An autism-causing calcium channel variant functions with selective autophagy to alter axon targeting and behavior

Common and rare variants of the CACNA1C voltage-gated calcium channel gene have been associated with autism and other neurodevelopmental disorders including schizophrenia, bipolar disorder and ADHD. However, little is known about how CACNA1C variants affect cellular processes to alter neurodevelopment. The Timothy syndrome mutation is a rare de novo gain-of-function variant in CACNA1C that causes autism with high penetrance, providing a powerful avenue into investigating the role of CACNA1C variants in neurodevelopmental disorders. Here, we use egl-19, the C. elegans homolog of CACNA1C, to investigate the role of voltage-gated calcium channels in autism. We show that an egl-19(gof) mutation that is equivalent to the Timothy syndrome mutation can alter axon targeting and affect behavior in C. elegans. We find that wildtype egl-19 negatively regulates axon termination. The egl-19(gof) mutation represses axon termination to cause axon targeting defects that lead to the misplacement of electrical synapses and alterations in habituation to light touch. Moreover, genetic interactions indicate that the egl-19(gof) mutation functions with genes that promote selective autophagy to cause defects in axon termination and behavior. These results reveal a novel genetic mechanism whereby a de novo mutation in CACNA1C can drive alterations in circuit formation and behavior.


Introduction
Variants in the CACNA1C voltage-gated calcium channel (VGCC) gene are common risk factors for autism and other neurodevelopmental disorders including schizophrenia, bipolar disorder and attention deficit hyperactivity disorder (ADHD). For example, genome wide association studies (GWAS) have associated common variants in CACNA1C to autism [1,2]. Moreover, statistical analysis of whole genome sequencing data indicates that rare variants in CACNA1C are also associated with autism [3][4][5][6][7][8][9][10]. Whereas the evidence is strongest for CAC-NA1C, variants in other VGCC subunit genes are also associated with autism [2,8,11,12]. Despite these insights from statistical analysis, little is currently known about how variants in VGCC genes affect cellular processes to disrupt neurodevelopment.
A major impediment to understanding how autism-associated variants affect cellular processes is that most variants have a small effect size. Because each variant only has a small effect, it is thought that multiple variants engage in genetic interactions that give rise to the neurodevelopmental defects underlying autism [13,14]. Therefore, a key goal in understanding the biological basis for autism is to understand genetic interactions between autism-associated variants. However, currently little is known about how autism-associated variants interact with each other. Moreover, in most cases the cellular mechanisms perturbed by each variant are also unknown.
Morphological abnormalities in axon development are associated with autism and other neurodevelopmental disorders [15][16][17][18][19]. Most mechanistic studies of autism-associated genes have focused on dendrite and synapse structure. However, imaging studies have suggested that alterations in axon targeting are a key feature of autism. For example, diffusion tensor imaging has revealed alterations in the Inferior Longitudinal Fasciculus in autistic individuals relative to healthy controls [15][16][17][18][19]. Moreover, functional MRI has revealed alterations in long range connectivity that can predict autism in individuals before the onset of symptoms [20][21][22]. These observations suggest that alterations in axon targeting are likely to underlie autism. However, little is currently known about how autism-associated variants can alter axon targeting.
In this work, we use the Timothy syndrome mutation as a platform to discover how autism-associated variants interact with each other to alter cellular processes and disrupt axon development. The Timothy syndrome mutation is actually a class of three rare de novo mutations in CACNA1C that encode either a G402R, G402S or G406R mutation in the CACNA1C protein. [10,23]. Although the Timothy syndrome mutation is rare with large effect, common variants with small effect in CACNA1C are also associated with autism [1,2]. As the mechanisms by which CACNA1C affects axon development are unknown, studies of the Timothy Syndrome variant may uncover genetic mechanisms that also apply to the more common CACNA1C risk variants.
Here, we investigate the role of egl-19, the C. elegans homolog of CACNA1C, in axon targeting. Like CACNA1C, EGL-19 is the pore-forming subunit for L-type voltage-gated calcium channels. We identify two egl-19 gain-of-function mutations that are equivalent to the Timothy syndrome G402R and G406R mutations. We find that each of these egl-19 gain-of-function mutations cause overextension of the PLM axon, leading to the misplacement of electrical synapses. Moreover, we find that egl-19 gain-of-function mutations interact genetically with a homolog of the autism-associated WDFY3 selective autophagy gene to disrupt axon development and alter behavior.

