Sensory Neuron Fates Are Distinguished by a Transcriptional Switch that Regulates Dendrite Branch Stabilization

SUMMARY

Sensory neurons adopt distinct morphologies and functional modalities to mediate responses to specific stimuli. Transcription factors and their downstream effectors orchestrate this outcome but are incompletely defined. Here, we show that different classes of mechanosensory neurons in C. elegans are distinguished by the combined action of the transcription factors MEC-3, AHR-1, and ZAG-1. Low levels of MEC-3 specify the elaborate branching pattern of PVD nociceptors, whereas high MEC-3 is correlated with the simple morphology of AVM and PVM touch neurons. AHR-1 specifies AVM touch neuron fate by elevating MEC-3 while simultaneously blocking expression of nociceptive genes such as the MEC-3 target, the claudin-like membrane protein HPO-30, that promotes the complex dendritic branching pattern of PVD. ZAG-1 exercises a parallel role to prevent PVM from adopting the PVD fate. The conserved dendritic branching function of the Drosophila AHR-1 homolog, Spineless, argues for similar pathways in mammals.

INTRODUCTION

Mammalian somatosensory neurons adopt specific dendritic architectures in defined layers of the skin to detect diverse stimuli, including touch, temperature, and injurious force (Basbaum et al., 2009; Delmas et al., 2011; Tsunozaki and Bautista, 2009). Simple organisms such as Drosophila and C. elegans also exhibit somatosensory neurons with distinctive topical dendritic arrays and polymodal responses and thus are useful models for elucidating the molecular genetic pathways that drive sensory neuron diversity (Chatzigeorgiou et al., 2010b; Hwang et al., 2007; Smith et al., 2010). Studies in Drosophila have established that specific sensory neuron types are defined by transcriptional mechanisms that regulate cell-intrinsic programs (Grueber et al., 2003; Karim and Moore, 2011; Kim et al., 2006; Parrish et al., 2006). Differential expression of a given transcription factor may distinguish between divergent sensory neuron fates. For example, the transcription factor Hamlet prevents a single-dendrite sensory neuron from adopting the alternative fate of its highly branched sister cell in which Hamlet is not expressed (Moore et al., 2002). Dendritic complexity can also depend on expression of a transcription factor at a specific level; sensory neurons with low levels of the transcription factor, Cut, adopt a simple morphology, whereas neurons showing high Cut expression display more complex dendritic arbors (Grueber et al. 2003; Jinushi-Nakao et al., 2007). Parallel roles for vertebrate Cut homologs in cortical neuron dendritic development argue for the likely conservation of Cut-regulated genes (Cubelos et al., 2010). Together, these results support the idea that precise regulation of both the abundance and cell-specific expression of transcription factors is required to define complex arrays of sensory neurons with unique morphologies and functions. Despite the prominent role of transcription factors in orchestrating sensory neuron differentiation, few downstream targets that contribute to these diverse outcomes are known (Jinushi-Nakao et al., 2007; Sulkowski et al., 2011).

In C. elegans, the LIM homeodomain transcription factor, MEC-3, specifies the fates of two discrete classes of somatosensory neurons with distinct architectures and sensory modalities. Neurons that detect light touch to the body display a simple, unbranched morphology (Chalfie et al., 1985; White et al., 1986). In mec-3 mutants, these touch receptor neurons (TRNs) adopt alternative fates and fail to respond to mechanical stimuli (Way and Chalfie, 1988). MEC-3 is also expressed in a highly branched polymodal nociceptor, the PVD neuron that detects harsh mechanical force and low temperature (Chatzigeorgiou et al., 2010b; Li et al., 2011; Way and Chalfie, 1989). mec-3 mutant PVD neurons fail to elaborate lateral branches and show defective function (Smith et al., 2010; Tsalik et al., 2003) (Husson et al., 2012). These observations raise the interesting question of how a single transcription factor can regulate distinctly different sensory neuron fates (Grueber et al., 2003).

Here, we report that MEC-3 functions in combination with the conserved transcription factors AHR-1 (aryl hydrocarbon receptor) and ZAG-1 (Zn finger/homeodomain protein) to distinguish between the light touch and PVD nociceptive neuron fates. Our results are consistent with a model in which low levels of MEC-3 specify a PVD-like neuron, whereas elevated MEC-3 activates a touch neuron developmental program. The neuronselective effect of this MEC-3 dose-dependent mechanism requires the dual function of AHR-1 in the AVM touch neuron. In the AVM cell, AHR-1 elevates MEC-3 expression as well as blocks downstream MEC-3 targets that result in traits normally reserved for PVD (e.g., lateral branching, sensitivity to low temperatures). Thus, AHR-1 is required for the twinned tasks of inducing the light touch fate while simultaneously preventing expression of nociceptor genes. We show that one of these targets, the claudin-like membrane protein HPO-30, acts in PVD to stabilize lateral dendrites. We hypothesize that HPO-30/claudin maintains PVD dendritic branches by mediating adhesive interactions with the adjacent epidermis. HPO-30 is ectopically expressed in the ahr-1 mutant AVM cell and is required for its PVD-like morphology. We note that this effect is remarkably similar to that of the mutant phenotype for the Drosophila AHR-1 homolog, Spineless, in which simple sensory neurons adopt more complex arbors, although the Spineless targets that effect this outcome are not known (Kim et al., 2006). The strong conservation of this role in dendritic branching suggests that the vertebrate Spineless homolog is likely to exercise a similar function, and thus that the downstream effector molecules that we have identified in C. elegans may also pattern the architecture of mammalian sensory neurons.

RESULTS

Mechanosensory Neurons Adopt Distinct Morphologies and Sensory Modalities

C. elegans responds to physical stimuli through a diverse array of mechanosensory neurons (Chatzigeorgiou et al., 2010b; Geffeney et al., 2011; Chalfie and Sulston, 1981; Hall and Treinin, 2011). Light touch to the body (posterior to pharynx) is mediated by six TRNs (AVM, PVM, PLML, PLMR, ALMR, and ALML), whereas a harsh mechanical stimulus to this region is detected by PVDL and PVDR (Figure 1) (Way and Chalfie, 1989). These neurons occupy unique locations and adopt distinct branching patterns. The touch receptor neurons display a simple morphology with unbranched longitudinal processes emanating from the cell soma. In contrast, the ‘‘harsh-touch’’ PVD neurons are highly branched with elaborate dendritic arbors that envelop the animal in a net-like array (Figure 1) (Halevi et al., 2002; Oren-Suissa et al., 2010; Smith et al., 2010; Tsalik et al., 2003). FLP neurons in the head, which also respond to harsh mechanical force (Chatzigeorgiou and Schafer, 2011), show a similar PVD-like pattern of orthogonal dendritic branches (Albeg et al., 2011; Smith et al., 2010). PVD displays additional sensory responses to temperature and hyperosmolarity (Chatzigeorgiou et al., 2010b) (shown later in Figure 4). The members of these subgroups of mechanosensory neurons are also distinguished by their developmental origins. The touch neurons ALMR, ALML, PLMR, and PLML are generated in the embryo (Sulston et al., 1983). AVM and PVM are each produced during the first larval (L1) stage by unique patterns of cell migration and division of Q-cell progenitors on the left (PVM) and right (AVM) sides of the body (Sulston and Horvitz, 1977). PVDL and PVDR arise from the ectodermal blast cell V5 during the second larval (L2) stage (Figure 1); the highly branched PVD dendritic arbor emerges during later larval (L3 and L4) development (Smith et al., 2010).

