UNC-30/PITX coordinates neurotransmitter identity with postsynaptic GABA receptor clustering

In the nervous system, neuronal communication depends on the proper transmission of signals through chemical and electrical synapses. In the context of chemical synapses, presynaptic neurons must be able to synthesize and package into synaptic vesicles specific chemical substances known as neurotransmitters (NTs), such as acetylcholine (ACh), gamma-aminobutyric acid (GABA) and glutamate (Glu). Upon secretion into the synaptic cleft, each NT molecule binds to its cognate receptor located at the postsynaptic cell membrane, thereby evoking postsynaptic electrical responses.

Genes encoding proteins for NT biosynthesis and packaging (e.g. enzymes, transporters) are co-expressed in specific neuron types. The co-expression of these proteins defines the NT identity (or NT phenotype) of individual neuron types (e.g. cholinergic, GABAergic, dopaminergic). Although instances of NT identity switching or multi-NT use by single neurons have been described (Dulcis et al., 2013; Pereira et al., 2015; Brunet Avalos and Sprecher, 2021; Pedroni and Ampatzis, 2019), it is generally the case that individual neuron types acquire a specific NT identity during development and maintain it throughout life, consistent with Dale's principle of ‘one neuron, one NT’ (Svensson et al., 2019). The continuous expression of NT identity genes is fundamental for the ability of a presynaptic neuron to signal to its postsynaptic targets. For efficient neurotransmission, however, it is equally important that cognate NT receptors cluster at postsynaptic domains precisely juxtaposed to presynaptic boutons (Gravielle, 2021; Mizumoto et al., 2023). Whether and how these two crucial processes, i.e. NT identity of the presynaptic neuron and NT receptor clustering at the postsynaptic cell, are coordinated remains poorly understood.

Genetic studies in nematodes (Caenorhabditis elegans), fruit flies (Drosophila melanogaster) and mice have revealed a phylogenetically conserved principle for the control of NT identity: transcription factors expressed in specific neuron types, termed ‘terminal selectors’, coordinate the expression of NT identity genes, thereby coordinating synthesis of enzymes and transporters necessary for NT biosynthesis and signaling (Hobert, 2008; Hobert and Kratsios, 2019). Terminal selectors broadly control batteries of genes encoding proteins essential for neuronal identity and function (e.g. ion channels, neuropeptides) (Hobert, 2011; Hobert and Kratsios, 2019). To date, terminal selectors have been predicted for 117 of the 118 C. elegans neuron types (Hobert, 2016; Reilly et al., 2022; Glenwinkel et al., 2021). Beyond C. elegans, terminal selectors have also been identified in D. melanogaster, cnidarians (Nematostella vectensis), marine chordates (Ciona intestinalis) and mice (Mus musculus) (Hobert and Kratsios, 2019), suggesting a deeply conserved role for these crucial regulators of NT identity. A defining feature of terminal selectors is their continuous expression – from development throughout adulthood – in specific neuron types (Hobert, 2008). Although the essential roles of terminal selectors in establishing NT identity during development are well-attested across model organisms, their involvement in maintaining NT identity in later life stages remains poorly examined (Achim et al., 2014), partially owing to the lack of genetic tools for inducible terminal selector depletion in late life stages.

In the case of GABAergic neurons, NT identity is defined by the co-expression of highly conserved proteins, including (1) the enzyme glutamic acid decarboxylase (GAD) (Sgado et al., 2011), which synthesizes GABA from its precursor, (2) the vesicular GABA transporter (VGAT), which packages GABA into synaptic vesicles, and (3) the GABA re-uptake transporter (GAT) (Fig. 1) (Gendrel et al., 2016). Importantly, reduced expression of these GABA identity determinants, as well as impaired GABA transmission, lead to a variety of neuropsychiatric diseases, including schizophrenia, autism, epilepsy and anxiety (Zhou and Bessereau, 2019; Bolneo et al., 2022).

Despite GABA being the most abundant inhibitory NT in both invertebrate and vertebrate nervous systems, it is poorly understood how the expression of GABA identity genes is controlled over time, from development through to adulthood, to ensure GABA neurotransmission. To date, a handful of studies in C. elegans and mice have identified terminal selectors in various GABAergic neuron types. Examples include homeodomain proteins (e.g. UNC-30/PITX) (Cinar et al., 2005; Eastman et al., 1999; Jin et al., 1994; Westmoreland et al., 2001), nuclear hormone receptors (e.g. NHR-67/NR2E1) (Gendrel et al., 2016) and GATA-type (GATA2/3) transcription factors, each necessary for expression of GABA identity genes during development (Kala et al., 2009; Lahti et al., 2016; Yang et al., 2010). However, whether any of these factors is required for maintaining GABA identity gene expression during post-embryonic life is unknown (Achim et al., 2014).

During neuronal development, GABA receptor (GABAR) clustering is fundamental for postsynaptic differentiation, a process primarily driven by synapse-organizing molecules that can be either secreted or bound to the cell membrane (Zhou and Bessereau, 2019). In mice, the cell adhesion molecule neuroligin 2, the scaffold protein gephyrin and the transmembrane protein β-dystroglycan act as synapse organizers to control GABAR clustering (Briatore et al., 2020; Essrich et al., 1998; Soykan et al., 2014). In C. elegans, the secreted molecule MADD-4 (Muscle Arm Development Defect-4) (ortholog of human punctin/ADAMTSL), acts as an anterograde synapse organizer at neuromuscular synapses (Pinan-Lucarre et al., 2014). Specifically, the short MADD-4 isoform (MADD-4B) activates UNC-40/DCC (deleted in colorectal cancer) signaling, recruiting an intracellular postsynaptic scaffold composed of FRM-3, a FERM domain protein, and LIN-2/CASK (Tu et al., 2015; Zhou et al., 2020). Moreover, MADD-4B controls GABAR positioning at synapses by recruiting the sole C. elegans neuroligin homolog, NLG-1, which binds to LIN-2 (Maro et al., 2015; Platsaki et al., 2020; Zhou et al., 2020; Tu et al., 2015). By contrast, the long madd-4 isoform, MADD-4L, promotes the clustering of levamisole-sensitive ACh receptors (L-AChRs) on muscle cells through the formation of an extracellular scaffold (Gally et al., 2004; Gendrel et al., 2009; Pinan-Lucarre et al., 2014; Rapti et al., 2011).