A VGCC mutation that causes autism in humans and disrupts axon termination in C. elegans
Both common and rare variants in the human CACNA1C gene have been associated with autism and other neurodevelopmental disorders [1,2,8,9,12,[24][25][26]. The Timothy syndrome mutations, G402R, G402S and G406R in human CACNA1C, are of particular interest because each causes autism with high penetrance [10,27,28]. To establish a model for investigating how the Timothy syndrome mutations affect neurodevelopment, we searched for an equivalent mutation in egl-19, the C. elegans orthologue of CACNA1C. We found that the egl-19 (n2368) mutation (hereafter referred to as egl-19(gof)) encodes a G365R variant of EGL-19 that is equivalent to the G402R Timothy syndrome variant in human CACNA1C (Fig 1A). In both human CACNA1C and C. elegans EGL-19, this Timothy syndrome mutation is a gain-of-function variant: it disrupts slow inactivation of the voltage-gated channel, thereby increasing calcium permeation [10].
To determine how the Timothy syndrome mutation affects neurodevelopment, we observed the PLM touch receptor neuron in egl-19(gof) mutants. The PLM cell body is located in the tail, with an axon extending along the lateral body wall. For this experiment we used two independently isolated egl-19(gof) alleles, egl-19(n2368) and egl- 19(tr134). Both of these alleles produce a G365R mutation in EGL-19, which is equivalent to the G402R Timothy syndrome mutation in humans ( Fig 1A) [29,30]. In wild-type animals, nearly all of the PLM axons terminate posterior to the ALM cell body (Fig 1B and 1D). However, in the egl-19(n2368) and egl-19 (tr134) G365R gain-of-function mutants, around 52% of the PLM axons terminate anterior to the ALM cell body (Fig 1C and 1D). By contrast, egl-19(lof) mutants exhibit normal PLM axon termination ( Fig 1D), with nearly all PLM axons terminating posterior to the ALM cell body. We also tested the egl-19(ad695) gain-of function mutation, which has previously been characterized as a weaker gain-of-function relative to egl-19(n2368) [30]. Consistent with these prior observations, we found that the egl-19(ad695) gain-of-function mutation causes axon termination defects with a lower penetrance relative to egl-19(n2368) and egl-19(tr134) (Fig 1D).
In humans, Timothy syndrome is also caused by a G406R mutation in CACNA1C [10]. Therefore, we tested egl-19(syb1243), a mutation that produces a G369R mutation in EGL-19, which is equivalent to G406R in human CACNA1C (Fig 1A). We found that the egl-19 (syb1243) mutation is homozygous lethal, with no maternal rescue. However, we observed the PLM axon in egl-19(syb1243) heterozygotes and found that around 18% of the PLM axons had axon termination defects ( Fig 1D). Together, these observations indicate that mutations equivalent to the Timothy syndrome mutations can cause defects in axon targeting.
To determine if the Timothy syndrome mutation functions cell autonomously to disrupt axon termination, we used Pmec-7::egl-19(gof) transgenes to express the EGL-19 G365R mutant protein specifically within touch receptor neurons (PLM, ALM, AVM and PVM) ( Fig  1E). We tested one transgene that was created by injecting Pmec-7::egl-19(gof) at a concentration of 5 ng/ul and another at 25 ng/ul. We found that both of these transgenes caused axon termination defects. These results suggest that EGL-19(GOF) functions cell autonomously in the PLM neuron to disrupt axon termination.
Since the egl-19(gof) mutation disrupts axon termination, it is possible that wild-type EGL-19 also negatively regulates axon termination. Alternatively, it is possible that egl-19(gof) acts in a neomorphic role not normally controlled by wild-type EGL-19. To determine the function of wild-type EGL-19, we conducted genetic analysis with a loss-of-function allele of egl-19. For this experiment, we used a null allele of the fsn-1 that causes axon termination defects [34,35]. In fsn-1(null) mutants, 53% of the PLM axons are overextended (Fig 2A). However, in fsn-1(null); egl-19(lof) double mutants, the phenotype is suppressed to only 33% of PLM axons overextended. These observations suggest that wild-type EGL-19 acts to negatively regulate axon termination.
EGL-19 is the pore forming subunit of the L-type VGCC. To further explore the role of voltage-gated calcium channels, we tested loss-of-function mutations in other genes that encode VGCC components. The unc-2 gene encodes the pore-forming subunit of the P/Qtype VGCC [36]. The human homolog of unc-2, CACNA1A, has also been associated with autism [1,37,38]. We found that a null allele of unc-2 could also suppress axon termination defects caused by a fsn-1(null) mutation (Fig 2A). We also tested a null mutation in unc-36, which encodes the alpha2-delta3 subunit that works with both the EGL-19 and UNC-2 poreforming subunits to modulate voltage dependence, activation kinetics, and calcium conductance [39][40][41]. The human homolog of UNC-36, CACNA2D3 has also been associated with autism [6,8,42]. We found that a null allele of unc-36 could also suppress axon termination defects caused by a fsn-1(null) mutation (Fig 2A). Together, these observations suggest that both L-type and P/Q-type voltage-gated calcium channels can negatively regulate axon termination. We note that the different levels of suppression observed between unc-36(null), unc-2 (null) and egl-19(lof) mutations are likely the result of differing roles played by each VGCC subunit or could reflect different strengths of the alleles that we used.

