C. elegans strains

All strains used in this study are listed in S1 Table. Strains were maintained as previously described [30].

C. elegans genome-engineered strains

Deletion alleles for hlh-32(ot1347), hlh-13(ot1388), hlh-13(ot1390), and hlh-15(ot1389) were generated using CRISPR/Cas9 genome engineering [31]. Multiple alleles encompassing the same genomic regions, as shown in Fig 1C, represent the same deletion generated in different backgrounds. Deletions were made using 2 crRNAs repaired with an ssODN, as described below:

hlh-32(ot1347): crRNAs (tttcagggtgtcgttagatt and catccgaaggagattagcgc), ssODN (cgctggtgatagattcctggatgttcaacaaattgcgctggtgatagattcctggatgttcaacaaattg)

hlh-13(ot1388) and hlh-13(ot1390): crRNAs (gaagctgtcatttacataag and acagagtttgtttaggcaat), ssODN (accaattacaattgtgaattcgagcagaaccacttGCCTAAacaaactctgtgtatgcgtaaacggcaga)

hlh-15(ot1389): crRNAs(ttgaatagtggaacacttgc and tgaattcacaaccttcacag), ssODN(atctgttactcgttttcctatcctctattccagcacagtggccaaccgatatatatatatatatagttcc)

A number of CRISPR/Cas9-engineered reporter and deletion strains were generated by SunyBiotech (indicated by the syb allele designation; see S1 Table). These include C-terminal gfp reporter insertions for all the 5 bHLH genes examined here and SL2::gfp::H2B reporter cassette insertions at the C-terminus of genes that serve as cell fate markers and/or potential targets of bHLH genes (nlp-49, kcc-3, col-105, pdf-1, flp-7, bcat-1). All other previously published cell fate markers used here are also listed in S1 Table.

C. elegans transgenic strains

To establish the sshk-1p::gfp reporter transgene, a marker for AUA neurons, the 370 bp intergenic region upstream of sshk-1 was fused to gfp followed by the 3′ untranslated region of unc-54, as previously described [32]. Primers were: sshk-1 promoter (370 bp): fwd CTGTTGACTAATCTCACAGC rev AGGAAAACTTTCAAATGAGAGG. The amplicon was injected together with a pha-1 wild-type copy into in a pha-1(e2123) mutant strain to generate otEx8199.

To generate flp-1p::genomic hlh-15::SL2::TagRFP, a 0.4 kb flp-1 promoter sequence from pCS169 (flp-1p(trc)::ICE) [33] attached to the full hlh-15 genomic sequence (flp-1p::hlh-15 genomic fragment containing exons and introns) was fully synthesized by Azenta Life Sciences and amplified using the following primers: fwd GGAAATGAAATCAGGAAACAGCTATGACCATGAGCTTAATTCCTAAAAACCC rev GGTGAAAGTAGGATGAGACAGCCGGCCGTTATTGAAGCAAGTTGTCTAGAAAAC. This amplicon was then ligated to the SL2::TagRFP backbone amplified from pCPL11 (srab-20p::lin-29a::SL2::TagRFP) using Gibson cloning. hlh-15(ot1389) mutant animals were then injected with the following concentrations of constructs: flp-1p::ghlh-15::SL2::tagRFP 25 ng/μl, Pttx-3::GFP 25 ng/μl, pBS 150 ng/μl to generate otEx8247 and otEx8248.

All other previously described transgenic strains used in this study are listed in S1 Table.

Neuronal identification using NeuroPAL

We crossed our bHLH reporter alleles to the NeuroPAL landmark strain (otIs669) to pinpoint which neuron types they are expressed [34]. Neuron types were identified based on the NeuroPAL color code they express and the position of their nuclei relative to other neurons. Detailed guides to neuronal identification using NeuroPAL can be accessed through https://www.hobertlab.org/neuropal/.

Automated worm tracking

Worm tracking was performed at room temperature using a WormTracker 2.0 imaging system [35]. Immediately before the experiment, NGM plates were seeded with a thin lawn (10 μl) of OP50, on which 5 young adult animals at a time were transferred and recorded for 5 min. Tracking videos were analyzed using the WormLab software (MBF Bioscience).

Defecation assay

Defecation assays were performed as described [36]. Well-fed, young adult worms were singled to NGM plates with a thin lawn of OP50. Before starting the assay, worms were left to acclimate for 10 min. Plates were mounted on a Nikon Eclipse E400 microscope and worms were observed using the 20X DIC objective. A posterior body muscle contraction (pBOC) of the worm signaled the start of the observation period. For 10 min, the pBOC and enteric muscle contraction (EMC) events were counted, including the first pBOC at the start of the assay.

Staining for lipids

Staining for lipids using Nile Red (NR) was performed following [37]. L4 worms raised at 20°C were harvested by washing with 1× PBST. Worms were pelleted by centrifugation at 560 × g for 1 min and supernatant discarded. To the pellet, 100 μl of 40% isopropanol was added and left to incubate for 3 min at room temperature. Worms were again centrifuged at 560 × g for 1 min and supernatant was carefully removed.

To prepare the NR stain, 6 μl of 5 mg/ml NR in 100% acetone was freshly added to 1 ml of 40% acetone. Working in the dark, 600 μl of NR stain was added and thoroughly mixed to each worm pellet by inverting the tubes. Samples were rotated in the dark for 2 h at room temperature. Worms were then pelleted at 560 × g for 1 min and the supernatant was removed. To remove excess NR stain, samples were incubated with 600 μl of PBST for 30 min in the dark. Samples were centrifuged at 560 × g for 1 min and all but 50 μl of supernatant was discarded. Worms were resuspended in the residual supernatant and imaged immediately after staining. Intensity of the lipid stain was measured in Fiji [38].

Microscopy

Worms were immobilized using 100 mM sodium azide (NaN₃) and mounted on 5% agarose pads on glass slides. Unless otherwise indicated, worms were imaged at the L4 or young adult stage. Images were acquired using either Zeiss LSM 980 confocal microscope or Axio Imager Z2 compound microscope at 40× magnification unless otherwise specified. Processing of images was done using the Zen (Zeiss) or Fiji [38] software. Fluorescence intensities of reporters were quantified using Fiji.

Microfluidic locomotory analysis

The time-lapse images (1 frame/second, 5 h recording) for behavioral assays were acquired using a high-resolution, multi-well imaging system [39]. We added 60 μl of liquid NGM buffer (same components as NGM but without agar or peptone) to each well of a 96-well plate (Corning, Inc.), then picked one Day 1 adult animal to each well. Where indicated, the NGM was supplemented with food E. coli OP50, and the bacteria were added to NGM to a final OD₆₀₀ ≈ 1. Multiple genotypes or conditions being compared were assayed on the same plate.

Image data processing and analysis were performed as in a previous study [40]. Consecutive images were subtracted to generate difference images. We normalized the difference image with the average pixel intensity of the 2 subtracted images. To reduce noise, we applied a Gaussian filter with SD equal to one pixel to the difference image. We then applied a binary threshold of 0.6 to the filtered difference image to determine whether movement occurred at each pixel. We summed up all pixels of the binarized difference image and used the resulting value to define the activity. To calculate the dwelling fraction of worms, we time-averaged the activity data with a smoothing kernel of 5 s and generated a histogram of the time-averaged activity for each worm. We then employed a nonlinear least-squares fitting algorithm to fit each individual worm’s activity histogram to 2 exponential terms (2 free parameters each) and a Gaussian term (3 free parameters). The zero bin on the activity histogram (corresponding to quiescence) was excluded from the histogram before the fitting process. The dwelling fraction of each worm was calculated as the area under the 2 exponential curve components of the fit.

Compensatory curvature response (CCR) assay

We employed a straight channel microfluidic device to perform the compensatory curvature response (CCR) behavioral assay as described in a previous study [41] (Fig 9A). The microfluidic device consists of 2,000 μm-wide open areas and 2 parallel straight channels (60 μm width, 200 μm length, 80 μm height). During the assay, we transferred day 1 adults to a food-free NGM buffer for approximately 5 min to remove bacteria, and then pipetted the animals from the washing buffer into a microfluidic chamber filled with NGM buffer containing 0.1% (by mass) bovine serum albumin (for preventing worms from adhering to chamber surfaces). We recorded behavioral videos (30 frames/second) of each animal in the chamber for about 3 min performing free locomotion in open areas or constrained locomotion with their mid-body in the narrow channels.