Figure 1. Mechanosensory Neurons in the Body Region.

(A–F) PVD neurons marked with F49H12.4::GFP (green) and touch receptor neurons (TRNs) labeled with mec-4::mCherry (red) on the (A–C) left and (D–F) right sides of an adult. Each highly branched PVD dendritic array envelops one side of the body. Unbranched TRNs include ALML, PVM, and PLML on the left side and ALMR, AVM, and PLMR on the right. An additional GFP-labeled neuron (arrow) in the head is shown in (A) and (D). Enlarged views are shown of ALM, PVM, and PVDL on the left sidein(B) and (C) and of PVDR and AVM on the right sidein(E) and (F). Arrows denote occasional GFP labeling of(C) the PDE neuron,(E) anteriorly directed AVM process, and (F) PVD axon.

(G and H) Schematics depict mechanosensory neurons on the (G) left and (H) right sides of the animal.

(I and J) Larval cell lineages are shown that generate (I) PVM and PVDL on the left and (J) AVM and PVDR on the right. Scale bars represent 10 mm.

Figure 4. AHR-1 and ZAG-1 Prevent Light Touch Neurons from Adopting Nociceptive Sensory Modalities.

(A) Assay for AVM-mediated response to light touch to the anterior body.

(B) Ninety-eight percent of wild-type animals show a normal light touch response, whereas a fraction of ahr-1 (ju145) (59%) and zag (rh315) (45%) mutants executed backward locomotion in response to this stimulus (n > 50).

(C and D) Harsh touch evokes calcium transients in (C) cAVM (ahr-1) and in (D) cPVM (zag-1), mimicking that of PVD. Average calcium transients for wild-type (WT) AVM in (C) and PVM in (D) are included for comparison.

(E) Acute exposure to cold temperature (10°C) activates a calcium signal in PVD, cAVM (ahr-1), and cPVM (zag-1) that is not detected in wild-type AVM or PVM.Error bars represent mean ± SEM.

(F) The PVD response to hyperosmolarity (1 M glycerol) is detected in cAVM but not in the wild-type AVM touch neurons. This effect is not statistically significant(p = 0.08) for cPVM versus PVM. Mutants were ahr-1(ju145) and zag-1 (rh315). Numbers labeling histograms (6–25) denote n experimental results. The p valueswere calculated by one-way analysis of variance followed by pairwise comparison using a Bonferroni t test. Error bars represent mean ± SEM.

See also Figure S3.

AHR-1 Prevents the Light Touch AVM Neuron from Adopting a PVD-like Nociceptor Fate

On the basis of a genetic screen for transcription factors that regulate PVD morphology, we initially reported that PVD displays extra dendritic branches in an ahr-1 mutant (Smith et al., 2010). A closer examination of ahr-1(ju145) animals revealed, however, that the additional PVD-like branches actually arise from another cell soma on the right side of the animal that expresses the PVD marker, F49H12.4::GFP (Watson et al., 2008) (Figure 2). A similar result was noted for the ahr-1(ia3) allele (Figure S1 available online). In most cases, this ectopic PVD-like cell is located anterior to the vulva, whereas PVD is positioned in the posterior body. In addition to mimicking the PVD pattern of dendritic branching, the extra PVD-like cell was ectopically labeled with additional green fluorescent protein (GFP) markers (ser2prom3 and egl-46) that are normally expressed in PVD (Table S1) (Tsalik et al., 2003; Wu et al., 2001). The PVD-like cell is unlikely to have arisen from a lineage duplication because we did not observe an additional PDE neuron (marked with dat-1::mCherry), which is normally produced in the cell lineage that gives rise to PVD (Figures 1I and 1J) (Table S1). We therefore considered the alternative possibility that the ectopic PVD-like cell was derived from a cell-fate conversion. The extra PVD-like neuron is located in an anterior lateral region normally occupied by AVM and its lineal sister SDQR (Figure 1). We noted that the light touch neuron-specific marker mec-4::mCherry was expressed in only five cells in ahr-1 mutants (86% of animals), whereas mec-4::mCherry marks all six light touch neurons in the wild-type (Table S1). In a small fraction of ahr-1 mutant animals (∼15%), mec-4::mCherry is expressed in a normal AVM cell, and SDQR adopts a PVD-like morphology (data not shown). These results suggest that AHR-1 function is required in AVM and SDQR and are also consistent with the known expression of AHR-1 in the Q-cell lineage (Qin and Powell-Coffman, 2004). In addition, we have shown that wild-type AVM morphology is restored by transgenic expression of functional AHR-1 protein in these cells (Figure S2). Together, these results suggest that AVM (and occasionally SDQR) is converted into a PVD-like cell in the absence of AHR-1 activity. We therefore refer to the ectopic PVD-like cell as a ‘‘converted AVM’’ cell (cAVM).

Figure 2. AHR-1/Spineless Regulates Mechanosensory Neuron Fate.

(A) The wild-type AVM neuron (arrowhead) displays a single anteriorly directed process in the ventral nerve cord (arrows).

(B–D) In ahr-1 (ju145) animals, cAVM adopts a highly branched dendritic arbor and a posteriorly directed axon, indicated by arrows in (B) and (D). PVDR shows an anteriorly directed ventral cord axon, indicated by arrowheads in (C) and (D).

(E) cAVM initiates lateral branching in the L2 larval stage in ahr-1(ju145) before PVDR (arrow) morphogenesis begins. An additional tail neuron expresses the PVD::GFP marker (arrowhead). See also Figure S1.