VGCC regulation of axon termination is specific to the FSN-1 pathway
FSN-1 functions in parallel to GLO-4 to promote PLM axon termination [34,43]. Although both pathways promote axon termination, they do so through distinct molecular mechanisms. Whereas FSN-1 is an F-box protein that regulates a MAP Kinase cascade [35], GLO-4 is a guanine nucleotide exchange factor for the GLO-1 Rab GTPase [43]. To determine if VGCCs regulate the GLO-4 pathway, we constructed glo-4(null);unc-2(null) and glo-4(null);unc-36(null) double mutants. We found that neither loss of unc-2 nor loss of unc-36 function suppresses the axon termination defects caused by loss of glo-4 function, suggesting that VGCCs do not regulate the GLO-4 pathway ( Fig 2B). To further explore the role of VGCCs within the context of the parallel FSN-1 and GLO-4 pathways, we constructed an fsn-1(null);glo-4(null);unc-36(null) triple mutant and a fsn-1(null);glo-4(null) double mutant ( Fig 2C). Consistent with prior studies, we found that fsn-1(null);glo-4(null) double mutants had termination defects in 86% of PLM axons [34]. In the triple mutant, loss of unc-36 function reduced this penetrance to 46%, which is similar to the glo-4 single mutants. Together, these observations suggest that VGCCs can negatively regulate axon termination in response to the FSN-1 pathway, but not GLO-4 pathway.
The RPM-1 ubiquitin ligase functions with the FSN-1 F-box protein to negatively regulate downstream proteins [34,35]. Loss-of-function mutations in these downstream proteins https://doi.org/10.1371/journal.pgen.1008488.g002 An autism-causing calcium channel variant functions with selective autophagy to alter axon targeting suppress the phenotype of loss-of-function mutations in FSN-1 and RPM-1. For example, FSN-1 and RPM-1 function together to negatively regulate the DLK-1 MAP Kinase [31]. Loss of DLK-1 function suppresses the phenotype that is caused by either loss of RPM-1 function or loss of FSN-1 function. Since loss of VGCC function can suppress the axon termination phenotype caused by loss of FSN-1 function, we considered the possibility that FSN-1 might function with RPM-1 to negatively regulate VGCCs. If this is true, loss of VGCC function should suppress the axon termination defect caused by loss of RPM-1 function. However, we found that axon termination defects caused by loss of RPM-1 function could not be suppressed by loss of function mutations in unc-36, unc-2 or egl-19 ( Fig 2B). These observations suggest that VGCCs are not downstream targets of RPM-1 and FSN-1.