The behavioral data from microfluidic experiments were analyzed using methods described previously [41]. For each worm, the normalized bending curvature K is calculated as the product of body curvature k and the worm body length L. To quantify the effect of straight-channel constraint on worm bending curvature amplitude, the whole-body curvature amplitude during constrained locomotion was computed and compared with that of free locomotion. Specifically, we analyzed worm bending curvature dynamics during free locomotion to generate an averaged curvature amplitude profile, A_free(s), as a function of body coordinate s defined such that s = 0 at the head and s = 1 at the tail. We divided video sequences of constrained movement into individual short sequences using a 3 s time window to analyze constrained locomotion. We manually marked the channel position in each image sequence to record the relative position of the constraint to the worm body. The normalized curvature change in response to mid-body constraint was calculated by considering only periods during which the anterior and posterior limits of the narrow channel were consistently within a body coordinate interval between 0.35 and 0.65. The resulting curvature dynamics were denoted as K_const, and the maximum value of |K_const(s,t)| in the time dimension was defined as the curvature amplitude profile of individual periods, denoted as . The normalized curvature change of each period was represented by . The normalized anterior curvature changes of individual periods was defined as where ( denotes the average of X over the interval [a, b]).

Lifespan and healthspan assays

Lifespan assays. Lifespan assays were performed as previously described [42]. One hundred (100) unhatched embryos were put on a plate, with 3 plates for each genotype. Throughout the experiment, plates were kept in dark boxes at 20°C as continuous exposure to light has been shown to decrease worm lifespan [43]. Embryos were left to develop for 3 days until Day 1 of adulthood, after which 15 young adults were gently transferred to each plate, with 7 plates for each genotype. The number of worms that were alive, dead, and censored were recorded every alternate day of adulthood. Censored worms are worms that die due to bagging, desiccation, contamination, or human error. Worms that are immobile and fail to respond to touch with a platinum pick are considered dead. Dead and censored worms were burned once they were recorded. For the first 6 days of adulthood, worms were transferred to a new plate every day to avoid contamination with their progeny. Once worms stopped producing progeny, usually after day 6 of adulthood, they were transferred to new plates every other day. Lifespan was recorded until death of the last worm, which usually occurred between 30 and 45 days of adulthood. In cases of contamination with mold or unwanted bacteria during the assay, all live worms were transferred to clean plates. Statistical analysis, Kaplan–Meier survival analysis, and Mantel–Cox log-rank test were performed using OASIS 2 software (https://sbi.postech.ac.kr/oasis2) [44].

Pharyngeal pumping rate. The assay NGM plate was seeded with 50 μl of OP50 1 day prior to the experiment. Ten minutes before the assay commenced, animals were transferred onto the assay plate. Pharyngeal pumping, a marker of health span [45]. was observed using a Nikon SMZ645 dissecting microscope. The pumping rate was calculated by counting the number of pumps over a 20-s interval, then multiplying this value by 3 to obtain the rate per minute.

Nucleolar size. To measure hypodermal cell nucleolar size, a prospective marker of lifespan [46], differential interference contrast (DIC) images of hypodermal cells of day 1 animals were acquired at 63× magnification. Nucleolar area was measured using Fiji, following.

Group A bHLH family genes in the nervous system of C. elegans

The C. elegans genome encodes members of all 6 groups of bHLH genes (Groups A to F) [8–10,47,48]. We focus here on members of group A since its striking expansion in metazoans [10] provides a clear indication of their role in cell type diversification. There are 19 group A genes in C. elegans, including paralogues of the phylogenetically deeply conserved Ato, ASC, NeuroD, Neurogenin, and MyoD proteins (Fig 1A). While several of these group A genes have been previously analyzed for their function within and outside the nervous system [16,17,49–54], others have remained uncharacterized.

Examination of group A bHLH transcript presence in the whole nervous system single-cell RNA atlas of C. elegans, established by the CeNGEN consortium [19], reveals that only a subset of group A proteins are expressed in the mature C. elegans nervous system, and each in a very small number of different mature neuron types (Fig 1A). These include several genes whose function in nervous system differentiation have not previously been analyzed. We chose the 5 most strongly expressed and previously little-studied genes from this category (Fig 1B), the PTF1a/NATO3 ortholog hlh-13, the NHLH1/2 ortholog hlh-15, and the 3 bHLHe22/e23 orthologs hlh-17, hlh-31, and hlh-32, for an in-depth analysis of their expression and function. First, we used CRISPR/Cas9 genome engineering to tag each of these 5 genes with gfp to analyze the expression of the protein product of these loci throughout the whole animal during all life stages, thereby significantly expanding previous transcript and reporter transgene analysis. Second, we used existing or generated new deletion alleles, again using CRISPR/Cas9 genome engineering, to rigorously probe their function. Reporter and mutant alleles for each examined locus are schematically shown in Fig 1C.

The bHLHe22 and bHLHe23 orthologs HLH-17 and HLH-32 proteins are expressed in neurons, but not glia

The C. elegans genome codes for 3 members of the Olig superfamily of transcription factors, hlh-17, hlh-31, and hlh-32. In vertebrates, this family is defined by 2 closely related subtypes of genes, the highly interrelated, paralogous Olig1, Olig2, and Olig3 genes and the 2 highly related bHLHb4 (now called bHLHe23) and bHLHb5 (now called bHLHe22) gene paralogs. While the 3 C. elegans genes hlh-17, hlh-31, and hlh-32 have previously been considered as orthologs of the Olig subbranch of bHLH genes [55], detailed sequence analysis, as well as orthology assignment tools demonstrate that hlh-17, hlh-31, and hlh-32 are more closely related to the bHLHe22/e23 subbranch than to the Olig subbranch [56–58] (Figs 1B and S1A–S1C). Conversely, another C. elegans bHLH gene, hlh-16, appears to be more closely related to the Olig genes than hlh-17, hlh-31, and hlh-32 (S1A–S1C Fig).

All 3 bHLHe22/e23 paralogs cluster within a small interval (<150 kb) on chromosome IV and hlh-17 and hlh-31 are direct neighbors (S1D Fig). The hlh-17 and hlh-32 genes display identical bHLH domain-encoding sequences, indicating that they have the same protein dimerization and DNA-binding properties (S1B Fig). All 3 genes are the result of species-specific gene duplications within the nematode phylum. Reciprocal BLAST searches show that C. remanei has at least 6 members of the bHLHe22/e23 subfamily while C. briggsae has only a single representative (Cbr-hlh-17). Nematodes from other clades also do not harbor hlh-17/31/32 duplications, leading us to conclude that an ancestral single bHLHb4/5 gene has undergone recent duplications within select members of the Caenorhabditis genus and then subsequently duplicated once in the vertebrate lineage, to generate bHLHe22 and bHLHe23. The related Olig-like gene hlh-16 has not undergone gene duplications in nematode genomes, but it has been lost in Drosophila, which contains instead a single bHLHe22/e23 representative, somewhat misleadingly called Oli [59] (Figs 1B and S1).

The fusion of 5′ promoter sequences of the hlh-17, hlh-31, and hlh-32 genes to a gfp reporter had shown that the hlh-17 reporter construct is expressed in the CEP sheath (CEPsh) glia cells and unidentified motor neurons in the ventral nerve cord, while the hlh-32 reporter fusion was expressed in an unidentified pair of head neurons; no expression could be discerned for the hlh-31 promoter fusion construct [55,60]. Promoter fusion constructs can miss cis-regulatory elements and, particularly in the case of the hlh-17/31/32 locus, the genomic rearrangements leading to the duplication of the ancestral gene may have resulted in complex arrangements of cis-regulatory elements. To eliminate these concerns, we tagged each of the 3 endogenous loci with gfp, using CRISPR/Cas9 genome engineering. The gfp tagged loci show expression patterns that are indeed different from the previously reported promoter fusions. We find that HLH-17::GFP and HLH-32::GFP protein are exclusively detectable in 3 neuron classes, the bilaterally symmetric glutamatergic AUA interneuron class and 2 B-class cholinergic ventral cord motor neurons, DB2 and VB2 (Fig 2). Together with a few other head and neck muscle-innervating motor neurons, these motor neurons are part of the retrovesicular ganglion (RVG), which is located at the anterior end of the ventral nerve cord. DB2 and VB2 are the most anteriorly located B-type motor neurons in the RVG and have lineage histories that are distinct from other DB and VB class members [61,62]. Based on scRNA-seq data [19,63], DB2 and VB2 are also molecularly subtly distinct from other DB and VB class members. Due to their location, the RVG motor neurons are analogous to the branchial motor neurons in the brain stem of vertebrates, but little is known about how RVG motor neuron subtypes are made to become different from other motor neurons of the same class.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Expression of hlh-17, hlh-31, and hlh-32 reporter alleles.