In vertebrates, there are two madd-4 orthologs: Adamtsl1 (previously known as punctin-1) and Adamtsl3 (previously known as punctin-2) (Hall et al., 2003; Hirohata et al., 2002). A recent study identified Adamtsl3 as an extracellular synapse organizer in the rodent hippocampus, where it supports glutamatergic and GABAergic synapse formation in vivo (Cramer et al., 2023). In the adult mouse brain, Adamtsl3 signals via DCC at GABAergic synapses and facilitates synapse maintenance, synaptic plasticity, and memory. In humans, ADAMTSL3 is widely expressed in the brain, and has been identified as a candidate gene for schizophrenia (Dow et al., 2011). Despite their well-established roles in GABAR clustering, the transcriptional mechanisms that control expression of synapse organizers remain poorly understood.

GABA neurotransmission relies on (1) the ability of the presynaptic neuron to continuously express GABA identity genes (e.g. GAD, VGAT, GAT) and (2) the ability of the postsynaptic neurons to cluster GABARs appropriately (Gravielle, 2021; Mizumoto et al., 2023). Whether these two processes, which occur in two synaptically connected cells, are coordinated remains poorly understood. In principle, at least three non-mutually exclusive models can be envisioned for the transcriptional control of a GABA synapse organizer (Fig. 1). GABAR clustering at the post-synaptic (target) cell could be achieved via the activity of a synaptic organizer (membrane-bound or secreted) produced in the post-synaptic cell. For example, gephyrin, a synapse organizer produced in the target cell, is essential for GABAR clustering (Fig. 1A, model 1). Alternatively, GABAR clustering in the post-synaptic cell may rely on secreted synaptic organizers, such as MADD-4/punctin, produced in the presynaptic GABAergic neuron (Fig. 1B,C, models 2 and 3). In that case, transcription of the synapse organizer gene may or may not require the activity of the terminal selector of the presynaptic neuron (model 2 versus 3). Our previous work in C. elegans provided support for model 3 in cholinergic neuromuscular synapses (Kratsios et al., 2015); the terminal selector UNC-3 (Collier, Ebf) is not only required for AChR clustering in the postsynaptic neuron, but also controls NT identity genes in the presynaptic cell. However, whether this principle of transcriptional coordination extends beyond cholinergic motor neurons (MNs) was unclear.

C. elegans has been a prime model to dissect molecular mechanisms underlying NT identity and synapse formation (Hobert, 2016; Mizumoto et al., 2023). Here, we show that the C. elegans terminal selector of GABAergic nerve cord MNs, UNC-30, is required for clustering of type A GABARs, a major type of inhibitory NT receptors (Ghit et al., 2021; Goetz et al., 2007). We find that UNC-30 acts directly to activate transcription of the synapse organizer madd-4B. Hence, the terminal selector UNC-30 coordinates GABAR clustering on postsynaptic muscle cells (via control of madd-4B) with acquisition of GABAergic identity in presynaptic MNs (Fig. 1, model 3). Further, we find that UNC-30 acts directly to maintain the expression of madd-4B and NT identity genes (e.g. unc-25, unc-47) in late larval and adult stages. Intriguingly, UNC-30 also represses transcription of the long madd-4 isoform (madd-4L), which is normally required for AChR clustering in postsynaptic muscle cells (Pinan-Lucarre et al., 2014). Hence, our work in GABA MNs highlights that NT receptor clustering, a central event of postsynaptic differentiation, is transcriptionally coordinated with acquisition and maintenance of NT identity, significantly extending previous observations made in C. elegans cholinergic MNs to other neuron types (Kratsios et al., 2015). Last, we uncovered additional target genes that are either positively or negatively regulated by UNC-30, indicating both activator and repressor functions. Such mechanistic insights may help us understand the molecular mechanisms underlying human genetic disorders caused by PITX gene mutations, such as Axenfeld–Rieger syndrome (Muzyka et al., 2023; Tumer and Bach-Holm, 2009; Tran and Kioussi, 2021).

C. elegans locomotion relies on both cholinergic and GABAergic MNs, the cell bodies of which intermingle along the ventral nerve cord (VNC) (Fig. 2A). Based on anatomical criteria, cholinergic and GABAergic MNs are respectively divided into six (VA, VB, DA, DB, AS, VC) and two (DD, VD) classes, which form en passant synapses along the ventral and dorsal nerve cords (Fig. 2A) (Mizumoto et al., 2023; Von Stetina et al., 2006). The coordinated activity of excitatory cholinergic and inhibitory GABAergic MNs generates sinusoidal locomotion, with each muscle cell receiving dual innervation from cholinergic and GABAergic MNs. Along the dorsal nerve cord (DNC) of adult animals, three cholinergic MN classes (DA, DB and AS) form dyadic synapses, providing excitatory input not only to dorsal muscles but also to VD GABAergic neurons, which in turn innervate and inhibit ventral muscles (Fig. 2A) (White et al., 1986). Along the VNC, another three cholinergic MN classes (VA, VB and VC) also form dyadic synapses with ventral muscles and DD GABAergic neurons, which innervate and inhibit dorsal muscles (Fig. 2A). Because each muscle cell receives both excitatory (ACh) and inhibitory (GABA) inputs, the C. elegans neuromuscular system represents a powerful model in which to study how different NT receptors precisely cluster in front of their corresponding neurotransmitter release sites (Mizumoto et al., 2023).

Within the C. elegans VNC, the transcription factor UNC-30 is specifically expressed in GABAergic (DD, VD) MNs (Jin et al., 1994), where it controls expression of GABA identity genes (Fig. 1) (McIntire et al., 1993; Cinar et al., 2005; Eastman et al., 1999; Jin et al., 1994; Westmoreland et al., 2001). Recent studies also implicated UNC-30 in synaptic remodeling, as it prevents premature synapse rewiring of DD cells and aberrant synapse rewiring of VD cells (He et al., 2015; Howell et al., 2015). However, whether UNC-30 is necessary for the postsynaptic differentiation of target muscle cells remained unknown.

We therefore investigated whether genetic loss of unc-30 affects GABAR clustering in C. elegans muscle cells innervated by GABAergic MNs. We used an endogenous RFP reporter for unc-49 (UNC-49::RFP), which encodes a type-A GABAR (GABA_AR) expressed in both ventral and dorsal body wall muscles (Fig. 2B) (Zhou et al., 2020). To visualize the presynaptic boutons of GABAergic MNs (DD, VD), we employed native and tissue-specific fluorescence (NATF) (He et al., 2019), resulting in GFP labeling of endogenous CLA-1 (Clarinet), an active zone protein (Xuan et al., 2017). Using these tools (Fig. 2), as well as an additional presynaptic marker (RAB-3::GFP) (Fig. S1A,B), we visualized in young adult (day 1) animals the juxtaposition of GABAergic presynaptic boutons and GABA_AR clusters in body wall muscles along the DNC (DD neuromuscular synapses) (Fig. 2B-E, Fig. S1A,B) and VNC (VD neuromuscular synapses) (Fig. 2F-I).

