Visualizing the organization and differentiation of the male-specific nervous system of C. elegans

It is generally appreciated that nervous systems are sexually dimorphic on a gross anatomical level. However, sex differences in nervous systems have been carefully mapped out, with single-cell resolution, in only very few animals. The nematode Caenorhabditis elegans is the only organism for which a complete cellular, lineage and anatomical map of the entire nervous system has been described for both sexes (Fig. 1) (Cook et al., 2019; Jarrell et al., 2012; Sulston et al., 1980; Sulston and Horvitz, 1977). With 387 neurons in total, the nervous system of the male is almost 30% larger than that of the hermaphrodite (302 neurons). Based on position, lineage, anatomy and molecular features, 294 neurons are shared between both sexes. Hermaphrodites, which are somatic females, contain an additional eight hermaphrodite-specific neurons that fall into two classes: the well-characterized HSN and VC motor neuron classes, both of which control egg laying behavior (Schafer, 2005). The male contains an additional 93 neurons that fall into 27 anatomically distinct classes (Fig. 1, Table S1) (Cook et al., 2019; Molina-Garcia et al., 2020; Sammut et al., 2015; Sulston et al., 1980), two located in the head (sensory neuron classes CEM and MCM), two in the ventral nerve cord (motor neuron classes CA and CP), and the remaining, heavily interconnected 23 classes located in the tail of the animal. With the exception of the four CEM sensory neurons in the head, which are born in the embryo and induced to die in hermaphrodites, all male-specific neurons are generated during postembryonic development from blast cells that proliferate, survive and differentiate in a male-specific manner (Fig. 1B) (Molina-Garcia et al., 2020; Sammut et al., 2015; Sulston et al., 1980). Based on cell division patterns, the 89 postembryonically generated male-specific neurons are generated at different larval stages. Each individual larval stage contributes to the generation of some of these postembryonic neurons (Sulston et al., 1980). However, when exactly these neurons terminally differentiate is poorly understood. Moreover, in his classic lineage studies Sulston et al. also noted that the number of two male-specific neuron classes, DX and EF, display a variable number of class members (Sulston et al., 1980). Similar variabilities in cell cleavage patterns have not been observed elsewhere within or outside the nervous system of C. elegans.

Despite many interesting aspects of the male nervous system, it has received little attention over the years compared with the nervous system of the hermaphrodite. A number of studies have illuminated aspects of the development and function of male-specific neurons, but those studies only dealt with a limited set of neurons (Barr et al., 2018; Emmons, 2014, 2018; Garcia et al., 2001; Garcia and Portman, 2016; Liu and Sternberg, 1995; Portman, 2017). Hence, many aspects of the development and function of the 93 male-specific neurons remain uncharted territory. With some notable exceptions, including the systematic mapping of neurotransmitter identities (Gendrel et al., 2016; Pereira et al., 2015; Serrano-Saiz et al., 2017), and marker analysis in the ray sensory neurons (Lints et al., 2004) and ventral nerve cord (Kalis et al., 2014), few molecular markers have been developed that label male-specific neurons. Single-cell transcriptome approaches have so far exclusively focused on the hermaphrodite (Cao et al., 2017; Packer et al., 2019; Taylor et al., 2021). This dearth of molecular markers complicates the means by which cellular expression patterns in the male tail can be unambiguously identified and therefore limits the ability to assess cell fate in specific mutant backgrounds.

Here, we address these shortcomings by showing that NeuroPAL, a previously described multicolor transgene that distinguishes all neuron classes in hermaphrodites (Yemini et al., 2021), can also be used to disambiguate the 93 neurons of the male nervous system. We find that the NeuroPAL transgene, which harbors more than 40 promoters that drive the expression of four distinct fluorophores, generates a color map that provides sufficient discriminatory power to identify all male-specific neurons reliably. We provide proof-of-principle examples that show how to use NeuroPAL to identify gene expression patterns in the nervous system, and we use the NeuroPAL color map to provide a number of insights into the development of the male-specific nervous system.

