To investigate the neurotransmitter identity of neurons throughout the entire C. elegans nervous system of both sexes, we consider here the expression pattern of the following 15 genetic loci (see also Figure 1A):

For all these 15 genetic loci, we compared scRNA transcriptome data from the CeNGEN scRNA atlas (at all four available stringency levels; Taylor et al., 2021) to previously published reporter and antibody staining data. As shown in Figure 1B and Supplementary file 1, such comparisons reveal the following: (a) scRNA data support the expression of genes in the vast majority of neurons in which those genes were found to be expressed with previous reporter gene approaches. In most cases, this is true even at the highest threshold levels for scRNA detection. (b) Vice versa, reporter gene expression supports scRNA transcriptome data for a specific neurotransmitter system in the great majority of cells. (c) In spite of this congruence, there were several discrepancies between reporter data and scRNA data. Generally, while valuable, scRNA transcriptome data cannot be considered the final word for any gene expression pattern assignments. Lack of detection of transcripts could be a sensitivity issue and, conversely, the presence of transcripts does not necessarily indicate that the respective protein is generated, due to the possibility of posttranscriptional regulation.

Hence, to consolidate and further improve neurotransmitter identity assignment throughout the entire C. elegans nervous system, and to circumvent potential limitations of multicopy, fosmid-based reporter transgenes on which previous neurotransmitter assignments have been based, we engineered and examined expression patterns of 16 knock-in reporter alleles of the 15 neurotransmitter synthesis, vesicular transport, and uptake loci listed above (Figure 1, Figure 2). For unc-17 and eat-4, we knocked-in a t2a::gfp::h2b (his-44) cassette right before the stop codon of the respective gene. For unc-25, we created two knock-in alleles with the t2a::gfp::h2b (his-44) cassette tagging isoforms a.1/c.1 and b.1 separately. For tdc-1, a gfp::h2b::t2a cassette was knocked into the N-terminus of the locus because of different C-terminal splice variants. The self-cleaving T2A peptide frees up GFP::H2B, which will be transported to the nucleus, thereby facilitating cell identification. For unc-46, unc-47, tph-1, bas-1, tbh-1, cat-1, cat-2, snf-3, and oct-1, we knocked-in a sl2::gfp::h2b cassette at the C-terminus of the locus. The SL2 sequence also provides for the separate production of GFP::H2B. Both types of reporter cassettes should capture posttranscriptional, 3’UTR-mediated regulation of each locus, e.g., by miRNAs and RNA-binding proteins (not captured by CeNGEN scRNA data). Since in each case the reporter is targeted to the nucleus, this strategy circumvents shortcomings associated with interpreting antibody staining patterns or dealing with too densely packed cytosolic signals. For mod-5, we analyzed a previously generated, non-nuclear reporter allele (Maicas et al., 2021). For all our neuronal cell identification, we utilized the neuronal landmark strain NeuroPAL (Tekieli et al., 2021; Yemini et al., 2021). The results of our neuronal expression pattern analysis are summarized in Figure 3 and detailed in Supplementary files 2 and 3. In the ensuing sections we describe these patterns in detail.

37 of the 38 previously reported neuron classes that express an eat-4 fosmid-based reporter (Serrano-Saiz et al., 2013) showed eat-4 transcripts in the CeNGEN scRNA atlas (Taylor et al., 2021) at all four thresholds of stringency, and 1/38 (PVD neuron) showed it in three out of the four threshold levels (Figure 1B, Supplementary file 1). However, scRNA transcripts were detected at all four threshold levels in three additional neuron classes, RIC, PVN, and DVA, for which no previous reporter data provided support. In a recent publication, we had already described that the eat-4 reporter allele syb4257 is expressed in RIC (Reilly et al., 2022) (confirmed in Figure 4A). We now also confirm expression of this reporter allele, albeit at low levels, in DVA and PVN (Figure 4B, Supplementary file 2).

Another neuron found to have some eat-4 transcripts, but only with the two lower threshold sets, is the I6 pharyngeal neuron. Consistent with our previous fosmid-based reporter data, we detected no I6 expression with our eat-4(syb4257) reporter allele. The eat-4 reporter allele also shows expression in the pharyngeal neuron M5, albeit very weakly (Figure 4A, Supplementary file 2), consistent with CeNGEN scRNA data. Weak expression of the eat-4 fosmid-based reporter in ASK and ADL remained weak, but clearly detectable with the eat-4(syb4257) reporter allele (Figure 4A, Supplementary file 2). Extremely dim expression in PHA can be occasionally detected. Whereas the PVQ neuron class displays eat-4 scRNA transcripts and was reported to show very dim eat-4 fosmid-based reporter expression, we detected no expression of the eat-4(syb4257) reporter allele in PVQ neurons (Figure 4B, Supplementary file 2). We also did not detect expression of eat-4(syb4257) in the GABAergic AVL and DVB neurons, in which a recent report describes expression of an eat-4 promoter fusion reporter (Li et al., 2023). An absence of eat-4(syb4257) expression in AVL and DVB is also consistent with the absence of scRNA transcripts in these neurons.

A few neurons were found to express eat-4 transcripts by the CeNGEN atlas, but only with lower threshold levels, including, for example, the RMD, PVM, and I4 neurons (Figure 1B, Supplementary file 1). We failed to detect reporter allele expression in RMD or PVM neurons, but occasionally observed very dim expression in I4. Lastly, we identified a novel site of eat-4 expression in the dopaminergic PDE neuron (Figure 4B, Supplementary file 2). While such expression was neither detected with previous reporters nor scRNA transcripts, we detected it very consistently but at relatively low levels.