Our assignment of the ectopic PVD-like cell to AVM is also consistent with the observation that cAVM shows PVD-like lateral branches in the L2 larvae soon after cAVM is generated in the L1, whereas PVD, which arises in the L2 stage, normally initiates branching later during the L3 larval period (Figure 2E) (Smith et al., 2010). For simultaneous observation of cAVM and PVD, we combined a mosaic PVD::mCherry marker with the integrated PVD::GFP label (see Experimental Procedures). We visualized the individual morphology of each neuron in randomly occurring animals that retain the PVD::mCherry marker in cAVM (mCherry + GFP) but not PVD (GFP only). This analysis confirmed that cAVM retains a PVD-like branching pattern in the adult (Figure 3A)in contrast to the normal AVM morphology of a single process that exits the cell soma, enters the ventral nerve cord, and projects anteriorly to the nerve ring (Figures 1 and 2A). The com- bination of the stable PVD::GFP marker with the mosaic PVD::mCherry label also revealed that cAVM branches rarely overlap with the PVD dendritic arbor, which appeared truncated and usually failed to enter the region occupied by cAVM in ahr-1 mutants (Figures 3A and 3B). In contrast, in wild-type animals, PVD dendrites may touch AVM as they extend anteriorly to envelop the entire body region (Figure 1). PVD branches, however, normally do not overgrow FLP, which shows a comparable dendritic branching pattern in the head (Albeg et al., 2011; Smith et al., 2010). We marked FLP with mec-3::GFP and cAVM with PVD::mcherry to confirm that cAVM and FLP show similar tiling behavior (15/16 animals; data not shown) (Figures 3C and 3D). Dendritic tiling is characteristic of sensory neurons with shared sensory modalities (Jan and Jan, 2010), but the mechanism of this effect is not known (Han et al., 2012). Our results are therefore consistent with a model in which the AVM touch neuron is converted into a harsh touch mechanosensory neuron resembling PVD and FLP in ahr-1 mutant animals.

Figure 3. cAVM Dendrites Tile with PVD and FLP Nociceptive Neurons in ahr-1 Mutant Animals.

(A) Merged image of PVDR (GFP) and cAVM (GFP + mCherry) in ahr-1(ju145) showing tiling (arrows).

(B) Schematic showing tiling between cAVM and PVDR dendrites (arrows) as well as an instance of overlapping branches (arrowhead).

(C) Merged image of cAVM (mCherry) and FLPR (GFP) in ahr-1 (ju145) mutants showing tiling (arrows). Arrowheads denote rare examples of overlapping FLPR and cAVM branches.

(D) Schematic showing tiling between cAVM and FLPR dendrites (arrows). (E and F) Schematics of mechanosensory neurons (right side) in (E) wild-type and (F) ahr-1 mutant backgrounds. Scale bars represent 10 µm.

See also Figure S2.

We noted an additional feature of cAVM morphology that is also indicative of this transformation. In wild-type animals, a single PVD axon turns anteriorly in the ventral nerve cord and terminates before reaching the vulval region (Figure 1D) (Smith et al., 2010; White et al., 1986). In the wild-type, the AVM axon shows a similar downward trajectory but enters the ventral nerve cord anterior to the vulva and projects into the nerve ring in the head (Figures 1 and 2A) (White et al., 1986). In ahr-1 mutants, the PVD axon appears normal (Figures 2C and 2D). However, the cAVM axon now extends posteriorly in the ventral nerve cord and grows toward the region occupied by the PVD axon (Figures 2B and 2D). These results suggest that cAVM has adopted an identity that changes its axonal guidance program to that of PVD. Furthermore, the convergent outgrowth of the cAVM and PVD axons toward a common destination in the ventral nerve cord is suggestive of a potential guidance cue originating from this region. Together, our results suggest that AHR-1 normally functions in the Q-cell lineage to prevent AVM from adopting a PVD-like fate.

cAVM Adopts Sensory Modalities Normally Displayed by PVD Neurons

In the wild-type animal, AVM mediates a characteristic response to ‘‘light touch’’; application of gentle physical stimulus (e.g., with an eyelash) to the anterior body region occupied by AVM evokes a backward locomotory escape response (Figure 4A) (Chalfie and Sulston, 1981). A majority (98%) of wild-type animals crawl backward after light touch to the anterior body, whereas only 59% (p < 0.01) of ahr-1 mutant animals reacted to this stimulus (Figures 4A and 4B). To test the idea that cAVM is specifically defective in light touch, we used a chameleon marker to visualize calcium transients in cAVM (Suzuki et al., 2003). This experiment revealed that cAVM neurons in ahr-1 mutant animals are less likely to respond to light mechanical stimuli than the wild-type AVM neuron (data not shown). Since the cAVM cell in ahr-1 mutants strongly resembles PVD, we next asked if cAVM also adopts PVD-like sensory modalities. We first established that harsh touch elicits a calcium transient in the cAVM cell in ahr-1 mutants similar to that of PVD neurons in wild-type animals (Figure 4C) (Chatzigeorgiou et al., 2010b). cAVM also displayed the normal response of PVD to cold temperature, which was not detected in wild-type AVM (Figure 4E). Last, we determined that 1 M glycerol stimulates PVD activity and that cAVM is also responsive to hyperosmolarity in an ahr-1 mutant, whereas AVM is not (Figure 4F). These data suggest that AHR-1 not only controls AVM morphology and axon guidance but also defines AVM sensory function. We therefore conclude that cAVM neurons are converted to a PVD-like fate in ahr-1 mutant animals.

PVD activation evokes an escape response in which the animal initiates a rapid crawling movement that depends on PVD output to the motor circuit in the ventral nerve cord (Husson et al., 2012). To test cAVM for this function, we used a light-activated Channelrhodopsin-2 (ChR2) for acute stimulation of cAVM (Figure S3). Selective activation of cAVM by this method in an ahr-1 mutant evoked a robust withdrawal response that was not observed in negative control ahr-1 animals that lacked the ChR2 trans-retinal chromophore. These results confirm that the ahr-1 mutant cAVM neuron regulates specific behavior and thus retains the capacity to signal other neurons in the motor circuit. These results also suggest that cAVM has adopted PVD-like morphology and sensory modalities but not the synaptic output of PVD, which preferentially activates interneurons in the forward locomotory circuit (Figure S3) (Husson et al., 2012)

ZAG-1/Zn Finger Homeodomain Protein Prevents the Light Touch PVM Neuron from Adopting a PVD-like Nociceptor Fate

We quantified the percentage of ahr-1 mutant animals with extra PVD-like cells in the anterior versus posterior regions that correspond to the locations of the two postembryonic touch neurons, AVM and PVM. Extra PVD-like cells were never observed in wild-type animals. In contrast, a majority (63%) of ahr-1 mutants show an ectopic PVD-like cell in the anterior region normally occupied by AVM. It is interesting that PVM was also converted to a PVD-like morphology but at a much lower frequency (Table S2). We therefore considered the possibility that AHR-1 functions primarily to specify the AVM cell fate but also exercises a minor parallel role in the PVM progenitor. This idea is substantiated by the finding that a null allele of the AHR-1 cofactor, AHA-1, resulted in a similarly biased transformation of AVM versus PVM to a PVD-like fate (Table S2). Thus, we hypothesized that an additional transcription factor could be primarily required for specifying the PVM cell fate.