In homozygous adult (day 1) animals carrying a strong loss-of-function (LOF) unc-30 allele, e191 (Jin et al., 1994; Brenner, 1974), we found that GABA_ARs are present on dorsal muscle (DNC), but no longer cluster opposite presynaptic GABA boutons of DD neurons (Fig. 2B,C, Fig. S1A,B). The levels of UNC-49::RFP expression in dorsal muscle cells were not affected in unc-30(e191) animals (Fig. 2D), suggesting that the GABA_AR clustering phenotype is not due to decreased UNC-49 expression. Similarly, we observed GABA_AR clustering defects and no effect on UNC-49 expression along the VNC of unc-30(e191) animals (Fig. 2F-H), indicating that VD neuromuscular synapses are also affected in the absence of unc-30. Although the correlation of GABA boutons and GABA_ARs was lower at both the DNC and the VNC of unc-30(e191) mutants (Fig. 2C-G), the density of GABA boutons was higher at the DNC and lower at the VNC, consistent with previously reported wiring defects in unc-30(e191) animals (Howell et al., 2015). Altogether, we conclude that unc-30 is required for GABA_AR clustering at neuromuscular synapses of GABAergic (DD, VD) neurons.

Because in control animals GABAergic (DD) and cholinergic (DA, DB, AS) neurons form en passant neuromuscular synapses with dorsal muscles at the DNC (Fig. 2A), we considered the possibility that, in unc-30(e191) mutants, GABA_ARs not only fail to cluster properly across GABA boutons (Fig. 2C,G), but are also inappropriately juxtaposed to presynaptic boutons of cholinergic (DA, DB, AS) MNs. To test the latter, we genetically labeled cholinergic presynaptic boutons with CLA-1::BFP and GABA_ARs with UNC-49::RFP, and found that GABA_ARs incorrectly localize opposite to cholinergic presynaptic boutons in unc-30(e191) mutants (Fig. 3A,C). Double immunofluorescence staining against UNC-49 and UNC-17 (VAChT/SLC18A3), another marker of cholinergic presynaptic boutons, yielded similar results at the DNC (Fig. S1C). Importantly, we also observed that GABA_ARs incorrectly localize opposite to cholinergic boutons at the VNC (Fig. 3E-G), indicating that the GABA_AR clustering defects are present both in VD and DD neuromuscular synapses. Notably, loss of unc-30 did not affect the density of cholinergic boutons (Fig. 3D,H) or UNC-49 expression (Fig. 3B,F) at the DNC and VNC.

We next wondered whether the aberrant GABA_AR clustering opposite to cholinergic boutons affects structural features of the cholinergic neuromuscular synapse. However, we found that the L-AChR UNC-29 normally localizes opposite to cholinergic boutons in unc-30(e191) mutants (Fig. S1D,F). Additional synaptic features, such as UNC-29::RFP expression on the muscle and density of presynaptic ACh boutons, were also not affected (Fig. S1E,G).

Altogether, unc-30 is necessary for the correct positioning of GABA_ARs at neuromuscular synapses of the DNC and VNC. Because unc-30 is present in GABAergic MNs but not expressed in body wall muscles or muscle progenitor cells (Figs S2, S3) (Jin et al., 1994), it is likely that unc-30 controls GABA_AR clustering in an indirect (non-cell-autonomous) manner.

We previously demonstrated that madd-4, a secreted synapse organizer, is required for GABA_AR and AChR clustering at C. elegans neuromuscular synapses (Pinan-Lucarre et al., 2014). The madd-4 locus generates two isoforms through alternative promoter usage (Fig. 4A) (Kratsios et al., 2015; Seetharaman et al., 2011). The long isoform (madd-4L) is produced by cholinergic MNs and required for L-AChR clustering at neuromuscular synapses (Fig. 4B) (Kratsios et al., 2015; Pinan-Lucarre et al., 2014). The short isoform (madd-4B) is required for GABA_AR clustering (Fig. 4B) (Pinan-Lucarre et al., 2014). Because madd-4B is produced by both GABAergic and cholinergic MNs (Kratsios et al., 2015; Pinan-Lucarre et al., 2014; Tu et al., 2015), it remained unclear whether madd-4B from GABAergic and/or cholinergic MNs is required for GABA_AR clustering at neuromuscular synapses.

To test this, we first analyzed animals specifically lacking madd-4B gene activity using the madd-4(tr185) allele (Fig. 4A). Confirming their previously reported synaptic phenotype (Tu et al., 2015; Zhou et al., 2020), we found that UNC-49::RFP fluorescence signal on the dorsal muscle of madd-4B(tr185) animals was no longer restricted to sites opposite GABA (DD) boutons (Fig. 4C). We indeed found that the correlation of GABA_AR localization and GABA boutons is lower in madd-4(tr185) animals (Fig. 4D). Importantly, GABA_ARs (visualized with UNC-49::RFP) were detected both at and between GABAergic presynaptic boutons along the DNC of madd-4B(tr185) animals (Fig. 4C). Because both GABAergic and cholinergic neuromuscular synapses are located en passant, the continuous distribution of UNC-49::RFP along the DNC suggests that GABA_AR clusters face both GABAergic (DD) and cholinergic (DA, DB, AS) presynaptic boutons (Fig. 2A). This is likely due to remaining MADD-4L expression in cholinergic MNs of madd-4B(tr185) mutants, leading to ectopic GABA_AR trapping at cholinergic neuromuscular synapses (Pinan-Lucarre et al., 2014). The aberrant clustering of GABA_ARs was not accompanied by an increase in UNC-49 expression levels on the muscle (Fig. 4E). Importantly, expression of madd-4B specifically in GABAergic MNs led to a complete rescue of the GABA_AR clustering defects at postsynaptic muscle cells, consolidating a non-cell-autonomous role for madd-4B.

