Cell-Autonomous and Non-Cell-Autonomous Regulation of a Feeding State-Dependent Chemoreceptor Gene via MEF-2 and bHLH Transcription Factors

Food and feeding-state dependent changes in chemoreceptor gene expression may allow Caenorhabditis elegans to modify their chemosensory behavior, but the mechanisms essential for these expression changes remain poorly characterized. We had previously shown that expression of a feeding state-dependent chemoreceptor gene, srh-234, in the ADL sensory neuron of C. elegans is regulated via the MEF-2 transcription factor. Here, we show that MEF-2 acts together with basic helix-loop-helix (bHLH) transcription factors to regulate srh-234 expression as a function of feeding state. We identify a cis-regulatory MEF2 binding site that is necessary and sufficient for the starvation-induced down regulation of srh-234 expression, while an E-box site known to bind bHLH factors is required to drive srh-234 expression in ADL. We show that HLH-2 (E/Daughterless), HLH-3 and HLH-4 (Achaete-scute homologs) act in ADL neurons to regulate srh-234 expression. We further demonstrate that the expression levels of srh-234 in ADL neurons are regulated remotely by MXL-3 (Max-like 3 homolog) and HLH-30 (TFEB ortholog) acting in the intestine, which is dependent on insulin signaling functioning specifically in ADL neurons. We also show that this intestine-to-neuron feeding-state regulation of srh-234 involves a subset of insulin-like peptides. These results combined suggest that chemoreceptor gene expression is regulated by both cell-autonomous and non-cell-autonomous transcriptional mechanisms mediated by MEF2 and bHLH factors, which may allow animals to fine-tune their chemosensory responses in response to changes in their feeding state.


Introduction
Animals modify their chemosensory behavior depending on their feeding-state, which allows them, for instance, to optimize their food-search strategy [1]. A simple strategy by which animals can rapidly alter their chemosensory behavior is by dynamically changing the gene expression levels of chemoreceptors localized in chemosensory neurons. This form of plasticity in chemoreceptor gene expression is observed across phyla [2][3][4][5][6][7], but how feeding state signals are translated into expression level changes of chemoreceptor genes is poorly understood. Transcriptional regulation of chemoreceptor genes is likely to be the primary mechanism by which certain animals can align their chemoreceptor repertoire and fine-tune their chemosensory responses with changes in their internal nutritional state and external environment such as food availability. Yet, the identity of these transcription factors and their mode of action remain largely unknown.
The nematode Caenorhabditis elegans provides an ideal system to identify transcriptional mechanisms underlying chemoreceptor gene regulation in specific chemosensory neurons, and has provided insight into the cis-and trans-regulatory logic for chemoreceptor gene expression. For example, the transcription factor HMBX-1, a C. elegans homolog of HMBOX1, regulates the srsx-3 chemoreceptor gene [8], which is asymmetrically expressed in AWC sensory neurons [9]. In the AWB sensory neurons, the KIN-29 salt-inducible kinase (SIK) regulates the expression of the str-1 chemoreceptor gene via the MADS-box transcription factor MEF-2 [10]. Moreover, promoter analysis of several AWB expressed chemoreceptors identified a shared bi-partite motif required for AWB specific expression [11]. A systematic analysis of all srh chemoreceptor gene family promoters found that a bi-partite E-box motif was sufficient to direct expression in ADL sensory neurons [12], suggesting that the E-box and its cognate binding proteins the basic Helix-Loop-Helix (bHLH) transcription factors may have important roles in the regulation of ADL-expressed chemoreceptor genes. bHLH transcription factors in C. elegans as well as in other multi-cellular organisms coordinate a number of developmental events in the nervous system, such as cell-specification and the differentiation of neurons. For example, the proneural HLH-2 factor in C. elegans is the sole ortholog of E/Daughterless proteins required for neural development [13][14][15] and distal cell-migration [16], while the Achaete-Scute (As/Sc) family protein, HLH-3, is needed for the differentiation of the hermaphrodite-specific neurons [17]. bHLH factors can form homodimers or heterodimers through interactions with other members of the bHLH family [18,19]. A recent comprehensive study determined the dimerization, spatiotemporal expression and Ebox (CANNTG) binding specificities of nearly all members of the C. elegans bHLH family. Similar to vertebrates, the E/Daughterless protein homolog, HLH-2, in C. elegans can form heterodimers with 14 other bHLH factors, including the As/Sc factor HLH-3 [13,20]. In comparison, the Max-like 3 bHLH protein MXL-3 and TFEB ortholog HLH-30 do not need bHLH partners, but likely target overlapping genes containing the same E-box sequence [20]. Indeed, MXL-3 and HLH-30 play antagonistic roles in the C. elegans intestine to regulate expression of the same lysosomal lipases (lipl) genes [21]. Interestingly, the expression levels of mxl-3 and hlh-30 are highly dependent on C. elegans feeding state conditions [21], suggesting that MXL-3 and HLH-30 act as transcriptional switches to couple an animals' feeding state to the regulation of metabolic gene expression in intestinal cells. Thus, many of the bHLH factors in C. elegans and other organisms have well-characterized roles in processes such as neural development and metabolic gene regulation, which could overshadow their role in orchestrating statedependent regulation of chemoreceptor genes in chemosensory neurons.
In order to better understand how feeding-state signals are translated into expression level changes of chemoreceptor genes in chemosensory neurons, we focused in this study on the candidate chemoreceptor gene, srh-234, which is specifically expressed in ADL sensory neurons of C. elegans. Although the chemical or subset of chemicals detected by this chemoreceptor remains to be determined, our recent work showed that the expression levels of srh-234 are highly dependent on feeding state conditions [22]. In fed animals, srh-234 is strongly expressed in ADL neurons, but when starved, the expression is rapidly down regulated. Multiple pathways acting in ADL sensory neurons including KIN-29 SIK [23] and the DAF-2 insulin-like receptor [24], as well as circuit inputs from the RMG interneuron mediated by the NPR-1 neuropeptide receptor [25,26] regulate expression of srh-234 in ADL [22]. We further showed that the MEF-2 transcription factor is required to reduce srh-234 expression during starvation [22], suggesting that MEF-2 functions in C. elegans as a feeding state-dependent transcriptional regulator for chemoreceptor genes. Interestingly, in mammals, members of the MEF2 family can interact directly with heterodimers formed between myogenic bHLH proteins and E/Daughterless proteins to cooperatively regulate muscle gene expression [27][28][29]. Similarly, the bHLH heterodimer consisting of the neurogenic MASH1 and E/Daughterless protein can cooperatively interact with MEF2 by allowing either type of factor to activate transcription of neuronal genes through the binding site of the other [30]. These observations raise the intriguing possibility that a similar transcriptional module consisting of MEF-2 and certain bHLH factors may operate on chemoreceptor genes in C. elegans.
Here, we show that bHLH transcription factors act together with MEF-2 to regulate expression of the chemoreceptor gene, srh-234, as a function of feeding state. We show that a MEF2 binding site is necessary and sufficient for starvation-induced down regulation of srh-234 expression, while an E-box motif in close proximity of the MEF2 site is required to direct srh-234 expression in ADL neurons. A systematic analysis of bHLH factors identified multiple bHLH genes as regulators of srh-234 expression levels. We show that mutations in hlh-2, hlh-3 and mxl-3 and hlh-4(RNAi) treated animals strongly reduce srh-234 expression in fed conditions, while mutations in hlh-10 weakly reduce srh-234 expression. In addition, we show that hlh-30 mutants similar to mef-2 suppress the starvation-induced down regulation of srh-234. We show that HLH-2, HLH-3, and HLH-4 are transiently expressed in, and act in ADL neurons to regulate srh-234 expression, whereas the expression levels of srh-234 in ADL is remotely regulated by MXL-3 and HLH-30 acting in the intestine. This non-cell-autonomous regulation of srh-234 expression mediated by MXL-3/HLH-30 requires DAF-2/DAF-16 insulin signaling in ADL. We also show that a subset of insulin-like peptides (ILPs) in the intestine, as well is in other neurons, are required to modulate srh-234 expression. Taken together, these findings suggest that MEF-2 and bHLH transcriptional factors link feeding state conditions via insulin signaling to the regulation of chemoreceptor genes in chemosensory neurons.