41 of previously described 52 neuron classes that show unc-17 fosmid-based reporter expression (Pereira et al., 2015) showed transcripts in the CeNGEN scRNA atlas at four out of four threshold levels, another seven neuron classes at three out of four threshold levels, and one at the lowest two threshold levels (Taylor et al., 2021). Only one neuron class, RIP, displayed scRNA levels at all four thresholds, but showed no corresponding unc-17 fosmid-based reporter expression (Figure 1B, Supplementary file 1). Using the unc-17(syb4491) reporter allele (Figure 1A), we confirmed expression in RIP (Figure 4C, Supplementary file 2). Of the additional neuron classes that show unc-17 expression at the lower stringency transcript detection levels (Figure 1B, Supplementary file 1), we were able to detect unc-17 reporter allele expression only in AWA (Figure 4C, Supplementary file 2).

Conversely, a few neurons display weak expression with previous multicopy, fosmid-based reporter constructs (RIB, AVG, PVN) (Pereira et al., 2015), but show no CeNGEN scRNA support for such expression (Taylor et al., 2021). The unc-17(syb4491) reporter allele confirmed weak but consistent expression in the PVN neurons as well as variable, borderline expression in AVG (Figure 4C and D). However, we failed to detect unc-17(syb4491) reporter allele expression in the RIB neurons.

We detected another novel site of unc-17 expression, albeit dim, in the glutamatergic AFD neurons (Figure 4C, Supplementary file 2). This expression was not reported with previous fosmid-based reporter or CeNGEN scRNA data. Consistent with AFD and PVN being potentially cholinergic, scRNA transcript reads for cha-1/ChAT, the ACh-synthesizing choline acetyltransferase, were also detected in AFD and PVN (Supplementary file 1).

Lastly, another notable observation is the lack of any unc-17 reporter expression or CeNGEN scRNA transcripts in the interneuron AVJ, but presence of CeNGEN scRNA transcript reads for cha-1/ChAT (Supplementary file 1), which shares exons with the unc-17/VAChT locus (Alfonso et al., 1994). Although no reporter data is available for cha-1/ChAT, such interesting mismatch between available unc-17 and cha-1/ChAT expression data could provide a hint to potential non-vesicular cholinergic transmission in the AVJ neurons in C. elegans, potentially akin to reportedly non-vesicular release of acetylcholine in the visual system of Drosophila (Yang and Kunes, 2004).

tph-1/TPH and bas-1/AAAD code for enzymes required for serotonin (5-HT = 5-hydroxytryptamine) synthesis (Figure 1A). scRNA transcripts for tph-1 and bas-1 are detected in previously defined serotonergic neurons at all four threshold levels (HSN, NSM, ADF) (Figure 1, Supplementary file 1). In addition to these well-characterized sites of expression, several of the individual genes show scRNA-based transcripts in a few additional cells: tph-1 at all four threshold levels in AFD and MI. Neither of these cells display scRNA transcripts for bas-1/AAAD, the enzyme that metabolizes the TPH-1 product 5-HTP (5-hydroxytryptophan) into serotonin (5-HT) (Figure 1A). To further investigate these observations, we generated reporter alleles for both tph-1 and bas-1 (Figure 2). Expression of the tph-1 reporter allele syb6451 confirmed expression in the previously well-described neurons that stained positive for serotonin, namely NSM, HSN, and ADF, matching CeNGEN data. While expression in AFD (seen at all four threshold levels in the CeNGEN scRNA atlas) could not be confirmed with the reporter allele, expression in the pharyngeal MI neurons could be confirmed (Figure 6A, Figure 6—figure supplement 1, Supplementary file 2).

We detected co-expression of the bas-1 reporter allele, syb5923, with tph-1(syb6451) in NSM, HSN, and ADF, in accordance with the previous reporter and scRNA data (Figure 6B, Supplementary file 2). However, bas-1(syb5923) is not co-expressed with tph-1 in MI (Figure 6A and B), nor is there CeNGEN-transcript evidence for bas-1/AAAD in MI (Figure 1, Supplementary file 1). Hence, TPH-1-synthesized 5-HTP in MI is not metabolized into 5-HT (serotonin), consistent with the lack of serotonin-antibody staining in MI (Horvitz et al., 1982; Sze et al., 2000).

We also detected tph-1(syb6451) reporter allele expression in the serotonergic VC4 and VC5 neurons (Figure 6A, Supplementary file 2), consistent with scRNA data (Figure 1, Supplementary file 1) and previous reporter transgene data (Mondal et al., 2018). This suggests that these neurons are capable of producing 5-HTP. However, there is no bas-1(syb5923) expression in VC4 or VC5, consistent with previous data showing that serotonin is taken up, but not synthesized by them (Duerr et al., 2001) (more below on monoamine uptake; Tables 1 and 2).

As expected from the role of bas-1/AAAD in dopamine synthesis (Hare and Loer, 2004), bas-1(syb5923) is also expressed in dopaminergic neurons PDE, CEP, and ADE. In addition, it is also expressed weakly in URB, consistent with scRNA data. We did not detect visible expression in PVW or PVT, both of which showed very low levels of scRNA transcripts (Figure 1, Supplementary file 1). Expression of bas-1/AAAD in URB may suggest that URB generates a non-canonical monoamine (e.g. tryptamine, phenylethylamine [PEA], or histamine), but since URB expresses no vesicular transporter (cat-1/VMAT, see below), we consider it unlikely that any such monoamine would be secreted via canonical vesicular synaptic release mechanisms.