Because both unc-30 and madd-4B mutants display defects in GABA_AR localization (Figs 2-4), we hypothesized that the transcription factor UNC-30 regulates madd-4B in GABAergic MNs. To test this, we employed CRISPR/Cas9 genome editing and generated an endogenous fluorescent reporter of madd-4B transcription. In agreement with transgenic madd-4B reporter expression (Kratsios et al., 2015), this endogenous 2xNLS::mScarlet::SL2::madd-4B transcriptional reporter (mScarlet::madd-4B hereafter) was expressed in both cholinergic and GABAergic MNs, although higher levels were observed in GABAergic MNs (Fig. 5A,B). To test the effect of unc-30 gene loss in madd-4B expression specifically in GABAergic MNs, we crossed a nuclear marker for cholinergic MNs (cho-1::SL2::YFP::H2B) to the mScarlet::madd-4B reporter in the context of control and unc-30(e191) animals (Fig. 5B). We observed a significant decrease in the number of GABAergic cells (defined by the absence of cho-1::SL2::YFP::H2B signal) expressing mScarlet::madd-4B in unc-30(e191) mutants at the fourth larval (L4) stage (Fig. 5B,C), i.e. all 13 GABAergic neurons (DD2-DD5, VD3-VD11) of the VNC expressed mScarlet::madd-4B in control animals, but only around ten neurons in unc-30(−) mutants (Fig. 5A,C). Importantly, the remaining mScarlet::madd-4B expression in these ten GABAergic neurons was also decreased, as revealed by quantification of mScarlet::madd-4B fluorescence intensity with single-cell resolution (e.g. VD3, DD2, VD4, VD5, DD3, VD6) (Fig. 5D). The remaining madd-4B expression suggests that additional, yet-to-be-identified factors cooperate with UNC-30 to activate madd-4B expression in these cells. In our analysis, we excluded six GABAergic MNs (DD1, DD6, VD1, VD2, VD13) because their location (outside the VNC) makes their identification less straightforward.

The single-cell resolution of our analysis indicates that unc-30 controls madd-4B transcription in both DD (e.g. DD2, DD3) and VD (e.g. VD3, VD4, VD5) neurons (Fig. 5D). To corroborate this, we quantified mScarlet::madd-4B expression at larval stage 1 (L1), a developmental stage at which only DD (not VD) neurons are present in the C. elegans nerve cord (Fig. S4A). Again, we found a significant decrease in madd-4B expression in DD neurons of unc-30(e191) mutants (Fig. S4B). In agreement with our endogenous transcriptional reporter (mScarlet::madd-4B), expression of a transgenic translational madd-4B reporter is also affected in unc-30 animals at L1 (Chen et al., 2021). Altogether, we conclude that unc-30 controls endogenous madd-4B transcription in GABAergic MNs, and this effect is observed both at early (L1) and late (L4) larval stages.

Because madd-4B expression is reduced in GABA MNs of unc-30(e191) animals (Fig. 5B-D), we investigated whether madd-4B is a direct target of UNC-30. Leveraging an available dataset of chromatin immunoprecipitation followed by sequencing (ChIP-Seq) (Yu et al., 2017), we identified UNC-30 binding at four genomic regions (peaks I-IV): peak I is upstream of madd-4L, whereas peaks II-IV surround the first exon of madd-4B (Fig. 5A). Within peaks I, II and IV, we identified a canonical UNC-30 binding site (TAATCC) (Cinar et al., 2005; Eastman et al., 1999). To test whether UNC-30 binding upstream of madd-4B is required for madd-4B expression, we employed CRISPR/Cas9 genome editing to delete a 506 bp-long region that spans peak II (Δ506 bp; Fig. 5A). This manipulation was conducted in animals carrying the endogenous mScarlet::madd-4B reporter. Similar to unc-30(e191) mutants, we observed a decrease in the number of GABAergic MNs expressing mScarlet in L4 stage animals homozygous for the 506 bp deletion (Fig. 5B,C), and in the levels of mScarlet expression in individual GABAergic MNs (Fig. 5D).

ChIP-Seq data and our analysis of mScarlet::madd-4B^Δ506bp animals strongly indicate that UNC-30 acts directly to activate madd-4B transcription. To further test this possibility, we examined transgenic animals carrying different transcriptional reporters of madd-4B (Fig. 5A). First, we found that reporters containing DNA sequences either 4.4 kb (madd-4B^4.4kb::GFP) or 1.9 kb (madd-4B^1.9kb::GFP) upstream of madd-4B (both containing peak II) drive GFP expression in GABA MNs (Fig. 5A,E,F), consistent with the endogenous madd-4B::mScarlet reporter (Fig. 5C). Second, madd-4B^4.4kb::GFP reporter expression depends on unc-30, evidenced by a reduction in the number of GABA MNs expressing GFP in unc-30(e191) mutants (Fig. 5E). Third, we found that mutation of the UNC-30 binding site II (wild type: TAATCC; mutated: GCGCGC) results in a significant decrease in the number of GABA MNs expressing madd-4B^1.9kb::GFP (Fig. 5F). Altogether, we conclude that UNC-30 acts directly to activate madd-4B transcription in GABA MNs (Fig. 5G).

The ChIP-Seq data also showed UNC-30 binding (peak I) upstream of exon 1 of madd-4L (Fig. 6A). Because madd-4L is known to be specifically expressed in cholinergic MNs (Kratsios et al., 2015; Pinan-Lucarre et al., 2014), we hypothesized that UNC-30 binds upstream of madd-4L to repress its transcription in GABA MNs. Supporting this notion, transgenic GFP animals carrying a 2.9 kb sequence upstream of madd-4L showed increased expression in GABA MNs of unc-30(e191) mutants (Fig. 6E,F). Next, we employed CRISPR/Cas9 genome editing and generated an endogenous mScarlet reporter for madd-4L (syb624[2xNLS::mScarlet::SL2::madd-4L]), referred to hereafter as mScarlet::madd-4L (Fig. 6A). We observed ectopic expression of mScarlet::madd-4L in GABA MNs of unc-30(e191) mutant animals both at L1 (Fig. S5A,B) and at L4 (Fig. 6B-D). We found that up to 13 GABA MNs of the VNC ectopically express mScarlet::madd-4L in unc-30(e191) mutants (Fig. 6B-D). However, mutating the endogenous UNC-30 binding sequence (WT site I: TAATCC site; MUT site I: GCGCGC) within peak I had no effect on the number of GABA MNs expressing mScarlet::madd-4L (Fig. S5C,D). We conclude that, in GABA MNs, UNC-30 controls two isoforms of the same synapse organizer in opposite ways: it directly activates madd-4B and indirectly represses madd-4L (Fig. 6G).