MEF2 and E-box binding sites regulate expression of the statedependent srh-234 chemoreceptor
We previously showed that the expression levels of a candidate chemoreceptor gene, srh-234, in the ADL sensory neuron type is dynamically modulated by the feeding state of C. elegans [22]. gfp expression driven by only 165 bp upstream sequence of the srh-234 transcriptional start site (srh-234p::gfp) is strongly expressed in ADL neurons in fed animals, but when animals were starved for long-periods of time (>6 hours), srh-234p::gfp was significantly down regulated [22]. This strongly suggests that the feeding state-dependent regulation of srh-234 occurs at the level of transcription.
To identify candidate transcription factor(s), we examined the 165 bp minimal promoter of srh-234 using Lasagna Search 2.0 software [31], and found a predicted MEF2 factor binding site (AGTTATATTTAA, -88 bp) [32,33] and a E-box motif (CACCTG, -52 bp) known to bind bHLH transcription factors [19] in close proximity of each other (Fig 1A). To directly test whether the identified MEF2 and E-box sequence motifs are important for srh-234 regulation, we generated gfp-reporter fusions of srh-234 promoters that carry mutations in the E-box and/ or MEF2 sites and monitored gfp expression in transgenic animals in fed and starved conditions. Mutating the E-box core of srh-234 (-E-box; CACCTG to CTGCAG), completely abolished expression levels of srh-234p::gfp in ADL in both fed and starved conditions (Fig 1B and  1C), suggesting that the srh-234 E-box motif is necessary to drive expression in ADL neurons. Considering that bHLH factors can bind as homo-or heterodimers to E-box motifs in C. elegans [18,19] with different bHLH family members having different preferences for the two central base pairs of the E-box [34], we also mutated either the left (-E-box L; CAC to TTC) or right (-E-box R; CTG to CAA) half-site of the E-box identified in the srh-234 promoter. Similar to mutating the E-box core, mutating both the left (-E-box L) and right E-box half site (-Ebox R) abolished expression levels of srh-234p::gfp in ADL neurons ( Fig 1C). Thus, each Ebox half-site is required for srh-234 expression in ADL neurons.
To further examine whether the srh-234 E-box motif is sufficient for expression in ADL neurons, we inserted the entire E-box sequence motif of srh-234 with a 3-bp flanking sequence (AATCACCTGTCC) into the upstream sequence of the AWA-expressed odr-10 chemoreceptor gene [35]. Animals carrying an odr-10p(+E-box)::gfp transgene showed expression of odr-10 in ADL neurons in addition to expression in AWA neurons, albeit weaker than AWA neurons (S1 Fig). It is possible that the relevant location of the E-box insert or additional DNA flanking of the E-box site is important to properly express odr-10 in ADL. Nevertheless, our finding is consistent with previous work showing that the E-box appears in ADL-expressed genes and functions as a potential ADL enhancer element [12]. Together, these results suggest that the srh-234 Ebox motif is both necessary and able to direct srh-234 expression in ADL neurons.
We next mutated the MEF2 site alone or together with the E-box motif (-E-box/MEF2), and again monitored srh-234p::gfp expression levels in both fed and starved conditions. Consistent with mef-2 mutants partially rescuing the starvation-induced reduction of srh-234p::gfp expression [22], we found that mutating the core of the MEF2 site (-MEF2; AGTTATATTTAA to AGTCGACTTTAA; Fig 1A) in the srh-234 promoter resulted in increased srh-234p::gfp expression levels during starvation, while in fed conditions the srh-234p::gfp expression levels are not significantly reduced (Fig 1B and 1C). Mutating the E-box and MEF2 motifs together (-E-box/MEF2) again abolished srh-234p::gfp expression similar to mutating the E-box core alone (-E-box; Fig 1C). These results suggest that the identified MEF2 binding site in the 165 bp upstream sequence of srh-234 is required to repress, but not to promote srh-234 expression in ADL during starved conditions. To further examine whether the MEF-2 site is sufficient for starvation-induced down regulation of srh-234 expression, we inserted the MEF2 site into the upstream regulatory sequence of the ADL-expressed chemoreceptor, sre-1, which is not regulated by feeding state conditions [22]. We generated reporters of the sre-1 promoter fused to gfp either with or without the srh-234 MEF2 site inserted upstream in close proximity of the predicted E-box of sre-1 (CACCTG, -69 bp ; Fig 2A), and examined these transgenic animals in both fed and starved conditions. Of note, the sre-1 promoter does not contain a predicted MEF2 binding site. We found that animals carrying the sre-1p(+MEF2)::gfp transgene reduced sre-1 expression levels in ADL in starved conditions compared to wild-type animals (Fig 2B and 2C). This starvation modulation is dependent on MEF-2 as mef-2 loss-of-function mutants carrying the sre-1p(+MEF2)::gfp transgene increased sre-1 expression in starved ADL neurons ( Fig 2C). Taken together, these results suggest that MEF-2 function and its candidate MEF2 binding site represses srh-234 expression when starved, while the predicted E-box motif directs expression of srh-234 expression in ADL.

