CFI-1 functions unilaterally to restrict gap junction formation in C. elegans

Across the nervous system, neurons mainly communicate with each other through chemical and electrical synapses (Jabeen and Thirumalai, 2018; Pereda, 2014). Chemical transmission involves the Ca²⁺-dependent release of vesicles containing neurotransmitters from a presynaptic cell (O'Rourke et al., 2012; Südhof, 2012). In contrast, electrical synapses form intercellular channels that enable communication between the interior of two adjacent neurons (Connors and Long, 2004). Electrical coupling is prevalent during developmental stages of the nervous system (Bruzzone and Dermietzel, 2006) and plays an important role in neuronal differentiation, cell death, cell migration, synaptogenesis and neuronal circuit formation (Chalfie et al., 1985; Curtin et al., 2002; Lopresti et al., 1974; Yeh et al., 2009; Dere and Zlomuzica, 2012; Talukdar et al., 2022). In mature nervous systems, electrical synapses account for nearly 20% of neuronal connections and broadly regulate neuron activity and neuronal circuits (Chalfie et al., 1985; Connors and Long, 2004; Cook et al., 2019; White et al., 1986; Zolnik and Connors, 2016). In comparison to chemical synapses, however, little is known of the mechanisms governing the formation of electrical synapses between a specific pair of neurons.

Gap junctions are the biophysical substrate for the neurophysiological component of electrical synapses (Connors and Long, 2004). Connexins and innexins form the channel-forming proteins in vertebrate and invertebrate gap junctions, respectively (Abascal and Zardoya, 2013). Although these proteins exhibit unrelated amino acid sequences, they share remarkably similar transmembrane topologies and morphological structures. Both connexins and innexins are formed by four transmembrane-spanning domains, two extracellular and one intracellular loop, and cytoplasmic C- and N-terminal endings (Söhl et al., 2005). Usually, six or eight connexins or innexins undergo oligomerization to form a hemichannel (Stauffer et al., 1991; Oshima et al., 2016). Hemichannels in adjacent neurons appose one another to form a gap junction channel (Walker and Schafer, 2020). Ectopically-expressing vertebrate connexin36 proteins could form electrical synapses and reprogram behavior in Caenorhabditis elegans (Choi et al., 2020; Rabinowitch et al., 2014). Meanwhile, innexin-containing gap junctions can function in vertebrates (Dykes et al., 2004; Phelan et al., 1998), suggesting a conserved mechanism for gap junction assembly. Recent studies further indicate that, with their powerful genetics and accessibility to live imaging, invertebrate model systems can be exploited to identify regulators of gap junction formation in mammals (Palumbos et al., 2021).

In cultured cells, overexpression of either connexins or innexins is sufficient to drive the assembly of gap junctions at the interface between random pairs of adjacent cells (Elfgang et al., 1995; Rabinowitch et al., 2014; Teubner et al., 2000). In vivo, however, gap junctions do not assemble between adjacent neurons that express compatible gap junction subunits (Bhattacharya et al., 2019; Fukuda, 2017; Greb et al., 2017; White et al., 1992; Yao et al., 2016). Thus, the assembly of neuron-specific electrical synapses requires additional regulatory mechanisms. Furthermore, in contrast to chemical synapses, which are primarily located at axonal boutons, electrical synapses can couple various neuronal compartments and processes (Alcami and Pereda, 2019), adding another layer of complexity to their functional impact. Gap junctions often occur as so-called gap junction plaques that can contain thousands of gap junction channels, and presence in aggregates has been an important criterion for the identification of gap junctions (Goodenough and Paul, 2009). A functional gap junction channel is formed by two opposing hemichannels located in the plasma membrane of targeting cells. It is unclear how individual electrically coupled cells contribute to the formation of gap junctions and whether the formation of gap junctions is controlled by one side or conjointly by both sides of the partner cells.

Here, we identified that the PVC interneuron forms gap junction connections with the PVR interneuron in C. elegans. The transcription regulator CFI-1 (ARID3) is specifically expressed in the PVC interneuron. Reducing cfi-1 expression in PVC is sufficient to enhance gap junction formation in its partner PVR neuron. Conversely, the ectopic expression of cfi-1 in PVR inhibits gap junction formation in PVC neuron. Beyond its rescue activity, providing cfi-1 in PVR leads to additional behavioral consequences, indicating that the unilateral regulation of gap junction formation could exert a distinct influence on neuron circuits. This influence is likely due to the unique connectivity of each electrical coupling cell in the neuronal network.