(A) Expression of CRISPR/Cas9-engineered reporter alleles, hlh-17(syb6303), hlh-31(syb6112), and hlh-32(syb6078) over the course of development (see Fig 1C for reporter allele). HLH-17::GFP and HLH-32::GFP can be detected in head neurons (red arrowheads), while no HLH-31::GFP signal could be detected anywhere at any stage. Asterisks indicate intestinal autofluorescence. Representative images for reporter alleles are shown with 10 μm scale bar. (B) Overlay with the NeuroPAL (otIs669) transgene demonstrates hlh-17(syb6303) and hlh-32(syb6078) expression in the AUA, VB2, and DB2 neurons. Representative close-up views for reporter alleles are shown with 10 μm scale bar.

https://doi.org/10.1371/journal.pbio.3002979.g002

Expression of HLH-17::GFP and HLH-32::GFP in AUA, DB2, and VB2 is observed throughout all larval and adult stages and commences when those neurons are generated (Fig 2). Its postembryonic expression is consistent with scRNA-seq data [19] (Fig 1A). Notably, however, we detected no HLH-17::GFP protein expression in CEPsh glia (or any other glia) at any developmental stage, indicating that hlh-17 transcripts, detected by scRNA-seq analysis [19] and inferred from transcriptional reporter constructs [55,60] may not become translated into detectable protein levels in the CEPsh glia. We have occasionally observed similar transcript and protein discordance for homeobox genes [12,19].

In contrast to HLH-32 and HLH-17, expression of HLH-31::GFP, encoded by the gene that directly neighbors HLH-17 (and which displays a somewhat degenerate bHLH domain sequence; S1 Fig), was not detectable in our analysis at any stage in any cell within or outside the nervous system. scRNA transcriptome analysis detects transcripts for hlh-31 in the DB2 and VB2 neurons (like its neighboring hlh-17 gene), and also in the CAN neurons [19] (Fig 1A). However, as with hlh-17 transcripts in the CEPsh cells (also detected by scRNAseq), no readily detectable protein appears to be made from these hlh-31 transcripts.

Removal of hlh-17, hlh-31, and hlh-32 does not obviously affect glia differentiation

To assess potential functions of hlh-17, hlh-31, and hlh-32, we generated single, double, and triple mutant strains using CRISPR/Cas9 genome engineering. Unlike the previously generated hlh-17, hlh-31, and hlh-32 mutant alleles, which left parts of each gene intact [55], our engineered alleles each deleted the entire respective locus (Fig 1C). Due to overlaps in gene expression, and the sequence identity of the bHLH domains (S1B Fig), we considered possible functional redundancies and therefore started our analysis by examining hlh-17/hlh-31/hlh-32 triple null mutant animals. These were generated in 2 consecutive genome editing rounds, first removing the 2 adjacent hlh-17 and hlh-31 loci (syb6635) and then removing the more distal hlh-32 locus (syb6773) in the background of the syb6635 allele. From here on, we refer to these hlh-32(syb6773)hlh-17hlh-31(syb6635) triple null mutants as “hlh-17/31/32^null”. These triple null mutant animals are viable, fertile and display no obvious morphological or behavioral abnormalities.

Consistent with the absence of HLH-17/31/32 protein expression in the CEPsh glia, we observed no effects of hlh-17/31/32^null mutants on CEPsh glia differentiation, as assessed by examination of CEPsh morphology and marker gene expression (S2A and S2B Fig). This is in line with a previous study that used partial deletions of the bHLH domains of hlh-17/31/32, displaying no effect on CEPsh morphology and marker expression [55].

hlh-17 and hlh-32 are required for AUA neuron differentiation

To examine AUA neuron identity specification, we used the NeuroPAL transgene, in which this neuron is marked with 2 distinct fate markers, eat-4/VGLUT and mbr-1 [34]. We find that the NeuroPAL color code of the AUA neuron pair is not obviously changed in hlh-17/31/32^null animals (Fig 3A). Examining other AUA markers, we first investigated expression of the neuropeptide-encoding flp-8 and flp-32 genes [19,64]. We found that flp-8, but not flp-32, expression is affected in triple null mutants (Fig 3B and 3C). Neither hlh-32(ot1347) single mutants nor hlh-17hlh-31(syb6635) double mutants recapitulated the flp-8 expression defects (Fig 3C), indicating that AUA-expressed hlh-32 and hlh-17 indeed act redundantly to specify flp-8 expression in AUA. This is not surprising given that both proteins are co-expressed and display an identical amino acid sequence of their bHLH domain.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. AUA cell fate analysis of hlh-17, hlh-31, and hlh-32 mutant animals.

(A, B) Triple hlh-17/31/32^null mutants do not show a change of NeuroPAL (otIs669) coloring of AUA (panel A) or expression of flp-32(syb4374) in AUA (panel B). In NeuroPAL, AUA neurons display a combination of mNeptune2.5 red pseudocolor driven by the eat-4 promoter and a dim mTagBFP2 blue pseudocolor driven by the mbr-1 promoter. In hlh-17/31/32^null mutants, NeuroPAL coloring of AUA is unchanged. Shown are representative images of wild-type and mutant animals with 10 μm scale bar. Statistical analysis was performed using unpaired t test. Individual points represent averages of the fluorescence intensities of the AUA neuron pair within an individual. The number of animals scored are enclosed in parentheses. Error bars indicate standard deviation of the mean. ns: not significant. (C) Expression of flp-8 reporter (ynIs78) in AUA is affected only in hlh-17/31/32^null triple mutants. Individual hlh-17/31(syb6635) and hlh-32(ot1347) mutants do not show loss of the flp-8 reporter. The flp-8 reporter is also expressed in the URX neurons, indicated by the asterisks. Shown are representative images of wild-type and mutant animals with 10 μm scale bar. Number of animals scored are shown within each bar. P-values were calculated using Fisher’s exact test. (D) A novel fate reporter for AUA was generated by fusing the promoter directly upstream sshk-1 with gfp (otEx8199). hlh-17/31/32^null triple mutants lose expression of sshk-1 reporter in AUA. The sshk-1 reporter is also expressed in the pharyngeal gland cells, indicated by the asterisks. Representative images of wild-type and mutant animals are shown with 10 μm scale bar. Number of animals scored are shown within each bar. P-values were calculated using Fisher’s exact test. Raw data for panels B–D can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3002979.g003

We generated another identity marker for the AUA neuron by interrogating the CeNGEN scRNA atlas for transcripts with selective, strong expression in AUA. We considered the T03D8.7 locus, which codes for a small secreted protein with a ShKT domain that, in other species, acts as a toxin against pathogens by inhibiting potassium channels [65]. We named this gene sshk-1 (for small ShKT domain) and found that a transcriptional reporter that encompasses the entire intergenic region to its preceding gene is indeed expressed exclusively in AUA within the entire nervous system; outside the nervous system expression is also observed in the exocrine pharyngeal gland cells, consistent with this protein having a potentially protective function against pathogens (Fig 3D). We found that expression of the sshk-1^prom reporter transgene is strongly affected in the AUA (but not gland cells) of hlh-17/31/32^null mutant animals (Fig 3D). In those animals where residual sshk-1^prom expression was visible, AUA neurite morphology appeared normal.

hlh-17 and hlh-32 diversify VB1/VB2 motor neuron subtype identity

We interrogated motor neuron identity specification using again hlh-17/31/32^null triple mutant animals because of (a) the redundant functions of co-expressed hlh-32 and hlh-17 we had revealed above in the AUA fate analysis; and (b) because of the theoretical possibility that even though hlh-31 showed no expression in wild-type animals, it may be up-regulated in hlh-32; hlh-17 double mutants. Between the motor neurons that express hlh-17 and hlh-32, VB2 and DB2, we focused our analysis on VB2 because more identity markers are available for VB2 than for DB2.

We first examined the effect of the hlh-17/31/32^null mutation on NeuroPAL, in which defects in neuronal specification can be detected through changes in neuron coloring conferred by specific sets of promoters [34]. We observed that in hlh-17/31/32^null animals, VB2 adopts NeuroPAL coloring that resembles another VB class member, VB1 (Fig 4A). The change in NeuroPAL coloring of VB2 can be attributed to loss of acr-5 expression, which normally distinguishes VB2 from VB1 [34]. Previous cell fate marker analysis from our lab [34,66,67], as well as recent scRNA-seq analysis [19,63] revealed a number of additional genes that are differentially expressed in VB2 versus VB1, which allowed us to test whether hlh-17/31/32 may indeed work to distinguish the identity of these neurons. We first interrogated markers shared by VB2 and VB1 but expressed at different levels, according to scRNA transcriptomes [19,63], namely the neuropeptides flp-32 and flp-7. Using CRISPR/Cas9-engineered reporter alleles, we found that the difference in expression level of these markers that exists between VB2 and VB1 in wild-type animals is abolished in hlh-17/31/32^null mutants (Fig 4B and 4C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. VB2 cell fate analysis of hlh-17, hlh-31, and hlh-32 mutant animals.