The CeNGEN scRNA atlas shows transcripts for the rate-limiting enzyme of dopamine synthesis encoded by cat-2/TH (Figure 1B, Supplementary file 1) at all four threshold levels in all three previously described dopaminergic neuron classes in the hermaphrodite, ADE, PDE, and CEP (Sulston et al., 1975; Sulston et al., 1980; Lints and Emmons, 1999). At lower threshold levels, transcripts can also be detected in the OLL neurons. A CRISPR/Cas9-engineered reporter allele for cat-2/TH, syb8255, confirmed expression in ADE, PDE, and CEP in adult hermaphrodites (Figure 7A, Supplementary file 2). As expected and described above, all three neuron classes also expressed bas-1/AAAD (Figure 6B) and cat-1/VMAT (Figure 6C, see below) (Supplementary file 2). We did not detect visible expression of cat-2(syb8255) in OLL. The OLL neurons also display no scRNA transcripts or reporter allele expression of bas-1/AAAD or cat-1/VMAT. No additional sites of expression of cat-2(syb8255) were detected in the adult hermaphrodite.

The invertebrate analogs of adrenaline and noradrenaline, tyramine and octopamine, are generated by tdc-1 and tbh-1 (Figure 1A; Alkema et al., 2005). Previous work had identified expression of tdc-1 in the hermaphrodite RIM and RIC neurons and tbh-1 in the RIC neurons (Alkema et al., 2005). Transcripts in the CeNGEN atlas match those sites of expression for both tdc-1 (scRNA at four threshold levels in RIM and RIC neurons) and tbh-1 (scRNA at four threshold levels in RIC neurons) (Figure 1B, Supplementary file 1). Much lower transcript levels are present in a few additional, non-overlapping neurons (Figure 1B). CRISPR/Cas9-engineered reporter alleles confirmed tdc-1 expression in RIM and RIC and tbh-1 expression in RIC (Figure 7B and C, Supplementary file 2). In addition, we also detected dim expression of tbh-1(syb7786) in all six IL2 neurons, corroborating scRNA transcript data (Figure 7C, Supplementary files 1 and 2). However, IL2 neurons do not exhibit expression of the reporter allele of tdc-1, which acts upstream of tbh-1 in the octopamine synthesis pathway, or of cat-1/VMAT, the vesicular transporter for octopamine (Figure 6C, see below). Hence, the IL2 neurons are unlikely to produce or synaptically release octopamine, but they may produce another monoaminergic signal (Table 2).

As the vesicular monoamine transporter, cat-1/VMAT is expected to be expressed in all neurons that synthesize serotonin, dopamine, tyramine, and octopamine (Figure 1A). Both scRNA data and a CRISPR/Cas9-engineered reporter allele, syb6486, confirm expression in all these cells (Figure 6C, Supplementary file 2). In addition, based on antibody staining and previous fosmid-based reporters, cat-1/VMAT is known to be expressed in neurons that do not synthesize serotonin but are nevertheless positive for serotonin antibody staining (VC4, VC5, and RIH) (Duerr et al., 1999; Duerr et al., 2001; Serrano-Saiz et al., 2017b). Again, both scRNA data and a CRISPR/Cas9-engineered reporter allele, syb6486, confirm expression in these cells (Figure 6C, Supplementary file 2).

In addition to these canonical monoaminergic neurons, the CeNGEN scRNA data shows cat-1/VMAT expression at all four threshold levels in RIR, CAN, AVM and, at a much lower threshold, eight additional neuron classes (Figure 1B, Supplementary file 1). Our cat-1/VMAT reporter allele, syb6486, corroborates expression in RIR and CAN, but not in AVM (Figure 6C, Supplementary file 2). We also observed expression of the cat-1 reporter allele in two of the neuron classes with scRNA transcripts at the lowest threshold level, ASI and variably, AVL (Figure 6C, Supplementary file 1). Interestingly, AVL does not express any other monoaminergic pathway genes (Supplementary file 2), therefore it may be transporting a new amine yet to be discovered. This scenario also applies for two male-specific neurons (more below). As previously mentioned, we detected no cat-1/VMAT expression in the tph-1/TPH-positive MI or the cat-2/TH-positive OLL neurons.

The cat-1/VMAT reporter allele revealed expression in an additional neuron class, the AUA neuron pair (Figure 6C, Supplementary file 2). Expression in this neuron is not detected in scRNA data; however, such expression may be consistent with previous CAT-1/VMAT antibody staining data (Duerr et al., 1999). These authors found the same expression pattern as we detected with cat-1/VMAT reporter allele, except for the AIM neuron, which Duerr et al. identified as CAT-1/VMAT antibody-staining positive. However, neither our reporter allele, nor a fosmid-based cat-1/VMAT reporter, nor scRNA data showed expression in AIM, and we therefore think that the neurons identified by Duerr et al. as AIM may have been the AUA neurons instead (see also Serrano-Saiz et al., 2017b). Additionally, a cat-1-positive neuron pair in the ventral ganglion, unidentified but mentioned by Duerr et al., 1999, is likely the tyraminergic RIM neuron pair, based on our reporter allele and CeNGEN scRNA data.

In addition to or in lieu of synthesizing monoamines, neurons can uptake them from their surroundings. To investigate the cellular sites of monoamine uptake in more detail, we analyzed fluorescent protein expression from engineered reporter alleles for the uptake transporter of serotonin (mod-5/SERT(vlc47)), the predicted uptake transporter for tyramine (oct-1/OCT(syb8870)), and that for betaine (snf-3/BGT1(syb7290)).

No comprehensive scRNA atlas has yet been reported for the nervous system of the male. Based on the expression of fosmid-based reporters, we had previously assembled a neurotransmitter atlas of the C. elegans male nervous system in which individual neuron classes are notoriously difficult to identify (Serrano-Saiz et al., 2017b). We have since established a NeuroPAL landmark strain that permits more reliable identification of gene expression patterns in both the hermaphrodite and male-specific nervous system (Tekieli et al., 2021; Yemini et al., 2021). We used NeuroPAL to facilitate the analysis of the expression profiles of our CRISPR/Cas9-engineered reporter alleles in the male, resulting in updated expression profiles for 11 of the 16 knock-in reporter alleles analyzed. As in the hermaphrodite, reasons for these updates vary. In addition to the improved accuracy of neuron identification provided by NeuroPAL, in some cases there are true differences of expression patterns between the fosmid-based reporters and reporter alleles. We elaborate on these updates for individual reporter alleles below.