PVM is located on the left side of the animal and adjacent to the PVD cell soma (Figure 1). Mutants of zag-1(rh315) (Wacker et al., 2003) showed an extra PVD-like cell in this location (Figure 5; Table S2) (Smith et al., 2010). In addition to displaying the highly branched morphology that is characteristic of PVD, the extra PVD-like cell also expressed multiple PVD markers (Table S1). We considered the possibility that this PVD-like cell could have arisen from duplication of the PVD lineage (Figure 1). However, the absence of an additional dat-1::mcherry-expressing PDE neuron in zag-1(rh315) excludes this model (data not shown). Because the PVD sister cell, V5Rpaapp, normally undergoes programmed cell death (Figure 1), we entertained the alternative idea that this cell survives in the zag-1 mutant and gives rise to a duplicate PVD neuron. This idea is ruled out, however, by the finding that the introduction of an egl-1 mutation to prevent V5Rpaapp apoptosis (Conradt and Horvitz, 1998) results in a third PVD-like cell on the left side in the zag-1; egl-1 double mutant (data not shown). Finally, expression of the light touch neuron-specific marker, mec-4::mCherry, was not detected in this region, therefore suggesting that the normal PVM cell is missing in the zag-1 mutant (Table S1). Based on these results, we conclude that the extra PVD neuron observed in zag-1 mutants arises from the conversion of PVM into a PVD-like cell. We refer to this converted PVM cell in zag-1 mutants as cPVM. Similar results were obtained for zag-1(ok214) and zag-1(zd86) (Clark and Chiu, 2003) (data not shown).

Figure 5. ZAG-1 Blocks Expression of the PVD Nociceptor Fate in the PVM Light Touch Neuron.

(A and B) In (A), the left side of an adult zag-1(rh315) mutant shows two PVD::GFP-labeled neurons in the posterior region with (B) PVD-like dendritic arbors.

(C) cPVM initiates ectopic branching in the L2 larval stage.

(D) PVD::mCherry labels cPVM and PDE (arrow). (E and F) A merged image is shown in (E) of PVDL (GFP) and PDE (arrow) with cPVM (GFP + mCherry) in zag-1(rh315) adult with (F) tracings depicting dendritic fields of PVDL and cPVM.

(G) Schematics of PVDL and PVM in wild-type versus zag-1 depict a representative example of a cPVM neuron occupying the posterior region normally enveloped by PVDL dendritic arbor. Scale bars represent 10 µm.

See also Table S1.

The timing at which cPVM initiates lateral branching is also consistent with the proposal that PVM is converted to a PVD-like fate in zag-1 mutants. PVM normally arises soon after hatching in the wild-type animal (Sulston and Horvitz, 1977) (Figure 1), and cPVM was initially observed in L1 zag-1 mutant animals. Also, as noted earlier for cAVM, the cPVM cell initiated a PVD-like branching pattern in L2 larvae in zag-1 mutants (Figure 5C), whereas the PVD neuron, which first appears in L2 animals, does not display lateral branches until later, in the L3 stage (Smith et al., 2010). We used transgenic animals expressing the mosaic PVD::mCherry marker to distinguish PVD versus cPVM lateral branches in later larval stages and in the adult. Random loss of the mCherry marker from PVD but not cPVM confirmed that the PVD-like branching pattern of the cPVM cell is retained during larval development (Figure 5) (see Experimental Procedures). This analysis also revealed that PVD (marked with PVD::GFP) showed a reduced number of lateral branches in the posterior region occupied by cPVM in the zag-1 mutant (Figure 5). The partial exclusion of PVD branches from the cPVM region is suggestive of tiling activity and is therefore consistent with a model in which ZAG-1 normally functions to prevent PVM from adopting a PVD-like mechanosensory neuron fate.

cPVM Neurons Display PVD-like Nociceptive Responses in zag-1 Mutants

Like AVM, the PVM neuron responds to gentle touch in wild-type animals (Chatzigeorgiou et al., 2010a), although PVM is not required for posterior touch avoidance behavior (Chalfie and Sulston, 1981). We therefore wondered if the zag-1 mutation would convert PVM from a gentle touch neuron to a harsh touch and cold-responsive neuron as previously observed for AVM. We used calcium imaging to confirm that cPVM neurons respond to harsh mechanical stimuli (Figure 4D). cPVM is significantly more responsive to cold shock than the native PVM neuron, which is insensitive to low temperature; comparable calcium transients were observed in the PVD cell in zag-1 mutants and in wild-type PVD cells (Figure 4E). It is interesting that both cPVM and PVD show variable cold-sensitive responses in zag-1 mutants potentially due to incomplete PVD and cPVM branch coverage (Figure 5). Although 1 M glycerol evokes a robust cPVM response, this effect is not significantly different from that of PVM in the wild-type animal (Figure 4F).

AHR-1 and ZAG-1 Function in Parallel Pathways to Specify Touch Neuron Fates

Our results indicate that most PVM neurons (∼95%) are converted into an extra PVD-like cell, cPVM, in zag-1 animals. Close inspection revealed that a smaller fraction (∼23%) of AVM neurons are also transformed into a PVD-like cell in zag-1 mutants (Table S2). This effect could contribute to the partial touch insensitivity of zag-1 mutants (Figure 4B). Because the ahr-1 mutant shows a reciprocal effect in which AVM adopts a PVD-like fate more frequently than PVM, we next asked if AHR-1 and ZAG-1 could function together to define the cell fate of both postembryonic light touch neurons. In zag-1;ahr-1 double mutants, 95% of animals showed conversion of both AVM and PVM into a PVD-like cell (Table S2). These results suggest that AHR-1 is principally required in AVM but also contributes to the PVM touch neuron fate. Conversely, ZAG-1 primarily defines the PVM fate but also functions with AHR-1 to specify AVM. Because our results show that AHR-1 is required in AVM to prevent the adoption of the PVD nociceptor fate, we next asked if AHR-1 interacts with MEC-3, a protein with dual roles in specifying both PVD and touch neuron fates.

AHR-1 Functions with MEC-3 to Specify Light Touch Mechanosensory Neuron Fate

mec-3 encodes a conserved LIM homeodomain transcription factor that is required for normal development of both PVD and light touch mechanosensory neurons (Way and Chalfie, 1988). Lateral branches are not generated in mec-3 mutant PVD neurons (Figure 6C), which suggests that MEC-3 activates a transcriptional cascade that promotes dendritic branching (Smith et al., 2010; Tsalik et al., 2003). Transgenic expression of MEC-3 in PVD restores lateral branching to a mec-3 mutant and therefore confirms the cell-autonomous function of MEC-3 in PVD (Figure S1). Because cAVM adopts a PVD-like morphology in ahr-1 mutant animals (Figure 6E), we wondered if mec-3 was also required for this elaborate dendritic branching pattern. To test this idea, we generated a double mutant of ahr-1;mec-3 and determined that cAVM neurons now resemble the simple, unbranched morphology of mec-3 mutant PVD neurons (Figure 6F). This finding confirms that mec-3 function is necessary for cAVM branching in the ahr-1 mutant. Two alternative models are consistent with this result: (1) AHR-1 normally limits MEC-3 expression in the touch neurons to prevent branching, and (2) AHR-1 functions downstream to block expression of MEC-3-dependent targets that drive the creation of PVD-like branches. To distinguish between these models, we first asked if AHR-1 regulates mec-3.

Figure 6. AHR-1 Regulates MEC-3 to Specify Mechanosensory Neuron Fate.

