreover, the ectopic expression of ahr-1 in RMED/V repressed the formation and elongation of the longitudinal processes along the nerve cords in both wild-type animals and in ahr-1(ju145) animals(Fig. 5B, Table 1). To further address whether the ectopic expression of ahr-1 causes RMED/V to adopt other differentiated features of RMEL/R cells, we analyzed the expression of P_lim-6rGFP in juEx467[P_unc-25AHR-1]transgenic animals. We found that in juEx467 animals, in addition to its normal expression in RMEL/R, RMED/V cells also expressed P_lim-6GFP (Fig. 5C-F, Table 2). Thus, the ectopic expression of ahr-1 in RMED/V cells can transform them into RMEL/R-like cells. These results strongly support the interpretation that ahr-1 normally functions in RMEL/R cells to regulate their cell-specific gene expression.

It has been well established that ahr-1 requires a nuclear cofactor ARNT (AHR nuclear translocator) to bind DNA(Reyes et al., 1992). The C. elegans ARNT AHA-1 is widely expressed(Powell-Coffman et al., 1998). A deletion mutation of aha-1(ia1) results in larval lethality that is due to its requirement in the pharynx. The larval lethality of aha-1(ia1) can be rescued by specific pharyngeal expression of aha-1 (H. Jiang, S. Wu and J.A.P.-C., unpublished).

To address whether AHA-1 is required for AHR-1 function in RMEL/R cell fate specification, we introduced P_unc-25GFP into aha-1(ia1) mutants that are viable because of the pharyngeal expression of aha-1. We found that aha-1(ia1) animals showed identical RMEL/R cell morphology phenotypes to ahr-1(ju145)(Fig. 6C, Table 1), indicating that aha-1 is required for RMEL/R cell fate. Furthermore, when we introduced into aha-1(ia1) mutants the juEx467 transgene that expresses ahr-1 ectopically in RMED/V cells, the RMED/V cells did not show inhibition of the longitudinal process extension(Fig. 6G,H; Table 1), supporting the conclusion that aha-1 is required for the effect of ectopic ahr-1.

Studies in yeast and mammalian cells have shown that the heat shock protein HSP90 is an essential chaperone to sequester AHRs in the cytoplasm(Pongratz et al., 1992). daf-21 encodes C. elegans HSP90 and has been implicated in larval viability and dauer formation(Birnby et al., 2000). A null allele of daf-21(nr2081) results in larval lethality, and an unusual allele daf-21(p673) is viable to adulthood but causes constitutive dauer formation. We found that in both daf-21(nr2081) and daf-21(p673) animals, RMEL/R cell morphology was normal(Fig. 6D, Table 1), indicating that daf-21 is unlikely to be required in RMEL/R neuron fate specification. Supporting this conclusion, daf-21(p673) showed no effects on the ectopic ahr-1 expression in RMED/V neurons(Fig. 6F, Table 1). It has been observed that a functional AHR-1::GFP is normally localized in the nucleus (H. Qin and J.A.P.-C., unpublished). These results suggest that unlike mammalian AHRs, C. elegans AHR-1 can be folded properly and localized to the correct cellular compartments independent of HSP90 chaperone function.

The four RME neurons are closely related neurons by function and neurotransmitter specificity. The data presented above show that ahr-1 regulates the cellular differentiation, but not GABA expression, of RMEL and RMER neurons. lim-6 is expressed in RMEL,RMER and three other GABA neurons, and has been shown to partially regulate the expression of unc-25 in AVL, DVB and RIS, but has no effect on RMEL/R (Hobert et al., 1999). To address whether ahr-1 might function together with lim-6to control GABA expression, we made ahr-1(ju145); lim-6(nr2073)double mutants. These double mutants expressed P_unc-25GFP and P_unc-47GFP normally (data not shown). This observation, together with that that ahr-1 regulates lim-6expression (Figs 3, 5), suggests that lim-6 acts downstream of ahr-1, and mediates other functions of ahr-1 in RMEL/R neurons.