The continuous expression of both unc-30 and madd-4B in GABAergic MNs, at larval stages and throughout adulthood, raises the question of whether UNC-30 is required continuously to activate madd-4B expression. We therefore generated an inducible unc-30 allele, leveraging the auxin-inducible degradation (AID) system (Towbin et al., 2012; Ashley et al., 2021; Zhang et al., 2015). Using CRISPR/Cas9, we introduced the mNG::3xFLAG::AID cassette before the unc-30 STOP codon (Fig. 7A). The resulting unc-30::mNG::3xFLAG::AID allele (syb2344) serves as an endogenous fluorescent (mNG, mNeonGreen) reporter of UNC-30, which can be degraded upon auxin treatment owing to the presence of the AID degron (Fig. 7B). We generated double-homozygous animals for unc-30::mNG::3xFLAG::AID and ieSi57 (Peft-3::TIR1::mRuby), the latter providing pan-somatic expression of TIR1 – an F-box protein that binds to AID in the presence of auxin, leading to proteasomal degradation of UNC-30::mNG::3xFLAG::AID. As proof of principle, we first assessed UNC-30::mNG::3xFLAG::AID levels in individual GABA MNs in ethanol-treated (control) or 4 mM auxin-treated animals for 2 days, from L3 to adult day 1 (Fig. 7B). Compared to ethanol-treated animals, auxin-treated animals showed a robust reduction in the levels of UNC-30::mNG::3xFLAG::AID fluorescence intensity, indicating efficient depletion (Fig. 7C-E). Auxin-treated animals also exhibited a significant reduction in mScarlet::madd-4B fluorescence intensity levels in all nerve cord GABAergic MNs (Fig. 7F-H, Fig. S6). Hence, UNC-30 is required during late larval and young adult stages to maintain madd-4B expression in GABAergic MNs (Fig. 7I). The continuous requirement of UNC-30 is likely to be essential to maintain GABA_AR clustering throughout life.

Prompted by our madd-4B observations, we next examined whether UNC-30 is continuously required to maintain the expression of additional target genes. A previous study using the strong LOF allele e191 showed that UNC-30 activates the expression of two GABA identity genes during development: unc-25 and unc-47 (Eastman et al., 1999). We obtained similar results by using the same unc-30(e191) allele (Fig. S7). Mutating the UNC-30 binding site (TAATCC) in transgenic unc-25 and unc-47 reporter animals resulted in reduced reporter expression in GABA MNs, strongly suggesting UNC-30 regulates these targets via direct binding (Eastman et al., 1999). Consistent with these previous findings, analysis of the UNC-30 ChIP-Seq dataset showed UNC-30 binding in the cis-regulatory regions of unc-25 and unc-47 endogenous loci (Fig. 8C,F).

Whether UNC-30 is required at post-embryonic stages to maintain the expression of these crucial determinants of GABAergic identity (e.g. unc-25, unc-47) and function is not known. We again employed the AID system in late larval stages, this time assessing the effect of UNC-30 depletion on expression levels of unc-25 and unc-47. We observed a significant reduction in their expression levels in all nerve cord GABAergic MNs (Fig. 8A-H, Fig. S8), suggesting that UNC-30 is not only required during early development to initiate expression of GABA biosynthesis genes, but also to maintain their expression during late larval stages (Fig. 8I).

Is UNC-30 also continuously required for normal animal behavior? Animals lacking unc-30 gene activity (homozygous strong loss-of-function mutants) display a characteristic locomotory phenotype nicknamed ‘shrinker’ (McIntire et al., 1993; Chalfie and Jorgensen, 1998), i.e. unc-30 mutants hypercontract their body wall muscles in response to touch owing to the lack of GABAergic MN inhibitory input to muscles. We indeed observed a striking and fully penetrant ‘shrinker’ phenotype in unc-30(e191) mutants compared with control animals (Fig. 8J). Importantly, auxin-mediated depletion of UNC-30 specifically at late larval/early adult stages also resulted in ‘shrinker’ animals (Fig. 8J). Because the auxin system does not fully eliminate UNC-30, as evidenced by quantification of UNC-30::mNG::3xFLAG::AID expression levels in individual GABAergic MNs (Fig. 7H), the ‘shrinker’ phenotype displayed variable expressivity (none, mild, strong) upon auxin treatment (Fig. 8J). In the control (ethanol) condition, we observed no shrinkers, suggesting that tagging the endogenous unc-30 gene with the mNG::3xFLAG::AID cassette does not result in detectable hypomorphic effects on locomotory behavior (Fig. 8J). We conclude that UNC-30 is continuously required for normal touch response.

A handful of UNC-30 target genes are known to date, including unc-25, unc-47, pde-4/PDE4B, acy-1/ADCy9, oig-1, flp-11, flp-13 and ser-2 (Table 1) (Cinar et al., 2005; Eastman et al., 1999; Howell et al., 2015; Shan et al., 2005; Yu et al., 2017). A unifying theme emerging from these studies is that UNC-30 acts as a transcriptional activator. Our findings on madd-4L (Fig. 6), however, suggest a repressive role for UNC-30 in GABA MNs. We therefore sought to identify new UNC-30 target genes to determine whether the duality in UNC-30 function (activator and repressor) is broadly employed.

First, we identified putative unc-30 targets by searching for UNC-30 binding peaks in genes that are normally expressed in GABA MNs (Smith et al., 2024; Taylor et al., 2021) (Table 1). In total, we tested six genes (tsp-7/Cd63, aman-1/Man2b1, nhr-49/Hnf4a, mab-9/Tbx20, nhr-40/NHR, ilys-4) by either generating new transgenic reporter animals (nhr-49, mab-9, nhr-40), or using available reporters (tsp-7, aman-1, ilys-4). Reporter expression for five of these genes (tsp-7, aman-1, nhr-49, mab-9, nhr-40) was significantly reduced in GABA MNs of unc-30 (e191) mutant animals (Fig. 9A,B, Table 1). Because ChIP-Seq shows UNC-30 binding to all four of these genes (Fig. 9A,B), it is likely that UNC-30 acts as a direct activator of tsp-7, aman-1, nhr-49, nhr-40 and mab-9 transcription (Fig. 9D).

Next, we aimed to identify genes that, like madd-4L, are repressed by UNC-30. We searched for UNC-30 binding peaks in genes that are not expressed in GABA MNs, but instead are normally expressed in cholinergic nerve cord MNs (Table 1). In total, we tested 11 genes, for which transgenic reporter animals were available. Two (unc-53/NAV1 and glr-5/GRIK4) of the 11 reporters showed ectopic expression in GABA MNs of unc-30(e191) mutant animals (Fig. 9C, Table 1). Interestingly, late larval depletion of UNC-30 with the AID system did not affect glr-5 expression, suggesting UNC-30 is not required at later stages to repress this gene in GABA MNs (Fig. S9).

Altogether, our work identified nine new UNC-30 target genes; six are activated (madd-4B, tsp-7, aman-1, nhr-49, mab-9, nhr-40) and three are repressed (madd-4L, unc-53, glr-5) by UNC-30 (Fig. 9D). This analysis significantly expands the known repertoire of UNC-30 target genes in the C. elegans nervous system (Table 1), consolidating its previously known activator role and uncovering a putative repressive function.