(A and B) Confocal images depicting PVD branching morphology in (A) wild-type and (B) ahr-1 adults.

(C) PVD neurons lack lateral branches in ahr-1(ju145); mec-3(e1338) animals.

(D) AVM neurons (mec-3::GFP) extend a single ventral process in wild-type (arrow denotes ALML anterior process)

(E) cAVM adopts a PVD-like morphology in ahr-1mutants.

(F) cAVM fails to extend lateral branches in ahr-1;mec-3 double mutants.

(G) mec-3::GFP is strongly expressed in AVM and in other light touch neurons; arrow marks dorsally located ALMR process. In ahr-1 (ju145) mutants, mec-3::GFP shows reduced expression in cAVM but not in ALMR.

(H) Quantification confirms reduced mec-3::GFP expression in cAVM in ahr-1 mutants (n = 54) versus wild-type (WT) (n = 67).

(I) The MEC-3 target, mec-4::mCherry is not detected in cAVM in ahr-1(ju145).

(J) Overexpression of MEC-3 (cAVM::MEC-3) in cAVM promotes expression of mec-4::mCherry. Scale bars represent 10 µm.

(K) Proposed transcriptional regulatory mechanism in AVM light touch neuron. See also Figure S4 and Table S4.

In wild-type animals, mec-3::GFP is normally expressed in the six light touch neurons and in the FLP and PVD neurons (Figure 6G) (Way and Chalfie, 1989). We noted that a mec-3::GFP reporter was strongly expressed in the touch neurons and in FLP but showed a consistently weaker signal in PVD (data not shown). In the ahr-1 mutant, mec-3::GFP expression was substantially reduced in cAVM in comparison to the wild-type AVM neuron (Figures 6G and 6H).We used fluorescence in situ hybridization (FISH) to confirm that mec-3 mRNA is expressed at a lower level in PVD and in cAVM than in wild-type AVM (Figure S4). These findings argued against the idea that AHR-1 inhibits mec-3 expression and favored the alternative possibility that AHR-1 activates mec-3 to specify touch neuron traits. We tested this hypothesis by examining the touch-neuron-specific marker mec-4::mCherry, which normally depends on mec-3 function for expression (Zhang et al., 2002). mec-4::mCherry is rarely detected in cAVM neurons (Figure 6I) but is restored by overexpression of MEC-3 in an ahr-1 mutant (Figure 6J; Table S1). It is also important to note that overexpression of MEC-3 did not prevent the formation of ectopic PVD-like branches or inhibit expression of the PVD-specific marker gene, F49H12.4::GFP in cAVM (Figure 6J). These results are consistent with a model in which MEC-3 must exceed a high threshold to activate expression of light touch neuron genes (e.g., mec-4) but also in which low levels of MEC-3 are sufficient to drive expression of transcripts that specify PVD-like traits (e.g., lateral branching). We therefore considered the hypothesis that AHR-1 negatively regulates PVD-like branching in AVM by inhibiting MEC-3 transcriptional targets (Figure 6K) and set out to identify these downstream genes.

MEC-3-Regulated Target Genes Are Required for Dendritic Branching

We hypothesized that MEC-3-regulated targets in PVD should include genes that promote branching since PVD neurons show a branchless phenotype in mec-3 mutants (Smith et al., 2010; Tsalik et al., 2003). To identify these genes, we used the mRNA tagging method to isolate PVD-specific transcripts from L2 stage larvae immediately prior to the period in which PVD lateral branching is first observed (Smith et al., 2010) (Figure S5). A comparison of wild-type versus mec-3 mutant PVD profiles revealed differentially expressed transcripts (see Experimental Procedures) (Table S4). We focused on the list of 185 downregulated genes in the mec-3 sample because MEC-3 is reported to function as a transcriptional activator (Xue et al., 1992). This analysis revealed several known mec-3-dependent genes (acp-2, des-2, deg-3, mec-7, mec-10, and mec-18) (Treinin et al., 1998; Zhang et al., 2002). Additional targets from this list include extracellular matrix proteins, transcription factors, and cell-surface receptors (Tables S3 and S4). A total of 66 mec-3-dependent transcripts were tested by RNAi to yield 17 hits with PVD branching defects (Table S4). These results were confirmed in mutants for a subset of conserved genes in this group. A mutation in acp-2 (acid phosphatase) results in a modest but significant reduction in PVD lateral branches (Figure S6). acp-2 was previously identified as a mec-3 target gene, but a role in mechanosensitive neuron morphogenesis was not reported (Zhang et al., 2002). The gene T24F1.4 encodes a short peptide (149 amino acids) with homology to tomoregulin, a vertebrate membrane protein that is highly expressed in the brain, where it is suggested to regulate dendrite morphogenesis (Siegel et al., 2002). A deletion mutant of T24F1.4 shows fewer 2° PVD branches (Figure S6) as well as a self-avoidance defect in which 3° branches overgrow one another (Smith et al., 2012) (data not shown). Our screen confirmed that egl-46 is regulated by mec-3 in PVD and promotes lateral branching (Smith et al., 2010). Last, a deletion mutation in the gene hpo-30 (claudin) showed the strongest PVD branching defect in our screen with fewer than half of the wild-type number of 2° branches (see Experimental Procedures). These lateral PVD branches are abnormally short in hpo-30(ok2047) and rarely show the highly stylized arbor that is characteristic of the wild-type PVD neuron (Figure 7A; Figure S6).

Figure 7. MEC-3 Promotes Expression of HPO-30/Claudin to Stabilize PVD Lateral Branches.

(A) PVD lateral branches are truncated and disorganized in hpo-30(ok2047). PVD (green) is marked with PVD::GFP, and the nervous system is labeled with pan-neural::dsRed.

(B and C) hpo-30::GFP expression in (B) wild-type (WT) versus (C) mec-3(e1338) PVD neurons. Scale bars represent 5 µm.

(D) Scatterplot shows reduced average intensity of hpo-30::GFP in mec-3 versus WT PVD neurons (n ≥ 8, p = 0.04). Error bars represent mean ± SEM.

(E) Transgenic expression of wild-type HPO-30 with the PVD promoter, F49H12.4 (PVD::hpo-30g) restores lateral branches to hpo-30(ok2047). Scalebar represents 10 µm.

(F) Quantification of PVD 2° branches in wild-type versus hpo-30 mutant animals, n > 10. Error bars represent mean ± SEM.

(G) PVD expression of C-terminal tagged HPO- 30::GFP fusion protein (PVD::HPO-30::GFP) res cues the Hpo-30 branching defect (F) and shows punctate staining (arrows) in PVD dendrites. Scale bar represents 5 µm.

(H and I) In (H), residual PVDR 2° branches (green) are largely associated with motor neuron commissures (red) (arrows) in hpo-30 mutant animals (scale bar represents 10 µm), whereas in (I) wild-type, the majority of PVD 2° branches do not fasciculate with motor neuron commissures (i.e., ‘‘pioneer’’ PVD 2° branches). ***p < 0.0001, n ≥ 15 animals. In (I), error bars represent mean ± SEM.