The gap junction protein innexin UNC-9 (Uncoordinated 9) is widely distributed in the C. elegans nervous system (Barnes and Hekimi, 1997). Using Punc-53 promoter-driven GFP-tagged UNC-9, we previously demonstrated its efficacy in highlighting gap junction connections established by BDU interneuron and PLM touch sensory neuron (Zhang et al., 2013). In addition to BDU, neurons expressing Punc-53 also include PVP, PVQ, ALN, PLN, and several unidentified neurons in the tail region (Fig. 1H) (Stringham et al., 2002). Hence, gap junctions formed by those neurons could also be revealed. PVQ neurons (PVQL and PVQR), with cell bodies situated in lumbar ganglia, extend neurites anteriorly and form gap junctions to each other around the anus region. Indeed, with Punc-53::UNC-9::GFP marker, distinctive fluorescence puncta could be observed in the anus region. Double labeling confirmed that these UNC-9::GFP puncta are indeed situated on PVQ neurons (Fig. S1A,B). In addition to the anus region, Punc-53-driven UNC-9::GFP puncta also appear on the posterior ventral cord (Fig. 1A,B). However, before our study, the specific neurons responsible for forming these gap junction connections had not been identified.

To explore the molecular mechanism underlying gap junction formation in vivo, we conducted a genetic screen to identify mutants exhibiting altered distribution of Punc-53::UNC-9::GFP signal. From this screen, we isolated the xd424 mutant. In xd424 animals, there was a noticeable increase in the number of UNC-9::GFP puncta in the posterior ventral cord region (Fig. 1A boxed region, B,C), while the distribution of UNC-9::GFP puncta in the head and tail region remained relatively unchanged (Fig. S1A,C,D). Through genetic mapping and whole-genome sequencing, we identified a nonsense mutation (Q362°chre) in the cfi-1 gene in the xd424 mutant genome (Fig. 1C). cfi-1 encodes a worm homolog of yeast DRI and mammalian ARID3A/B proteins. All ARID family members feature a DNA-binding domain initially identified for its interaction with AT-rich DNA elements. Functioning as transcription regulators, ARID proteins have been implicated in regulating cell growth, differentiation, and development (An et al., 2010; Joseph et al., 2012; Ratliff et al., 2016). However, whether the ARID family plays a role in gap junction formation is unclear. Other cfi-1 alleles, including cfi-1(ot786), cfi-1(ky651) and cfi-1(ky650) mutants, display increased UNC-9::GFP puncta phenotype (Fig. 1A-C) similar to xd424 animals. Complementary tests further showed that xd424 is a cfi-1 allele (Fig. 1A). In addition to UNC-9, the gap junction protein UNC-7 is also broadly expressed within the worm nervous system. When we introduced the Punc-53::UNC-7::GFP marker into xd424 mutant animals, we observed an increase in the number of UNC-7::GFP puncta (Fig. 1D,E). Similarly, we examined the expression of the neuronal innexin protein INX-13 and found that, in xd424 mutant animals, the number of Punc-53::INX-13::GFP puncta was also elevated compared to wild-type controls (Fig. 1F,G). unc-1 encodes the stomatin-like protein and has been shown to regulate gap junctions in C. elegans (Chen et al., 2007). In wild-type animals, we found that the co-expressed (Punc-53 driven) UNC-1::wrmScarlet signal is adjacent to UNC-9::GFP puncta (Fig. S1E). When cfi-1 mutant was introduced into the double labeling marker, we found that the UNC-1::wrmScarlet signal is markedly increased. Meanwhile, the UNC-9::GFP signal remains spatially adjacent to the elevated UNC-1::wrmScarlet signal (Fig. S1F). These findings suggest that the xd424 mutation affects the overall abundance of gap junction connections rather than specifically altering UNC-9::GFP aggregates.

The augmented presence of gap junction puncta in cfi-1 mutant animals may potentially arise from an elevation in Punc-53-expressing neurons. To test this possibility, we examined the Punc-53-driven GFP marker and found that the number of Punc-53-expressing neurons or the intensity of Punc-53-driven GFP is not increased by the cfi-1(ot786) mutation (Fig. 1H-J). This suggests that it is unlikely that cfi-1 regulates gap junction formation through altering the expression of Punc-53 promoter.

Introducing a wild-type copy of cfi-1 into Punc-53-expressing neurons suppressed the increased gap junction puncta phenotype of xd424 (Fig. 2A,B), indicating that loss of cfi-1 function is responsible for the increased gap junction formation defect. Additionally, cfi-1 could act cell-autonomously within Punc-53-expressing neurons to regulate gap junction formation.

Surprisingly, however, the cfi-1 gene appears not to be expressed in Punc-53-expressing neurons. In the CRISPR/Cas9-created CFI-1::GFP knock-in (KI) line, CFI-1::GFP KI positive cells are mainly distributed in pharyngeal muscles and neurons in the head and ventral cord region (Fig. 2C; Fig. S2A,B) (Kerk et al., 2017). In the tail region, cfi-1 is exclusively expressed in two PVC (PVCL and PVCR) and two LUA (LUAL and LUAR) neurons (Fig. 2D; Fig. S2A). Double-labeling analysis further indicated that the nuclear localized CFI-1::GFP signal does not overlap with any Punc-53::mCherry-expressing cells (Fig. 2E). Pan-neuronally expressing cfi-1 could rescue the excessive gap junction phenotype of the cfi-1 mutant (Fig. 2A), while Pbnc-1 or Punc-129 promoter-driven cfi-1 expression in cholinergic motor neurons and/or body wall muscles in or surrounding ventral nerve cord could not suppress the increased UNC-9::GFP puncta phenotype in cfi-1(xd424) mutants (Fig. S2C). Intriguingly, a wild-type cfi-1 driven by a Pcfi-1 promoter, the expression of which is limited to PVC and LUA neurons in the tail region (Fig. S3A), could significantly suppress the increased UNC-9::GFP puncta phenotype of cfi-1 mutants (Fig. 2A). Hence, endogenous cfi-1 may function in LUA and/or PVC to regulate gap junction formation in Punc-53-expressing neurons.