bHLH factors regulate the expression of srh-234
Transcription factors of the bHLH protein family recognize DNA binding sites containing the E-box motif (CANNTG) [36]. The C. elegans genome contains at least 42 bHLH factors [37] that can bind E-box and/or E-box-like sequences as obligatory homo-or heterodimers with different DNA binding specificity [19,20]. To identify bHLH factors that regulate srh-234 expression in ADL neurons, we examined srh-234p::gfp expression levels in existing mutant animals or via RNAi-mediated interference (RNAi) of bHLH genes in both fed and starved conditions. In particular, we focused on bHLH factors that are known or predicted to bind the E-box identified in the srh-234 promoter. Of the 18 bHLH genes examined, we found that hlh-2, hlh-3 and mxl-3 mutants strongly reduced but did not abolish srh-234p::gfp expression levels in ADL during well-fed conditions, while hlh-6 and hlh-10 mutants and hlh-4(RNAi) treated animals reduced srh-234p::gfp expression (Figs 3A, 3C and S2). We further confirmed the reduced srh-234 expression levels observed in mxl-3 mutants by qRT-PCR (S2 Fig). Since we previously showed that ADL sensory inputs are required for srh-234 expression [22], it is possible that the reduced srh-234p::gfp expression levels in ADL in these bHLH mutants results from altered sensory cilia of ADL neurons. However, with the exception of hlh-6 mutants (30% of animals dye-fill ADL, n = 20), ADL neurons in hlh-2, hlh-3, hlh-10, and mxl-3 mutants, as well as in hlh-4(RNAi) treated animals show wild-type dye-filling (S2 Fig), ruling out a cilia defect of ADL neurons in these mutants. Moreover, as a measure of functional integrity of ADL neurons, hlh-2 and hlh-3 mutants show normal avoidance responses to known chemicals sensed by ADL (e.g. CuCl 2 , glycerol, SDS and octanol) when compared to wild-type (S3 Fig).
The E/Daughterless bHLH ortholog HLH-2 is essential for early neural development [13], and can form heterodimers with 14 other bHLH factors, including HLH-3, HLH-4 and HLH-10 [20], through an E-box sequence motif (CACCTG) identical to the identified E-box motif in the srh-234 cis-regulatory region. However, although mutations in the srh-234 E-box motif fully abolished expression of srh-234p::gfp in ADL neurons (Fig 1B and 1C), animals homozygous for hlh-2(tm1768), a hypomorphic allele that is fully sterile at 25°C and partially sterile at 20°C [16], or hlh-3(ot354) null mutants alone, did not abolish expression of srh-234p::gfp images are lateral views of ADL sensory neurons with the arrow pointing to the cell body; anterior is left. Images were acquired at the same exposure time at room temperature. (C) Relative expression of mutated srh-234 regulatory sequences in the ADL cell body compared to wild-type adults when fed or starved. Data shown is the average of least three independent transgenic lines for each genotype with n>25 adult animals for each transgenic line. * indicates values that are different from that of wild-type adult animals at P<0.001. Error bars denote the SEM.
We next asked whether HLH-2 and HLH-3 regulate additional ADL-expressed chemoreceptor genes. The food-independent sre-1 promoter is specifically expressed in ADL and contains a predicted E-box identical to the srh-234 E-box (CACCTG), but its promoter sequence does not seem to contain a candidate MEF2 binding site. We found that hlh-2 (tm1768) and hlh-3(ot354) singles as well as hlh-2(tm1768);hlh-3(ot354) double mutants do not reduce the expression levels of sre-1p::gfp (S4 Fig), suggesting that HLH-2 and HLH-3 do not regulate sre-1 expression, even though both the cis-regulatory regions of srh-234 and sre-1 contain the same E-boxes. It is possible that DNA sequences flanking the E-box [39,40], or the MEF2 binding site known to function in the recruitment of bHLH factors to E-box-dependent genes [28,30], play an important role in binding specificity of bHLH factors. When we examined expression of sre-1p(+MEF2)::gfp in either hlh-2 or hlh-3 mutants, the expression in ADL was not reduced (S4 Fig). These results suggest that the introduction of the identified MEF2 site into the sre-1 promoter sequence does not confer HLH-2 and HLH-3-mediated regulation. However, when we examined expression of odr-10p(+srh-234 E-box)::gfp (containing the srh-234 E-box plus direct flanking DNA sequences; S1 Fig) in ADL in hlh-2 mutants, we find that ADL expression is reduced, albeit weakly, suggesting that the srh-234 E-box and direct DNA flanking sequence is likely important for HLH-2-mediated regulation.
In addition to feeding conditions, we examined whether each of these bHLH factors also regulate srh-234p::gfp expression levels in starved conditions. We found that hlh-30 mutants, and to a lesser extent hlh-34 mutants, suppressed the starvation-induced down regulation of srh-234p::gfp similar to mef-2 mutants, while in fed conditions hlh-30 and hlh-34 mutants resulted in increased expression levels of srh-234p::gfp (Figs 3B, 3C and S2). Interestingly, hlh-30 is known to regulate the expression of the same target genes as mxl-3 in an antagonistic manner [21]. Our results suggest a similar antagonistic role for mxl-3 and hlh-30 in regulating gene expression such that mxl-3 mutants reduce srh-234p::gfp expression when fed, while hlh-30 mutants increase srh-234p::gfp expression when starved. In contrast, mxl-3 and hlh-30 mutants do not alter expression levels of the food-independent chemoreceptor gene, sre-1 (S4 Fig), consistent with the model that HLH-30 and MXL-3 are required to regulate srh-234 expression levels as a function of feeding state. Taken together, these results suggest that HLH-2 may partner with HLH-3 and HLH-4 to promote srh-234 expression in ADL via the E-box, while MXL-3 and HLH-30 are necessary for state-dependent regulation of srh-234.

MEF-2-mediated regulation of srh-234 is dependent on bHLH function
bHLH proteins can interact with members of the MEF2 family to cooperatively regulate gene expression in muscle and neuronal cells [27][28][29][30]. Having implicated a candidate MEF2 binding site as well as loss of MEF-2 function in the starvation-induced down regulation of srh-234 expression in ADL [22], we examined the genetic interactions between mef-2 and bHLH genes on the regulation of srh-234. When we combined loss-of-function mutations in mef-2 with hlh-3, the double mutants showed a reduced srh-234p::gfp expression phenotype in fed conditions similar to single hlh-3 mutants (Fig 3C), suggesting that hlh-3 acts downstream of, or in parallel to, mef-2 to regulate srh-234 expression. However, when we combined mutations in mef-2 with mxl-3, double mutants exhibited expression levels of srh-234p::gfp similar to mef-2 single Representative expression of ADL sensory neurons driven by wild-type srh-234 regulatory sequences fused to gfp (srh-234p::gfp) in adult wild-type and mutant animals of the indicated bHLH genes when either well-fed (A) in the presence of E. coli OP50 food, or starved (B) in the absence of food for 12 hours. Confocal images are lateral views of ADL sensory neurons with the arrow pointing to the cell body; anterior is left. Images were acquired at the same exposure time at room temperature. (C) Relative expression of srh-234p::gfp in adult single and double mutants of the indicated genotypes in well-fed and starved conditions. Adult animals (n>20) for each genotype were examined at the same exposure time. * and ** indicates values that are different from that of wild type adult animals at P<0.001 and P<0.01, respectively. # and ## between the genotypes compared by brackets at P<0.001 and P<0.01, respectively. Error bars denote the SEM. ns indicates not significant. All strains contain stably integrated copies of a srh-234p::gfp transgene (oyIs56).