Here, we describe a molecular mechanism that coordinates two spatially separated processes that are essential for the function of chemical synapses: NT biosynthesis in the presynaptic cell and NT receptor clustering at the postsynaptic cell. Using C. elegans neuromuscular synapses as a model, we show that the terminal selector-type transcription factor UNC-30 is required continuously to maintain expression of GABA identity genes (e.g. unc-25, unc-47) in presynaptic GABAergic MNs, thereby ensuring GABA synthesis and release. In postsynaptic target muscle cells, UNC-30 acts non-cell-autonomously to control clustering of GABA_ARs – the most prominent inhibitory NT receptors in animal nervous systems (Ghit et al., 2021; Goetz et al., 2007). Mechanistically, we propose that UNC-30 directly regulates the production of MADD-4B, a secreted synapse organizer. Hence, UNC-30 coordinates GABA_AR clustering on postsynaptic muscle cells with the acquisition and maintenance of GABAergic identity of presynaptic cells (Fig. 1C, model 3), safeguarding GABA neurotransmission.

UNC-30 is required in late larval and adult stages to maintain expression of GABA biosynthesis genes (e.g. unc-25, unc-47), consolidating its role as a terminal selector of GABA MN identity (Eastman et al., 1999; Jin et al., 1994). Further, UNC-30 acts directly to activate and maintain transcription of madd-4B, a secreted synapse organizer necessary for GABA_AR clustering on target muscle cells (Pinan-Lucarre et al., 2014). Because UNC-30 is continuously required in GABA MNs, this simple co-regulatory strategy of unc-25, unc-47 and madd-4B by a terminal selector may ensure that key features of a functional synapse will continue to appear together throughout life (Fig. 1C). Hence, the presynaptic neuron will continue to synthesize and release GABA (ensured by continuous unc-25 and unc-47 expression) and the postsynaptic neuron will constantly have the means to receive GABA via cognate receptor clustering (ensured by continuous madd-4B expression). Because UNC-30 orthologs are expressed in planarian (Currie and Pearson, 2013; Marz et al., 2013), fly (Vorbruggen et al., 1997), zebrafish (Shi et al., 2005) and mouse (Bifsha et al., 2017; Luk et al., 2013) nervous systems, the co-regulatory principle described here may be broadly applicable across species.

The only other known example of a terminal selector that operates in an analogous manner is UNC-3, the sole C. elegans ortholog of the COE (Collier/Olf/EBF) family of proteins (Dubois and Vincent, 2001). In nerve cord cholinergic MNs, UNC-3 acts as a terminal selector, directly regulating scores of effector genes (e.g. ACh biosynthesis proteins, ion channels) (Kratsios et al., 2012; Li et al., 2020). Like unc-30, unc-3 is not expressed in C. elegans muscles, yet it is required for AChR clustering on muscle cells (Kratsios et al., 2015). In cholinergic MNs, UNC-3 not only directly activates madd-4B (function of which in these cells is discussed in the next section), but also madd-4L, which is required for AChR clustering (Kratsios et al., 2015). By contrast, we find that UNC-30 activates madd-4B but represses madd-4L in GABA MNs, thereby ensuring expression of the appropriate madd-4 isoform (madd-4B). Altogether, NT receptor clustering in C. elegans neuromuscular synapses is achieved by two different terminal selectors regulating, in distinct ways, the two isoforms of the same synapse-organizing molecule: UNC-3 activates both madd-4B and madd-4L in cholinergic MNs, whereas UNC-30 activates madd-4B but represses madd-4L in GABA MNs.

Besides NT receptor clustering, additional synaptic connectivity defects have been reported in MNs of unc-3 and unc-30 mutant animals (Barbagallo et al., 2017; He et al., 2015; Howell et al., 2015; Kratsios et al., 2015; Philbrook et al., 2018). Specifically, cholinergic MN input onto GABA MNs is disrupted in unc-3 mutants (Barbagallo et al., 2017). In this case, UNC-3 controls nrx-1/neurexin, a synapse organizer necessary for AChR localization onto dendrites of GABA MNs (Philbrook et al., 2018). UNC-30 has been implicated in C. elegans synaptic remodeling, as it is necessary to prevent premature synapse rewiring of DD neurons and aberrant synapse rewiring of VD neurons (He et al., 2015; Howell et al., 2015). This is achieved by UNC-30 directly regulating OIG-1, a single immunoglobulin domain protein that functions as a synaptic organizer (He et al., 2015; Howell et al., 2015). Consistent with a recent review (Hobert and Kratsios, 2019), this work and the aforementioned studies provide strong evidence for expanding the definition of terminal selector genes; they not only regulate effector genes required for NT biosynthesis and neuronal signaling (e.g. ion channels), but also control synaptic connectivity via the regulation of distinct synapse organizers.

It is tempting to speculate that mammalian terminal selectors may operate in an analogous manner. For example, the terminal selector of mouse spinal MNs, Isl1 (Cho et al., 2014), may control transcription of agrin, an MN-derived synapse organizer necessary for AChR clustering in mouse skeletal muscles (Sanes and Lichtman, 1999).

Synapse organizers are cell adhesion or secreted molecules that control synapse formation and/or maintenance (Johnson-Venkatesh and Umemori, 2010; Mizumoto et al., 2023). Their adhesive and signaling properties mediate uni- or bidirectional signaling, enabling pre- and/or postsynaptic differentiation (Sanes and Zipursky, 2020). Understanding the spatiotemporal regulation of synapse organizers is important because synapses must be built at the right place and time. However, we know very little about the transcriptional mechanisms that control synapse organizer expression, in part because these molecules usually have multiple isoforms (e.g. neurexins, neuroligins, agrin, MADD-4/punctin) (Fukai and Yoshida, 2021; Gomez et al., 2021; Yamagata et al., 2015). Multiple isoforms can be produced via either alternative RNA splicing or promoter usage. To date, substantial research has focused on alternative splicing of synapse organizers (e.g. neurexin isoforms) (Traunmuller et al., 2023), leaving their transcriptional mechanisms poorly understood.

MADD-4L is only produced by cholinergic MNs (Kratsios et al., 2015; Pinan-Lucarre et al., 2014). Upon secretion, it promotes clustering of L-AChRs by an extracellular scaffold composed of LEV-10 (LEVamisole resistant-10), LEV-9 and OIG-4 (One ImmunoGlobulin domain-4) (Gally et al., 2004; Gendrel et al., 2009; Pinan-Lucarre et al., 2014; Rapti et al., 2011). By contrast, MADD-4B is produced by both cholinergic and GABAergic MNs. At GABAergic neuromuscular synapses, MADD-4B promotes GABA_AR clustering on muscle cells through binding to NLG-1/neuroligin and activation of UNC-40/DCC signaling (Zhou et al., 2020; Tu et al., 2015; Maro et al., 2015). At cholinergic neuromuscular synapses, MADD-4B inhibits the attraction of GABA_A receptors by MADD-4L (Pinan-Lucarre et al., 2014). Hence, spatial (neuron type-specific) regulation of MADD-4 isoform expression is crucial for the formation and function of excitatory (ACh) and inhibitory (GABA) synapses in C. elegans. Our previous work identified UNC-3 as a critical activator of both madd-4 isoforms in cholinergic MNs (Kratsios et al., 2015). Here, we show in GABA MNs that UNC-30 controls the two madd-4 isoforms in opposite ways; it provides direct and positive input to the madd-4B promoter and negative input to the madd-4L promoter, thereby ensuring proper GABA_AR clustering on target muscle cells.