Providing cfi-1 in non-Punc-53-expressing LUA and PVC neurons successfully rescued the cfi-1 mutant phenotype, indicating a potential cell-non-autonomous function of cfi-1. How does this cell-non-autonomous rescue occur? We hypothesized that PVC and/or LUA might form gap junction connections with certain Punc-53-expressing neurons (Fig. 2F) and the loss of cfi-1 function in PVCs and/or LUAs resulted in defects in gap junction formation in the partner cells of PVCs and/or LUAs, as revealed by the Punc-53::UNC-9::GFP marker. If this hypothesis holds, which Punc-53-expressing neurons  – the PVC and/or LUA – form gap junctions? To address this question, we first examined whether LUA or PVC neurons form gap junctions on the posterior ventral cord. LUA neurons extend relatively short neurites anteriorly, terminating at the anus region (lumbar commissures), and do not reach the posterior ventral cord region where UNC-9::GFP puncta are observed. In contrast, PVC neurites extend to the nerve ring region in the head (White et al., 1986). We therefore suspected that the UNC-9::GFP puncta on the posterior ventral cord are more likely to form on partner cells of PVC neurons rather than LUA neurons. Indeed, when we co-expressed Punc-53::UNC-9::GFP with the PVC-specific marker (Pnmr-1 promoter-driven mCherry) (Shaham and Bargmann, 2002), we found that the UNC-9::GFP puncta on the posterior ventral cord are aligned with PVC neuronal processes (Figs 3A, 4G).

We further tested whether cfi-1 function in PVC neurons is required for the formation of gap junctions in Punc-53-expressing neurons. To address this, we used the auxin-inducible degradation (AID) system (Zhang et al., 2015). In the C. elegans genome, we integrated the wrmScarlet::degron DNA sequence into the 3′-terminal end of the cfi-1 coding region. We then expressed the TIR1 E3 ligase in PVC neurons using the Pnmr-1 promoter. Upon exposure to auxin, the wrmScarlet signal was selectively reduced in PVC neurons but remained unchanged in LUA neurons, indicating the specific depletion of CFI-1 protein in PVC neurons. When cfi-1 function was specifically reduced in PVCs, we observed a substantial increase in the number of Punc-53::UNC-9::GFP puncta (Fig. 3C-H). Together, the above data collectively suggest that cfi-1-expressing PVC neurons form gap junctions with Punc-53-expressing neurons, and cfi-1 function in PVCs is crucial for suppressing gap junction formation in Punc-53-expressing neurons.

Next, we set out to identify the Punc-53-expressing neurons with which the PVC neurons make gap junctions. The known Punc-53-expressing neurons include ALNs, PLNs, PVPs, and PVQs. Co-labeling experiments with markers including Plad-2::mCherry (ALN and PLN), Podr-2b::mCherry (PVP), and Psra-6::mCherry (PVQ), confirmed the expression of Punc-53 in these neurons (Chou et al., 2001; Troemel et al., 1995; Wang et al., 2008) (Fig. S3B-D). Additionally, the Punc-53::GFP signal was observed in unidentified neurons in the tail region. Through further co-labeling experiments, we identified these unknown Punc-53-expressing neurons as PVWs (labeled by Pflp-7) (Kim and Li, 2004), PVR (labeled by Pflp-10), and PQR (also labeled by Pflp-10) (Kim and Li, 2004) (Fig. S3E,F), but not PLM (labeled by Pmec-7::mCherry) (Mitani et al., 1993) or PVN (labeled by Ppdf-1::mCherry) (Barrios et al., 2012) neurons (Fig. S3G,H).