HLH-2/3/4 act in ADL neurons to regulate srh-234 expression
Heterodimers of HLH-2/HLH-3 and HLH-2/HLH-4 are known to physically bind an Ebox sequence motif (CACCTG) [20], which is identical to the E-box identified in the 165 bp cis-regulatory region of srh-234. This suggests that HLH-2 and its partners HLH-3 and HLH-4 might directly modulate expression levels of srh-234 in ADL neurons. We therefore determined whether HLH-2 function in ADL was sufficient for srh-234 regulation by selectively rescuing the expression of hlh-2 in ADL in a hlh-2 mutant background, and measured its effect on srh-234 expression in fed conditions. For this, we used the promoter of svh-1 encoding a growth factor whose ADL-specific expression [41] is not reduced in hypomorphic hlh-2(tm1768) mutants (S5 Fig). We found that specific expression of hlh-2 in ADL (ADL::hlh-2) fully restored the reduced srh-234p::gfp expression phenotype of hlh-2(tm1768) mutants, whereas expression in the intestine (intestine::hlh-2) did not result in a rescue ( Fig 4A). These results suggest that HLH-2 function is necessary in ADL to regulate srh-234 expression.
Next, because we showed that HLH-2 and its partners HLH-3 and HLH-4 act in ADL to regulate srh-234 expression, and because these bHLH factors are co-expressed in head neurons and other tissues in a temporal manner [20], we performed co-expression analyses to determine when they are expressed in ADL. We therefore created animals carrying either hlh-2p:: mCherry::his-11, hlh-4p::gfp, or a rescuing hlh-3p::hlh-3::SL2::gfp construct together with an ADL-specific reporter (srh-234p::gfp or sri-51p::mCherry) (see Material and Methods), and observed that hlh-2, hlh-3, and hlh-4 are all expressed in ADL neurons but in a temporal specific manner (Figs 4C, 4D and S6). While hlh-2 is expressed in ADL throughout development from the embryo to the adult consistent with previous reports [13], we find that both hlh-3 and hlh-4 are transiently expressed in ADL (Fig 4C, 4D and 4E). Expression of hlh-3 in ADL is observed in the late embryo before hatching, but not in L2 larvae or in adults (Figs 4D and S6). However, it remains possible that hlh-3 is very lowly expressed in ADL. In contrast, we observed that hlh-4 is expressed in ADL neurons in L1 larvae and in adults, but we did not observe hlh-4 expression in ADL in late-staged embryos before hatching (Figs 4D and S6). Since HLH-4 may share target genes with HLH-10 in the same cell(s) as previously proposed [20], and because mutations in hlh-10 reduce srh-234p::gfp expression in fed conditions, albeit MXL-3 and HLH-30 act in the intestine to remotely regulate srh-234 expression in ADL Our systematic analysis of bHLH loss-of-function mutants suggests that MXL-3 and HLH-30 regulate srh-234 expression as a function of feeding state. We next determined the site(s) of action of MXL-3 and HLH-30 for regulation of srh-234. Expression of mxl-3 or hlh-30 under control of its endogenous promoter (mxl-3::mxl-3 or hlh-30::hlh-30) fully rescued the reduced srh-234 expression phenotype of mxl-3 and hlh-30 mutants as a function of feeding state (Fig 5A and 5B). However, when we selectively expressed mxl-3 or hlh-30 in ADL using the sre-1 promoter (ADL::mxl-3 or ADL::hlh-30), we did not observe any rescue of the srh-234 expression phenotypes of mxl-3 and hlh-30 mutants, respectively (Fig 5A and 5B). This lack of rescue could not be attributed to MXL-3 or HLH-30 regulating sre-1, because mutations in mxl-3 and hlh-30 did not reduce expression levels of sre-1p::gfp in ADL (S4 Fig). Moreover, selective down regulation of mxl-3 in ADL by feeding RNAi also did not change the srh-234 expression levels (S7 Fig). Thus, MXL-3 and HLH-30 are required in cells/tissues other than ADL neurons to regulate srh-234 expression.
Since both MXL-3 and HLH-30 act in the intestine to regulate transcription of lysosomal lipase (lipl) genes in response to starvation [21], we sought to determine whether MXL-3 and HLH-30 may act in the intestine to remotely regulate srh-234 expression in ADL. Indeed, selective expression of mxl-3 in intestinal cells using the ges-1 promoter (intestine::mxl-3), as well as expression of hlh-30 cDNA in the intestine (intestine::hlh-30) partially rescued the srh-234p:: gfp expression phenotype of mxl-3 and hlh-30 mutants in fed and starved conditions, respectively (Fig 5A and 5B). Consistent with these findings, tissue-specific inactivation of mxl-3 in the intestine using a promoter-driven hairpin RNAi strategy (Fig 5A) [42], or feeding mxl-3 (RNAi) in sid-1(pk3321); srh-234p::gfp mutant animals carrying a intestine::sid-1 transgene reduced srh-234p::gfp expression levels in ADL (S7 Fig). However, expression of mxl-3 or hlh-30 under control of its endogenous promoter (mxl-3::mxl-3 or hlh-30::hlh-3) fully rescued the srh-234 expression phenotype of mxl-3 and hlh-30 mutants in fed and starved conditions, respectively ( Fig 5A and 5B), suggesting that additional mxl-3-and hlh-30-expressed cells and/or tissues may contribute to srh-234 regulation. Because MXL-3 is expressed in the AWC sensory neuron [21], which is a key neuron for dauer regulation as a function of feeding state [43], we tested whether mxl-3 expression in AWC is required to regulate srh-234. However, expression of mxl-3 specifically in the AWC neurons under control of the str-148 promoter (AWC::mxl-3) did not restore the mxl-3-mediated reduction of srh-234 expression in fed conditions (Fig 5A). Collectively, these results support the model that both MXL-3 and HLH-30 function is required in the intestine, at least in part, to non-cell-autonomously regulate srh-234 expression in ADL.

MXL-3 and HLH-30-mediated regulation of srh-234 is dependent on insulin signaling
We next asked whether certain nutritional signaling pathways are required for this MXL3-and HLH-30-mediated regulation of srh-234 expression in ADL. As insulin signaling is coupled to the feeding and nutritional status of many animals including C. elegans [44], and because we previously showed that the function of the DAF-2 insulin-like receptor and its downstream target DAF-16 FOXO transcription factor are specifically required in ADL to regulate srh-234 expression [22], we asked whether MXL-3/HLH-30-mediated regulation of srh-234 in ADL is dependent on insulin signaling. Consistent with our previous findings [22], srh-234 expression is strongly reduced in well-fed daf-2 mutants (Fig 5D). When we combined mutations in daf-2 or daf-16 with mxl-3, the double mutants exhibited a srh-234p::gfp expression phenotype in fed conditions similar to daf-2 and daf-16 single mutants, respectively (Fig 5C and 5D), suggesting that the daf-2/daf-16 insulin signaling pathway acts genetically downstream of, or in parallel to mxl-3, (C) Selective expression of the daf-2 insulin-like receptor in ADL suppresses the reduced srh-234 phenotype of adult mxl-3; daf-2 double mutants. Shown is the relative expression of srh-234p::gfp in adult single and double mutants of mxl-3 and daf-2 compared to wild-type adults during well-fed conditions. daf-2 (e1307) is a temperature sensitive allele. For these experiments, animals were raised at 15°C (permissive temperature) and shifted to the 25°C (restrictive temperature) as L4-staged larvae. (D) Mutations in daf-2 and daf-16 suppress the srh-234 expression phenotypes of hlh-30 and mxl-3 mutants, respectively, whereas hlh-2 suppresses daf-16 in regulating the expression of srh-234. Shown is the relative expression of srh-234p::gfp in the ADL cell body in adult single and double mutants of hlh-30, daf-2 or daf-16 compared to wild-type adults in well-fed and starved conditions with >25 adult animals for each genotype. (A-D) data shown is the average of at least two independent transgenic lines with n>25 adult animals for each line. * and ** indicates values that are different from that of wild-type adult animals at P<0.001 and P<0.01, respectively, and # and ## between the genotypes compared by brackets at P<0.01 and P<0.05, respectively. Error bars denote the SEM. (E, F) Quantification (E) and representative images (F) of subcellular localization of DAF-16::GFP in ADL neurons of adult wild-type and mxl-3 mutants in either fed or starved conditions. n>20 adult animals for each genotype and condition. * indicates values that are different from fed wild-type distribution at P<0.001 as determined by the χ 2 test.
To further examine the connection between MXL-3 function in the intestine and the transcriptional pathways that act in ADL neurons (i.e. MEF-2, DAF-16 and HLH-2) to regulate srh-234 [22], we first asked whether expression levels of mef-2 are changed in response to starvation and to changes in MXL-3 and DAF-2 signaling. Interestingly, we found that both mxl-3 and daf-2 mutants as well as prolonged starvation increases the expression of a previously reported translational MEF-2::GFP fusion transgene [10] in many head neurons (>2.7x fold increase in mef-2p::mef-2::gfp expression in starved wild-type, mxl-3 and daf-2 mutants compared to fed wild-type animals; S8 Fig). No detectable expression changes are observed for hlh-2 between fed and starved wild-type animals, as well as between wild-type and daf-2 or daf-16 mutants (S8 Fig). Thus, MEF-2 function is dependent on starvation and DAF-2 and MXL-3 signaling, consistent with our previous data showing that MEF-2 may act as a state-dependent transcriptional regulator of srh-234 expression.
We next asked whether MXL-3 function changes the subcellular localization of DAF-16 to regulate srh-234 expression. Previous work has shown that DAF-16::GFP shuttles between the cytoplasm and the nucleus in response to changes in environmental conditions, such as starvation and heat shock [45]. Moreover, reduced daf-2 insulin signaling results in the nuclear localization of DAF-16::GFP [46]. If mutations in mxl-3 reduce srh-234 expression in fed conditions due to reduced daf-2 signaling, we would predict that DAF-16::GFP subcellular localization in ADL neurons is similarly affected in mxl-3 mutants by shuttling DAF-16::GFP into the nucleus of ADL. We therefore generated transgenic animals that express a DAF-16::GFP fusion protein controlled by the sre-1 promoter (ADL::daf-16::gfp) in a hlh-2p::mCherry::his transgenic background. In these transgenic adult animals, DAF-16::GFP localizes specifically to ADL neurons with mCherry fluorescence in the nucleus of ADL. In fed conditions, we found that DAF-16::GFP is mostly present in the cytoplasm in wild-type animals (Fig 5E and 5F). However, after prolonged starvation, DAF-16::GFP shuttles (at least in part) to the nucleus (Fig 5E), suggesting that the DAF-16::GFP transgene appears to be functional. When we examined this transgene in a mxl-3 mutant background, the DAF-16::GFP becomes similarly localized in the nucleus as in starved conditions (Fig 5E and 5F), although not all of the DAF-16::GFP fusion proteins are fully excluded from either the cytoplasmic or nuclear compartment, possibly due to overexpression of the DAF-16::GFP transgene in ADL neurons. These results suggest that starvation regulates DAF-16 subcellular localization in ADL, and that MXL-3 function may be important for subcellular localization of DAF-16 in ADL. Taken together, these results support the model that MXL-3 regulates both MEF-2 and DAF-16-mediated mechanisms to regulate srh-234 expression.