Two transcription factors have been reported to be differentially expressed in RMED and RMEV neurons. ceh-10 is expressed in RMED, and in ceh-10 mutants RMED does not express GABA(Forrester et al., 1998). lim-4 is expressed in RMEL(Sagasti et al., 1999). In a lim-4 mutant, RMEL expresses GABA and has normal morphology(Fig. 7B, and data not shown). As ectopic expression of ahr-1 is able to cause RMED/V neurons to adopt an RMEL/R-like fate, we addressed whether ceh-10 and lim-4 might repress the expression of ahr-1 in RMED/V neurons. A functional AHR-1::GFP is expressed in the nuclei of a selected group of neurons including RMEL/R (H. Qin and J.A.P.-C., unpublished). We found that the expression of an integrated transgene expressing AHR-1::GFP was unaffected in ceh-10(ct78) and lim-4(ky403) animals(Fig. 7A). We further investigated how ahr-1 might interact with ceh-10 or lim-4 by examining the RME neurons in pair-wise double mutants(Fig. 7B). We found that double mutants of ahr-1(ju145); ceh-10(ct78) showed an additive phenotype such that RMEL/R displayed ectopic nerve processes and RMED frequently failed to express P_unc-25GFP to the same degree as ceh-10(ct78) alone. Double mutants of ahr-1(ju145);lim-4(ky403) exhibited only RMEL/R defects as did ahr-1(ju145)alone. Furthermore, in ceh-10(ct78); lim-4(ky403) double mutants,RMEV was normal, and RMED frequently failed to express P_unc-25GFP as did ceh-10 alone. Thus, these observations are consistent with a conclusion that ahr-1, ceh-10 and lim-4 function independently of each other to control the development or function of the four RME neurons.

AHRs and the partner ARNTs are evolutionarily conserved sequence-specific DNA binding proteins, and are thus far extensively studied for their functions in toxin response in mammals (Hahn,2002). C. elegans ahr-1 is expressed in selected groups of neurons (H. Qin and J.A.P.-C., unpublished), two of which we identify as RMEL and RMER neurons. We show that in ahr-1 loss-of-function mutants, RMEL/R cells extend ectopic axons and fail to express RMEL/R specific markers, although the expression and vesicular uptake of GABA in these neurons are normal. The altered RMEL/R neurons assume a morphology that resembles that of RMED/V neurons, and a RMED/V specific marker P_avr-15GFP is ectopically expressed in RMEL/R cells in ahr-1 mutants. These results led us to conclude that loss of ahr-1 function causes RMEL and RMER neurons to adopt a RMED/V-like fate. Furthermore, we show that the ectopic expression of ahr-1 in RMED/V cells represses neurite extension and induces the expression of a RMEL/R-specific marker,P_lim-6rGFP. These observations indicate that ahr-1 acts as a cell-type-specific factor to control RMEL/R fate specification.

This function of AHR-1 requires its co-factor AHA-1. aha-1 is ubiquitously expressed (Jiang et al.,2001). We find that loss of function in aha-1 causes morphological defects in RMEL/R neurons identical to those in ahr-1mutants, and that the ectopic effect of ahr-1 requires aha-1. An independent study has found that several ahr-1expressing neurons exhibit aberrant cell morphology and cell migration defects in ahr-1 mutants (H. Qin and J.A.P.-C., unpublished data). Thus, it is the cell-type-specific expression of ahr-1 that allows the AHR-1/AHA-1 heterodimer to exert specific effects on neuronal differentiation.

The twenty-six GABAergic neurons in C. elegans fall into five classes based on morphology and function(McIntire et al., 1993). The data presented here, together with previous published reports(Jin et al., 1994; Forrester et al., 1998; Hobert et al., 1999), classify them into additional subtypes and reveals different themes of regulatory network underlying the specification of each subtype. The ventral cord type D motor neurons are related in cell lineage, morphology and function. They express the UNC-30 homeodomain protein(Jin et al., 1994). UNC-30 co-regulates the expression of unc-25 and unc-47, along with other unknown targets, to control multiple differentiation events in the type D neurons (Eastman et al.,1999).

By contrast, the specification of the other seven GABA neurons depends on both cell-specific and common factors. The LIM-6 transcription factor is expressed in five GABA neurons, but has differential effects on their differentiation. In lim-6 mutants, RMEL and RMER exhibit no discernible abnormality, but AVL and DVB display variable defects in neurite outgrowth and GABA expression, but not GABA vesicular packaging(Hobert et al., 1999). AHR-1 acts as a cell-type-specific determinant to control the cellular morphogenesis, but not neurotransmitter expression and packaging, of RMEL and RMER. The function of AHR-1 depends on its ubiquitously expressed cofactor AHA-1. lim-6 is probably a downstream target of ahr-1, but lim-6 is unlikely to mediate all functions of ahr-1 in RMEL/R neurons because neither glr-1 nor GABA expression is altered in lim-6 mutants (Hobert et al.,1999) (this study).

Our observation that ectopic ahr-1 expression can transform RMED/V to a RMEL/R-like fate further suggests that RMED/V cells may express some cellular determinant(s) that represses ahr-1 expression and promotes their fate specification. Although ceh-10 and lim-4 are expressed in RMED and RMEV, respectively(Forrester et al., 1998; Hobert et al., 1999), we find that they do not function by repressing ahr-1 expression in RMED/V. The analysis of double mutants among ahr-1, ceh-10 and lim-4is also consistent with the conclusion that the specification of the four RME neurons involves parallel signaling pathways that may use multiple transcription cascades.