The egl-19(gof) mutation alters PLM axon connectivity
To determine if the egl-19(gof) mutation affects connectivity of the PLM axon, we examined its chemical and electrical synapses. In wild-type animals, the PLM axon extends a synaptic branch that forms a cluster of chemical synapses onto axons in the ventral nerve cord [32]. We used a mec-7::rfp transgene [44] to visualize the PLM synaptic branch and found that it appears normal in egl-19(gof) mutants (Fig 3A and 3B). We also used a mec-7::gfp::rab-3 transgene [45] to visualize synaptic vesicles and found that these also appear normal in egl-19(gof) mutants (Fig 3A and 3B). We measured the length of the synaptic vesicle clusters in wildtype and egl-19 (gof) mutants and found no significant difference (t-test, p>0.05): egl-19(gof) = 4.7±0.11μm (n = 50); wildtype = 5.09±0.18μm (n = 50). Moreover, consistent with prior findings [34], we found that about 15% of fsn-1 null mutants were missing the PLM branch ( Fig 3C). However, this missing branch phenotype was not suppressed in fsn-1(null); egl-19(lof) double mutants ( Fig 3C). These observations suggest that the EGL-19(GOF) mutation does not affect the PLM synaptic branch or its chemical synapses. Moreover, wildtype EGL-19 does not affect the PLM branch. However, we cannot rule out the possibility that the synaptic branch and chemical synapses are affected in more subtle ways.
We next asked if the egl-19(gof) mutation affects electrical synapses. In wild-type PLM axons, electrical synapses are clustered in two distinct zones [46]. Zone 1 electrical synapses are located close to the PLM cell body, whereas zone 2 electrical synapses are located at the PLM axon tip, which is posterior to the ALM cell body (Fig 3D). Since the egl-19(gof) mutation causes overextension of the PLM axon, it is possible that it could also cause misplacement of zone 2 electrical synapses to a location anterior to the ALM cell body. Alternatively, it is possible that egl-19(gof) causes axon overextension, but leaves the zone 2 synapses in their normal location posterior to the ALM cell body. To differentiate between these two possibilities, we used a mec-7::unc-9::gfp transgene [46] to visualize the UNC-9 Innexin, a marker for electrical synapses [46]. We found that the egl-19(gof) mutation caused misplacement of the zone 2 electrical synapses to a point anterior to the ALM cell body (Fig 3E). Misplacement of the electrical synapse occurred in 52.9±7.0% of egl-19(gof) PLM axons, but only in 7.1±3.4% of wildtype PLM axons (n = 50 for both genotypes, significantly different by z-test for proportions, p<0.0001). Therefore, the PLM axon overextension caused by egl-19(gof) is also associated with misplacement of zone 2 electrical synapses.