We analyzed eat-4/VGLUT (syb4257), unc-17/VAChT (syb4491), unc-25/GAD (ot1372), and unc-47/VGAT (syb7566) expression in the male-specific nervous system using NeuroPAL landmark strains (otIs696 for eat-4 and otIs669 for all others)(Figures 10 and 11). Of all those reporter alleles, unc-25/GAD (ot1372) was the only one with no updated expression. Specifically, in addition to confirming presence of expression of the unc-25(ot1372) reporter allele in CP9, EF1/2, EF3/4, we also confirmed its lack of expression in anti-GABA-positive neurons R2A, R6A, and R9B (Gendrel et al., 2016; Serrano-Saiz et al., 2017b; Figure 11A, Supplementary file 3).

In the preanal ganglion, we observed weak expression of unc-17(syb4491) in DX3/4 (Figure 10B, Supplementary file 3), hence assigning previously unknown neurotransmitter identity to these neurons. Related to DX3/4, we also confirmed expression of unc-17 in DX1/2 in the dorsorectal ganglion, consistent with fosmid-based reporter data (Supplementary file 3; Serrano-Saiz et al., 2017b). In the lumbar ganglion, we detected novel expression of unc-17(syb4491) in five pairs of type B ray neurons, namely R1B, R4B, R5B, R7B, and R9B (Figure 10B, Supplementary file 3). Expression in all these neurons is low, possibly explaining why it is not observed with an unc-17 fosmid-based reporter (Serrano-Saiz et al., 2017b).

In the ventral nerve cord, we found additional, very weak expression of eat-4(syb4257) in CA1 to CA4 (Figure 10A, Supplementary file 3), as well as weak expression of unc-17(syb4491) in CP1 to CP4 (Figure 10B, Supplementary file 3), all undetected by previous analysis of fosmid-based reporters (Serrano-Saiz et al., 2017b). Conversely, two neurons lack previously reported expression of fosmid-based reporters; CP9 does not show visible unc-17(syb4491) expression (Figure 10B) and neither does CA9 show visible expression of unc-47(syb7566) expression (Figure 11C). We also realized that the neuron identifications of CA7 and CP7 were previously switched (Serrano-Saiz et al., 2017b), due to lack of proper markers for those two neurons. With NeuroPAL, we are now able to clearly distinguish the two and update their classic neurotransmitter reporter expression: CA7 expresses high levels of eat-4(syb4257) (Figure 10A, Supplementary file 3), very low levels of unc-17(syb4491) (Figure 10B), and no unc-47(syb7566) (Figure 11C); CP7 expresses no eat-4(syb4257) (Figure 10A, Supplementary file 3), very low levels of unc-17(syb4491) (Figure 10B), and very low levels of unc-47(syb7566) as well (Figure 11C). Taken together, the analysis of reporter alleles reveals a remarkable diversity of CA and CP neurons, summarized in Figure 10C.

In the head, we detected expression of unc-47(syb7566) in the male-specific neuron class MCM (Figure 11B, Supplementary file 3), previously not observed with fosmid-based reporters. Consistent with fosmid-based reporter data, the other male-specific head neuron class, CEM, shows expression of unc-17(syb4491) (Supplementary file 3) and unc-47(syb7566) (Figure 11B, Supplementary file 3) reporter alleles.

We analyzed the expression of reporter alleles for genes involved in monoamine biosynthesis and uptake in the male-specific nervous system: cat-1/VMAT (syb6486), tph-1/TPH (syb6451), cat-2/TH (syb8255), bas-1/AAAD (syb5923), tdc-1/TDC (syb7768), tbh-1/TBH (syb7786), mod-5/SERT (vlc47), oct-1/OCT (syb8870), and snf-3/BGT1 (syb7290). As in the hermaphrodite nervous system, we used the NeuroPAL reporter landmark (otIs669) for neuron ID (Tekieli et al., 2021). We found novel expression patterns in all male-specific ganglia (Figures 12 and 13, Supplementary file 3).

In vertebrates, glia can produce various signaling molecules, including neurotransmitters (Araque et al., 2014; Savtchouk and Volterra, 2018). There is some limited evidence for neurotransmitter synthesis in C. elegans glia. In males, it had been reported that the socket glia of spicule neurons synthesize and utilize dopamine, based on their expression of cat-2/TH and bas-1/AAAD (Lints and Emmons, 1999; Hare and Loer, 2004; LeBoeuf et al., 2014). We confirmed this notion with cat-2/TH and bas-1/AAAD reporter alleles (Figure 14A). Additionally, we detected expression of the cat-1/VMAT reporter allele in these cells (Figure 14A), indicating that these glia secrete dopamine by canonical vesicular transport. We did not detect cat-1/VMAT in other glial cell types. In addition to the spicule socket glia, we also observed bas-1(syb5923) reporter allele expression in cells that are likely to be the spicule sheath glia (Figure 14A), as well as in additional glial cell types in the head and tail (Figure 14B). We detected no glial expression of other monoaminergic synthesis machinery.

We detected no expression of vesicular transporters or biosynthetic synthesis machinery for non-aminergic transmitters in glia of either sex. This observation contrasts previous reports on GABA synthesis and release from the AMsh glial cell type (Duan et al., 2020; Fernandez-Abascal et al., 2022). We were not able to detect signals in AMsh with anti-GABA staining, nor with an SL2- or T2A-based GFP-based reporter allele for any unc-25 isoform (Gendrel et al., 2016) (M Gendrel, pers. comm.; this paper).