All vertebrate AHRs bind dioxins through their PAS domains, and normally reside in the cytoplasm in a complex with HSP90 and XAP2 chaperonins(Petrulis and Perdew, 2002). Dioxin binding induces nuclear translocation, allowing the formation of AHR/ARNT heterodimer (Hahn,2002). Supporting the role of AHRs in toxin response,AHR-deficient mice are resistant to many of the deleterious effects of AHR-activating pollutants(Fernandez-Salguero et al.,1995; Schmidt et al.,1996; Shimizu et al.,2000; Matikainen et al.,2001). However, these mice also show defects in liver, heart,ovary, vascular and immune systems, and have reduced growth, reproduction and survival, suggesting that AHR has important developmental functions, of which little is understood. In zebrafish embryos, exposure to TCDD increases apoptosis in the dorsal midbrain (Dong et al., 2001). In rat, AHR is expressed in the preoptic area of the brain and co-localizes with GAD67, suggesting that the GABAergic neurons may be the cellular targets of TCDD-induced AHR function(Hays et al., 2002).

The C. elegans AHR-1 and Drosophila AHR Spineless (SS)have a PAS domain with divergent sequences and do not bind dioxin(Powell-Coffman et al., 1998; Butler et al., 2001). In Drosophila ss mutants, the distal antenna is transformed to distal leg, and most of the tarsal region of each leg is deleted(Duncan et al., 1998). SS is expressed in the tissues that are affected. Ectopic expression of sscan induce ectopic antennal structures from some tissues, indicating that ss functions as a tissue-specific factor to control distal antennal identity. Thus, studies in Drosophila and those reported here reveal an evolutionarily conserved function of AHRs as cell-type-specific determinants in animal development.

How might AHR signaling have evolved? At least two AHR-1 partners have been identified. One is ARNT, a ubiquitously expressed nuclear bHLH-PAS protein,and the other is the cytoplasmic chaperonin HSP90(Hahn, 2002). The ARNT knockout mice died early, precluding the analysis of its involvement in AHR developmental function (Kozak et al.,1997). The Drosophila ARNT Tango is required for controlling antennal identity like ss(Emmons et al., 1999). We show that C. elegans AHA-1 is required for AHR-1 in controlling RMEL/R cell identity. Therefore, the AHR and ARNT are evolutionarily conserved functional partners in regulating developmental processes.

Studies in yeast and mammalian cells have implicated two roles of HSP90 in AHR function (Pongratz et al.,1992). HSP90 sequesters AHRs in the cytoplasm in the absence of ligands and also aids the proper folding of AHRs for ligand binding. The in vivo functional involvement of HSP90 in AHR toxin response has not been established. Although C. elegans AHR-1 can bind rabbit HSP90 in vitro(Powell-Coffman et al., 1998),AHR-1 is normally localized to nucleus (H. Qin and J.A.P.-C., unpublished). We find that daf-21/Hsp90 is not required for RMEL/R cell identity, nor is it required for ectopic AHR-1 function, implying that AHR-1 can fold properly in the absence of HSP90.

Drosophila SS and Tango are both normally localized to the nucleus, and the nuclear localization of Tango depends on SS and other bHLH-PAS factors (Ward et al.,1998; Emmons et al.,1999). Similarly, the nuclear localization of C. elegansAHA-1 in intestinal cells is dependent on the co-expression of a bHLH-PAS dimerization partner HIF-1 (Jiang et al.,2001). The studies in invertebrates thus support a conclusion that the subcellular localization of AHRs and ARNT is achieved through co-expression of protein binding partners, not by ligand-induced nuclear translocation (Crews and Fan,1999). Interestingly, expression studies in yeast show that the domains of C. elegans AHR-1 predicted to bind HSP90 and ligand exert a repressive function to inhibit nuclear translocation or transcriptional activation of AHR-1 (Powell-Coffman et al., 1998). Thus, one may speculate that the toxin response of AHRs in vertebrate animals evolved through a recently acquired interaction with HSP90 or other chaperonins, which confers the toxin-inducible feature of the mammalian AHRs.

We thank O. Hobert for otIs37 and otEx406, J. Kaplan for nuIs25, C. Johnson and J. Thomas for daf-21 strains, A. George for clh-6GFP, A. Fire for GFP vectors, H. Qin for pHT101 plasmid, and the C. elegans Genome Consortium for the sequences. We appreciate valuable discussions with O. Hobert, and thank A. Chisholm for comments on the manuscript. Some of the strains used here were obtained from the Caenorhabditis Genetics Center, which is supported by the NIH. Y.J. is an assistant investigator of HHMI. This work was supported by NSF grant 9874456 to J.A.P.-C. and NIH grant NS35546 to Y.J.