With the exception of neurotransmitter pathway genes (Gendrel et al., 2016; Lints and Emmons, 1999; Pereira et al., 2015; Serrano-Saiz et al., 2017), few molecular markers have been comprehensively described throughout the entire male-specific nervous system (www.wormbase.org). For several related neuron classes, for example the ray neurons, molecular markers are available, but they do not provide sufficient resolution to distinguish between all individual class members (Lints and Emmons, 1999; Lints et al., 2004). We set out to test whether the NeuroPAL transgene that we previously described for the C. elegans hermaphrodite (Yemini et al., 2021) would provide a similarly information-rich molecular map of the male-specific nervous system.

The NeuroPAL transgene was designed to provide color codes to all neurons of the C. elegans hermaphrodite (Yemini et al., 2021). This was achieved through the judicious use of four fluorophores with separable emission spectra (mTagBFP2, CyOFP1, Tag-RFP-T, mNeptune2.5), expressed under the control of a set of 43 different promoters with overlapping expression profiles [39 neuron type-specific promoters plus a synthetic ultra-pan-neuronal (UPN) driver made up of the cis-regulatory elements of four distinct, but fused, pan-neuronal promoters] (Yemini et al., 2021). Promoter choices were dictated by the goal of having neighboring neurons display distinct color codes, thereby unambiguously discriminating between neighboring neuron identities.

We found that three NeuroPAL transgenes (independently integrated transgenes otIs696, otIs669 and otIs670) distinguish all neighboring male-specific neurons from one another. This is illustrated in the whole-animal overview in Fig. 2A along with large-scale images of all regions of the C. elegans nervous system that contain male-specific neurons (Fig. 2B-D, Fig. S1). The origin of the color code for each neuron is listed in Table S1. For several of the neuron classes, we do not know from which driver the fluorophore color derives, but this is not relevant for providing disambiguation between neighboring neurons. NeuroPAL not only provides color codes for neuron classes for which few or no molecular markers were previously available, but it also distinguishes neuronal subclasses that could previously not be discriminated. For example, individual subclasses of A- and B-type ray neurons subtypes can all be distinguished based on color code and position (Fig. 2, Table S1).

We first used NeuroPAL to address questions that relate to stereotypy of the male-specific nervous system. In the original lineage analysis of the male tail, an unusual phenomenon, not observed anywhere else in the entire organism was reported: descendants of the U ectoblast produce variable numbers of DX and EF neurons, a notion indicated by stippled lines in the original lineage diagram (Sulston et al., 1980) (redrawn here in Fig. 1B). This contrasts the complete stereotypy and deterministic nature of all other C. elegans cell lineages, both neuronal and non-neuronal (Sulston et al., 1980; Sulston and Horvitz, 1977). Moreover, this variability was reported to be restricted to the EF and DX neurons that descend from the U neuroblast and that are located in the preanal ganglion (the EF3 and 4, and DX3 and 4 neurons). In contrast, the DX and EF neurons that are produced from the F neuroblast (EF1 and 2, DX1 and 2), located in the dorsal rectal ganglion, were generated in an apparently invariant manner (Sulston et al., 1980) (Fig. 1B). However, no quantification of this observation was provided. Also, because the lineage analysis relied on cleavage pattern alone, the extent to which the variably produced DX and EF neurons acquire a differentiated state was also not clear.

Using the distinctive NeuroPAL color codes for DX and EF neurons, we examined 22 young adult males and found variability in the presence of fully differentiated EF and DX neurons in the preanal ganglion, as assessed by wild-type expression of NeuroPAL colors in these neurons (Fig. 3A). Within the F-derived dorsorectal ganglion, 22/22 animals invariably showed two fully differentiated DX (DX1 and DX2) and two fully differentiated EF neurons (EF1 and EF2), corroborating the classic lineage report (Sulston et al., 1980). In the U-derived preanal ganglion, 19/22 animals showed one DX and one EF neuron (DX3 and EF3), 1/22 had one additional EF (EF4), and 2/22 had one additional EF (EF4) and one additional DX (DX4) neuron.