Among the Punc-53-expressing neurons, ALN and PLN neuronal processes are positioned laterally, away from the ventral cord (Fig. S3B), making it unlikely for them to form gap junctions with PVCs. PVW neurons form gap junctions with PVC neurons but in the distal anterior region of the ventral cord (White et al., 1986). Apart from that, there is limited information about which Punc-53-expressing neurons may form gap junctions on PVCs. Therefore, to identify the partner cells of PVCs in gap junction formation, we further employed the Punc-53::UNC-9::GFP marker and examined the distribution of GFP puncta on PVP, PVQ, PVW, PVR, or PQR neurons. Our survey revealed that Punc-53-driven UNC-9::GFP puncta are absent on PVQs (Fig. 4A,G), PVWs (Fig. 4B,G), or PVPs (Fig. 4C,G). In contrast, on the posterior ventral cord, the UNC-9::GFP puncta are exclusively localized on the PVR neuron (Fig. 4D,G). We further co-labeled the PVR neuron and PVC gap junctions and found that PVCs indeed form gap junctions with the PVR neuronal process (Fig. 4E,G). Conversely, PVR gap junctions are specifically situated on PVC neurons (Fig. 4F,G). Apart from UNC-9, PVR also expresses innexin protein UNC-7 (https://cengen.shinyapps.io/CengenApp/), which could partner with UNC-9 to form heterotypic gap junctions (Shui et al., 2020; Starich et al., 2009). To further test the gap junction partnership between PVR and PVC, we expressed UNC-7 in PVR and UNC-9 in PVC. We found that the gap junction proteins from PVR and PVC are adjacent to each other (Fig. 4H). Taken together, PVCs and PVRs form gap junctions with each other (Fig. 4L).

Does cfi-1 influence the gap junction formation between PVC and PVR neurons? Co-labeling experiments with the PVR neuron revealed that the excessive Punc-53::UNC-9::GFP puncta in cfi-1 mutants were localized on PVR (Fig. 4I,K), rather than on PVQ, PVP, or PVW neurons (Fig. 4J,K). Conversely, when examining PVC neurons, we found that the surplus UNC-9::GFP puncta were positioned on PVC neurons (Fig. 3B). Therefore, cfi-1 specifically regulates the formation of gap junctions between PVC and PVR neuronal pairs.

So far, we had been using promoter-driven UNC-9::GFP to examine the spatial arrangement of gap junction connections in C. elegans neurons. To address potential concerns regarding overexpression, we further employed the Native and Tissue-specific Fluorescence (NATF) technique. NATF combines CRISPR/Cas9-mediated genome editing with the split-GFP method, facilitating the specific labeling of proteins within tissues at their endogenous levels (He et al., 2019). Basically, we divided the GFP molecule into two fragments, GFP1-10 and GFP11. The DNA sequence encoding the GFP11 fragment was inserted into the 3′-end of the unc-9 gene via the CRISPR/Cas9 method. In UNC-9::GFP11 KI worms, no green fluorescence signal was detectable. We then introduced the tissue-specific-expressing promoter-driven GFP1-10 fragment into UNC-9::GFP11 KI worms. This allowed for the characterization of endogenous gap junctions within specific tissues or cells (Fig. 5A,B). In comparison to Punc-53-driven UNC-9::GFP, the signal from NATF-mediated Punc-53-driven UNC-9::split-GFP was dimmer, with fewer visible UNC-9::split-GFP puncta (Fig. S4A,B). Nevertheless, consistent with Punc-53-driven UNC-9::GFP, UNC-9::split-GFP revealed the presence of PVQ gap junctions in the anus region (Figs S1B, S4A,B boxed region). Similarly, puncta of UNC-9::split-GFP were observable in the posterior ventral cord region, indicating a similar distribution pattern of gap junctions in vivo between Punc-53::UNC-9::GFP and Punc-53::UNC-9::split-GFP.

Upon introducing the PVC marker into the Punc-53::UNC-9::split-GFP expression line, we found that the UNC-9::split-GFP puncta on the posterior ventral cord are specifically located on PVC neurons (Fig. 5F). This observation supports the idea that PVCs form gap junctions onto Punc-53-expressing PVR neurons. Notably, aside from PVC-PVR gap junctions, no other discernible UNC-9::split-GFP puncta were evident in the posterior ventral cord region. In cfi-1 mutant animals, the number of UNC-9::split-GFP puncta along the posterior ventral cord was increased (Fig. 5C,I). Through double staining, we confirmed that the surplus UNC-9::split-GFP puncta in cfi-1 mutants were formed between PVC (Fig. 5F) and PVR neurons (Fig. 5G). Additionally, we generated a 3D reconstruction image, illustrating that the green Punc-53::UNC-9::split-GFP puncta are embedded within PVC neurons (Fig. 5H). Could CFI-1 function in either PVC or PVR to regulate the formation of endogenous PVC-PVR gap junctions? Using the UNC-9::split-GFP marker, we found that introducing the wild-type cfi-1 gene into PVR or PVC could lead to a significant rescue effect in cfi-1 mutants (Fig. 5D,E,I). Together, these findings further support the notion that cfi-1 operates bidirectionally in either PVC or PVR to regulate gap junction formation between PVC and PVR neurons (Fig. 4L).

Given the broad expression of cfi-1 along the ventral cord, we further investigated its impact on endogenous gap junction formation in motor neurons through the NATF approach. We found that cfi-1 mutation does not impede the formation of gap junction connections in DD/VD neurons (Fig. S4C,D). Thus, cfi-1 may selectively regulate gap junction formation in the PVC-PVR neuronal pair.