Insulin-like peptides regulate srh-234 expression, which is dependent on MXL-3 function
To further test the model that MXL-3 function in the intestine acts through insulin signaling to regulate srh-234 expression in ADL neurons, we sought to identify insulin-like peptides (ILPs) that may target the DAF-2 insulin-like receptor acting in ADL. Since a semi-dominant mutation (sa191) in daf-28 that is suggested to block other agonistic ILPs for DAF-2 [47] reduces srh-234 expression in fed conditions [22], we first tested whether DAF-28 itself may be required for srh-234 regulation. However, daf-28 null mutants do not appear to alter srh-234 expression in fed or starved conditions (S9 Fig), suggesting that other ILPs are required to regulate srh-234 expression in ADL. To identify these ILPs, we decided again to use the intestinespecific knockdown RNAi approach in the sid-1 mutant background [38] by feeding RNAi directed against individual ILP genes (S9 Fig). Of the 29 ILP genes examined, we found that selective down regulation of ins-3(RNAi) in the intestine reduced srh-234 expression in ADL, whereas ins-7(RNAi), ins-21(RNAi), and ins-28(RNAi) increased srh-234 expression (Figs 6A, 6B and S9). Thus, selective down regulation of a subset of ILP genes in the intestine alters the expression of srh-234 in ADL. Since ins-3 and ins-7 expression in the intestine is dependent on feeding state conditions [48,49], and because mxl-3 transcription is down regulated by starvation [21], we next asked whether the expression of these ILP genes are regulated by mxl-3. We focused on the ins-3 ILP because the promoter sequence of the ins-3 gene contains candidate MXL-3 and HLH-30 sites as determined by Lasagna 2.0 software (S1 Table) [31]. Interestingly, we found that intestinal expression but not neuronal expression of ins-3p::gfp is reduced, albeit weakly, in mxl-3(ok1947) mutants (S10 Fig) (1.8x fold decrease of ins-3p::gfp in the intestine in mxl-3 compared to wild-type), suggesting that MXL-3 may directly regulate ins-3 expression in the intestine. No detectable expression differences between wild-type and mxl-3(ok1947 mutants were observed for ins-4p::gfp and ins-5p::gfp reporters (S10 Fig). Of note, selective down regulation of ins-4 and ins-5 in the intestine by RNAi do not appear to alter srh-234 We further show that the expression of srh-234p::gfp in ADL neurons is increased in lossof-function mutants of ins-26 ILP in both fed and starved conditions when compared to wildtype animals (S9 Fig), while selective down regulation of ins-26 by RNAi in the intestine does not affect srh-234p::gfp expression in ADL (S9 Fig). The ins-26 ILP is known to be expressed in ASE, ASI and AWC sensory neurons in addition to the intestine and other cells, but is not known to be expressed in ADL [48]. Recent work indicated that the expression of ins-26 in AWC neurons is dependent on food availability; ins-26 expression in AWC is reduced upon starvation [43]. When we examined ins-26p::gfp expression in a mxl-3 mutant background in fed conditions, we observed that ins-26 expression was strongly increased in cells tentatively identified as ASE neurons (and possibly ASI) (S10 Fig). In wild-type animals, this ins-26p::gfp transgene is weakly expressed in head neurons [50]. Interestingly, intestine-specific expression of mxl-3 (intestine::mxl-3) could rescue this observed increase of ins-26p::gfp expression in ASE in mxl-3 mutants back to near wild-type levels (only 3% animals carrying the intestine::mxl-3 transgene show strong expression of ins-26p::gfp in ASE compared to 84% animals without the transgene). Thus, MXL-3 plays a non-cell-autonomous role for regulation of srh-234 expression levels in ADL, either by regulating ILP genes in the intestine (e.g. ins-3), or more indirectly by regulating the expression of ILP genes in neurons other than ADL (e.g. ins-26). Together, these results support the model that MXL-3 function in the intestine acts through insulinmediated signaling to remotely regulate srh-234 expression in ADL.

Discussion
In this study, we investigated the transcriptional mechanisms by which feeding-state information is translated into expression level changes of chemoreceptor genes, using the srh-234 promoter of C. elegans as a model system. The results reveal an important role for MEF2 and basic helix-loop-helix (bHLH) transcription factors in the regulation of srh-234 expression as a function of feeding state. We identified transcriptional module(s) consisting of MEF-2 and bHLH factors that act together in ADL neurons in a temporal manner to properly regulate srh-234 expression. We also showed that the expression levels of srh-234 in ADL are regulated remotely by bHLH factors acting in the intestine, through an insulin-mediated signaling pathway. These results suggest that both cell-autonomous and non-cell-autonomous transcriptional mechanisms mediated by MEF-2 and bHLH factors regulate srh-234 gene expression levels (Fig 7), providing a sensory neuron-gut interaction for modulating chemoreceptor gene expression as a function of feeding state.