(A) In hlh-17/31/32^null mutants, VB2 adopts a VB1-like NeuroPAL (otIs669) coloring. (B) In hlh-17/31/32^null mutants, flp-32(syb4374) reporter allele expression in VB2 is decreased, reaching the same level of expression as that in VB1. (C) The difference in flp-7(syb5413) reporter allele expression levels between VB2 and VB1 is abolished in hlh-17/31/32^null mutants. (D) Expression of the wgIs707 fosmid-based reporter for transcription factor sptf-1 is normally restricted to VB1 but becomes derepressed in VB2 in hlh-17/31/32^null animals. (E) The nlp-45(ot1032) reporter allele marks VB1 alone in wild type but is ectopically expressed in VB2 in hlh-17/31/32^null animals. (F) The col-105(syb6767) reporter allele is a marker for VB1 that is ectopically expressed in VB2 in hlh-17/31/32^null animals. (G) The reporter allele for transcription factor bnc-1(ot845) loses expression in VB2 in hlh-17/31/32^null animals. (H) hlh-17/31/32^null mutants retain strong expression of unc-17(syb4491) in VB2 and VB1. (I) Expression of the pdf-1(syb3330) reporter allele in VB2 and VB1 remain unchanged in hlh-17/31/32^null mutants. (J) Schematic summarizing the relative expression of markers expressed in either VB2, VB1, or both, in wild-type and hlh-17/31/32^null animals. Differential expression of markers between VB2 and VB1 is abolished in hlh-17/31/32^null animals. The acr-2 and acr-5 reporters are part of the NeuroPAL transgene. Representative images of wild-type and mutant animals are shown with 10 μm scale bar. For panels D, E, and F, quantitative comparison of the fluorescence between VB2 and VB1 in hlh-17/31/32^null animals is in S2C Fig. For panels B, C, and I, statistical analysis was performed using one-way ANOVA with Tukey multiple comparisons test. Each point represents the corresponding neuron from one individual and the number of animals scored are enclosed in parentheses. Error bars indicate standard deviation of the mean. ****p ≤ 0.0001, *p ≤ 0.05, ns: not significant. For panels D–G, statistical analysis was performed using Fisher’s exact test, with the number of animals scored shown within each bar. Raw data for panels B–G and I can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3002979.g004

If VB2 indeed acquires VB1 fate, markers exclusive to VB1 should become ectopically expressed in the VB2 neurons of hlh-17/31/32^null mutants. To examine such a potential VB2-to-VB1 transformation, we sought markers normally present only in VB1 but not VB2. A fosmid-based reporter transgene of the sptf-1 zinc finger transcription factor had previously been shown to be expressed in VB1, but not VB2 [19] (Fig 4D). Moreover, we found that an nlp-45 neuropeptide reporter allele, previously described to be in either VB1 or VB2 [67], is expressed in VB1, not VB2 (Fig 4E). In addition, based on scRNAseq data [19,63], we CRISPR/Cas9-engineered a reporter allele for the collagen col-105, and confirmed expression in VB1, not VB2 (Fig 4F). Armed with these 3 VB1(+) VB2(-) markers, we indeed found that all 3 (sptf-1, nlp-45, and col-105) become ectopically expressed in the VB2 neuron of hlh-17/31/32^null triple mutant animals at the same intensity as VB1 (Figs 4D–4F and S2C). Lastly, a zinc finger transcription factor, bnc-1, normally expressed in VB2 (and all other B-type motor neurons), but not VB1 [66], fails to be properly expressed in VB2 in hlh-17/31/32^null mutants (Fig 4G). In contrast, markers that are expressed at similar levels between VB2 and VB1, such as reporter alleles for the pan-cholinergic marker unc-17/VAChT, the pdf-1 neuropeptide or the acr-2 AChR, are unaffected in hlh-17/31/32^null mutants (Fig 4H and 4I; acr-2 is located on the NeuroPAL transgene; Fig 4A). Taken together, hlh-17/31/32 are neither required for the generation of VB2 nor for the adoption of B-motor neuron fate but differential gene expression that normally exists between VB2 and VB1 in wild-type animals is abolished in hlh-17/31/32^null animals, with VB2 taking on a VB1-like gene expression profile (Fig 4J). In other words, hlh-17/31/32 instruct VB2 to become different from VB1 and an absence of these genes results in a homeotic neuronal identity transformation from VB2 to VB1, based on molecular markers (Fig 4J).

Behavioral analysis of hlh-17 hlh-31 hlh-32 triple mutant animals

hlh-17 mutants were previously reported to display defects in the defecation motor program, which was ascribed to hlh-17 function in the CEPsh glia [68]. We used our more unambiguous hlh-17 null allele, in combination with the hlh-31 and hlh-32 null alleles, to avoid any potential issues of genetic redundancies among these genes, to confirm these defects. Testing both the posterior body wall contraction (pBOC) and the enteric muscle contraction (EMC) step of the defecation motor program, we observed no defects in hlh-17/31/32^null triple mutant animals (S3A Fig).

We also examined locomotory behavior of hlh-17/31/32^null mutant animals using an automated worm tracker system [35] and observed a number of defects, including in reversal behavior and speed (S3B Fig). We have not pursued the question of whether those defects are due to AUA and/or VB2 differentiation defects. We note, however, that whole brain activity recordings have shown striking correlations of the activity patterns of AUA with the command interneuron AVA [34], which controls several locomotory features that we find to be defective in hlh-17/31/32^null mutant animals.

Exclusive expression of the PTF1a ortholog HLH-13 in the octopaminergic RIC interneurons

The C. elegans genome encodes a single ortholog of the 2 paralogous vertebrate bHLH genes PTF1a and NATO3 (aka FERD3L), called hlh-13 [8,9,69] (Fig 1B). A previous analysis of C. elegans hlh-13/PTF1a suggested expression in dopamine neurons, based on a small reporter transgene that contained only 2.1 kilobases upstream of the gene [70,71]. However, the reported position of the GFP(+) cells in larval/adult animals is not fully consistent with expression in dopamine neurons and scRNA-seq data also suggests expression in, at most, a subset of dopaminergic neurons, plus other neurons that showed no expression of the previous reporter transgene [19]. To resolve these issues, we tagged the endogenous hlh-13/PTF1a gene locus with gfp, using CRISPR/Cas9 engineering (Fig 1C). Using the NeuroPAL cell ID tool [34], we observed strong HLH-13 protein expression exclusively in the sole octopaminergic neuron class in C. elegans, RIC [72], throughout all larval and adult stages (Fig 5A). This is consistent with scRNA-seq data of L4 stage animals [19] (Fig 1A). RIC is also among the very few neuron classes that continuously express the E/Da protein HLH-2 throughout postembryonic life (Fig 1A) [15,73]. Together with their documented capacity to interact in yeast 2-hybrid assays [48], this co-expression indicates that like its vertebrate ortholog [74], HLH-13/PTF1A may heterodimerize with the E protein HLH-2 in vivo.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. PTF1a and NATO3 ortholog HLH-13 is expressed in RIC and controls its differentiation.