Electrical coupling through gap junctions dynamically influences neuronal circuitry and behavioral outcomes. In C. elegans, PVC serves as a major command interneuron for mediating forward locomotion triggered by posterior touch (Fig. S5A) (Chalfie et al., 1985). PVR forms gap junctions with touch-sensing neurons ALM and PLM (White et al., 1986), which are responsible for receiving anterior and posterior touch stimuli, respectively (Chalfie et al., 1985). Here, we discovered that PVR also formed gap junctions with PVC interneurons (Fig. S5A), and mutations in cfi-1 led to excessive gap junction formation rather specifically between PVC and PVR neurons (Fig. S5B). Notably, cfi-1 mutant animals have been shown to exhibit defects in touch response (Shaham and Bargmann, 2002). Specifically, when gentle touch stimuli were applied around the posterior region, the anterior movement of cfi-1 mutant animals was inhibited (Fig. 5K), while touch stimuli applied around the anterior region resulted in posterior locomotion similar to wild type (Fig. 5J). This observation is consistent with the role of cfi-1 in restricting electrical coupling between PVC and PVR neurons. Introduction of wild-type cfi-1 into PVC neurons successfully rescued the posterior touch response defect in cfi-1 mutant animals, highlighting the crucial role of PVCs in posterior touch response (Fig. 5K; Fig. S5A). Excessive gap junction formation between PVC and PVR neurons could be inhibited by expressing cfi-1 in PVR neurons. Remarkably, introducing wild-type cfi-1 into PVR neurons also suppressed the posterior touch response defect of cfi-1 mutants (Fig. 5K). We noticed that when cfi-1 was expressed only in PVR but not in PVC, some animals exhibited abnormal backward movement upon posterior touch (Fig. 5K). Given the distinct connectivity patterns of PVC and PVR, it is possible that individual partner cells of the electrically coupled neuron pair may have distinct outputs to the touch circuit (Fig. S5C and Discussion). Nevertheless, the rescuing activity of cfi-1 in either PVC or PVR supports the notion that unilateral bidirectional regulation of gap junction formation could profoundly influence neuron function and, subsequently, the behavioral outcomes controlled by relevant neural circuits.

The nearly completed connectome of the simple worm nervous system offers a distinctive and pivotal model for elucidating the principles governing gap junction formation in vivo. Here, we revealed that gap junction formation can be orchestrated by either one of the partner cells. Given that a functional gap junction is constituted by hemichannels originating from opposing cells, this unilateral bidirectional regulatory mechanism could exert intricate and profound effects on the neuronal network.

The C. elegans genome contains a single ARID3 ortholog CFI-1, whereas mammals contain three ARID3 subfamily members (ARID3A, ARID3B and ARID3C) (Kortschaket al., 2000; Shaham and Bargmann, 2002; Wilsker et al., 2005). Previous studies have indicated that ARID3 members participate in, for example, embryonic patterning, cell lineage, cell cycle, apoptosis, stem cell differentiation and tumorigenesis (An et al., 2010; Herrscher et al., 1995; Joseph et al., 2012; Numata et al., 1999; Ratliff et al., 2014, 2016; Roy et al., 2014; Shandala et al., 1999; Valentine et al., 1998; Webb et al., 2011). In C. elegans, cfi-1 was originally cloned based upon its role in the development of neuronal diversity (Shaham and Bargmann, 2002). Through the generation of an in vivo binding map on the C. elegans genome, Li et al. recently identified that the majority of CFI-1-targeting genes encode markers associated with neuronal terminal differentiation (Li et al., 2023). Specifically, the cfi-1 mutation results in the downregulation of several key effector genes in PVC, including the vesicular acetylcholine transporter UNC-17 and the choline transporter CHO-1 (Glenwinkel et al., 2021). This alteration in gene expression may consequently impact the functional dynamics of PVC. Therefore, CFI-1 may act a terminal selector, orchestrating the terminal differentiation of distinct neuron types. In line with this study, we found a tight correlation between cfi-1 expression in PVCs and gap junction formation in the PVC-PVR neuronal pair. Notably, electrical coupling emerges as a crucial feature in neuronal terminal differentiation. Here, cfi-1 does not appear to regulate the specificity of gap junction formation; instead, it modulates the quantity or abundance of gap junctions formed between appropriate partner cells. Thus, terminal selectors may also exert control over the electrical coupling strength of specific neuronal subtypes.