There is, however, abundant evidence for neurotransmitter uptake by C. elegans glial cells, mirroring this specific function of vertebrate glia (Henn and Hamberger, 1971). We had previously shown that one specific glia-like cell type in C. elegans, the GLR glia, take up GABA via the GABA uptake transporter SNF-11 (Gendrel et al., 2016). We did not detect unc-47/VGAT fosmid-based reporter expression in the GLRs (Gendrel et al., 2016) and also detected no expression with our unc-47/VGAT reporter allele. Hence, these glia are unlikely to release GABA via classic vesicular machinery. Other release mechanisms for GABA can of course not be excluded. Aside from the snf-11 expression in GLR glia (Gendrel et al., 2016), we detected expression of the putative tyramine uptake transporter oct-1/OCT in a number of head glial cells (Figure 8K), as well as broad glial expression of the betaine uptake transporter snf-3/BGT1 in the head, midbody, and tail (Figure 9E and F). These results indicate tyramine and betaine clearance roles for glia.

We detected expression of a few neurotransmitter pathway genes in cells outside the nervous system. The most prominent sites of reporter allele expression are located within the gonad. We detected expression of tdc-1(syb7768) and tbh-1(syb7786) reporter alleles in the gonad of hermaphrodite as well as tdc-1(syb7768) expression in the neuroendocrine uv1 cells (Figure 14C; Figure 14—figure supplement 1), as previously reported (Alkema et al., 2005). Intriguingly, while cat-1(syb6486) is expressed in a midbody gonadal cell posterior to the vulva, likely the distal valve (Figure 6C, Figure 14—figure supplement 1), we observed no expression of cat-1(syb6486) in the gonad or the uv1 cells (Figure 14C). This suggests alternative release mechanisms for tyramine and octopamine. A vertebrate homolog of the putative tyramine uptake transporter, oct-1, has been found to be located presynaptically and to co-purify with synaptosomes (Berry et al., 2016; Matsui et al., 2016), therefore indicating that this transporter may have the potential to also act in tyramine release, at least in vertebrate cells. However, we observed no expression of our oct-1 reporter allele in uv1 or gonadal cells.

In the male, tdc-1(syb7768), tbh-1(syb7786), cat-1(syb6486), and oct-1(syb8870) animals also show reporter expression in the somatic gonad: while all four genes are expressed in the vas deferens, cat-1 and tbh-1, but not tdc-1 or oct-1, are expressed in the seminal vesicle (Figure 14C, Figure 8K). A similar pattern of cat-1(+); tbh-1(+); tdc-1(-); oct-1(-) is detected in several male-specific neurons and may indicate the usage of a novel transmitter (e.g. PEOH, see Discussion) by these cells. snf-3/BGT1 is also expressed in male somatic gonad cells, indicating that these cells could also use betaine for signaling (Figure 9E).

The AAADs tdc-1 and bas-1 are also prominently expressed in the intestine, where bas-1 has been shown to be involved in generating serotonin-derived glucosides (Yu et al., 2023). bas-1, but not tdc-1, is also expressed in the hypodermis and seam cells, as is the betaine uptake transporter snf-3 (Figure 9, Figure 14—figure supplement 1). The tph-1 reporter allele expresses in a subset of pharyngeal non-neuronal cells during the L1 to L4 larval stages of development (Figure 6—figure supplement 1), which is consistent with low levels of tph-1 transcripts detected in pharyngeal muscles in the CeNGEN scRNA dataset (Taylor et al., 2021). Additionally, we observed previously uncharacterized eat-4/VGLUT expression in muscle cells in both sexes (Figure 14D).

Direct detection of neurotransmitters through antibody staining has shown that at least two neurotransmitters, GABA and serotonin, are present in some neurons that do not express the synthesis machinery for these transmitters (Tables 1 and 2). Instead, these neurons acquire GABA and serotonin through uptaking them via defined uptake transporters, SNF-11/BGT1 for GABA (Mullen et al., 2006) and MOD-5/SERT for serotonin (Ranganathan et al., 2001; Jafari et al., 2011). A combination of CeNGEN scRNA transcriptome and our reporter allele data corroborates the absence of synthesis machinery in these presumptive uptake neurons (Tables 1 and 2). One interesting question that relates to these uptake neurons is whether they serve as ‘sinks’ for clearance of a neurotransmitter or whether the taken-up neurotransmitter is subsequently ‘recycled’ for synaptic release via a vesicular transporter. Previous data, as well as our updated expression profiles, provide evidence for both scenarios: ALA and AVF do not synthesize GABA via UNC-25/GAD, but they stain with anti-GABA antibodies in a manner that is dependent on the uptake transporter SNF-11 (Gendrel et al., 2016). ALA expresses unc-47, hence it is likely to synaptically release GABA, but AVF does not, and it is therefore apparently involved only in GABA clearance. Similarly, RIH, AIM, and PGA express the serotonin uptake transporter mod-5/SERT and stain for serotonin in a MOD-5-dependent manner (Jafari et al., 2011) (this study), but only RIH and PGA, not AIM, expresses the vesicular transporter cat-1/VMAT, suggesting RIH and PGA are likely serotonergic signaling neurons whereas AIM is a clearance neuron.

Some neurons do not obviously fall into the synthesis or uptake category, most notably, the anti-GABA-antibody-positive AVA and AVB neurons (both of which conventional cholinergic neurons). None of these neurons express unc-25/GAD, nor the snf-11/BGT1 uptake transporter, yet unc-25/GAD is required for their anti-GABA-positive staining (Gendrel et al., 2016). This suggests that GABA may be acquired by these neurons through non-canonical uptake or synthesis mechanisms. Also, the AVA and AVB neurons do not express UNC-47 (Gendrel et al., 2016; Taylor et al., 2021) (this study); hence, it is not clear if or how GABA is released from them. A member of the bestrophin family of ion channels has been shown to mediate GABA release from astrocyte glia in vertebrates (Lee et al., 2010) and, more recently, from C. elegans glia (Cheng et al., 2024; Graziano et al., 2024). However, while there are more than 20 bestrophin channels encoded in the C. elegans genome (Hobert, 2013), they do not appear to be expressed in the AVA or AVB neurons, based on CeNGEN scRNA data (Taylor et al., 2021).