The egl-19(gof) mutation interacts with selective autophagy genes to disrupt axon termination
As part of an ongoing effort to test autism-associated genes for roles in axon development, we identified a genetic interaction between egl-19(gof) and wdfy-3, a homolog of the autism-associated WDFY3 selective autophagy gene. For this experiment, we used the wdfy-3(ok912) deletion allele, hereafter called wdfy-3(lof). This allele is likely to be a null or strong loss of function because it creates a frameshift that disrupts the FYVE domain, WD repeat domain and nearly all of the beach domain. We found that the PLM axon was normal in wdfy-3(lof) single mutants. In egl-19(gof); wdfy-3(lof) double mutants, the axon termination defects observed in egl-19(gof) single mutants were almost completely suppressed ( Fig 4A). Moreover, PLM axon termination defects caused by transgenic expression of EGL-19(GOF) in touch receptor neurons were also suppressed by wdfy-3(lof). The PLM neurons are likely to express WDFY-3 because RNAseq on purified touch receptor neurons identified wdfy-3 mRNA transcripts [47]. The wdfy-3 gene is an orthologue of the human autism-associated WDFY3 gene that encodes a protein required for cargo selection during selective autophagy [48,49]. Therefore, these observations establish a genetic pathway between two autism-associated genes that regulates axon termination.
The epg-7 gene encodes an additional component required for selection of cargo for autophagy [50], and the cup-5 gene encodes a scaffold protein that promotes lysosome biogenesis [51,52]. We found that axon termination defects caused by the egl-19(gof) mutation could be suppressed by either a likely null mutation in epg-7 or a hypomorphic mutation in cup-5 (Fig 4A). These observations suggest that the egl-19(gof) mutation causes axon termination through a mechanism that requires selective autophagy. An autism-causing calcium channel variant functions with selective autophagy to alter axon targeting

WDFY-3 negatively regulates PLM axon termination
Having found that wdfy-3 can negatively regulate axon termination in the egl-19(gof) mutant, we wanted to ask if wdfy-3 could also regulate axon termination independently of the gain-offunction egl-19 mutation. For this experiment, we used a fsn-1(null) mutation to induce axon termination defects. We found that loss of wdfy-3 function completely suppresses the axon termination defects caused by the fsn-1(null) mutation (Fig 4B). Like the VGCC loss-of-function mutants, we also found that the wdfy-3(lof) mutant does not suppress axon termination defects caused by rpm-1(lof). Together, these observations suggest that WDFY-3, like VGCCs, can regulate axon termination signaling downstream of FSN-1, but is not a downstream target of FSN-1 and RPM-1.

The egl-19(gof) mutation interacts with the wdfy-3 selective autophagy gene to regulate habituation to light touch
The PLM neuron is a mechanosensory neuron that is responsible for sensing light touch in the posterior of C. elegans [53]. When light touch is applied to the tail, the animal responds by moving forward [54][55][56]. However, after repeated touches, the animal habituates and becomes
We conducted a touch assay to determine if the egl-19(gof) mutation affects the response to light touch. Each animal was subjected to ten eyelash touches alternating between the head and tail (Fig 5). For the initial touch, there was no significant difference in the response rate between egl-19(gof) mutants and wild-type animals. However, for each subsequent touch, the egl-19(gof) mutants had a significantly lower response rate relative to wild-type animals. These observations suggest that the egl-19(gof) mutation enhances habituation to light touch.
Because we found that the egl-19(gof) mutation interacts with wdfy-3 to disrupt axon termination, we next asked if this genetic interaction can also affect habituation to light touch. If the egl-19(gof) mutation functions with wdfy-3 to alter habituation, we expect that loss of wdfy-3 function will reduce the effect of the egl-19(gof) mutation on habituation. Indeed, we found that in egl-19(gof);wdfy-3(lof) mutants, the habituation to light touch was not significantly different relative to wild-type animals. Together, these observations suggest that the egl-19(gof) mutation acts through wdfy-3 to affect both axon termination and habituation to light touch.

Discussion
The Timothy syndrome mutation in CACNA1C has the unusual property of being causative for autism with high penetrance, providing an opportunity to discover the downstream cellular processes that are perturbed to cause autism. However, the cellular processes that interact with this mutation to give rise to autism have remained unknown. To address this question, we created a disease model in C. elegans that utilizes a mutation equivalent to the Timothy  -3(lof). Asterisks indicate statistically significant difference compared to wild type worms, Z-test for proportions ( � p<0.0001). Error bars represent the standard error of the proportion. For each genotype, the assay was repeated 3 times by 3 different observers who were blind to the genotype. Each experiment included 20 worms for a total n of 60 for each genotype. Alleles: egl- 19

(gof) is egl-19(n2368), wdfy-3(lof) is wdfy-3(ok912).
https://doi.org/10.1371/journal.pgen.1008488.g005 An autism-causing calcium channel variant functions with selective autophagy to alter axon targeting syndrome mutation in humans. Our results reveal that selective autophagy genes interact with the Timothy syndrome mutation to disrupt axon termination and alter behavior. Because common variants of CACNA1C are associated with autism, it is likely that this mechanism will be broadly applicable to autism in humans.