The EF and DX neurons are also the neurons with the greatest inter-animal variability in their relative positioning, as inferred by closely considering the overall variability of positioning of both sex-shared and sex-specific neurons in the tail of the animal. We had previously measured positional variability for neurons in the hermaphrodite head, where the vast majority of neurons are generated embryonically (Yemini et al., 2021); here, we observed a similar extent of variability in the male head, despite the addition of six male-specific neurons (Fig. 3B,C, Fig. S2A-C, Table S2). However, in the tail, where the vast majority of the postembryonically added male-specific neurons are located, there was substantially more positional variability, both in the sex-shared neurons as well as in the sex-specific neurons (Fig. 3D-H, Table S2). The EF and DX neurons stand out in the extent of variability in their positioning. It will be interesting to investigate whether the inter-animal variability in neuronal soma position in the male tail also translates into variability in neuronal process adjacency, and hence connectivity, between individual animals.

Compared with the hermaphrodite nervous system, there is a remarkable scarcity of molecular markers for neurons in the male tail. The vast majority of reporter transgenes that researchers generate to analyze the expression of their gene of interest are usually only examined in hermaphrodites. One reason for the reluctance of identifying sites of reporter gene expression in the male tail has been the absence of reliable landmark reporters for most male-specific neurons. The fluorescence emission properties of NeuroPAL are designed to be separable from those of GFP signals. This allows researchers to overlay a GFP signal from a reporter gene of interest onto the neuron-specific color barcodes of NeuroPAL, thus identifying the sites of expression of GFP (or CFP/YFP)-based reporter transgenes. As a proof of principle, we analyzed CRISPR/Cas9-engineered gfp-based reporter alleles of three neuropeptide-encoding genes that were previously uncharacterized in males (flp-3, flp-27 and nlp-51). To do so, we crossed these gfp-tagged alleles into a NeuroPAL background. We found that flp-3::T2A::3xNLS::gfp was expressed in the male-specific CA1, CA2, CA3 and CA4 neurons, located in the ventral nerve cord, as well as the male-specific R4A and SPV neurons in the tail (Fig. 4A). We found that flp-27::T2A::3xNLS::gfp was expressed in the male-specific neurons CEMV, CEMD, CA8, CA9, PGA and R7A, and was dimly and variably expressed in R6B as well as the sex-shared neuron ASG (Fig. 4B). nlp-51::SL2::GFP::H2B was expressed in the male-specific PVX, R3B and PHD neurons in the tail and variably in R4B (Fig. 4C); in addition, non-dimorphic expression was observed in the sex-shared neurons AIM, RIP and PVN (Fig. 4C). These expression patterns corroborate the molecular diversity of members of the CA-type ventral nerve cord motor neurons, and that of the ray sensory neurons, as noted previously with other markers (Kalis et al., 2014; Lints et al., 2004).

As described above, NeuroPAL is an indicator of expression for 39 neuron-type specific genes, as well as four pan-neuronal genes (the enhancers of which are fused together in the ‘UPN’ construct), marking all male-specific neurons. These reporter genes measure a wide variety of phenotypic features of a neuron, including neurotransmitter synthesis and transport, neurotransmitter receptors, neuropeptides, sensory receptors from various families, and pan-neuronal features (Table S1) (Yemini et al., 2021). The markers therefore provide a panoramic view of the differentiated state of all individual neurons, and this state can be probed for proper execution in mutant backgrounds.

We illustrate the utility of NeuroPAL for such mutant analysis using three prominent patterning genes: a miRNA (lin-4) (Lee et al., 1993), a HOX cluster gene (egl-5/AbdB) (Chow and Emmons, 1994) and a proneural bHLH gene (lin-32/Atonal) (Zhao and Emmons, 1995). The functions of these genes have been reported for only select parts of the male-specific nervous system (Chalfie et al., 1981; Chow and Emmons, 1994; Zhao and Emmons, 1995). We sought to assess whether their reported defects can be recapitulated and better characterized with NeuroPAL. Furthermore, we anticipated identifying novel defects in these mutants in previously unexamined parts of the male-specific nervous system. Both of these expectations were fulfilled in all three cases examined, as described in the following sections.