As a transcription regulator, how could CFI-1/ARID3 achieve inhibition of gap junction formation from one of the partner cells? In the PVC-PVR pair, endogenous cfi-1 exhibits expression solely in PVC neurons, not in the PVR neuron. Within the PVCs, CFI-1 may directly modulate the gene expression of gap junction proteins, thereby exerting control over the quantity or abundance of gap junction connections. The whole genome binding map did not identify innexin unc-9 as a direct target gene of CFI-1 (Li et al., 2023). However, given the limited number of neurons (confined to PVC and PVR) in which gap junction formation is influenced by CFI-1, it is difficult to rule out the possibility that CFI-1 could directly regulate the gene expression of UNC-9 or other gap junction proteins. Alternatively, CFI-1 is involved in the transcription regulation of genes associated with the stability of gap junctions. While numerous proteins in cultured cells have been identified as regulators of gap junction homeostasis, few studies have investigated how the abundance of gap junction connections is determined in an intact nervous system. Recent research has demonstrated that the conserved CASPR family nematode protein NLR-1 anchors the F-actin at the gap junction plaque, and plays a crucial role in gap junction assembly (Meng and Yan, 2020). The cAMP-dependent signaling pathway regulates the trafficking of gap junction proteins, thereby controlling the subcellular localization of C. elegans gap junction connections (Palumbos et al., 2021). Neurobeachin, a conserved protein with multiple protein-binding domains, localizes to electrical synapses and regulates the robust localization of neuronal connexins postsynaptically (Martin et al., 2023; Miller et al., 2015). However, the specific mechanistic involvement of CFI-1 in modulating the trafficking or stability of gap junction proteins within established cellular pathways, leading to the inhibition of gap junction formation, remains elusive. How does the formation of gap junctions in PVC neurons influence the corresponding process in PVR neurons? We suspect that CFI-1 may regulate gap junction formation in opposing cells indirectly by modulating the transcription of unidentified extracellular signaling molecules. Alternatively, the formation of connexin/innexin hemichannel clusters in one cell may directly impact the clustering of connexin/innexin hemichannels in another cell. This direct effect likely arises from the stability of gap junctions facilitated by interaction between juxtaposed hemichannels (Sosinsky and Nicholson, 2005). It is noteworthy, however, that the expression of compatible gap junction subunits does not guarantee the assembly of gap junction connections between adjacent neurons in vivo (Bhattacharya et al., 2019; Fukuda, 2017; Greb et al., 2017; White et al., 1992; Yao et al., 2016). Thus, while the formation of connexin/innexin hemichannel clusters in one cell may influence the clustering of connexin/innexin hemichannels in another cell, the assembly of neuron-specific electrical synapses likely involves additional regulatory mechanisms.

cfi-1 is expressed in diverse motor neurons within the ventral cord. However, the overall distribution of gap junctions on ventral cord motor neurons does not appear to be affected by cfi-1 mutations. This highlights the cell-context dependent nature of cfi-1 in regulating gap junctions. Consistent with this observation, CFI-1 engages in collaborative interactions with multiple transcription regulators to influence the differentiation of distinct neuronal cell types. In partnership with the POU homeodomain transcription factor UNC-86, CFI-1 directly activates terminal differentiation genes specific to IL2 neurons (Shaham and Bargmann, 2002). Concurrently, CFI-1 collaborates with two distinct homeodomain proteins, UNC-42 (PROP1) and CEH-14 (LIM), to govern the terminal differentiation processes of AVD and PVC interneurons, respectively (Glenwinkel et al., 2021; Berghoff et al., 2021). The transcription factor UNC-3, an ortholog of human EBF1, EBF2, and EBF4, regulates expression of cfi-1 in motor neurons (Kerk et al., 2017). Within ventral nerve cord motor neurons, CFI-1 functions as a repressor of the glutamate receptor gene glr-4 (GRIK4). The glr-4 gene undergoes positive regulatory control from three conserved transcription factors, namely UNC-3, LIN-39 (HOXA5), and MAB-5 (HOXB8/HOXC8), while experiencing negative regulation from CFI-1 (Kerk et al., 2017). Therefore, CFI-1 likely collaborates with distinct sets of transcription regulators in various neuronal subsets, and its functional requirement in gap junction formation may only be evident in selective neuronal populations.

The touch response could be restored when placing cfi-1 in its non-native expressing cell, PVR, indicating that the PVC functional deficiency in cfi-1 mutants is likely due to over-coupling between PVC and PVR neurons. Intriguingly, the expression of cfi-1 in PVR neurons, but not in PVC neurons, leads to abnormal backward movement in response to posterior touch, indicating that, despite their electrical coupling, PVC and PVR neurons do not function as a unified entity. Rather, the entire and unique connectivity of a given neuron determines its distinctive function. So, how can we explain the abnormal backward movement? In a simple model (Fig. S5C), we might suggest that when a wild-type copy of cfi-1 is introduced into PVR, the electrical coupling between PVR and PVC is selectively reduced. Consequently, the electrical coupling between PVC and AVA may be strengthened, allowing more mechanical signal to flow into DA/AS motor neurons, thus causing the backward movement. Of course, one could explore different neuronal connections to achieve a similar behavioral outcome by altering the activity of other neurons. Our current understanding of gap junction formation is far from complete. The complex interactions among diverse terminal selectors suggest that there will likely be significant divergence in the regulatory mechanisms governing gap junction formation in individual neurons. To add further complexity, most neurons usually express more than one gap junction gene (Evans and Martin, 2002). Additionally, one side of an electrical synapse is not necessarily considered the mirror of the other (Rash et al., 2013). Therefore, determining whether cfi-1-mediated PVR-PVC gap junction formation would influence gap junctions between other neurons and PVC or PVR is rather challenging.