In humans, PITX gene mutations cause various congenital defects and cancer (Tran and Kioussi, 2021). Pitx genes belong to the PAIRED (PRD) class of highly conserved homeobox genes. In mice, Pitx genes play crucial roles in the development of the nervous system, craniofacial structures, and limbs (reviewed by Tran and Kioussi, 2021). Pitx2 and Pitx3 are expressed in discrete cell populations of the mouse midbrain and spinal cord (Bifsha et al., 2017; Luk et al., 2013; Zagoraiou et al., 2009). Pitx3 is essential for the survival of dopaminergic neurons of the substantia nigra, a key cellular substrate of Parkinson's disease (Luk et al., 2013). Importantly, human variants of PITX2 or PITX3 affect eye development (Tran and Kioussi, 2021; Semina et al., 1998). PITX2 variants cause Axenfeld–Rieger syndrome, a disorder that affects primarily the eyes, whereas PITX3 variants are associated with congenital cataracts (Muzyka et al., 2023; Tumer and Bach-Holm, 2009).

Mechanistically, functional assays showed that human PITX2 and PITX3 gene variants result in reduced transcriptional activity (Wang et al., 2013; Zhao et al., 2015; Verdin et al., 2014). However, the transcriptional targets of PITX proteins remain poorly defined and whether they act as transcriptional activators and/or repressors is not well defined. Our study contributes to these knowledge gaps in three ways. First, we identify nine new UNC-30 target genes (activated: madd-4B, mab-9, nhr-49, nhr-40, tsp-7, aman-1; repressed: madd-4L, unc-53, glr-5), significantly expanding the list of PITX targets in the nervous system (Table 1). Second, consistent with its previously described direct mode of activation of genes involved in GABA biosynthesis and neuronal rewiring (Eastman et al., 1999; Yu et al., 2017; Howell et al., 2015; He et al., 2015), our mutational analysis indicates that UNC-30 acts directly to activate madd-4B. Last, we propose that, in GABA MNs, UNC-30 acts both as an activator and repressor of distinct sets of genes. A similar dual role for UNC-30 has recently been described in C. elegans glia, where it promotes GLR glia morphology and represses alternative mesodermal fates (Stefanakis et al., 2024).

Future studies are needed to dissect the molecular mechanism underlying the dual role of UNC-30 in GABA MNs. It is likely that cooperation with distinct transcription factors shifts its transcriptional activity from an activator to a repressor. Candidates include LIN-39/HOX, a known transcriptional activator in GABA MNs (Feng et al., 2020), and UNC-55/NR2F, a known transcriptional repressor in these cells (Shan et al., 2005; Yu et al., 2017). Another limitation relates to maintenance of GABA_AR clustering. Although we showed that UNC-30 is required to maintain madd-4B transcription in late larval/early adult stages, it remains unknown whether inducible UNC-30 depletion at these stages affects maintenance of GABA_AR clustering. Alternatively, it is possible that MADD-4B secretion is only required to initiate GABA_AR clustering, but subsequent maintenance of clustering may rely on other factors intrinsic to the postsynaptic cell.

Although GABA_ARs undergo similar ectopic clustering at cholinergic neuromuscular junctions in both madd-4(tr185) and unc-30(e191) animals, there is a difference: GABA_ARs partially cluster at GABAergic neuromuscular junctions in the madd-4(tr185) mutant, whereas they fail to properly cluster at these synapses in the unc-30(e191) mutant (Figs 2 and 4). This suggests that, in addition to madd-4B, other yet-to-be identified UNC-30 target genes may also act as GABA synapse organizers. In the madd-4(tr185) mutant, these additional genes may operate in parallel with MADD-4L, which has been shown to trap GABA_ARs at cholinergic neuromuscular junctions (Pinan-Lucarre et al., 2014). Last, our work is focused on neuromuscular synapses. Notably, punctin (ADAMTSL3) and other secreted synapse organizers (e.g. cerebellins, pentraxins, Sema3F, BDNF) are expressed in the mammalian brain (Dow et al., 2011; Uemura et al., 2010; Sigoillot et al., 2015; Cramer et al., 2023). Hence, similar co-regulatory strategies to the one described here may operate in neuron–neuron or neuron–glia synapses in the central nervous system.

Worms were grown at 15°C, 20°C or 25°C on nematode growth media (NGM) plates seeded with bacteria (Escherichia coli OP50) as food source. All C. elegans strains used in this study are listed in Table S1.

Reporter gene fusions for cis-regulatory analysis were made using either PCR fusion or Gibson Assembly Cloning Kit (NEB, 5510S) (Hobert, 2002). Targeted DNA fragments were fused (ligated) to tagrfp or gfp coding sequence, followed by the unc-54 3′ UTR. Mutations of UNC-30 binding sites were introduced by PCR mutagenesis. The product DNA fragments were either injected into young adult pha-1(e2123) hermaphrodites at 50 ng/µl using pha-1 (pBX plasmid) as co-injection marker (50 ng/µl) and further selected for survival, or injected into young adult N2 hermaphrodites at 50 ng/µl (plus 50 ng/µl pBX plasmid) using myo-2::gfp as co-injection marker (3 ng/µl) and further selected for GFP signal. Primer sequences used for reporter construct generation are provided in Table S2.

Single-copy insertion alleles krSi92 [Punc-47::T7::madd-4S::GFP] and krSi342 [Punc-30::gfp1-10] were generated by the miniMos method (Frokjaer-Jensen et al., 2014). Worms were injected with 15 ng/μl plasmid of interest containing the promoter and the open reading frame fused to fluorescent proteins, 50 ng/μl pCFJ601 (Mos1 transposase), 10 ng/μl pMA122 (negative selective marker Phsp16.2::peel-1) and 2.5 ng/μl pCFJ90 (Pmyo-2::mCherry). Neomycin (G418) was added to plates 24 h after injection at 1.5 μg/μl final concentration. Candidate plates were heat-shocked for 2 h at 34°C. Worms with an insertion were isolated and subsequently maintained as homozygous carriers of krSi92 or krSi342.