(A) Expression of the hlh-13 CRISPR/Cas9-engineered reporter allele syb7685 over the course of development. Overlay with the NeuroPAL transgene otIs669 shows expression in RIC neurons. (B) Representative pictures and quantification showing loss of NeuroPAL (otIs669) coloring, but retention of panneuronal signal, in the RIC neurons of hlh-13/PTF1a mutants. Animals were scored at the young adult stage. Statistical analysis was performed using Fisher’s exact test. N is indicated within each bar and represents number of animals scored. (C) Representative pictures and quantification showing loss of RIC cell fate markers in hlh-13/PTF1a mutants. tbh-1 expression in the gonadal sheath is unaffected by the loss of hlh-13/PTF1a. cat-1 expression of the dopaminergic CEP neurons is not affected in hlh-13/PTF1a mutant animals. Reporter genes used are promoter fusion transgene for tbh-1(nIs107) and CRISPR/Cas9-engineered reporter alleles for tdc-1(syb7768), cat-1(syb6486), and flp-32(syb4374). Because hlh-13 and cat-1 are linked, the exact same mutation as hlh-13(ot1388) (Fig 1C) was generated in the background of the cat-1(syb6486) reporter; this allele is hlh-13(ot1390). Animals were scored at the young adult stage. Statistical analysis was performed using Fisher’s exact test. N is indicated within each bar and represents number of animals scored. Raw data for panels B and C can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3002979.g005

In the embryo, we observe HLH-13 protein expression in 3 postmitotic neuron pairs that resolves to expression in a single neuron pair (RIC) before hatching (Fig 5A). Since hlh-13/PTF1a scRNA transcripts are observed in embryonic (but not postembryonic) CEP neurons [19,75], we crossed the hlh-13/PTF1a reporter allele with a transgenic line that expresses RFP in dopaminergic neurons including CEP, but a lack of overlap of the fluorescent signals ruled out that these neurons are the embryonic CEP neurons. Based on embryonic lineaging data of an hlh-13/PTF1a reporter transgene by the EPIC project [76], these HLH-13(+) cells may be the sister of the RIC neurons that are fated to die during embryogenesis and the 2 cousins of the RIC neurons, the SIBD neurons.

Octopaminergic RIC neurons fail to properly differentiate in hlh-13/PTF1a mutant animals

We used 2 mutant alleles to analyze hlh-13/PTF1a function: (a) a deletion mutant, tm2279, generated by the National BioResource Project (NBRP) knockout consortium, which deletes most of the locus, including most of its bHLH domain; and (b) ot1388, a complete locus deletion that we generated by CRISPR/Cas9 genome engineering (Fig 1C). Animals carrying either allele are viable, fertile and display no obvious morphological abnormalities. We assessed the function of hlh-13/PTF1a in neuronal differentiation by again using the NeuroPAL transgene. The RIC neuron class, the only cell type with strong and consistent postembryonic expression, display a loss of both markers on the NeuroPAL transgene, ggr-3 and mbr-1 (Fig 5B). Persistence of the NeuroPAL panneuronal signal in the RIC neurons of hlh-13/PTF1a mutant animals demonstrates that the RIC neurons are still generated (Fig 5B).

We extended our analysis of the differentiation defects in RIC interneurons by examining additional markers of their octopaminergic identity, including the 2 biosynthetic enzymes TDC-1 and TBH-1 [72]. TDC-1, an aromatic amino acid decarboxylase, is exclusively expressed in the octopaminergic RIC and tyraminergic RIM neurons to convert tyrosine to tyramine and TBH-1, a tyramine hydroxylase, is exclusively expressed in RIC to convert tyramine to octopamine [72]. We found that expression of both a tbh-1 reporter transgene and a tdc-1 reporter allele [77] is eliminated in the RIC neurons of hlh-13/PTF1a mutants (Fig 5C). Moreover, a CRISPR/Cas9-engineered reporter allele of the monoaminergic vesicular transporter CAT-1/VMAT [77] also fails to be expressed in the RIC neurons of hlh-13/PTF1a mutant animals (Fig 5C). Expression of cat-1/VMAT is unaffected in dopaminergic neurons in hlh-13/PTF1a mutants (Fig 5C), consistent with hlh-13/PTF1a not being expressed in dopaminergic neurons.

The RIC neuron pair also expresses a unique signature of neuropeptide-encoding genes [19]. We used a CRISPR/Cas9-engineered reporter allele of one of them, flp-32 [78], and found that its expression is eliminated in hlh-13/PTF1a null mutants (Fig 5C).

We examined the fate of other neurons that are predicted to express much lower levels of hlh-13/PTF1a mRNA based on embryonic or postembryonic scRNA-seq data, including all dopaminergic neurons, and found no differentiation defects using NeuroPAL. The abovementioned flp-32 RIC marker is also expressed in ALA (in which hlh-13 transcript but no HLH-13::GFP protein can be detected), but we observed no flp-32 expression defects in ALA. We conclude that hlh-13/PTF1a selectively affects RIC neuron differentiation.

Physiological and behavioral consequences of hlh-13/PTF1a loss

The hlh-13-expressing RIC neuron pair is the only neuron class that produces octopamine [72], an endocrine monoaminergic regulator, analogous to vertebrate norepinephrine, that links nutrient cues to lipolysis to maintain energy homeostasis [79]. We confirmed that loss of tbh-1 results in an up-regulation of lipid stores in the intestine, both under well-fed and starved conditions (Fig 6A), as previously reported [79]. Consistent with hlh-13/PTF1a controlling RIC differentiation, and hence octopamine synthesis and release, we find that hlh-13/PTF1a mutant animals also display improper fat accumulation in the intestine, both under well-fed and starved conditions (Fig 6A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Physiological consequences of loss of the PTF1a and NATO3 ortholog hlh-13.

(A) hlh-13(ot1388) and tbh-1(n3247) mutant animals display comparable nile red lipid staining defects in well-fed and in starved animals. Statistical analysis was performed using one-way ANOVA with Dunnett’s multiple comparisons test. Error bars indicate standard deviation of the mean. Number of animals scored for each genotype is enclosed in parentheses. ****p ≤ 0.0001, ***p ≤ 0.001, *p ≤ 0.05. (B, C) Two independent hlh-13/PTF1a mutant alleles do not phenocopy the previously reported [40] (and herein repeated) locomotory defects of tbh-1(n3247) mutant animals, indicating that the cellular source of octopamine required for proper locomotory behavior may not be the RIC neurons. Error bars indicate mean ± SEM. ***P < 0.001, **P < 0.01, ns: not significant, compared with wild-type, Dunnett’s multiple comparison tests. Raw data for panels A–C can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3002979.g006

As another adaptation to changes in food availability, octopamine also controls dwelling behavior. Quantitative locomotory analysis of feeding C. elegans showed that tbh-1 mutant animals exhibit a larger fraction of dwelling and smaller fraction of roaming compared to wild-type worms [40]. Interestingly, this difference is not recapitulated in hlh-13/PTF1a mutant animals (Fig 6B and 6C), which exhibit roaming and dwelling at rates similar to those of wild-type worms. Since octopamine is produced not only in the RIC neurons, but also in gonadal sheath cells [72], and since we find gonadal expression of tbh-1 to be unaffected by loss of hlh-13/PTF1a (Fig 5C), we surmise that octopamine from the gonadal sheath cells may be sufficient for modulation of dwelling behavior.

Exclusive expression of the NHLH1 and NHLH2 ortholog HLH-15 in 2 neuron classes

Previous phylogenetic analysis has shown that the 2 paralogous vertebrate bHLH genes NSCL1/NHLH1/HEN1 and NSCL2/NHLH2/HEN2 have a single C. elegans ortholog, hlh-15 [8,9,69] (Fig 1B). A promoter-based transgene of hlh-15/NHLH was previously reported to be expressed in the tail DVA neuron and either the RIF or RIG head interneurons [48]. We used the CRISPR/Cas9 system to tag the hlh-15/NHLH locus with gfp (Fig 1C) and again used the NeuroPAL cell ID tool to assess the sites of hlh-15::gfp expression [34]. Throughout all postembryonic stages and adulthood, we observe HLH-15/NHLH protein expression exclusively in 3 neurons, the 2 peptidergic bilateral AVK interneurons AVKL and AVKR as well as the unpaired cholinergic tail interneuron DVA (Fig 7A–7C). The exclusive expression in these 2 neuron classes commences at around the comma/1.5-fold stage and persists throughout all larval and adult stages. Both the embryonic and postembryonic reporter allele expression are consistent with scRNA-seq data from embryos and L4 stage animals [19,75] (Fig 1A). We re-examined the previously published promoter-reporter fusion construct for hlh-15 and found that this construct is indeed also expressed in DVA and the AVK neurons (and not in RIF or RIG, as previously described; [48]). AVK and DVA are also among the very few neuron classes that continuously express the E/Da homolog HLH-2 throughout postembryonic life (Fig 1A) [15,73]. This indicates that HLH-15/NHLH may, like its vertebrate homolog [80], heterodimerize with the E protein HLH-2, a notion supported by HLH-15::HLH-2 protein interactions in yeast 2-hybrid assays [48].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. NHLH1 and NHLH2 ortholog HLH-15 is expressed in AVK and DVA neurons.