The co-expression of a specific uptake transporter and a vesicular transporter corroborates the potential usage of betaine as a neurotransmitter. Betaine is known to be synthesized in C. elegans but is also taken up via its diet (Peden et al., 2013; Hardege et al., 2022). Betaine has documented effects on C. elegans behavior and acts via activation of several betaine-gated ion channels (Peden et al., 2013; Hardege et al., 2022). Expression of biosynthetic enzymes suggests betaine production in at least the RIM neuron class, which also expresses the vesicular transporter cat-1/VMAT, capable of transporting betaine (Hardege et al., 2022). The expression of the betaine uptake transporter snf-3/BGT1 in CAN, AUA, RIR, ASI, and male-specific neuron PDC, coupled with their co-expression of cat-1/VMAT, suggests that several distinct neuron classes in different parts of the nervous system may uptake betaine and engage in vesicular betaine release via CAT-1/VMAT to gate betaine-activated ion channels, such as ACR-23 (Peden et al., 2013) or LGC-41 (Hardege et al., 2022). Additionally, we detected the snf-3/BGT1 reporter allele in several other neuron classes that do not co-express cat-1/VMAT. This indicates that these neurons could function as betaine clearance neurons.

Lastly, based on sequence similarity and expression pattern, we predict that the ortholog of the OCT subclass of SLC22 family, oct-1, could serve as a tyramine uptake transporter in C. elegans. We identified RIM as the only neuron expressing an oct-1 reporter allele, suggesting that like several other monoaminergic neuron classes, RIM both synthesizes its monoaminergic transmitter and reuptakes it after release.

We identified neurons that do not express any conventional, well-characterized vesicular neurotransmitter transporter families, namely UNC-17/VAChT, CAT-1/VMAT (the only SLC18 family members), UNC-47/VGAT (only SLC32 family member), or EAT-4/VGLUT (an SLC17 family member). Six sex-shared neurons (AVH, BDU, PVM, PVQ, PVW, RMG) and one male-specific neuron (SPD) fall into this category. Most of these neurons exhibit features that are consistent with them being entirely neuropeptidergic. First, electron microscopy has revealed a relative paucity of clear synaptic vesicles in most of these neurons (White et al., 1986; Cook et al., 2019; Witvliet et al., 2021). Second, not only do these neurons express a multitude of neuropeptide-encoding genes (Taylor et al., 2021), but they also display a dense interconnectivity in the ‘wireless’ neuropeptidergic connectome (Ripoll-Sánchez et al., 2023).

That said, electron microscopy shows that some of the neurons devoid of conventional neurotransmitter pathway genes generate synapses with small, clear synaptic vesicles, indicative of the use of non-peptidergic transmitters (e.g. the sex-shared RMG and PVM neurons or the male-specific SPD neurons) (White et al., 1986; Cook et al., 2019; Witvliet et al., 2021). It is therefore conceivable that either conventional neurotransmitters utilize non-conventional neurotransmitter synthesis and/or release pathways, or that completely novel neurotransmitter systems remain to be discovered. Although the C. elegans genome does not encode additional members of the SLC18A2/3 (cat-1/VMAT, unc-17/VAChT) or SLC32A1 (unc-47/VGAT) family of vesicular neurotransmitter transporters, it does contain a number of additional members of the SLC17A6/7/8 (VGLUT) family (Hobert, 2013). These may serve as non-canonical vesicular transporters of more uncommon neurotransmitters or, alternatively, may be involved in modulating release of glutamate (Serrano-Saiz et al., 2020; Choi et al., 2021). Uncharacterized paralogs of bona fide neurotransmitter uptake transporters (SLC6 superfamily) may also have functions in neurotransmitter release rather than uptake. However, based on CeNGEN scRNA data, no robust or selective expression of these SLC17 or SLC6 family members is observed in these ‘orphan neurons’.

Our analysis expands the repertoire of neurons that co-transmit multiple neurotransmitters (Figure 3). Neurotransmitter co-transmission has been observed in multiple combinations in the vertebrate brain (Wallace and Sabatini, 2023). In C. elegans, the most frequent co-transmission configurations are a classic, fast transmitter (acetylcholine or glutamate) with a monoamine. Co-transmission of two distinct monoaminergic systems also exists. In several cases, however, it is not clear whether the second neurotransmitter is indeed used for communication or whether its presence is merely a reflection of this neuron being solely a clearance neuron. For example, the glutamatergic AIM neuron stains positive for serotonin, which it uptakes via the uptake transporter MOD-5, but it does not express the vesicular monoamine transporter cat-1/VMAT (Figures 3, 6, and 8, Tables 1 and 2).

Co-transmission of small, fast-acting neurotransmitters (glutamate, GABA, acetylcholine) does exist, but it is rare (Figure 3). The most prominent co-transmission configuration is acetylcholine with glutamate, but acetylcholine can also be co-transmitted with GABA. There are no examples of co-transmission of glutamate and GABA, as observed in several regions of the vertebrate brain (Wallace and Sabatini, 2023). There are also examples of possible co-transmission of three transmitters (Figure 3).

Interestingly, co-transmission appears to be much more prevalent in the male-specific nervous system, compared to the sex-shared nervous system (Figure 3, Supplementary file 3). This may relate to male-specific neurons displaying a greater degree of anatomical complexity compared to the hermaphrodite nervous system, both in terms of branching patterns and extent of synaptic connectivity (Jarrell et al., 2012; Cook et al., 2019). Given that all co-transmitting neurons display multiple synaptic outputs (Cook et al., 2019), it appears possible that each individual neurotransmitter secretory system is distributed to distinct synapses. Based on vertebrate precedent (Wallace and Sabatini, 2023), co-release from the same vesicles is also possible.