An understanding of genetic interactions between variants is key to understanding autism
Only a small fraction of autism cases are thought to be caused by a single variant. Rather, most cases of autism are thought to be caused by genetic interactions between variants [9,14,57]. For example, likely gene-disrupting (LGD) mutations have been associated with 15-20% of autism cases. In addition, autistic individuals carry more missense mutations in autism-associated genes relative to healthy controls [9,58].
However, both LGD and missense mutations are rare and therefore almost always heterozygous. Therefore, in most cases, it is thought that each of these mutations have little or no effect on their own. Thus, it is likely that the disorder arises from genetic interactions between autism-associated variants. Indeed, statistical analysis of sequencing data suggests that autism arises from the combined action of multiple variants [9,14,59].
Although the heritability of autism has been estimated at 83% [60], the complexity of the genetic interactions that give rise to autism make it difficult to predict and diagnose autism from whole genome sequencing data. In fact, with current knowledge, no genetic cause can be found from whole genome sequencing data for most cases of autism. The solution to this challenge could come from genetic analysis. For example, in most cases, single heterozygous null mutations have no phenotype. However, many cases exist to show that an animal carrying two heterozygous null mutations can exhibit a phenotype when each of the mutated genes function in the same genetic pathway [61,62]. Thus, if an individual is heterozygous for two LGD variants in each of two autism-associated genes that function in a pathway, this individual would carry a higher risk for autism. Therefore, knowledge of the pathways that link autism-associated genes will help promote our understanding of the genetic basis of autism.