(A–C) Expression of the hlh-15/NHLH CRISPR/Cas-9-engineered reporter allele syb7688 over the course of development. Overlay with the NeuroPAL transgene otIs669 shows expression in AVK and DVA neurons. (D) hlh-15/NHLH autoregulates its own expression in AVK and DVA neurons. Top panel shows hlh-15 locus. The whole upstream intergenic region (from ATG to −777 bp) was fused to gfp to generate wwEx42 [48]. Representative pictures and quantification showing expression of hlh-15/NHLH promoter fusion (wwEx42) in hlh-15/NHLH mutants. Animals were scored at the L1 (middle panel) and young adult stage (bottom panel). Statistical analysis was performed using Fisher’s exact test. N is indicated within each bar and represents number of animals scored. Raw data for panel D can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3002979.g007

Continuous expression of transcription factors throughout the life of a cell often reflect transcriptional autoregulation [81,82]. To ask whether hlh-15/NHLH autoregulates, we utilized a previously described transcriptional reporter of the hlh-15/NHLH locus in which its intergenic 5′ promoter region was fused to gfp [48]. This reporter recapitulates continuous expression in the AVK and DVA neurons (Fig 7D). In an hlh-15/NHLH null mutant background, expression of the reporter is initially normally observed in AVK and DVA, but it fades during postembryonic life to result in complete absence at the adult stage (Fig 7D). This observation demonstrates autoregulation of the hlh-15/NHLH locus.

The peptidergic AVK neuron class fails to differentiate properly in hlh-15/NHLH mutant animals

We used 2 mutant alleles to analyze hlh-15/NHLH function: (a) a deletion mutant, tm1824, generated by the NBRP knockout consortium, which deletes most of the locus, including most of its bHLH domain; and (b) a complete locus deletion, ot1389, that we generated by CRISPR/Cas9 genome engineering (Fig 1C). Animals carrying either allele are fully viable and display no obvious morphological abnormalities. Unlike their vertebrate counterparts [28], hlh-15/NHLH mutants display no obvious fertility defects.

We assessed the potential function of hlh-15/NHLH in controlling fate specification of the AVK and DVA neurons by first using NeuroPAL, the marker transgene in which all neurons express specific codes of cell fate markers [34]. In NeuroPAL, the AVK neurons are marked with the flp-1 neuropeptide gene and DVA is marked with the nlp-12 neuropeptide and the 2 ionotropic glutamate receptors nmr-1 and glr-1. NeuroPAL colors in hlh-15/NHLH mutants indicate a loss of the AVK color code (flp-1 expression), but no effect on the color code of the DVA neuron (Fig 8A and 8B). We confirmed the proper differentiation of DVA in hlh-15/NHLH mutants using another DVA fate marker, a reporter allele for the acetylcholine vesicular transporter unc-17/VAChT, whose expression we found to be unaffected as well (Fig 8A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. hlh-15/NHLH gene affects the differentiation of the AVK interneurons.

(A) Representative pictures and quantification showing unaffected NeuroPAL (otIs669) coloring in DVA of hlh-15/NHLH mutants (left panel). The expression of the CRISPR/Cas9-engineered reporter allele unc-17(syb4491) in DVA is also not affected (right panel). (B) Representative pictures and quantification showing loss of NeuroPAL (otIs669) coloring in AVK (left panel) in hlh-15/NHLH mutants. Despite loss of NeuroPAL coloring, AVK is still generated in hlh-15/NHLH mutants, as visualized by DIC microscopy (right panel). (C) Representative pictures and quantification showing loss of cell fate markers in hlh-15/NHLH mutants. Reporter transgenes used are flp-1(bwIs2), dop-3(vsIs33), and CRISPR/Cas9-engineered reporter alleles nlp-49(syb3320), nlp-50(syb6148), nlp-69(syb4512), tkr-1(syb4595), twk-47(bab385), and unc-42(ot986). Animals were scored at the young adult stage. Statistical analysis was performed using Fisher’s exact test. N is indicated within each bar and represents number of animals scored. Raw data for panels A–C can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3002979.g008

We confirmed the AVK differentiation defects of hlh-15/NHLH null mutants using another transgene to measure flp-1 expression, bwIs2 [83] (Fig 8C). Given that AVK is a unique neuropeptidergic hub interneuron, receiving and sending a multitude of neuropeptidergic signals [84], we further probed the acquisition of AVK’s neuropeptidergic identity in hlh-15/NHLH mutant animals. Specifically, we probed the peptidergic identity of AVK by CRISRP/Cas9-engineered reporter alleles for 3 additional neuropeptides, predicted by scRNA-seq data to be expressed in AVK, nlp-49, nlp-50, and nlp-69 [19]. nlp-49 was previously also reported to be required in AVK to control motor behavior [33,85] and is, next to flp-1, the neuropeptide that is most selectively expressed in AVK [19]. We found that expression of nlp-49 in AVK is strongly affected in hlh-15/NHLH mutant animals, while expression in other nlp-49(+) neurons is unaffected (Fig 8C). Similarly, expression of the nlp-50 and nlp-69 reporter alleles is also selectively lost in the AVK neurons of hlh-15/NHLH mutants (Fig 8C).

Apart from transmitting neuropeptidergic signals, AVK also expresses a multitude of neuropeptidergic receptors [19,84]. We examined one of them, the tachykinin receptor tkr-1, expressed in AVK and many other head neurons [84], and found its expression in AVK to be disrupted in hlh-15/NHLH mutant animals (Fig 8C). AVK also receives dopaminergic signals via the AVK-expressed dopamine receptor dop-3 [41], and we find dop-3 expression to also be disrupted in hlh-15/NHLH mutant animals (Fig 8C).

The nervous system-wide scRNA atlas of C. elegans predicts another unique identity marker of the AVK neurons, a 2 pore TWIK-type potassium channel, twk-47 [19]. Its selective expression in AVK was recently validated with a promoter-based transgene construct [86] and is further validated with a CRISPR/Cas9-genome engineered, wrmScarlet-based reporter allele (kindly provided by T. Boulin). We find that expression of this twk-47 reporter allele is also eliminated from the AVK neurons of hlh-15/NHLH mutants (Fig 8C).

Contrasting effects on neuron type-specific terminal “function” genes, we found that hlh-15/NHLH null mutant animals show normal expression of a previously described terminal selector of AVK differentiation, the unc-42 homeobox gene [87,88], and also retain AVK as visualized by DIC microscopy (Fig 8B and 8C). These observations indicate that (a) hlh-15/NHLH is not acting as a proneural gene to determine the generation of the AVK neurons; and (b) that hlh-15/NHLH may work together, perhaps in a heteromeric, combinatorial manner, with unc-42 as a terminal selector of AVK identity (see Discussion).

Proprioceptive defects in hlh-15/NHLH mutant animals

The AVK neurons were previously shown to act downstream of dopaminergic neurons to control proprioceptive behavior, a process mediated by the AVK-released and hlh-15-controlled flp-1 gene [41]. Given that hlh-15/NHLH controls flp-1 expression in AVK (the only C. elegans neuron that expresses flp-1), we predicted that hlh-15/NHLH mutants might display similar defects in proprioceptive behavior. We tested this prediction by measuring the AVK- and flp-1-dependent CCR in animals carrying either of the 2 available hlh-15/NHLH deletion alleles and found a severe reduction of the response, mimicking the genetic ablation of AVK (Fig 9A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Behavioral and physiological analysis of hlh-15/NHLH mutants.