Electrical coupling could occur dendro-dendritically, somato-somatically, or between axons (Alcami and Pereda, 2019). In PVC and PVR neurons, gap junctions form in the middle of their anterior neuronal processes. The excess gap junction connections resulting from cfi-1 loss-of-function also occur in this region. What signifies this particular region? In the ventral cord, PVC neurons receive synaptic input from various neurons including AVA, PHB, PHC, VA12, LUA, PVM, PVN, DVA, and PVD, while primarily sending synaptic output to motor neurons such as VBs and DBs. Notably, PVD and DVA neurons are presynaptic to both forward and backward interneurons, contributing to both the anterior and posterior touch circuits. Conversely, PVR in the ventral cord receives synaptic input from PVM, DVA and PVM, and through gap junctions is coupled with DVA and PLM. Along with ALM and PLM, AVM neurons sense gentle mechanical stimuli to the body and provide input to command interneurons. This positioning of PVC-PVR gap junctions likely facilitates sensory-motor integration. However, how this specific subcellular localization of gap junctions is determined during development is completely unknown. Similarly, the unique roles of each neuron in specifying the location and abundance of gap junctions within the PVC-PVR pair are not yet understood. Our studies have only scratched the surface of the intricate regulatory mechanisms governing electrical synapse formation in vivo. Further investigation is needed to fully elucidate how characteristic electrical coupling is achieved within specific neuronal pairs in the nervous system.

C. elegans strain maintenance and genetic manipulation were performed under standard conditions as previously described (Brenner, 1974). Mutants and transgenic fluorescence reporters used in this study were: LGI, cfi-1(xd424), cfi-1(ot786), cfi-1(ky650), cfi-1(ky651), xdKi79 (CFI-1::GFP KI), xdKi83 (CFI-1::wrmScarlet::degron KI), LGII, xdIs174 (Punc-53::UNC-9::GFP, Pmyo-2::RFP), LGX, xdKi13 (UNC-9::GFP11 KI). Additional transgenic lines were: xdEx276 (Punc-53::GFP, Podr-1::RFP), xdEx2366 (Pcfi-1::GFP, Podr-1::RFP), xdEx2373 (Punc-53::CFI-1, Podr-1::GFP), xdEx2383 (Punc-53::CFI-1::mCherry, Podr-1::GFP), xdEx2386 (Punc-53::GFP, Pnmr-1::mCherry, Podr-1::GFP), xdEx2442 (Punc-53::GFP, Pflp-7::mCherry, Podr-1::GFP), xdEx2445 (Punc-53::GFP, Pflp-10::mCherry, Podr-1::GFP), xdEx2448 (Punc-53::GFP, Plad-2::mCherry, Podr-1::GFP), xdEx2462 (Punc-53::GFP, Psra-6::mCherry, Podr-1::GFP), xdEx2465 (Punc-53::GFP, Podr-2 2b::mCherry, Podr-1::GFP), xdEx2489 (Punc-53::GFP, Pcfi-1::MYR::mCherry, Podr-1::GFP), xdEx2495 (Punc-53::CFI-1, Podr-1::GFP), xdEx2509 (Punc-53::GFP1-10, Podr-1::RFP), xdEx2837 (Pcfi-1::CFI-1::mCherry, Podr-1::GFP), xdEx2915 (Punc-53::mCherry, Podr-1::RFP), xdEx2918 (Pcfi-1::CFI-1, Podr-1::GFP), xdEx3051 (Pnmr-1::mCherry; Pmyo-2::RFP), xdEx3071 (Pflp-7::mCherry; Podr-1::GFP), xdEx3072 (Podr-2 2b::mCherry; Podr-1::GFP), xdEx3073 (Pflp-10::mCherry; Pnmr-1::UNC-9::GFP; Podr-1::RFP), xdEx3076 (Pnmr-1::mCherry; Pflp-10::UNC-9::GFP; Podr-1::RFP), xdEx3079 (Pflp-10::mCherry; Podr-1::GFP), xdEx3080 (Pnmr-1::mCherry; Podr-1::GFP), xdEx3081 (Psra-6::mCherry; Podr-1::GFP), xdEx3117 (Pnmr-1::TIR1; Podr-1::GFP), xdEx3519 (Punc-53::UNC-1B::wrmScarlet; Podr-1::GFP), xdEx3523 (Punc-53::UNC-7::GFP; Podr-1::RFP), xdEx3533 (Punc-53::INX-13::GFP; Podr-1::RFP) and xdEx3564 (Punc-53::UNC-7::wrmscarlet; Pnmr-1::UNC-9::GFP; Podr-1::GFP).

The cfi-1(xd424) mutant was isolated from xdIs174 animals treated with ethylmethane sulfonate (EMS). We screened 10,000 mutagenized haploid genomes, and 16 mutations were isolated from this screen.