See also Figures S5–S8 and Tables S3 and S5.

HPO-30/Claudin Is Regulated by mec-3 and Functions in PVD and FLP to Promote Lateral Branching

hpo-30 encodes a predicted protein with four transmembrane domains and topological similarity to members of the claudin-like family of membrane proteins (Figure 7G). We used a GFP reporter containing a 3 kb region upstream of the hpo-30 coding sequence to assay hpo-30 expression in vivo. This experiment confirmed that phpo-30::GFP is highly expressed in PVD (Figure 7B; Figures S7A and S7C). Our microarray results show that hpo-30 transcript levels are reduced in PVD in mec-3 mutants (Table S4). This observation is consistent with the finding that phpo-30::GFP intensity was significantly lower in PVD in mec-3 mutant animals in comparison to wild-type (Figures 7C and 7D). PVD-specific expression of a genomic clone spanning the wild-type HPO-30 coding region (PVD::hpo-30 g) restored lateral PVD branching in an hpo-30 mutant and thus confirmed that HPO-30 functions cell autonomously in PVD (Figures 7E and 7F). Expression of a chimeric protein in which GFP was fused to the HPO-30 C terminus produced distinct GFP puncta that were largely restricted to dendrites and rarely observed in the PVD axon (Figure 7G). This pattern of localization may reflect the in vivo distribution of native HPO-30 because the HPO-30::GFP protein rescues the Hpo-30 branching defect and is therefore functional (Figure 7F).

In addition to expression in PVD, the hpo-30::GFP reporter was also detected in the FLP neuron and in a subset of additional head and tail neurons and in the ventral nerve cord. This finding is consistent with microarray data that also detected hpo-30 expression in FLP (Topalidou and Chalfie, 2011). hpo-30::GFP was not detected in touch neurons (Figure S7). A mec-3::GFP reporter confirmed that lateral branching is deficient in FLP in an hpo-30 mutant (Figure S7E). In contrast, touch neurons, which also express mec-3::GFP, do not show obvious hpo-30-dependent defects (data not shown). These results suggest that HPO-30 is required for the elaborate pattern of dendritic branching adopted by the PVD and FLP nociceptors but is not necessary for normal touch neuron morphogenesis.

HPO-30/Claudin Is Required for Stabilizing Lateral PVD Dendrites

To understand the mechanism by which hpo-30 regulates dendritic branching, we used time-lapse imaging to visualize dendritic outgrowth. In wild-type animals, 2° dendritic growth is highly dynamic with active extension and retraction of lateral filopodia during the early L3 larval stage when 2° branches are initiated (Smith et al., 2010). hpo-30 mutants show active levels of branch initiation but significantly fewer lateral dendrites in the adult (Figure 7; Figure S8). In the wild-type, each 2° branch adopts an orthogonal trajectory as it extends from the 1° process to grow out along the circumferential axis. Each 2° process then turns at a sublateral nerve cord and gives rise to 3° branches that project along the anterior-posterior axis and sprout 4° processes (Smith et al., 2010). In contrast, in hpo-30 mutants, lateral branches adopt a wide array of angles with respect to the 1° process and rarely reach the sublateral nerve cord (Figure 7A; Figure S8). These observations suggest that hpo-30 is not necessary for PVD lateral branch initiation but may be required for stabilizing nascent 2° dendrites.

We have previously shown that PVD 2° dendrites may either fasciculate with circumferential motor neuron commissures or show pioneer outgrowth along the inner surface of the epidermis (Smith et al., 2010). A mechanism that depends on fasciculation likely predominates on the right side, which contains the majority of motor neuron commissures (Smith et al., 2010; White et al., 1986). This idea is supported by the results of a genetic experiment in which the elimination of GABAergic motor neuron commissures selectively reduces the number of PVD 2° branches on the right side but not on the left (Figure S8). This effect is also consistent with time-lapse movies showing that nascent lateral branches that come into contact with commissures tend to stabilize as 2° dendrites (Figure S8) (Smith et al., 2010). hpo-30 mutants display a striking asymmetric defect in which the majority of PVD lateral branches are restricted to the right side (Figure 7H), and most of these fasciculate with motor neuron commissures (Figure 7I). Thus, hpo-30 appears to function largely in commissure-independent stabilization of lateral branches. This analysis defines two mechanisms of dendrite stabilization, one that requires HPO-30 and is not associated with the commissures and a separate pathway that utilizes a different protein for fasciculation with motor neuron commissures (Smith et al., 2010). HPO-30 is also likely to support higher order PVD branching since the residual 2° branches in hpo-30 mutants do not result in recognizable menorahs with a full complement of 3° and 4° dendrites (Figures 7 and 8). The frequent occurrence of overlapping PVD dendrites in the hpo-30 mutant (Figures 7A and 7H) is suggestive of an additional role in dendrite self-avoidance.

Figure 8. HPO-30/Claudin Is Required for Lateral Branching of Touch Neurons.

(A–C) Representative schematic drawings of PVD::GFP marker in (A) ahr-1(ju145), (B) hpo-30(ok2047), and (C) ahr-1(ju145);hpo-30(ok2047). (D) PVD::GFP marks cAVM and PVDR in ahr-1(ju145);hpo-30(ok2047).

(E and F) In (E), mec-4::mCherry marks an unbranched wild-type ALMR touch neuron, and (F) reveals ectopic lateral branches (arrows) with forced expression of HPO-30 in PLMR from mec-4::hpo-30 g. Scale bar represents 2 µm.

(G) Model of MEC-3-dependent specification of PVD versus TRN fate. MEC-3 is expressed at low levels in PVD to specify the nociceptor fate. AHR-1 elevates MEC-3 expression in AVM above a threshold that activates TRN-specific genes and simultaneously blocks expression of MEC-3-acti-vated PVD-specific (e.g., hpo-30) transcripts.

See also Figure S7 and Table S2.

AHR-1 Blocks Expression of HPO-30 in AVM to Prevent Ectopic Lateral Branching

Because HPO-30 is required for PVD dendritic branching, we hypothesized that HPO-30 is also necessary for branching of the extra PVD-like cell, cAVM, in ahr-1 mutants. This idea was substantiated by the finding that cAVM lateral branches were largely eliminated in ahr-1;hpo-30 double mutants (Figures 8C and 8D). To ask if AHR-1 regulates HPO-30, we visualized hpo-30::GFP in an ahr-1 mutant background and confirmed that hpo-30::GFP is ectopically expressed in cAVM (Figure S7). These results indicate that AHR-1 blocks expression of HPO-30 to prevent touch neurons from adopting the lateral branching architecture of the PVD neuron (Figure 6K). Reduced hpo-30::GFP expression in cAVM in an ahr-1;mec-3 double mutant confirmed that mec-3 function is necessary for ectopic hpo-30::GFP expression in cAVM in an ahr-1 mutant background (data not shown). This effect is also consistent with our finding that mec-3 promotes hpo-30::GFP expression in PVD (Figure 7D). Thus, our results are indicative of a transcriptional mechanism in touch neurons (Figure 6K) in which ahr-1 activates mec-3 while simultaneously blocking expression of hpo-30, a mec-3 target gene that promotes lateral branching.