(A) Measurement of the CCR. C. elegans partially constrained within a microfluidic channel [41]. a: anterior region, p: posterior region, c: constrained middle region. For each strain, the normalized anterior curvature change is equal to the anterior curvature amplitude during microfluidic constraint minus the anterior curvature amplitude of freely moving worms, divided by the anterior curvature amplitude of freely moving worms. hlh-15(tm1824) and hlh-15(ot1389) null mutants display defects in their CCR. These defects are similar to those observed upon genetic ablation of AVK using a transgenic array, zxIs28[flp-1p(trc)::ICE; myo-2p::mCherry], in which the ICE caspase was driven by an AVK-specific flp-1 promoter [33]. N ≥ 10 animals per condition. Error bars indicate mean ± SEM. ***P < 0.001, Dunnett’s multiple comparison tests. (B) hlh-15(tm1824) and hlh-15(ot1389) null mutants show increased lifespan. Survival plot of N2, hlh-15(tm1824), and hlh-15(ot1389) animals. (C) hlh-15/NHLH null mutations attenuate the reduction in pharyngeal pumping rate associated with aging. Quantification of pharyngeal pumping rate for N2, hlh-15(ot1389), and hlh-15(tm1824) at day 1 young adult and day 11 aged adult animals [45]. (D) Day 1 adult hlh-15/NHLH null mutants show reduced size of hypodermal cell nucleoli, a prospective hallmark of longevity. Representative DIC images of hypodermal nucleoli (yellow stippled lines; within nucleus, circled with red stippled line) and measurements of nucleolar size for N2, hlh-15(ot1389), and hlh-15(tm1824) are shown with 10 μm scale bar. The quantification panel on the left shows the size of individual hypodermal nucleoli while the panel on the right shows average hypodermal nucleolar size for each animal. (E) Lifespan extension defects of hlh-15(ot1389) mutants are rescued by expressing hlh-15/NHLH selectively in the AVK neurons (flp-1p::hlh-15). Survival plot of N2, hlh-15(ot1389), and 2 independent transgenic lines (otEx8247 and otEx8248) of hlh-15(ot1389) animals expressing genomic hlh-15 DNA specifically in AVK. This assay could not be performed with the AVK::ICE lines because animals had a sick appearance. (F) Removal of neuropeptide processing in AVK neurons extends lifespan. Survival plot of floxed egl-3(nu1711) animals with the transgene (kyEx6532) or without the transgene (siblings) that expresses Cre recombinase specifically in AVK. For panels B, G, and H, statistics were performed using Log-Rank Test followed by Bonferroni corrections. N = 105 for each strain. All lifespan assays were replicated twice, with a repeat experiment for each shown in S4 Fig. Mean and 75% mortality lifespan tables can be found in S2 Table. For panel C, two-way ANOVA was performed, while for panels E and F, one-way ANOVA was performed, followed by Bonferroni correction. Raw data for panels A–F can be found in S1 Data. CCR, compensatory curvature response; DIC, differential interference contrast.

https://doi.org/10.1371/journal.pbio.3002979.g009

hlh-15/NHLH acts in the AVK interneuron class to regulate life- and health span

RNAi knockdown of hlh-15/NHLH, as well as knockdown of its presumptive, class I heterodimerization partner hlh-2, has previously been reported to extend lifespan of C. elegans [89,90]. We found that the reported lifespan extension effects of hlh-15(RNAi) animals are recapitulated in animals that carry either of the 2 independent deletion alleles of the hlh-15/NHLH locus (Fig 9B). We extended the lifespan analysis of hlh-15 mutants by considering health span of hlh-15/NHLH mutants. Health span can be measured by examining feeding behavior (pharyngeal pumping) in aged, but still alive worms [45]. We find this measure of health span to also be improved in both hlh-15/NHLH mutant strains (Fig 9C). Lastly, we tested a prospective hallmark of longevity, smaller nucleolar size [46] and found that in both mutant hlh-15/NHLH alleles, this parameter also shows the expected feature of long-lived animals (Fig 9D).

The previous RNAi-based analysis of hlh-15/NHLH [89] had not considered the potential cellular focus of action of hlh-15/NHLH for lifespan control. Since loss of hlh-15/NHLH only affects the differentiation of the AVK neurons, and not the only other neuron (DVA) in which hlh-15/NHLH is expressed, we tested the possibility that hlh-15/NHLH acts in AVK to control animal lifespan. Specifically, we asked whether the extended lifespan of hlh-15/NHLH is restored to normal wild-type–like lifespan by resupplying hlh-15/NHLH selectively in the AVK interneurons. We found that transgenic animals that re-express hlh-15/NHLH from an AVK-specific promoter in an hlh-15/NHLH null mutant background indeed displayed a rescue of the lifespan extension effect of hlh-15 mutants (Figs 9E and S4), indicating that a single interneuron has the capacity to control animal lifespan.

Previous work has shown lifespan extension of animals in which the expression of an enzyme involved in branched-chain amino acid metabolism, bcat-1 (branched-chain amino acid transferase-1), was reduced by RNAi [89]. Moreover, animal-wide RT-qPCR analysis suggested that levels of bcat-1 transcripts are reduced in hlh-15(RNAi) animals [89]. A putative HLH-15/NHLH binding site in the bcat-1 promoter made these authors suggest that HLH-15/NHLH directly controls bcat-1 expression. We sought to independently probe the potential link of hlh-15/NHLH and bcat-1 with alternative approaches. To this end, we first tagged the bcat-1 locus with gfp using CRISPR/Cas9 genome engineering. As expected from a metabolic enzyme, we found very broad (albeit not entirely ubiquitous) and robust expression of bcat-1 across all major tissue types (S5 Fig). However, in an hlh-15/NHLH null mutant background, bcat-1 expression appeared unchanged, in both young and aged adult animals (S5 Fig). While this experiment does not entirely rule out that the lifespan extending effects of hlh-15/NHLH may be mediated by bcat-1, it indicates that hlh-15/NHLH and AVK may act through other means to control animal lifespan.

One set of possible lifespan modulating effector genes of hlh-15/NHLH are neuropeptides secreted by AVK. Based on scRNA-seq data, as well as reporter gene data, AVK shows highly selective expression of at least 8 neuropeptide-encoding genes, some of which are deeply conserved throughout animal phylogeny [19,85,91,92]. As demonstrated above, the expression of at least some, if not all AVK-expressed neuropeptides is under hlh-15/NHLH control (Fig 8C). Rather than venturing into testing lifespan effects of individual neuropeptide-encoding genes, we eliminated all neuropeptide processing selectively in the AVK neurons, using a conditional, floxed allele of the egl-3 proprotein convertase required for neuropeptide processing [93,94]. We found that animals in which we removed egl-3 in AVK also showed an increased lifespan (Fig 9F). We conclude that the lifespan-extending effect of hlh-15/NHLH can be explained via hlh-15/NHLH controlling the neuropeptidergic phenotype of AVK and, more generally, that a single interneuron class has the capacity to influence the lifespan of an animal.

Revising expression patterns of conserved bHLH proteins in C. elegans

We have discovered functions of several, phylogenetically conserved bHLH proteins during terminal differentiation of postmitotic neurons, thereby expanding our understanding of bHLH gene function in the nervous system. The 3 clustered bHLH genes hlh-17/31/32 were originally considered orthologs of Olig genes, which are important regulators of glia and motor neuron differentiation in vertebrates [95–97]. However, by sequence homology, HLH-17, HLH-31, and HLH-32 are the orthologs of the vertebrate bHLHe22 and bHLHe23 proteins rather than the closely related vertebrate Olig proteins. Unlike the vertebrate Olig proteins, vertebrate bHLHe22/e23 have no reported function in glia cells, but rather act in neuronal differentiation in select neuron classes in the retina, telencephalon, and spinal cord [20–26,59]. The Drosophila bHLHe22/e23 homolog also functions in motor neurons rather than glia [59]. Similarly, we find here that the C. elegans bHLHe22/e23 orthologs HLH-17 and HLH-32 are expressed and function in neurons, including motor neurons, but not glia. We cannot exclude that HLH-17/31/32 proteins are expressed below the levels of reporter allele-based detectability in the CEPsh glia cells. However, given the robust transcription of the hlh-17/31/32 genes in CEPsh glia (based on scRNA-seq and promoter-fusion reporters), we consider posttranscriptional repression the most parsimonious explanation. Such posttranscriptional regulation is evident upon comparison of mRNA transcripts and homeodomain protein expression in several distinct neuron types [12,19].

We also found that the expression pattern of a reporter allele of the PTF1a homolog hlh-13 is different from the previously reported expression of a multicopy hlh-13/PTF1a reporter transgene [70,71]. However, in this case, we do not infer posttranscriptional regulation as a source for differences, since our reporter allele data is consistent with scRNA-seq data. We rather consider previous reporter genes to either produce incorrect expression and/or the sites of expression have been incorrectly identified, a problem remedied by our usage of the neuronal landmark strain NeuroPAL.

Lastly, NeuroPAL has also been instrumental in defining the proper identity of hlh-15/NHLH expressing cells, which had only been incompletely achieved with previous reporter transgenes [48]. Taken together, our expression pattern analysis of these 5 conserved bHLH genes illustrated the importance of using the best possible tools (reporter alleles and cell ID tools) to properly infer protein expression patterns.