Animals lacking the lin-4 miRNA display an iteration of cellular fates normally executed at the first larval stage (Chalfie et al., 1981; Lee et al., 1993), as inferred mainly from analysis of the ectodermal V and T ectoblasts in the male. Specifically, based on cellular cleavage patterns and neuron-like nuclear morphologies, L1 stage-specific T-cell neurons appear to be duplicated, whereas T-derived ray neurons that are normally generated at late larval stages do not appear to be generated (Chalfie et al., 1981) (redrawn in Fig. 5B). We extended these previous findings through our ability to visualize neuronal differentiation programs with greater detail using NeuroPAL. We confirmed that T-derived ray neurons are not generated in lin-4 mutant animals whereas neurons displaying the color code of the T cell-derived PHC, PHD, PLN and PVW appear to be duplicated (Fig. 5C). Similarly, the V5 lineage-derived ray neurons R1A/B and R2A/B, normally generated at the L4 stage, failed to be generated in lin-4 mutants (Fig. 5D).

Cleavage defects in other lineages that produce male-specific neurons (B, Y, U, F) had been noted in lin-4 mutant males (Chalfie et al., 1981), but whether and to what extent neuronal fate specification is disrupted in these lineages remained unclear. We observed no cells with color codes representative of neurons derived from the B, Y and F-blast cells, which are normally born at late larval stages (Fig. 5C,E). This parallels the effect of lin-4 on the late-born neurons in the T and V lineage and underscores that in lin-4 mutants development is arrested in a juvenile state. The color patterns in the preanal ganglion, where U cell descendants are normally located, was too complex to interpret, and therefore we cannot infer the nature of defects in this lineage.

We also found that the fate of all late-born, male-specific neurons that are generated by the P neuroblast are lost in lin-4 mutant animals. In the tail, male-specific neuronal cell fates derived from the P10 and P11 lineages (HOA, HOB, etc.) appeared to be lost. Lastly, in the ventral nerve cord, we observed a loss of color codes of the CA8-9 and CP8-9 neurons (Fig. 5F). These neurons are normally generated by a cell division event at the L3 stage, and this division is possibly absent in lin-4 mutants. In contrast, P cell-derived ventral nerve cord motor neurons, which are generated in both sexes at early larval stages, differentiated normally in lin-4 null mutants (Fig. 5F).

The B, Y, U and F ectoblasts, which divide exclusively in males, express the HOX cluster protein EGL-5/AbdB (Ferreira et al., 1999). In egl-5 mutants, these ectoblasts fail to undergo proper divisions and generate no neurons (Chisholm, 1991). This conclusion was based on light-microscopy criteria, i.e. absence of characteristic dense neuronal nuclei. Using the neuronal cell fate markers present in the NeuroPAL transgene, we further corroborated this notion: none of the colored neurons that descend from B, Y, U and F (Fig. 1B) could be observed in egl-5 mutants (Fig. 6).

Previous work has revealed an anterior-posterior patterning role of the HOX genes mab-5 and egl-5 in the ray lineage (Emmons, 2005). Ray neuron 2 is MAB-5 positive and EGL-5 negative, whereas ray neurons 3 to 6 are EGL-5 positive (Lints et al., 2004). It was suggested that in egl-5 mutants, rays 3, 4 and 5 homeotically transform to the fate of ray 2 neurons. With the limited markers available at the time, this suggestion remained tentative (Lints et al., 2004). We verified this suggestion by confirming that color code changes in NeuroPAL are consistent with a ray 2 neuron identity transformation (Fig. 6B).

Using NeuroPAL, we discovered an additional homeotic identity transformation in the posterior-most, male-specific CA motor neurons. The CA8 and CA9 neurons adopted similar color codes to the more-anterior CA7 neuron (Fig. 6C). These transformations are conceptually similar to the homeotic transformations observed in the sex-shared, posterior-most DA and AS neurons in egl-5 mutants where the most posterior class member also transforms its identity to that of the more anteriorly located neuron (Kratsios et al., 2017).