Having shown that hpo-30 function is required for the PVD-like dendritic morphology of cAVM (Figure 8D), we next asked if hpo-30 expression was sufficient to induce lateral branching in wild-type light touch neurons. Normally, touch neurons adopt a simple, unbranched morphology (Figures 1 and 8). Ectopic expression of HPO-30 in PLM with the mec-4 promoter, however, resulted in the appearance of aberrant lateral branches that are not observed in the wild-type (Figure 8F). AVM and PVM did not show ectopic branches in this experiment, but their longitudinal processes are located in the ventral nerve cord (Figure 1) and thus are not in contact with the epidermal region in which HPO-30 normally promotes PVD branching. These results suggest that HPO-30 is expressed in PVD and FLP, where it contributes to dendritic branching but is excluded from the touch neurons to preserve a characteristically simple, unbranched morphology. However, the shorter length and relative infrequency of these HPO-30-induced ectopic branches, in comparison to the elaborate PVD arbor, are consistent with our finding that HPO-30 is not required for branch initiation in PVD (Figure S8A). In agreement with the idea that additional factors, regulated by mec-3, may promote branching in PVD-like neurons, lateral PVD branches were not restored by forced expression of HPO-30 in mec-3 mutants (data not shown).

DISCUSSION

Sensory neurons display a wide range of morphological motifs and functional modalities that serve to transduce diverse types of external stimuli into specific physiological responses (Delmas et al., 2011). Transcription factors define both the identity and number of each type of sensory neuron and thus are critical determinants of organismal behavior (Jan and Jan, 2010). The downstream pathways that distinguish the architectural and functional properties of different sensory neuron classes are largely unknown, however. Here, we show that the conserved transcription factors MEC-3, AHR-1 and ZAG-1, function together to define distinct sensory neuron fates in C. elegans and identify downstream targets that are necessary for these roles.

A Transcription Factor Code Distinguishes the Fates of Different Classes of Mechanosensory Neurons

The MEC-3 LIM homeodomain protein is expressed in both touch receptor neurons (TRNs) and in PVD (Way and Chalfie, 1989) but is responsible for distinctly different sets of characteristics displayed by these separate classes of mechanosensory neurons. In PVD neurons, MEC-3 promotes the creation of a highly branched dendritic arbor and nociceptive responses to harsh stimuli, whereas in the TRNs, MEC-3 is necessary for light touch sensitivity and for the adoption of a simple, unbranched morphology. Genetic ablation of mec-3 or its upstream regulator, the POU domain protein UNC-86, disrupts the function and morphological differentiation of both of these types of mechanosensory neurons (Husson et al., 2012; Smith et al., 2010; Tsalik et al., 2003; Way and Chalfie, 1989). How are these different MEC-3-dependent traits produced? Our results (Figure 6) suggest that low levels of MEC-3 are sufficient to specify the PVD fate, whereas elevated MEC-3 drives TRN differentiation. The existence of this threshold effect is also supported by the finding that overexpression of MEC-3 induces TRN-specific gene expression in the PVD-like FLP neuron (Topalidou and Chalfie, 2011). This simple model is not sufficient, however, to explain why PVD nociceptor genes, which are turned on by low levels of MEC-3, are actually repressed in the TRNs as MEC-3 expression is elevated. Our findings now provide a mechanism for this effect. In the light touch AVM neuron, AHR-1 elevates MEC-3 expression while simultaneously blocking downstream MEC-3 targets that drive PVD branching and nociceptor function (Figure 6K). We suggest that ZAG-1 may exercise a similar role in PVM (Figure 5). This mechanism is robust because each of these TRNs is effectively transformed into a functional PVD-like neuron when either ahr-1 orzag-1 is genetically eliminated (Figures 2, 4, 5, and S3). Thus, our work has revealed the logic of alternative genetic regulatory pathways in which a single type of transcription factor (e.g., MEC-3) can specify the differentiation of two distinct classes of mechanosensory neurons (Figure 8G). A related mechanism accounts in part for the dose-dependent effects of the homeodomain transcription factor Cut on the branching complexity of larval sensory neurons in Drosophila (Grueber et al., 2003). The transcription factor Knot/Collier is selectively deployed in Type IV da neurons to antagonize expression of Cut targets that produce the dendritic spikes that are characteristic of Type III da neurons. In this case, however, Knot does not regulate Cut expression but functions in a parallel pathway (Jinushi-Nakao et al., 2007). Our finding that the Zinc-finger transcription factor ZAG-1 is required to prevent the PVM touch neuron from adopting a PVD nociceptor fate mirrors the recent observation that genetic ablation of the mammalian ZAG-1 homolog Zfhx1b (Sip1, Zeb2) results in cortical interneurons adopting the fate of striatal GABAerigic cells (McKinsey et al., 2013). Our results are suggestive of a potentially complex regulatory mechanism in which AHR-1 and ZAG-1 inhibit expression of nociceptor genes (e.g., hpo-30) whereas MEC-3 activates transcription of these targets. Additional upstream regulators of mec-3, UNC-86, and ALR-1, are also likely involved in this pathway (Topalidou et al., 2011; Xue et al., 1992).

MEC-3 Regulates Expression of a Claudin-like Membrane Protein that Stabilizes Dendritic Branches

Although transcription factors are well-established determinants of sensory neuron fate, the downstream pathways that they regulate are largely unknown (Jan and Jan, 2010; Jinushi-Nakao et al., 2007; Parrish et al., 2006; Sulkowski et al., 2011). As a solution to this problem for MEC-3, we used a cell-specific profiling strategy (Petersen et al., 2011; Spencer et al., 2011; Von Stetina et al., 2007) to detect mec-3-regulated transcripts in the PVD neuron. We used a combination of RNAi and mutant analysis to identify the subset of targets that affect PVD branching morphogenesis (Figure S6; Tables S3 and S5). Additional experiments with one of these hits, the claudin-like protein HPO-30, revealed a key role in the generation of PVD branches. We note that HPO-30 is expressed in the FLP neuron (Figure S7), where it also mediates the higher order branching morphology shared by FLP and PVD (Smith et al., 2010; Topalidou and Chalfie, 2011) (Figure S7). Time-lapse imaging has revealed that PVD lateral or 2° branches may adopt either of two different modes of outgrowth along the inside surface of the epidermis: (1) fasciculation with existing motor neuron commissures or (2) independent extension as noncommissural or ‘‘pioneer’’ dendrites (Smith et al., 2010). Our results show that the principle role of HPO-30 is to stabilize pioneer 2° branches (Figure 7) and, thus, that additional unknown factors may drive fasciculation with motor neuron commissures (Smith et al., 2010). Because claudins serve as key constituents of junctions between adjacent cells (Simske and Hardin, 2011; Steed et al., 2010; Tsukita and Furuse, 2000), it seems likely that HPO-30 functions in this case to link growing 2° dendrites with the nematode epidermis. We note that an additional mem- brane component, the LRR protein DMA-1, displays a mutant PVD branching phenotype strongly resembling that of Hpo-30 and therefore could also function in this pathway (Liu and Shen, 2012). The intimate association of topical sensory arbors with the skin (Delmas et al., 2011; Han et al., 2012; Kim et al., 2012) and the broad conservation of junctional proteins across species (Labouesse, 2006; Steed et al., 2010) point to the likelihood that homologs of HPO-30/Claudin and similar components could be widely utilized to pattern sensory neuron morphogenesis.