MEF-2 and HLH-2/3/4 act cell-autonomously in ADL through MEF2 and E-box sites to regulate srh-234 expression
Our results reveal a new role for the E/Daughterless ortholog HLH-2 and Achaete-scute homolog HLH-3 in the regulation of chemoreceptor genes in addition to their well-characterized function in nervous system development, such as in the establishment of neuronal cell fates [13][14][15][16][17]. HLH-2 is known to form heterodimers with both HLH-3 and HLH-4 in vivo [13,20], and we find that they all act in ADL sensory neurons to similarly reduce srh-234 expression levels in fed conditions. While hlh-2 is expressed throughout development in ADL neurons, hlh-3 and hlh-4 appear to be transiently expressed in ADL, with hlh-3 being expressed in ADL in early embryo but not in adults, while hlh-4 is expressed in ADL in adults but not in embryo. If hlh-3 is not expressed in adult ADL neurons, then how can hlh-3 regulate srh-234 expression in the adult? It is possible that hlh-3 is important to first initiate srh-234 expression in ADL in the early embryo, such that the following sequential steps to maintain srh-234 expression into adulthood do not occur. These later steps could involve hlh-4 and/or hlh-10, or alternatively hlh-3 may regulate other signaling genes required to maintain srh-234 expression through development. Although the exact mechanisms underlying srh-234 regulation mediated by these bHLH factors in ADL neurons is not yet clear, we speculate that srh-234 may be regulated by a combination of different bHLH heterodimer pairs in a temporal-specific manner. Switching has been shown to occur among heterocomplexes of bHLH factors. For instance, the switch from Myc:Max to Mad:Max bHLH pairs whom all recognize the same E-box motif results in a change in transcriptional regulation of target genes required for cell proliferation [51]. A similar switching mechanism may operate between bHLH pairs in C. elegans in order to properly initiate and maintain expression of srh-234 in ADL neurons throughout development.
Mutational analysis of the cis-regulatory region of srh-234 identified an E-box site known to physically bind bHLH heterodimers, such as HLH-2/3 and HLH-2/4 [13,20]. In C. elegans, the E-box is highly enriched in the srh-family of chemoreceptor genes, and is able to drive expression in ADL neurons [12]. Consistent with these findings, introduction of the srh-234 Ebox motif in the regulatory sequence of the AWA-expressed odr-10 chemoreceptor gene resulted in odr-10 expression in both AWA and ADL neurons. Thus, the identified E-box in the srh-234 regulatory sequence is able to direct expression in ADL. However, we did not detect any requirement for HLH-2 and HLH-3 in directing the expression of srh-234 in ADL neurons, such that levels of srh-234 expression are not fully abolished in single or double mutants between hlh-2 and hlh-3, and/or via ADL-specific down regulation of these bHLH genes by feeding RNAi. Thus, although residual activity of bHLHs cannot be ruled out in these experiments, we suggest that additional factors are required to initiate expression of srh-234 in ADL neurons.
In contrast to srh-234 gene regulation, we find that mutations in hlh-2 and hlh-3 do not alter the expression levels of a food-independent chemoreceptor, sre-1, in ADL. Thus, HLH-2 and HLH-3 do not appear to regulate additional ADL-expressed chemoreceptors, even though the core of the predicted sre-1 E-box is identical to the srh-234 E-box (CACCTG). Nucleotide sequences directly flanking the core E-box motif are known to influence the DNA binding specificity and affinity of bHLH factors [39,40], which could partly explain the expression differences observed for sre-1 and srh-234 in hlh-2 and hlh-3 mutants. Consistent with these observations, expression of odr-10p(+srh-234 E-box)::gfp in ADL neurons in hlh-2 mutations is reduced, suggesting that the srh-234 E-box site and direct flanking sequence seems to play an important role in HLH-2-mediated regulation.
In addition to the srh-234 E-box, we show that the cis-regulatory sequence of srh-234, but not that of sre-1, contains a putative MEF2 binding site in close proximity to its E-box. We find that this MEF2 site is necessary to modulate srh-234 expression levels upon starvation. However, MEF-2 does not itself dictate expression in ADL neurons, as loss-of-function mutations in mef-2 or mutation of its putative MEF2 binding site did not reduce srh-234 expression in ADL in fed conditions [22]. Thus, the MEF-2 transcription factor acts to repress srh-234 expression when starved, but it is likely not required to promote srh-234 expression when fed. Intriguingly, introduction of the srh-234 MEF2 site into the regulatory sequence of sre-1 close to its predicted E-box motif (i.e. sre-1p(+MEF2)::gfp) was sufficient to confer starvationinduced down regulation via MEF-2. It is known that bHLH factors can be recruited to activate E-box-dependent genes when they also contain a MEF2 binding site [28,30]. However, when we examined expression of sre-1p(+MEF2)::gfp in either hlh-2 or hlh-3 mutants, the sre-1 expression in ADL was not reduced. These results suggest that introduction of a MEF2 binding site to the sre-1 promoter sequence in close proximity to the sre-1 E-box motif likely does not recruit HLH-2 and HLH-3 to the cis-regulatory region of the sre-1 chemoreceptor gene. Based on these findings, the question arises how introduction of the MEF-2 site into the sre-1 promoter results in starvation-dependent regulation of sre-1 without HlH-2/3. One possibility is that in response to starvation MEF-2 can repress sre-1 through other bHLH factors that bind to the sre-1 E-box, suggesting that MEF-2 may act as a state-dependent regulator more generally.
Our results on the cis-regulatory analysis and genetic experiments of srh-234 suggest that MEF-2 may act as a transcriptional co-regulator for bHLH heterodimers in ADL neurons to temporarily regulate srh-234 expression levels as a function of feeding state (Fig 7). In this model, feeding state-dependent expression changes in srh-234 expression are directed by multiple transcriptional modules, where one module consists of a MEF-2 transcription factor and its MEF2 binding site, and the other module consists of a bHLH heterodimer (e.g. HLH-2/3, and HLH-2/4) and its bi-partite E-box binding site. When fed, MEF-2 activity is low and is prevented from repressing HLH-2/3/4 factors, and expression of srh-234 in ADL is increased. When starved, MEF-2 is no longer inhibited, allowing binding to the MEF2 site, which in turn represses HLH-2/3/4-mediated activation of srh-234 at the E-box site, ultimately leading to the decrease in srh-234 expression.