The single ortholog of the proneural Atonal gene in C. elegans, lin-32, was initially identified and characterized based on its proneural role in V5 and V6 ectoblast-derived ray neurons (Zhao and Emmons, 1995). Using NeuroPAL, we observed variable loss of neuronal fate specification in the ray lineage in lin-32(tm1446) mutants in the V5- and V6-derived ray neurons, as evidenced by missing color codes in these neurons, including the pan-neuronal color (Fig. 6E,F). The most-posterior ray neurons, generated by the T lineage (Fig. 1B), were unaffected in lin-32 null mutants, as were all of the other T-derived neurons (Fig. 6E,F). Y-, U- and F-derived neurons, as well as P cell-derived, male-specific VNC motor neurons (see lineage diagram in Fig. 1B) were also unaffected in lin-32 mutants (Fig. 6G,H). However, within the B lineage, we discovered a proneural role of lin-32 mutants. Whereas the neurons generated by the posterior daughter of the B ectoblast cell (DVE and DVF, located in the dorsorectal ganglion) differentiated normally, the neuronal cell fates generated by the anterior daughter of B (mostly spicule neurons) could not be detected (Fig. 6E,G).

We next used NeuroPAL as a tool to provide a panoramic view of timing of neuronal differentiation in the male tail. As illustrated in Fig. 1B, male-specific neurons are generated at multiple distinct developmental stages. Some male-specific neurons are generated in the embryo (CEM neurons), more are generated at the first larval stage (e.g. PVX of the P cell lineage), the L2 stage (e.g. PCB), the L3 stage (B and P cell descendants) and the L4 stage (mostly ray sensory neurons) (Sulston et al., 1980). Strikingly, we found that, despite their distinct generation time, the expression of the more than 40 terminal differentiation markers (located on the NeuroPAL transgene) are tightly coordinated over a relatively small window during the mid-to-late L4 stage (Fig. 7, Fig. S3). At early L4, color codes are not yet established (Fig. 7), nor are they at earlier larval stages (Fig. S3). The mid-to-late L4 stage is concomitant with the beginning of male tail retraction, a sexually dimorphic process that results in the generation of male-specific mating organs (Emmons, 2005; Nguyen et al., 1999; Sulston et al., 1980). The extent of this coordination is striking, not only because of the substantial number of cell types over which we observed this coordination, but also because of the breadth of the distinct molecular features that are covered by these molecular markers. Among the components of this marker set are the cis-regulatory elements from four distinct pan-neuronal genes (unc-11, rgef-1, ehs-1, ric-19). Like the neuron-class-specific marker genes, all these elements only begin to drive reporter gene expression during the mid-to-late L4 stage, concomitant with male tail retraction.

None of the markers for which we observed a coordinated, delayed onset during the L4 stage in sex-specific neurons displayed a delay in sex-shared and embryonically born neurons. That is, all NeuroPAL markers turn on after sex-shared neuron birth and remain stable throughout all larval stages (Fig. S3). The best illustration of the specificity of the coordinated expression wave in male-specific neurons are sex-shared neurons that are born in both sexes at similar stages of larval development. We consistently found that, in all cases, the NeuroPAL color code turned on shortly after the generation of these neurons (i.e. after their terminal cell division) (Fig. S3). For example, the sex-shared PDB and VD13 neurons, both close lineal relatives to the male-specific PVX neuron, were born at about the same time as the male-specific PVX (Fig. 1B). Whereas PVX generated its color code only at the L4 stage, PDB and VD13 generated their color code much earlier, at the late L1 stage that occurs soon after their birth (Fig. S3). Similarly, the L2-generated, male-specific PCB neuron class turned on its color code at the mid-to-late L4 stage, whereas the L2-generated, but sex-shared RMF neuron class and the L2-generated neurons of the postdeirid lineage generated their color codes shortly after their birth (Fig. 7).

In order to not be entirely reliant on the NeuroPAL transgene in assessing the timing of neuronal differentiation, we examined whether the delayed, coordinated onset of molecular differentiation features can be observed with other reporters as well. To this end, we utilized fosmid-based reporters for three genes, the pan-neuronally expressed rab-3 gene (otIs498), the synaptic organizer oig-1 (otIs450) and the vesicular transporter unc-47 (otIs564), as well as three reporter alleles in which endogenous genes were tagged with gfp through CRISPR/Cas9 genome engineering, namely the neuropeptide-encoding genes flp-27 and nlp-51 and the vesicular glutamate transporter eat-4. Whereas rab-3 was expressed in all male-specific neurons, oig-1, unc-47, flp-27, nlp-51 and eat-4 were expressed in a select subset of male-specific neurons, generated before the L4 stage. We found that all of these genes only turned on expression at the L4 stage (Fig. 8). The delay in onset of expression compared with generation of the neuron is too long to be explained by fluorescent protein maturation alone.