Function of bHLH genes in neuronal identity specification

We discovered terminal neuronal differentiation defects for each of the bHLH subfamilies that we examined in this study. The bHLHe22/e23 orthologs hlh-17 and hlh-32 function as “terminal subtype selectors” that diversify motor neuron subclass identity in the retrovesicular ganglion that harbors neurons akin to vertebrate branchial motor neurons. Loss of these factors results in a homeotic subtype identity transformation akin to those observed in C. elegans HOX mutants at the posterior end of the ventral nerve cord [98,99]. In another neuron class, AUA, which is potentially also involved in motor control [34], hlh-17 and hlh-32 affect select, but not all aspects of the differentiation program. The function of hlh-17 and hlh-32 in motor neuron differentiation is reminiscent of the function of Drosophila and vertebrate homologs of these genes, which play a role in various aspects of motor neuron specification [59,95,100,101], but we added here some intriguing granular detail to such motor neuron function. While anatomical reconstructions [102] and, more recently, scRNA-seq analysis [19,63], have revealed that the most anteriorly located B-type motor neurons within the retrovesicular ganglion are morphologically and molecularly distinct from other B-type motor neurons along the ventral nerve cord, no molecular mechanisms for their subclass diversification have been previously identified. We found that hlh-17 and hlh-32 do not affect the expression of features shared by B-type neuron types, but instead promote the expression of features that are characteristic of one subtype (VB2), while repressing features characteristic of another subtype (VB1). Generally, all motor neuron markers, including B-type motor neuron markers and at least some, if not all subtype-specific markers described here, are directly controlled by the terminal selector unc-3, an EBF-type transcription factor [66,103]. Akin to other subtype diversifiers (e.g., BNC-1) [66], we propose that HLH-17/32 act as subtype diversifiers to either directly or indirectly promote or antagonize the ability of UNC-3 to turn on select target genes.

hlh-13/PTF1a and hlh-15/NHLH appear to act in a canonical terminal selector-type manner in 2 different neuron classes, hlh-13/PTF1a in the octopaminergic RIC interneuron class and hlh-15/NHLH in the peptidergic AVK interneuron class. In both neuron classes, loss of the respective bHLH gene affects the expression of all markers tested, but they do not affect the generation of the neurons and they are continuously expressed throughout the life of the neuron to possibly maintain their identity. Such continuous expression is likely ensured by transcriptional autoregulation, as we have explicitly shown for hlh-15/NHLH.

The terminal selector roles of hlh-13/PTF1a and hlh-15/NHLH appear similar to the role of the ASC-type hlh-4 bHLH as a validated terminal selector of the ADL sensory neuron class, where HLH-4 likely operates in conjunction with the common heterodimerization partner, the E/Da-homolog HLH-2, to co-regulate scores of terminal effector genes through direct binding to E-box motifs [17]. In both the AVK and RIC neuron classes, the HLH-15/NHLH and HLH-13/PTF1A protein, respectively, are also likely to heterodimerize with the common class I E/Da HLH-2 protein, based on: (a) co-expression of HLH-13/PTF1A and HLH-15/NHLH with HLH-2, which is expressed in 8 neuron classes, including the RIC and AVK neurons (ADL, ASH, PHA, PHB, RIC, AVK, DVA, and PVN) [15,73]; (b) physical interaction of HLH-15::HLH-2 and HLH-13::HLH-2 observed in protein-protein interaction assays [48]; (c) similar interactions of the vertebrate homolog of HLH-13 (PTF1a) and HLH-15 (NHLH1/2) with vertebrate E-proteins [74,80]. hlh-2 is not expressed in the hlh-17/31/32-expressing AUA, DB2, and VB2 neurons, but this is consistent with Olig-related proteins acting as strong homodimers [104]. It is notable that for some of the neurons that express the common E/Daughterless bHLH component HLH-2 postembryonically, no expression of other group A bHLH proteins is observed (Fig 1A). This indicates that in these neurons (e.g., ASH or PHB), hlh-2 may act as a homodimer, as it does in other non-neuronal cells [105]. A role of a terminal selector HLH-2 homodimer is conceivable because E-box motifs are enriched in the ASH and PHB terminal gene batteries [19].

Vertebrate homologs of hlh-13 and hlh-15 may have maintained their function as regulators of neuronal differentiation. The vertebrate bHLH PTF1a and its paralogue NATO3 have multiple functions both early and late during neuronal development [27,29]. The vertebrate hlh-15 homologs NHLH1 and NHLH2 control the proper differentiation, and perhaps also maintenance, of several peptidergic hormone-producing cells in the hypothalamus [28,106,107], reminiscent of the role of hlh-15/NHLH in controlling the identity of the peptidergic AVK neurons. Roles in initiating and maintaining terminal differentiation programs as putative terminal selectors have, however, not yet been explicitly investigated for those vertebrate homologs, something that should be tested through temporally controlled removal of vertebrate homologs specifically during or after terminal differentiation.

In conclusion, based on our mutant and expression analysis, the bHLH genes described here do not act as neural fate-determining classic proneuronal bHLH genes in progenitors, but rather act as drivers of terminal differentiation programs in postmitotic neurons.

Discovery of a lifespan-controlling peptidergic hub neuron

The very restricted expression of hlh-15/NHLH in exclusively 2 neuron classes, combined with its striking function in controlling animal lifespan led us to discover a function of a single interneuron class, AVK, in controlling lifespan of the animal. The impact of the C. elegans nervous system on animal lifespan has long been appreciated [108,109], but has largely been restricted to the sensory periphery, which releases distinct types of neuropeptides, mostly insulins, but also other types of neuropeptides to modulate animal aging [110–112]. We discover here a different type of neuron that controls lifespan, the AVK interneurons.

The AVK neuron class has been shown to regulate responses to various physiological states to control locomotion and behavior [33,41,85,93,113], but AVK has not previously been implicated in lifespan control. Moreover, AVK displays fundamental differences to neurons previously implicated in lifespan control. It is neither a sensory neuron, nor a main postsynaptic target of sensory neurons [102]. Its expression of a large number of neuropeptide receptors as well as of neuropeptides that act through receptors distributed throughout the animal nervous system has revealed AVK to be a central neuropeptidergic hub neuron [114] that likely controls internal states of the animal. We hypothesize that AVK may serve as an integrator of various internal states, resulting in the release of neuropeptide(s) that limit animal lifespan. It will be interesting to determine whether the release of these neuropeptides occurs continuously or only at specific time points during the life of the organism.

Both in terms of its peptidergic nature and also its developmental specification, AVK is conceptually remarkably similar to peptidergic neurons in the mammalian hypothalamus. Peptidergic hypothalamic neurons control internal states of mammals and have been implicated in lifespan control as well [115–117]. Strikingly, the mammalian homologs of the AVK-specifying hlh-15 gene, NHLH1 and NHLH2, control the proper specification of several classes of peptidergic neurons of the hypothalamus [28,106,107]. It will be intriguing to define with better resolution the exact type of hypothalamic neurons that continuously express and require NHLH1/2 during adulthood as this may help to more precisely determine—in analogy to HLH-15/NHLH expression and function—the nature of the neurons that may control lifespan in mammals. Intriguingly, the presently completely uncharacterized sole homolog of NHLH1/2 and hlh-15 in Drosophila, the HLH4C gene, appears to be expressed in the pars intercerebralis, which is considered to a hypothalamus-like structure in Drosophila [118].

bHLH genes and homeobox gene as regulators of neuronal identity

The bHLH proteins that we have characterized here are unlikely to act in isolation but can rather be expected to act in combination with other transcription factors to control terminal neuron differentiation as terminal selectors. Their most likely partners are homeodomain transcription factors. Each neuron class in C. elegans is defined by a unique combination of homeodomain proteins and mutant analysis has corroborated their widespread deployment as drivers of terminal neuron differentiation programs [12,13]. Based on previous mutant analysis, HLH-15/NHLH (likely in combination with HLH-2) may cooperate with the Prop1-type homeodomain protein UNC-42 in the AVK neurons, where UNC-42 was previously shown to display similar differentiation defects to those that we have described here [83,88]. Similarly, in the octopaminergic RIC neurons, HLH-13/PTF1A likely acts together with the Pbx-type homeodomain protein UNC-62, whose loss also results in RIC differentiation defects [13]. And even though hlh-17/31/32 may only control aspects of AUA differentiation, a potential cofactor is yet another homeobox gene, the POU homeobox gene ceh-6, which controls AUA differentiation [119]. Direct partnerships between bHLH and homeodomain proteins have been observed in other systems as well [120–122].

Taking a step back and comparing the relative functional deployment of the 42 bHLH family members and 102 homeobox genes in inducing and maintaining terminal differentiation programs in C. elegans, it is apparent that homeobox genes cover the differentiating nervous system much more broadly than bHLH factors do. Mutant analyses of many bHLH and homeobox genes support the notion that with notable exceptions (such as those described here), bHLH genes tend to be more prominently involved in early patterning events, while homeobox genes appear to be biased toward controlling terminal differentiation events in the nervous system. It will be fascinating to see whether this is an evolutionary conserved theme.