MXL-3/HLH-30 act non-cell-autonomously in the intestine to regulate srh-234 expression
In addition to bHLH factors (i.e. HLH-2/3/4) and the MEF-2 transcription factor, we showed that both the TFEB ortholog HLH-30 and Max-like 3 homolog MXL-3 regulate chemoreceptor gene expression as a function of feeding state. Genetic epistasis experiments suggest that mxl-3 and hlh-30 act upstream of mef-2, hlh-2 and hlh-3 in the regulation of srh-234 expression. Using cell-specific rescue and selective RNAi feeding experiments, we showed that the function of MXL-3 and HLH-30 is required, at least in part, in intestinal cells but not in ADL to regulate the expression of srh-234 in ADL. In both C. elegans and mammals, HLH-30 TFEB coordinates a transcriptional program by regulating the expression of autophagy in response to nutritional stress [21,52], as well as the expression of host-defense genes [53]. Moreover, HLH-30 acts together with MXL-3 as a transcriptional switch to antagonistically regulate lysosomal lipase genes in the C. elegans intestine [21]. We find a similar antagonistic role for MXL-3 and HLH-30 acting in the intestine such that srh-234 expression in ADL neurons is strongly enhanced in starved hlh-30 mutants, while mxl-3 mutants reduce srh-234 expression levels in fed conditions. In contrast, in starved mxl-3 mutants, srh-234 expression is similarly reduced as in starved wild-type animals, suggesting that MXL-3 does not appear to regulate srh-234 expression in response to starvation. How then may MXL-3 and HLH-30 activity regulate srh-234 expression as a function of feeding state? Previous elegant work [21] showed that the transcriptional activity of mxl-3 and hlh-30 is highly regulated by feeding state conditions, such that upon starvation, mxl-3 transcription in the intestine is rapidly down regulated, and at the same time, hlh-30 transcription is induced in the C. elegans intestine. In this scenario, it is possible that in starved mxl-3 mutants, HLH-30 activity is high and can therefore continue to repress and decrease srh-234 expression to starved wild-type levels. Conversely, in starved hlh-30 mutants, the loss in repression of srh-234 expression mediated by HLH-30, turns on and increases srh-234 expression to near fed wild-type levels. In order for this to happen, both MXL-3 and HLH-30 likely act antagonistically on the same E-box site of shared target genes, similar as proposed for lysosomal lipase gene regulation in the C. elegans intestine [21]. Taken together, these results suggest a non-cell-autonomous antagonistic role for MXL-3 and HLH-30 in the regulation of srh-234 gene expression as a function of feeding state.
An intestine-to-neuron model for regulation of srh-234 as a function of feeding state How might the action of bHLH factors (i.e. MXL-3 and HLH-30) in a remote tissue such as the intestine regulate the expression of a MEF-2 and bHLH-dependent chemoreceptor gene in the ADL neuron? Previous work has shown that communication between neurons and the intestine via insulin signaling is important for the regulation of metabolism, longevity and dauer formation of C. elegans [47,49,54]. The intestine, in particular, appears to be a major site for transcriptional regulation of insulin-like peptide (ILP) genes in response to feeding state conditions (fed vs starved) [48]. Our results reveal that insulin signaling from the intestine is similarly important for the regulation of srh-234 expression in ADL. First, we show that daf-2/daf-16 insulin signaling pathway acts downstream of mxl-3 and hlh-30 in the regulation of srh-234. Moreover, rescue experiments suggest that MXL-3 and HLH-30 act specifically in the intestine to remotely regulate srh-234 expression in ADL. Second, MXL-3-mediated regulation of srh-234 is dependent on DAF-2 signaling acting in ADL. Third, mxl-3 mutants appear to alter the subcellular localization of DAF-16::GFP in ADL, suggesting that MXL-3 function is required to repress DAF-16 in ADL. Fourth, we show that starvation as well as reduced mxl-3 and daf-2 signaling increases the expression of mef-2, and this higher MEF-2 activity may repress the bHLH-mediated activation of srh-234. The specific mechanism by which DAF-2/DAF-16 signaling regulates MEF2 activity remains to be deciphered, but interestingly, the DAF-16 ortholog FOXO has been shown to work in concert with MEF2 family members during cardiovascular development [55]. Lastly, we show that mxl-3 mutants reduce, albeit weakly, the expression of a candidate agonistic ins-3 ILP in the intestine, which when specifically knocked down by RNAi in the intestine weakly reduces srh-234 expression. Thus, these data suggest a non-cell-autonomous intestinal role for a mxl-3-dependent ins-3 ILP in the regulation of srh-234.
It is likely that the ins-3 ILP does not act individually on the DAF-2 pathway in ADL. We found that expression of daf-2 in ADL neurons can rescue the reduced srh-234 expression phenotype of mxl-3; daf-2 double mutants (Fig 5C), suggesting that the ins-3 ILP does not act as the limiting factor for activation of DAF-2 in ADL to regulate srh-234 expression. Moreover, ins-3 is also expressed in sensory neurons, ASI and ASJ, all of which likely do not express mxl-3 [21], suggesting that mxl-3-independent mechanisms likely exist to drive ins-3 expression. Consistent with these findings, daf-2 mutants show a significantly stronger reduced srh-234 expression than mxl-3 mutants, suggesting that there must be additional factors that may regulate ILPs besides MXL-3. In addition to the ins-3 ILP, we also found that other ILPs from the intestine (e.g. ins-7, ins-21, ins-28) as well as ILPs from other neurons (e.g. ins-26 ASE) negatively regulate srh-234 expression in fed conditions. The combined action of the ILPs from different tissues (the intestine) and neurons (e.g. ASE) on srh-234 regulation is not yet clear. ILPs are known to regulate each other transcriptionally in complex networks in a combinatorial and coordinated fashion [54,56]. Moreover, a complex network of ILPs secreted from head neurons are known to both promote and oppose dauer entry via intestinal DAF-2 activity [47]. In this regard, perhaps ILPs secreted by the intestine and other neurons mediated by mxl-3-dependent and independent transcriptional mechanisms could both promote and antagonize DAF-2 activity in ADL to dynamically modulate srh-234 expression as a function of feeding state.

Concluding remarks
Plasticity in chemoreceptor gene expression may be a simple strategy by which an animal can modulate its chemosensory responses in changing external and internal state conditions. Our results describe cell-autonomous and non cell-autonomous transcriptional mechanisms involving MEF-2 and basic helix-loop-helix (bHLH) factors and their cognate binding sites, respectively, by which C. elegans may integrate and translate feeding-state information into proper expression changes of individual chemoreceptor genes. We expect that continued investigation of the mechanisms by which the expression of individual chemoreceptor genes are changed by internal state and external conditions such as food availability will lead to insights into the general principles of behavioral plasticity. Moreover, understanding chemoreceptor gene regulation could provide important insight into the mechanisms by which feeding-state dependent host-seeking and leaving behaviors in mosquitoes, and possibly parasitic nematodes, are accompanied by modulation of chemoreceptor gene expression.
Measurement and quantification of srh-234 and sre-1 gfp-reporter expression levels Transgenic animals carrying promoter::gfp reporters were grown at 20°C on NGM plates seeded with E. coli OP50 as the bacterial food source. Young adult animals were washed with M9 buffer (to remove any bacteria in the gut) and transferred onto plates with E. coli OP50 bacterial food (fed) or without food for 12-24 hours (starved) unless indicated otherwise. Levels of promoter::gfp expression (e.g. srh-234, sre-1) were imaged and measured under a microscope equipped with epifluorescence as previously described [22]. For each promoter::gfp construct, expression from the same extra chromosomal array was examined in wild-type and mutants. Briefly, we mounted animals on agarose pads containing 3 mM sodium azide unless indicated otherwise, and visualized these on a Leica DM5500 compound microscope or a Leica TCS SP8 confocal equipped with epifluorescence. Microscope and camera settings were kept constant between images of different genotypes and conditions, unless indicated otherwise. The mean pixel intensity of gfp fluorescence in the entire cell-body of ADL was quantified using Volocity software. Prior to measurement, images of ADL were cropped for srh-234 expression analysis. Statistical analyses of srh-234 expression were performed using the nonparametric two-tail Mann-Whitney test.  [21], and the ADL-specific reporter sri-51p(3 kb)::mCherry was a kind gift from Cori Bargmann [12].

Expression constructs of bHLH genes and generation of transgenic animals
For generating the hlh-10p::gfp and svh-1p::gfp reporters, we fused about 1.5 kb and 1.6 kb promoter sequence of hlh-10 and svh-1, respectively, to gfp as previously described [58]. The sre-1p::gfp construct was generated by inserting about 1 kb promoter sequence of sre-1 into the pPD95.77 vector (AddGene). Transgenic animals were generated using either the unc-122p:: dsRed (50-100 ng/μl) or the pRF4 rol-6(su1006) co-injection markers injected at 150 ng/μl using a standard microinjection protocol [59]. Expression constructs were injected at 20-30 ng/μl. All amplified products in the generated constructs were sequenced to confirm the absence of errors generated via the amplification procedure.