Lastly, we investigated whether the delayed differentiation onset of male-specific neurons depends on the sexual identity of the animal. To this end, we prevented cell death of the CEM neurons in hermaphrodites, using a canonical ced-3 mutant allele (Ellis and Horvitz, 1986), and assessed the onset of expression of the CEM differentiation marker pkd-2::gfp in these animals. If the proper timing of pkd-2 induction would require the animal to have a male identity (for example, if male-specific cells are required to induce pkd-2 expression), pkd-2::gfp expression should either not be induced or be initiated at an improper stage (e.g. right after the birth of the neurons). However, we found that pkd-2::gfp expression is still timed to the L4 stage of ced-3 mutant hermaphrodites (Fig. S4). Hence, male-specific cues are not required to time CEM differentiation properly.

We conclude that earlier born male-specific neurons delay their differentiation until the fourth larval stage and then differentiate in a coordinated manner, concomitantly with the differentiation of sexual organs.

The nervous system of the C. elegans male contains almost 30% more neurons than the hermaphrodite. The male-specific nervous system is structurally complex and controls the many intricate steps of male-mating behavior (Garcia et al., 2001; Liu and Sternberg, 1995; Barr et al., 2018; Portman, 2017). From a developmental perspective, it is fascinating to ask how complex interconnected circuitry is established during postembryonic development and integrated with an already existing, sex-shared nervous system. To address the many questions relating to the development and function of the male-specific nervous system, it is important to be able to characterize gene expression patterns and gene function, and to visualize male neuronal activity patterns. The tool that we present here, NeuroPAL, presents a major stepping stone to achieve these goals. NeuroPAL enables rigorous analysis of gene expression patterns and gene function in the male-specific nervous system. Moreover, the ability to combine NeuroPAL with GCaMP-based neuronal activity recordings – as recently shown in the hermaphrodite nervous system (Yemini et al., 2021) – will pave the way to decipher neuronal activity patterns reliably in the male nervous system.

Owing to its ability to visualize the expression of more than 40 distinct genes that reveal the live differentiated state of neurons throughout the entire nervous system of both sexes, we have been able to use NeuroPAL to gain insights into the development of the male-specific nervous system. We corroborated and extended the findings of an unusual non-stereotypic variability in the generation of a specific set of neurons, the EF and DX neurons. We further refined and also revealed the patterning roles of three gene regulatory factors: a Hox cluster gene, a miRNA and a bHLH transcription factor. Perhaps most interestingly, we used NeuroPAL to reveal that, despite their diverse birth dates, male-specific neurons coordinate the acquisition of terminal identity features to within a specific window of time, the mid-to-late fourth larval stage. At this time, other non-neuronal mating structures – including fans, rays and spicules – are generated (Emmons, 2005; Nguyen et al., 1999; Sulston et al., 1980). We term this coordinated differentiation ‘just-in-time’ differentiation, to illustrate that neurons only acquire their functional properties once all the ‘effector systems’ of the male-specific nervous system (i.e. all the end organs innervated by the male-specific neurons) come into existence, and thus only once the mating process becomes physically possible as a result of the generation of such mating organs. It is important to emphasize the two reasons why just-in-time differentiation cannot merely be a reflection of a delay in maturation of the fluorophores with which we measure differentiation programs. First, fluorescent signals are visible in sex-shared neurons immediately after their generation at early larval stages, whereas male-specific neurons that are born concurrently show a delay of up to several larval stages (>24 h later); in contrast, fluorophore maturation times are known to operate on a much faster scale, with most maturing within <1 h (Balleza et al., 2018; Cranfill et al., 2016). Second, the birth dates of different male-specific neurons are distinct, yet the onset of fluorophore expression is coordinated to occur at the same time.