CRISPR/Cas9 genome editing was performed by SunyBiotech or SEGICel following standard procedures (Dickinson and Goldstein, 2016). The unc-30 endogenous reporter allele syb2344 [unc-30::mNG::3xFlag::AID] was generated by SunyBiotech via CRISPR/Cas9 by inserting the mNG::3xFLAG::AID cassette immediately before the unc-30 termination codon. The endogenous madd-4L reporter allele syb624 [2xNLS::mScarlet::SL2::madd-4L] was generated by inserting the 2xNLS::mScarlet::SL2 cassette immediately after the ATG of madd-4L. The endogenous madd-4B reporter allele syb623 [2xNLS::mScarlet::SL2::madd-4B] was generated by inserting the 2xNLS::mScarlet::SL2 cassette immediately after the ATG of madd-4B. The mScarlet sequence is preceded by two copies of a nuclear localization signal (2xNLS) and followed by the SL2 trans-splicing element (Fig. 5A). Hence, the 2xNLS::mScarlet sequence and endogenous madd-4B are transcribed as one mRNA, but each is translated independently as a result of the SL2 element. The endogenous madd-4B reporter allele syb3561 [2xNLS::mScarlet::SL2::madd-4B^Δ506 bp] was generated by creating a 506 bp-long deletion (−1433 bp to −927 bp from the madd-4B ATG) in the background strain carrying the endogenous madd-4B reporter allele syb623 [2xNLS::mScarlet::SL2::madd-4B]. The endogenous madd-4L reporter alleles kas31 and kas32 were generated by mutating the canonical UNC-30 binding site I (TAATCC) to GCGCGC in the background strain carrying the endogenous madd-4L reporter allele syb624 [2xNLS::mScarlet::SL2::madd-4L].

The cla-1 bab462 [cla-1::spgfp11×7] knock-in allele was generated by inserting the spgfp11×7 cassette immediately before the STOP of cla-1.

AID-tagged proteins are conditionally degraded when exposed to auxin in the presence of TIR1 (Ashley et al., 2021; Zhang et al., 2015). Animals carrying auxin-inducible alleles of unc-30 (syb2344[unc-30::mNG::3xFLAG::AID]) IV were crossed with ieSi57 animals that express TIR1 pan-somatically. Auxin (indole-3-acetic acid; Alfa Aesar, A10556) was dissolved in ethanol (EtOH) to prepare 400 mM stock solutions which were stored at 4°C for up to 1 month. NGM agar plates were poured with auxin or EtOH added to a final concentration of 4 mM and allowed to dry overnight at room temperature. Plates were seeded with OP50 bacteria. To induce protein degradation, worms of the experimental strains were transferred onto auxin-coated plates and kept at 20°C. As a control, worms were transferred onto EtOH-coated plates instead. Auxin solutions, auxin-coated plates and experimental plates were shielded from light.

For Figs 2-4 and  Fig. S1, young adult C. elegans were mounted on 2% agarose (w/v in water) dry pads immersed in 2% polystyrene beads (0.1 mm diameter, Polyscience, 00876-15) diluted in M9 buffer. Images were taken using a Nikon-IX86 microscope (Olympus) equipped with an Andor spinning disk system (Oxford Instruments), a 60×/NA 1.42 oil immersion objective and an Evolve EMCCD camera. For each animal (Figs 2-4, Fig. S1), an image of the DNC or VNC at the first quarter of the worm was acquired as a stack of optical sections (0.2 µm apart). Pearson's coefficient was calculated as described (Tu et al., 2015). Pearson's r values are indicated. For the remaining figures, worms were anesthetized using 100 mM sodium azide (NaN₃) and mounted on a 4% agarose pad on glass slides. Images were taken using an automated fluorescence microscope (Zeiss, Axio Imager.Z2). Several z-stack images (each ∼1 µm thick) were acquired with a Zeiss Axiocam 503 mono using the ZEN software (Version 2.3.69.1000, Blue edition). Representative images are shown following maximum projection of 1-8 µm z-stacks using the maximum intensity projection type. Image reconstruction was performed using ImageJ/Fiji software (Schindelin et al., 2012).

MNs were identified based on a combination of the following factors: (1) colocalization with fluorescent markers with known expression pattern, (2) invariant cell body position along the VNC, or relative to other MN subtypes, (3) MN birth order, and (4) number of MNs that belong to each subtype.

For Fig. S1, immunofluorescence staining was performed as described (Tu et al., 2015). Images were acquired using a Leica 5000B microscope equipped with a spinning disk CSU10 (Yokogawa) and a Coolsnap HQ2 camera.

To quantify FI of individual MNs in the VNC, images of worms from different genetic backgrounds were taken with identical parameters through full-thickness z-stacks that covered the entire cell body. Image stacks were then processed and quantified for FI using Fiji. The focal plane in z-stacks that had the brightest FI was selected for quantification. Background signal was minimized by using Fiji's background subtraction feature (rolling ball at 50 pixels). The cell outline was manually selected, and Fiji was used to quantify the FI and area to get the mean value for FI.

For quantification, box and whisker plots were adopted to represent the quartiles in graphs. The box includes data points from the first to the third quartile value with the horizontal line in box representing the median value. Upper and lower limits indicate the maximum and minimum, respectively. Unpaired t-test with Welch's correction was performed and P-values were annotated. Visualization of data and P-value calculation were performed using GraphPad Prism Version 9.2.0 (283). Each experiment was repeated twice.

For experiments presented in Figs 2-4 and Fig. S1, box and whisker plots show median, lower and upper quartiles, and whiskers represent s.d. For Pearson's correlation coefficient, fluorescence intensity and boutons density quantifications, Mann–Whitney tests (comparison of two genotypes) and Kruskal–Wallis followed by Dunn's post-tests (comparison of more than two genotypes) were performed if data did not follow a normal distribution (Shapiro normality test) or did not show equality of variances (Bartlett test). If data were normal and showed equality of variances, unpaired, two-tailed Student's t-tests were performed.

In all figures, box and whisker plots show median, lower and upper quartiles, whiskers represent minimum and maximum. Black circles depict values.

We thank the Caenorhabditis Genetics Center (CGC), funded by NIH Office of Research Infrastructure Programs (P40 OD010440), for providing strains. We thank members of the Kratsios lab (Mira Antonopoulos, Jayson Smith, Filipe Marques, Anthony Osuma, Manasa Prahlad) for providing feedback, and Yihan Chen and Jihad Aburas for technical assistance. We thank Laure Granger and Driss Laabid for technical assistance. We thank Le Centre d'Imagerie Quantitative Lyon-Est (LyMIC-CIQLE, Lyon, France) imaging facility for support and access to equipment. One strain was generated by SEGiCel (SFR Santé Lyon Est CNRS UAR 3453, Lyon, France).

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202733.reviewer-comments.pdf