ADL-mediated avoidance assays
For examining avoidance responses to CuCl 2 , glycerol and SDS, we used the drop-test assay as previously described [60,61]. Briefly, we picked L4-staged larvae of either wild-type, hlh-2 (tm1768) or hlh-3(ot354) mutants about 20 hours before the assay to NGM plates with (onfood) or without food (off-food). For on-food plates, 25 μl of overnight culture of E. coli OP50 in LB was spread on each plate. For off-food plates, 25 μl of LB was spread on each plate. All plates were allowed to dry for 1 hour without lids. Animals were picked using an eyelash pick to a plate without food for less than a min (to prevent food being transferred), before animals were transferred to on-food and off-food assay plates. Animals were allowed to settle for 10 min, and then assayed using a capillary to deliver the repellent drop. About 5 animals were assayed on each plate at room temperature, all animals were stimulated every 60 seconds and the percentage of responding animals were recorded. The avoidance index is the number of positive responses divided by the total number of trials. The tested repellents were 2 mM CuCl 2 (Acros Organics), 0.5 M glycerol (Sigma) and 0.1% SDS (Thermo Scientific). All repellents were dissolved and diluted in M13 buffer (30 mM Tris, 100 mM NaCl, 10 mM KCl).
For examining the avoidance response to octanol, we used the smell-on-a-stick assay as previously described [62,63]. Briefly, the blunt end of a paintbrush hair (Loew-Cornell, 9000 Kolinsky 7 size) was taped to a Pasteur pipette, and dipped in either 30% or 100% 1-octanol and placed in front of either a forward-moving wild-type, hlh-2(tm1768) or hlh-3(ot354) mutant animal. The time to respond (seconds) is the amount of time it took for the animal to initiate backward movement at room temperature. Octanol (Sigma) was diluted and freshly made in ethanol.

RNAi feeding of bHLH genes
Feeding RNAi of selected bHLH genes was performed in the rrf-3(ok1426) background to enhance neuronal RNAi [64]. Briefly, L4-staged larvae of rrf-3(pk1426) mutant animals carrying a stably integrated srh-234p::gfp transgene (oyIs56) were placed on an NGM agar plate containing 1 mM IPTG and seeded with E. coli HT115 bacterial food producing dsRNA directed against the selected bHLH genes. Bacteria expressing dsRNAs directed against selected bHLH genes were taken from the C. elegans feeding RNAi library (Source BioScience). For hlh-2 (RNAi) and hlh-4(RNAi), we cloned the wild-type genomic sequence of hlh-2 and hlh-4 in the L4440 vector (Addgene) because these bacterial RNAi clones in the feeding RNAi library were incorrect. For fed conditions, animals fed with bHLH(RNAi) were scored as adults in the next generation (F1), and expression of srh-234p::gfp was measured and quantified as described above. For starved conditions, animals in the F1 generation were transferred as L4-staged larvae to RNAi plates without food and srh-234p::gfp expression was determined about 12 hours later. The L4440 empty vector was used as the RNAi control. All RNAi bacterial clones used were cultured as previously described [65], and confirmed by sequencing.

Tissue-specific RNAi of bHLH and ILP genes
For tissue-selective knockdown of bHLH and ILP genes, we employed a previously described feeding RNAi strategy [38]. Briefly, we amplified and cloned the wild-type genomic sequence of sid-1 in the pSM SL2::mCherry::unc-54 3'UTR vector (a kind gift from Cori Bargmann) using the ADL-specific promoter, sre-1, and the intestinal-specific promoter, ges-1, resulting in constructs: pMG57 sre-1p::sid-1 genomic::SL2::mCherry (ADL::sid-1), and pMG66 ges-1p::sid-1 genomic::SL2::mCherry (intestine::sid-1). Transgenic animals were generated in sid-1(pk3321) mutant animals carrying the stably integrated srh-234p::gfp transgene (oyIs56) using a standard microinjection protocol [59]. Expression constructs were injected at 30 ng/μl together with the pRF4 rol-6(su1006) as a co-injection marker at 150 ng/μl. To selectively enhance RNAi in either ADL neurons or in the intestine (but not in other tissues), L4-staged larvae of transgenic animals carrying either ADL::sid-1 or intestine::sid-1 were placed on an NGM agar plate containing 1 mM IPTG and freshly seeded with E. coli HT115 bacterial food producing dsRNA directed against individual bHLH genes or ILP genes as previously described [65]. Expression of srh-234p::gfp was measured and quantified in transgenic animals fed with either bHLH (RNAi) or ILP(RNAi) in the next generation (F1) at the adult stage unless indicated otherwise. The L4440 empty vector was used as the RNAi control, and all RNAi bacterial clones used were confirmed by sequencing.

Subcellular localization of DAF-16::GFP
For measuring the fraction of DAF-16::GFP in the compartment (cytoplasmic:nuclear) of ADL, we generated transgenic animals that carry both a sre-1p::daf-16a::gfp reporter as an extra chromosomal array (ADL::daf-16::gfp] and the integrated hlh-2p::mCherry::his reporter. Transgenic fed animals were grown at room temperature until young adults, and then transferred onto plates with E. coli OP50 bacterial food (fed) or without food for 12 hours (starved). Next, animals were mounted on glass slides containing 1 mM levamisol to paralyze the animals, and visualized as described above. Fluorescent images were taken at the same exposure time using Volocity software. The fraction of DAF-16::GFP in the subcellular compartment of ADL neurons was measured by counting the fraction of animals that show nuclear fluorescence in either the red (cytoplasm > nuclear), orange (nucleus = cytoplasm), or yellow color (nucleus > cytoplasm). Statistical analyses of the distribution between genotypes and conditions was performed using the χ 2 test.

Dye-filling of ADL sensory neurons
Dye-filling experiments were performed similarly as previously described [66]. A stock dye solution containing 5 mg/μl red fluorescent lipophilic dye DiI (Sigma Aldrich) was diluted in M9 buffer by 10,000 times for optimal signal intensity. Animals were soaked in the fluorescent dye solution for one hour and then rinsed with M9 buffer twice. Stained animals were recovered for 1 hour on NGM plates seeded with E. coli OP50 bacterial food before examination of ADL sensory neurons. Overview of tissue-specific knockdown of bHLH genes by RNAi. (B-C) Representative expression of srh-234p::gfp in adult animals by feeding either hlh-2(RNAi), hlh-3(RNAi), hlh-4(RNAi), mxl-3(RNAi) or control RNAi. Feeding RNAi was performed in sid-1(pk3321)him-5(e1409) mutants with the stably integrated oyIs56 transgene carrying either a ADL::sid-1 (B) or intestine::sid-1 (C) extra chromosomal transgene to selectively enhance RNAi in ADL neurons or in the intestine, respectively, but not in other tissues (see Material and Methods). Images are lateral views of ADL with the arrow pointing to the cell body; anterior is left. Images were acquired at the same exposure time at room temperature. (D-E) Relative expression of srh-234p::gfp in the ADL cell body of adult animals when fed either hlh-2(RNAi), hlh-3(RNAi), hlh-4(RNAi), or mxl-3(RNAi) compared to adult animals fed with control RNAi. Data shown is the average of at least two independent transgenic lines with n>20 adult animals for each line. Ã indicates values that are different from that of wild-type adult animals at P<0.001, and # between the genotypes compared by brackets at P<0.001. Error bars denote the SEM. , ins-4p::gfp and ins-5p::gfp (B) and ins-26p::gfp (C) in adult wild-type (left panels) and mxl-3 mutant animals (right panels) during fed conditions. White box represents intestinal area used for comparative analysis of expression. In all experiments, images were taken at the same exposure time at room temperature. The same transgene was examined in each genotype. Arrow indicates ASE neuron. (EPS) S1 Table. Predicted MXL-3 and HLH-30 binding sites in insulin-like peptide promoters. A search of possible MXL-3 and HLH-3 binding sites present in promoter sequences of insulinlike peptide genes using Lasagna 2.0 (http://biogrid-lasagna.engr.uconn.edu/lasagna_search/) with a cut-off value of P<0.01. (XLSX)