Selective autophagy functions with EGL-19(GOF) to alter axon development and behavior
A key finding of our study is the identification of a genetic interaction between the homologs of two autism-associated genes, egl-19 and wdfy-3. Our genetic analysis indicates that wdfy-3 and other selective autophagy genes are required for the egl-19(gof) mutation to disrupt axon termination. Moreover, we also find that wdfy-3 can negatively regulate axon termination independently of the egl-19(gof) mutation. These genetic interactions could be explained by a few different molecular models. First, it is possible that EGL-19 functions in a pathway with WDFY-3 to negatively regulate axon termination. Alternatively, it is possible that EGL-19 and WDFY-3 function in two separate parallel pathways that can both negatively regulate axon termination. We favor the former model, because both egl-19(lof) and wdfy-3(lof) are able to suppress axon termination defects caused by loss of fsn-1 function, but not by loss of rpm-1 function. Moreover, the strength of the suppression of the egl-19(gof) phenotype by the wdfy-3 (null) mutation is most consistent with the idea that EGL-19 and WDFY-3 function together in a pathway. If EGL-19 and WDFY-3 do function together in a pathway, it is possible that the Timothy syndrome mutation induces excessive selective autophagy that causes a disruption of axon termination. As an alternative, it is also possible that selective autophagy could function upstream of the EGL-19(GOF) mutant protein. In this scenario, selective autophagy could promote the function of the EGL-19(GOF) protein by affecting its turnover, stability or localization.
The genetic interaction between egl-19 and wdfy-3 also regulates habituation to light touch. It is possible that this genetic interaction affects behavior by functioning in the developing nervous system to regulate connectivity. Alternatively, it is possible that the genetic interaction between egl-19 and wdfy-3 functions in the mature nervous system to regulate neural function. Although our data cannot distinguish between these two possibilities, recent work on mice favor the former possibility [63]. Conditional knockout of CACNA1C in forebrain neurons during development results in anxiety in adult mice, whereas knockout of CACNA1C during adulthood does not. These observations lend support to the possibility that CACNA1C acts during development to alter circuit formation, which in turn affects behavior in the adult.
The interaction between the Timothy syndrome mutation and wdfy-3 provides biological evidence for a role of selective autophagy in autism. A major challenge in autism genetics is to confirm and characterize the roles of autism candidate genes. For example, WDFY3 is a candidate gene for autism because whole genome sequencing has found that 3 out of 6707 sequenced autism genomes contain a heterozygous de novo likely-gene-disrupting mutation in WDFY3 [6,8,42,64]. However, despite this association, it is not possible to determine if WDFY3 variants contribute to the cause of autism. Our results place genes that promote selective autophagy in a pathway with a mutation that is causative for autism in humans, thereby providing the first biological evidence for a role of selective autophagy in autism.
Selective autophagy is also required for normal axon development. Aside from its function with EGL-19(GOF) in inducing axon defects, our results suggest that WDFY-3 also functions independently of EGL-19(GOF) to regulate axon development. Consistent with this idea, loss of WDFY3 causes the disorganization and loss of many commissural axon tracts in mice [65]. Loss of WDFY3 in mice also attenuates the response to guidance cues in vitro, suggesting that selective autophagy could regulate the response to guidance cues. Despite these insights, the mechanism through which selective autophagy regulates axon targeting is currently unknown. Based on our genetic analysis, we propose a mechanism whereby selective autophagy functions with voltage-gated calcium channels to regulate the response to axon targeting cues.
Selective autophagy and bulk autophagy may have distinct functions in neurodevelopment. Whereas our results suggest a mechanism for selective autophagy in the negative regulation of axon termination, prior studies have reported a role for bulk autophagy in promoting synapse development and inhibiting axon growth. In cultured mouse neurons, knockdown of an autophagy gene promotes axon growth, whereas induction of autophagy inhibits axon growth [66]. In Drosophila, autophagy promotes development of the neuromuscular junction [67]. In C. elegans, autophagosomes form at synaptic sites and are required for presynaptic assembly [68,69]. This role for autophagy in synaptogenesis is specific to bulk autophagy, since mutations in selective autophagy genes do not affect synaptogenesis [69]. Interestingly, our results suggest that selective autophagy may regulate axon termination without affecting synaptogenesis.
Selective autophagy could function with VGCCs to regulate other aspects of autism-related pathology. Whereas our study focuses on the role of the Timothy syndrome mutation in misregulating axon termination and behavior, previous work has found that the Timothy syndrome mutation can promote activity-dependent dendrite retraction in cultured mouse neurons and can inhibit the elaboration of mouse dendrites in vivo [70]. The downstream cellular mechanisms for this effect on dendrites are not yet known, but it is possible that selective autophagy could also be involved in this process. Alternatively, it is possible that the Timothy syndrome mutation functions through distinct mechanisms to affect dendrite development and axon development.

Potential role for VGCCs in regulating the RPM-1 pathway
Our results identify specific genetic interactions between VGCC genes and RPM-1 pathway genes. These genetic interactions indicate that loss of VGCC function suppresses axon termination events that are caused by loss of fsn-1 function, but not loss of glo-4 function. Moreover, we find that loss of VGCC function does not suppress defects caused by loss of rpm-1 function, but can partially suppress defects caused by the double loss of fsn-1 and glo-4. Taken together, these observations suggest that VGCCs specifically regulate axon termination signaling downstream of FSN-1, but are not themselves downstream targets of RPM-1 and FSN-1.
The genetic interactions between the VGCC genes and RPM-1 pathway genes could be explained by a model where VGCCs negatively regulate an unknown protein that functions with RPM-1 to enhance signaling events that promote axon termination downstream of FSN-1, but not GLO-4 (S1 Fig). In fact, prior work has found that FSN-1 binds to RPM-1 and promotes axon termination by negatively regulating the DLK-1 MAP kinase signaling pathway [31,35]. Moreover, the PPM-1 phosphatase also binds to RPM-1 and promotes axon termination by negatively regulating DLK-1 MAPK signaling [71]. Thus, it is possible that EGL-19 might repress axon termination by negatively regulating PPM-1, or another protein that plays a similar role (S2 Fig).