Conservation of Aryl Hydrocarbon Receptor Function in Dendrite Morphogenesis

ahr-1 encodes a member of the bHLH-PAS family of transcription factors and is the nematode homolog of the aryl hydrocarbon receptor (AHR) protein. In mammals, AHR is activated by the xenobiotic compound dioxin to trigger a wide range of pathological effects (Wilson and Safe, 1998). Invertebrate AHR proteins are not activated by dioxin, which suggests that this toxin-binding function represents an evolutionary adaptation unique to vertebrates (Hahn, 2002; Powell-Coffman et al., 1998). An ancestral role for AHR is suggested by AHR mutants in C. elegans and Drosophila that display distinct developmental defects in which a given cell type or tissue adopts an alternative fate (Huang et al., 2004; Struhl, 1982). For example, stochastic expression of the Drosophila AHR homolog, Spineless, promotes the adoption of one specific photoreceptor sensory neuron identity at the expense of another (Wernet et al., 2006). Our results parallel these findings with the demonstration that AHR-1 function is required in C. elegans to distinguish between alternative types of mechanosensory neurons; in ahr-1 mutants, the unbranched light touch neuron, AVM, is transformed into a functional homolog of the highly branched PVD nociceptor. This role for ahr-1 in C. elegans is particularly notable because the AHR-1 homolog, Spineless, also regulates branching complexity in Drosophila. In spineless (Ss) mutants, Class I and II sensory neurons, which normally display simple branching patterns, adopt more complex dendritic arbors (Kim et al., 2006). This phenotype resembles our finding in C. elegans that the simple morphology of the AVM neuron is transformed into the highly branched architecture of the PVD nociceptor in ahr-1 mutants. Ss mutants in Drosophila also show the opposite phenotype of more complex class III and class IV da neurons assuming simpler branching patterns, which could therefore reflect an additional role for spineless in this context of promoting the creation of dendritic branches. On the basis of these results, we suggest that the striking conservation of the shared role of AHR homologs in regulating sensory neuron fate and branching complexity in nematodes and insects argues that this function is evolutionarily ancient and, thus, that the downstream effectors that we have identified in C. elegans may also pattern the dendritic architecture of vertebrate sensory neurons.

EXPERIMENTAL PROCEDURES

Nematode Strains and Genetics

Strains, genetics, molecular biology, and optogenetic methods are described in the Supplemental Experimental Procedures.

Confocal Microscopy

Nematodes were immobilized and imaged as previously described (Smith et al., 2010) (see Supplemental Experimental Procedures). We collected z stacks of cAVM (labeled with F49H12.4::mcherry) and FLP (marked with uIs22 (mec-3::GFP)) in ahr-1(ju145), and FLP and cAVM branches in each focal plane were examined for contact; 15 of 16 animals did not show overlapping FLP/cAVM branches. Time-lapse movies were obtained as described elsewhere (Smith et al., 2010, 2012). Confocal scans were generated from strain NC2440, which carries an F49H12.4::mCherry-marked extrachromosomal array in an ahr-1(ju145); wdIs52 background to obtain differentially labeled cAVM (mCherry + GFP) versus PVD (GFP) in mosaic animals. A similar strategy used strain NC2517 to obtain images of differentially labeled cPVM (mCherry + GFP) versus PVD (GFP).

Microarray Analysis to Identify Candidate mec-3-Dependent Transcripts in PVD

The mRNA tagging method was used (Smith et al., 2010) to isolate PVD-specific transcripts from synchronized populations of L2 stage wild-type (NC1981) and mec-3 mutant (NC2228) transgenic lines (Figure S5). RNA was amplified and hybridized as labeled cDNA to Affymetrix C. elegans tiling arrays (Spencer et al., 2011). Microarray data were quantile normalized, and probe-specific effects were reduced by robust-multichip average, omitting the background adjustment step (Bolstad et al., 2003; Irizarry et al., 2003). PVD-specific transcripts isolated from mec-3 mutant animals were compared to PVD-specific transcripts from wild-type animals. Differentially expressed genes were determined using a linear model and Bayes-moderated t statistic (Smyth, 2004). Transcripts with ≥ 1.5-fold change and ≤ 1% FDR were called differentially expressed. Expression profiles were also generated for the wild-type and mec-3 mutant whole animal RNA samples that were initially generated for the immunoprecipitation step (Figure S5). These reference data sets were used to exclude differentially expressed transcripts arising from the contribution of variations in developmental age or sample preparation to background RNA. In this case, transcripts that were detected as differentially expressed between the wild-type and mec-3 reference samples were removed from the list of significantly different PVD-specific transcripts to produce the final data set of PVD-specific mec-3-regulated transcripts (Table S4).

RNAi Screen for PVD Morphological Defects

We used eri-1(mg366);wdIs52 animals for RNAi screening of candidate mec-3 targets (Earls et al., 2010; Smith et al., 2010). Confocal z stacks were collected for each RNAi clone. We used z projections to count 2° dendrites in each animal. The 2° dendrites were scored as PVD lateral branches that reached the location of either the dorsal or the ventral sublateral nerve cord (Smith et al., 2010). Other defects in PVD development were also noted. A positive hit was defined as any RNAi clone that resulted in PVD defects in more than one animal in at least two replicates. The experimenter was blind to the identity of RNAi clones for all screens.

Calcium Imaging and Nociceptive Modality

Calcium transients generated by harsh touch and cold temperature were measured with optical recordings as previously described (Chatzigeorgiou et al., 2010b) (see Supplemental Experimental Procedures). For the glycerol experiments, animals were placed under the microscope in a perfusion chamber (RC-26GLP, Warner Instruments) under constant flow rate (0.4 ml/min) of ‘‘neuronal buffer’’ (100 mM NaCl, 1 mM MgSO₄, 10 mM HEPES-NaOH [pH 7.1]) using a perfusion pencil (AutoMate). Outflow was regulated using a peristaltic pump (Econo Pump, Biorad). Repellents were delivered with manually controlled valves. Glycerol was dissolved in M13 buffer (30 mM Tris-HCl [pH 7.0] 100 mM NaCl, 10 mM KCl) (Wood, 1988) to a final concentration of 1 M.