The ‘just-in-time’ terminology is adapted from ‘just-in-time’ specific transcriptional programs in metabolic pathways (Zaslaver et al., 2004). Genes that code for specific proteins in the metabolic production machinery display temporal dynamics which ensure that, when a metabolic production pipeline is being ramped up under specific conditions, proteins are generated only when needed in the production pipeline. This allows the machinery to reach a production goal with minimal total enzyme production (Zaslaver et al., 2004).

Coordinated, just-in-time differentiation programs are apparent throughout all male-specific neurons. For one of them, the CEM neurons, the delayed onset of differentiation had been previously noted before. The CEM neurons are born in the embryo (and die in hermaphrodites) but were reported to initiate expression of several molecular features, including the putative sensory receptor pkd-2 and its cholinergic-neurotransmitter phenotype, only by the L4 stage (Lawson et al., 2019; Pereira et al., 2015; Wang et al., 2010). CEM neurons only synapse onto sex-shared neurons that were generated and already differentiated in the embryo (Cook et al., 2019), thus the just-in-time differentiation of CEMs at the L4 stage cannot simply relate to the appearance of sex-specific effector cells at the L4 stage. The reason that CEM differentiation is delayed until the L4 stage likely lies in their function: CEM neurons sample mating cues (Narayan et al., 2016; Srinivasan et al., 2008) and hence are not required to operate until the male is sexually mature.

Just-in-time differentiation is not unique to male-specific neurons. Hermaphrodite-specific neurons, of which there are only two classes, the HSN and VC neuron classes, had already been reported to acquire their fully differentiated state only at the L4 stage. In the HSN neurons, which are embryonically born, this is best evidenced by the acquisition of its serotonergic neurotransmitter identity, which they acquire at the late L4 stage (Desai et al., 1988). In the VC neurons, which are generated by the late first-larval stage, cholinergic marker gene expression only becomes induced at the L4 stage as well (Pereira et al., 2015). The logic of the just-in-time differentiation of the HSN and VC neurons is evident: they innervate vulval musculature that only becomes generated and properly placed during late-larval development (Sulston and Horvitz, 1977).

How is this coordinated, just-in-time differentiation wave genetically specified? For the proper timing of differentiation of the male-specific CEM and hermaphrodite-specific HSN neurons, the heterochronic pathway has been implicated (Lawson et al., 2019; Olsson-Carter and Slack, 2010). This pathway is composed of a series of sequentially activated gene-regulatory factors, including transcription factors, regulatory RNAs and translational regulators (Rougvie and Moss, 2013). However, the effects of this pathway on CEM and HSN timing was shown to be only partial (Lawson et al., 2019; Olsson-Carter and Slack, 2010), suggesting the involvement of other regulatory factors. For example, it can be envisioned that target-derived signals from synaptic partners help to coordinate the timing of just-in-time differentiation. Perhaps sex-specific neurons are under a ‘differentiation arrest’ that actively inhibits the execution of terminal differentiation. This notion is again inferred from the CEM and HSN neuron cases, the key identity specifier of which, the terminal selector UNC-86 (Lloret-Fernandez et al., 2018; Shaham and Bargmann, 2002), is already present in the CEM and HSN neurons since their birth (Finney and Ruvkun, 1990). Although UNC-86 is required for the expression of HSN and CEM differentiation markers that become induced at the L4 stage (Lloret-Fernandez et al., 2018; Shaham and Bargmann, 2002), it is apparently unable to induce these features until the time is right. Correctly timed induction might thus be achieved by an inhibitory mechanism that prevents UNC-86 function or, alternatively, by the absence of an essential UNC-86 co-factor, expression of which is temporally controlled.