The observation of sexual dimorphisms in neurotransmitter abundance in specific regions of the mammalian brain has been one of the earliest molecular descriptors of neuronal sex differences in mammals (McCarthy et al., 1997). However, it has remained unclear whether such differences are the result of the presence of sex-specific neurons or are indications of distinctive neurotransmitter usage in sex-shared neurons. Using C. elegans as a model, we have been able to precisely investigate (a) whether sex-specific neurons display a bias in neurotransmitter usage and (b) whether there are neurotransmitter dimorphisms in sex-shared neurons (Pereira et al., 2015; Gendrel et al., 2016; Serrano-Saiz et al., 2017b) (this paper). We found that male-specific neurons display a roughly similar proportional usage of individual neurotransmitter systems and note that male-specific neurons display substantially more evidence of co-transmission, a possible reflection of their more elaborate morphology and connectivity. We also confirmed evidence for sexual dimorphisms in neurotransmitter usage in sex-shared neurons (Supplementary file 4), which are usually correlated with sexual dimorphisms in synaptic connectivity of these sex-shared neurons (Cook et al., 2019).

Neurotransmitter uptake is a classic function of glial cells across animal phylogeny (Henn and Hamberger, 1971), and such uptake mechanisms are observed in C. elegans as well. Previous reports demonstrated glutamate uptake by CEPsh (Katz et al., 2019) and GABA uptake by GLR glia (Gendrel et al., 2016). We now add to this list betaine uptake by most glia, as inferred from the expression pattern of SNF-3/BGT1 (Figure 9, Supplementary file 1).

Studies in vertebrates have also suggested that specific glial cell types synthesize and release several neurotransmitters (Araque et al., 2014; Savtchouk and Volterra, 2018). For example, astrocytes were recently shown to express VGLUT1 to release glutamate (de Ceglia et al., 2023). Evidence of neurotransmitter synthesis and release also exists in C. elegans glia. Previous work indicated that glia associated with male-specific spicule neurons synthesize (through cat-2/TH and bas-1/AAAD) the monoaminergic transmitter dopamine to control sperm ejaculation (LeBoeuf et al., 2014). Our identification of cat-1/VMAT expression in these glia indicate that dopamine is released via the canonical vesicular monoamine transporter. We also detected expression of bas-1/AAAD in additional male and hermaphrodite glia, indicating the production of other signaling substances released by these glia. bas-1 has indeed recently been shown to be involved in the synthesis of a class of unconventional serotonin derivatives (Yu et al., 2023).

There have been previous reports on GABA synthesis and release from the AMsh glial cell type (Duan et al., 2020; Fernandez-Abascal et al., 2022). We were not able to detect AMsh with anti-GABA staining, nor with reporter alleles of unc-25/GAD. However, since very low levels of unc-25 are observed in the AMsh scRNA datasets (Taylor et al., 2021; Purice et al., 2023), the abundance of GABA in AMsh may lie below conventional detection levels.

Outside the nervous system, the most prominent and functionally best characterized usage of neurotransmitters lies in the hermaphrodite somatic gonad, which has been shown to synthesize octopamine and use it to control oocyte quiescence (Alkema et al., 2005; Kim et al., 2021). Intriguingly, we also detected tbh-1, tdc-1, and cat-1 expression in the somatic gonad of the male, specifically the vas deferens, which is known to contain secretory granules that are positive for secretory molecular markers (Nonet et al., 1993). The presence of octopamine is unexpected because, unlike oocytes, sperm are not presently known to require monoaminergic signals for any aspect of their maturation. It will be interesting to assess sperm differentiation and function of tbh-1 or tdc-1 mutant animals. The usage of monoaminergic signaling systems in the gonad is not restricted to C. elegans and has been discussed in the context of sperm functionality and oocyte maturation in vertebrates (Mayerhofer et al., 1999; Ramírez-Reveco et al., 2017; Alhajeri et al., 2022).

Our analysis also provides an unprecedented and systematic comparison of antibody staining, scRNA transcript data, reporter transgene expression, and knock-in reporter allele expression. The bottom-line conclusions of these comparisons are: (1) Reporter alleles reveal more sites of expression than fosmid-based reporters. It is unclear whether this is due to the lack of cis-regulatory elements in fosmid-based reporters or issues associated with the multicopy nature of these reporters (e.g. RNAi-based gene silencing of multicopy arrays or squelching of regulatory factors). Another factor to consider is that neuron identification for most fosmid-based reporters was carried out prior to the introduction of NeuroPAL. Consequently, errors occasionally occurred, as exemplified by the misidentification of neuron IDs for CA7 and CP7 in previous instances (Serrano-Saiz et al., 2017b). (2) The best possible reporter approaches (i.e. reporter alleles) show very good overlap with scRNA data, thereby validating each approach. However, our comparisons also show that no single approach is perfect. CeNGEN scRNA data can miss transcripts and can also show transcripts in cells in which there is no independent evidence for gene or protein expression. Conversely, antibody staining displays vagaries related to staining protocols and protein localization, which can be overcome with reporter approaches, but the price to pay with reporter alleles is that if they are based on SL2 or T2A strategies, they may fail to detect additional levels of posttranslational regulation, which may result in protein absence even in the presence of transcripts. The existence of such mechanisms may be a possible explanation for cases where the expression of synthesis and/or transport machinery expression does not match up (e.g. tdc-1(-); tbh-1(+) neurons).