DNA fragments were inserted into the pSM, ΔpSM, pPD95.77 or pPD95.75 vector using the Gibson assembly method (pEASY-Uni Seamless Cloning and Assembly Kit, TransGen Biotech). For cfi-1 tissue specific rescue experiments, DNA constructs were injected into young adult animals at a concentration of 5 ng/µl. For split GFP labeling, DNA constructs containing the GFP1-10 fragment were injected into young adult animals at a concentration of 1 ng/µl. To label various neurons, the corresponding DNA constructs were injected into worms at a concentration of 50 ng/µl. To label gap junctions, the corresponding DNA constructs were injected into worms at a concentration of 5 ng/µl. The co-injection marker was Pmyo-2::RFP, Podr-1::GFP, or Podr-1::RFP injected at a concentration of 50 ng/µl. Integrated strains were obtained using the trimethylpsoralen/ultraviolet (TMP/UV) method (Gengyo-Ando and Mitani, 2000).

The KI strains were generated by CRISPR/Cas9-mediated genome editing, constructed by SunyBiotech and verified with PCR and sequencing. The sequences of sgRNA and corresponding primers are listed here. To generate the xdKi13 (UNC-9::GFP11 KI), sgRNAs (CGTATGGTTGCAACTCACGCCGG and CCGGAGAACTACCCTGTTACGAG) and primers (forward primer GGTGTTTTCCTACTTCGTATGGTTG and reverse primer GACGACTACACCCATTGACGAC) were used. To generate xdKi79 (CFI-1::GFP KI) and xdKi83 (CFI-1::wrmScarlet::degron KI), sgRNAs (TCAGTATCAATGGAAATCAACGG and ATCAACGGAATCACCTATCAAGG) and primers (F/s1: ATGGTGCATCGAGTATGAGGA; R/s2: CAAATTGCGATCACCGAGA; xdKi79-mid-F/s3: ATGCCCGAAGGTTATGTACAGG; xdKi83-mid-F/s3: CAGCCGACATCCCAGACTACTA; xdKi83-mid-R/s4: TTGAAGTCGGCGAGGTAA) were used.

Animals were placed on 2.5% agar pads in M9 buffer with 1.4% 1-phenoxy-2-propanol. Fluorescence images of nematodes were captured using the Leica SP8 confocal microscope. The photographs were taken at the young adult stage unless specifically indicated. Confocal stacks were projected into a single image. The number of UNC-9::GFP puncta with the gray value of ≥40 was counted. Relative UNC-9::GFP distribution on each neuron is defined as the GFP puncta number at the mCherry region on a single layer of image divided by total GFP puncta number along the ventral nerve cord region. The posterior ventral cord region is defined as extending from 50 µm posterior to the vulval region to 190 µm posterior to it. The tail region is characterized as ranging from 20 µm anterior to the anus region to 120 µm anterior to it. For Fig. 5B-E, the straightened ventral cord images were obtained using ImageJ. The 3D reconstruction of the co-localization analysis between Punc-53::GFP1-10; UNC-9::GFP11 and PVC neuron marker was performed using Imaris software.

To compare multiple groups, one-way ANOVA was used with an appropriate multiple comparisons post hoc test (the test used is stated in each figure legend). *P<0.05; **P<0.01; ***P<0.001; NS, not significant.

Temporally controlled protein degradation depletion using the AID system was adapted for C. elegans (Zhang et al., 2015). The degron-tagged CFI-1::wrmScarlet was generated using CRISPR/Cas9, and it was conditionally degraded in PVC neurons when exposed to auxin in the presence of PVC neuron-specific Pnmr-1::TIR1. The indole-3-acetic acid (IAA) was dissolved in DMSO to prepare a 500 mM store solution, and was preserved in a dark place at 4°C. The IAA solution was added to nematode growth medium (NGM) agar plates to a final concentration of 1 mM, and the plates were shielded from light at 4°C for 4 weeks. Concentrated OP50 was seeded on the auxin NGM plates and allowed to dry overnight at room temperature. To induce CFI-1 degradation, L4 stage Pnmr-1::TIR1; CFI-1::wrmScarlet::degron worms were transferred onto the IAA-coated plates and kept at 22°C. The worms were transferred to new IAA plates every 3 days. The CFI-1 expression pattern and UNC-9::GFP distribution were characterized in F1 young adults.

Gentle touch was delivered with an eyelash affixed to a pipet tip. The posterior or the anterior part of the worm was touched with the eyelash edge in a top-down fashion. Young adult worms were placed on NGM plates with a thin layer of OP50 lawn. Responses to stimulation were recorded within a 3 s window. Each worm was tested once. Each trial encompassed the evaluation of over 17 animals per genotype, with three independent replicates. All data are shown as mean±s.d.

We thank the Caenorhabditis Genetics Center (CGC, USA) and the National BioResource Project (NBRP, Japan) for providing strains. We also thank Dr Cornelia I. Bargmann for providing the cfi-1(ky650) mutant strain.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202955.reviewer-comments.pdf