An intrinsic feature of terminal differentiation programs of many, and perhaps all, C. elegans neurons may explain why terminal differentiation of sex-specific neurons appears to be an all-or-nothing event. In many, and perhaps all, C. elegans neurons, gene expression programs within a neuron are highly coordinated by the activity of terminal-selector transcription factors that become active right after the birth of a neuron in order to initiate terminal differentiation (Hobert, 2016). Triggering the entire differentiation program of a neuron prematurely, i.e. before needed, and thus not coordinated with the differentiation of other neurons, may send uninformative or even conflicting signals to the sex-shared nervous system. Just-in-time differentiation, coordinated over multiple cell and tissue types, ensures that individual components of a nervous system only go online once every individual component is set in place.

The following mutant alleles were used in this study: lin-4(e912), egl-5(u202), ced-3(n717) and lin-32(tm1446). The following reporter strains were used: NeuroPAL strains otIs669; him-5 and otIs670; him-8 (Yemini et al., 2021). The following fosmid-based reporters were used: otIs450[oig-1(fosmid)::SL2::GFP], otIs564[unc-47(fosmid)::SL2::mCherry::H2B] and otIs498[rab-3(fosmid)::SL2::NLS::YFP::H2B]. The following CRISPR/Cas9-generated reporter strains were generated by SunyBiotech through insertion of a reporter cassette at the 3′ end of the respective gene: flp-3(syb2634[T2A::3xNLS::GFP]), flp-27(syb3213[T2A::3xNLS::GFP]), nlp-51(syb3936[SL2::GFP::H2B]) and eat-4(syb4257[SL2::GFP::H2B]). The following promoter-based reporter was used: myIs4[Ppkd-2::GFP+Punc-122::GFP] him-5(e1490)V (Bae et al., 2006).

Analysis of NeuroPAL color codes of egl-5, lin-32 and lin-4 mutants was performed using a him-5 mutation in the background. To rule out the contribution of the him-5 mutation to color code changes in mutants, we compared the NeuroPAL color codes in naturally induced males versus him-5 males. We examined all neurons, both sex shared and male specific, and found no differences in color code.

Worms were anaesthetized using 20 mM sodium azide and were mounted on 5% agarose pads. Images were acquired on a Zeiss LSM880 confocal microscope, equipped with seven laser lines (405, 458, 488, 514, 561, 594 and 633 nm) and processed using ImageJ software. Gamma was adjusted for maximal color distinction for all images. All reporter and mutant strains were imaged at 40× magnification.

To obtain worms staged throughout larval development, adult hermaphrodites were allowed to lay eggs for 1 h. After 1 h, the hermaphrodites were removed from the plates. The plates were then stored at 20°C until the time points corresponding to each larval stage when they were imaged. To capture the early, mid (when the tail hypodermal cells retract), and late L4 stages, worms were imaged at ∼44 h, ∼52 h and ∼55 h, respectively, after egg laying and storage at 20°C. As there is some variability even within carefully staged worms, each L4 stage was confirmed by the male tail morphology.

To generate the statistical atlases of male neuron positions, their variability, and colors, we used our previously published strategy (Yemini et al., 2021). Briefly, we initialized the atlas to be the point cloud of one of the worms, then iteratively aligned each animal's neuron position point cloud to the current atlas, until all animals were aligned. These point clouds, which represent neuron positions and colors, were extracted from images of 12 male heads and 13 male tails that were manually annotated. Males were age-matched to those used in our previously published hermaphrodite atlas (Yemini et al., 2021). For each neuron, the atlas includes a mean position, a mean color and a covariance matrix representing the variability for the neuron's position and color across the population of worms. We computed the positions of the male and hermaphrodite neurons using the determinant of the neuron's positional covariance matrix as an estimate of their aligned spatial occupancy volume (Table S2). We used these volumes to compute differences in neuron position variability between hermaphrodites and males. To account for color variability resulting from any changes in the imaging hardware (e.g. aging equipment, such as excitation lasers), we affine-transformed the male atlas colors to those of the hermaphrodite using the sex-shared neurons. Color alignment and histogram matching have been previously shown to be a crucial step for downstream analysis in such computations as atlas creation and neural segmentation (Nejatbakhsh et al., 2020; Varol et al., 2020).

We thank Rene Garcia for help in distinguishing DVE and DVF, Maureen Barr for pkd-2::gfp, Laura Pereira for help in generating the male NeuroPAL map, and members of the Hobert lab for comments on the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199687