Our detailed analysis of reporter allele expression has uncovered several cases where expression of a neurotransmitter pathway gene in a given neuron class appears very low and variable from animal to animal. Such variability only exists when expression is dim, thus one possible explanation for it is that expression levels merely hover around an arbitrary microscopical detection limit. However, we cannot rule out the other possibility that this may also reflect true on/off variability of gene expression. Taking this notion a step further, we cannot exclude the possibility that expression observed with reporter alleles misses sites of expression. This possibility is raised by our inability to detect unc-25/GAD reporter allele expression in AMsh glia (Duan et al., 2020; Fernandez-Abascal et al., 2022) or eat-4 reporter allele expression in AVL and DVB neurons, in which some (but not other) multicopy reporter transgenes revealed expression of the respective genes (Li et al., 2023). Functions of these genes in the respective cell types were corroborated by cell-type-specific RNAi experiments and/or rescue experiments; whether there is indeed very low expression of these genes in those respective cells or whether drivers used in these studies for knock-down and/or rescue produce very low expression in other functionally relevant cells remains to be resolved.

In conclusion, we have presented here the most complete neurotransmitter map that currently exists for any animal nervous system. Efforts to map neurotransmitter usage on a system-wide level are underway in other organisms, most notably, Drosophila melanogaster (Deng et al., 2019; Eckstein et al., 2024). The C. elegans neurotransmitter map presented here comprises a critical step toward deciphering information flow in the nervous system and provides valuable tools for studying the genetic mechanisms underlying cell identity specification. Moreover, this neurotransmitter map opens new opportunities for investigating sex-specific neuronal differentiation processes, particularly in the male-specific nervous system, where a scarcity of molecular markers has limited the analysis of neuronal identity control. Lastly, our analysis strongly suggests that additional neurotransmitter systems remain to be identified.

While the gene expression patterns delineated here enable informed predictions about novel neuronal functions and neurotransmitter identities, further investigations involving genetic perturbations, high-resolution imaging, complementary functional assays, and analyses across developmental stages are needed to shed further light on neurotransmitter usage. Nonetheless, this comprehensive neurotransmitter map provides a robust foundation for deciphering neural information flow, elucidating developmental mechanisms governing neuronal specification, exploring sexual dimorphisms in neuronal differentiation, and potentially uncovering novel neurotransmitter systems awaiting characterization.

Knock-in reporter alleles were generated either by SunyBiotech (syb alleles) or in-house (ot alleles) using CRISPR/Cas9 genome engineering. Most genes were tagged with a nuclear-targeted gfp sequence (gfp fused to his-44, a histone h2b gene) at the 3' end of the locus to capture all isoforms, except tdc-1 which was tagged at the 5' end. For unc-25, both isoforms were individually tagged since a single tag would not capture both. Transgene schematics are shown in Figure 2.

Reporter alleles generated in this study:

Since we did not detect fluorophore signals in the hdl-1(syb1048[hdl-1::gfp]) strain, we attempted to amplify low-level signals, by inserting Cre recombinase at the C-terminus of the hdl-1 locus (hdl-1(syb4208[hdl-1::t2a::3xnls::cre])). We crossed this strain to the recently published ‘Flexon’ strain (arTi361[rps-27p::gfp"flexon"-h2b::unc-54–3'UTR]) (Shaffer and Greenwald, 2022). Even low expression of hdl-1 should have led to Cre-mediated excision of the flexon stop cassette, which is designed to abrogate gene expression by a translational stop and frameshift mutation, and subsequently can result in strong and sustained gfp expression under the control of the rps-27 promoter and thereby providing information about cell-specific hdl-1 expression. However, no robust, consistent reporter expression was seen in hdl-1(syb4208[hdl-1::t2a::3xnls::cre]); arTi361[rps-27p::gfp"flexon"-h2b::unc-54–3'UTR] animals.

Three of the reporter alleles that we generated were already previously examined in specific cellular contexts:

One of the reporter alleles was obtained from the Caenorhabditis Genetics Center (CGC):

For adult animal imaging, 15–25 (exact number depending on the difficulty of neuron ID) same-sex L4 worms were grouped on NGM plates 6–9 hr prior to imaging to control for accurate staging and avoid mating. Young adult worms were then anesthetized using 50–100 mM sodium azide and mounted on 5% agarose pads on glass slides. Z-stack images were acquired with ZEN software using Zeiss confocal microscopes LSM880 and LSM980 or a Zeiss Axio Imager Z2 and processed with ZEN software or FIJI (Schindelin et al., 2012) to create orthogonal projections. Brightness and contrast, and in some cases gamma values, were adjusted to illustrate dim expression and facilitate neuron identification.

Neuron classes were identified by crossing the gfp reporter alleles with the landmark strain ‘NeuroPAL’ (allele otIs669 or otIs696, for bright reporters and dim reporters, respectively) and following published protocols (Tekieli et al., 2021; Yemini et al., 2021) (also see ‘lab resources’ at hobertlab.org). For neuron identification of the eat-4(syb4257), unc-46(syb7278), and unc-47(syb7566) reporter alleles in hermaphrodites, the reporter alleles were also crossed into the fosmid-based reporter transgenes of the same gene [eat-4(otIs518), unc-46(otIs568), and unc-47(otIs564)] as a ‘first-pass’ to identify potential non-overlapping expression of the two alleles. For tph-1(syb6451) analysis, an eat-4 fosmid-based reporter (otIs518) was also used. For identification of VC4, VC5, HSN, and uv1, an ida-1p::mCherry integrant (LX2478, lin-15(n765ts) X; vsls269[ida-1::mCherry]) was also used in some cases (Fernandez et al., 2020). For phasmid neurons, dye-filling with DiD (Thermo Fisher Scientific) was sometimes used to confirm neuron ID. For glial expression, a panglial reporter otIs870[mir-228p::3xnls::TagRFP] was used. For hypodermal cells identification, a dpy-7p::mCherry reporter stIs10166 [dpy-7p::his-24::mCherry+unc-119(+)] was used (Liu et al., 2009).