Role for VGCC-mediated calcium transients in axon growth
Our study focuses on the genetic mechanisms that mediate the role of VGCC genes in axon termination, but does not address how alterations in calcium permeation might be involved in this process. Interestingly, a recent study of cultured prenatal mouse neurons has revealed that VGCCs function during axon outgrowth to produce calcium transients that have very different properties compared to those produced during synaptic transmission [72]. These transients have been named Spontaneous Regenerative Calcium Transients (SRCaTs) and are mediated by Ca V 1.2, which includes the homolog of EGl-19, CACNA1C. Unlike its function in adult neurons, Ca V 1.2 appears to open near resting potential, suggesting that the Ca V 1.2 channel may open spontaneously in developing axons. Knockout of CACNA1C in these cultured neurons causes a decrease in axon growth. Thus, Ca V 1.2 functions in axons to regulate axon growth, using a mechanism that is very different than how it functions in synaptic transmission. The role of SRCaTs in regulating axon growth are unknown. However, our results suggest the possibility that these SRCaTs may regulate signaling downstream of FSN-1 (FBXO45), utilizing a mechanism that involves selective autophagy.

Potential role for common CACNA1C variants in affecting selective autophagy and axon development in autism
We propose that the effect of CACNA1C variation in altering axon development is not limited to the Timothy syndrome mutation, but rather extends to the other autism-associated CAC-NA1C variants. This idea is supported by our genetic analysis suggesting that the effect of the egl-19(gof) Timothy syndrome mutation on axon termination is not neomorphic, but rather reflects an increase in the normal function of egl-19. Therefore, other gain-of-function and loss-of-function variants in CACNA1C could contribute to autism by altering axon development. Consistent with this idea, statistical analysis has identified some candidate variants in VGCC genes that are likely to be gain-of-function and others that are likely to be loss-of-function [6,10,11,42,73]. Therefore, we speculate that both under-activation and over-activation of the signaling pathways that promote axon termination could contribute to autism.
The Timothy syndrome mutation is a very rare de novo mutation, and is therefore only responsible for a very tiny fraction of autism cases. However, several common variants in CACNA1C have also been associated with autism [1,2,74]. For example, the A genotype at the rs1006737 locus in CACNA1C confers risk for autism and is present in about 33% of the human population. This A genotype at rs1006737 is located within a large intron and is thought to cause CACNA1C gain-of-function because neurons with the risk genotype have higher levels of CACNA1C mRNA and increased L-type calcium currents relative to neurons with the non-risk genotype [75]. Therefore, this risk variant may be associated with a gain-offunction of CACNA1C that could disrupt axon development in a way analogous to the Timothy syndrome mutation. However, the small effect size of the rs1006737 locus suggests that this is a relatively weak CACNA1C gain-of-function.
Although common alleles have a small effect size relative to the Timothy syndrome mutation, they could interact with risk variants in other genes that function in a genetic pathway with CACNA1C. For example, the rs1006737 risk variant could provide a weak gain-of-function in CACNA1C that does not cause autism on its own. However, the rs1006737 risk variant could synergize with a gain-of-function risk variant in WDFY3 to contribute to autism. Alternatively, a weak loss of function in CACNA1C could synergize with a weak loss-of-function in WDFY3 to give rise to autism.

Analysis of phenotypes
For analysis of axon termination phenotypes, animals were mounted on a 5% agarose pad and observed with a 40x objective. For PLM axon termination, an axon was scored as defective if it grew anterior to the ALM cell body. PLM neurons were visualized with the muIs32 transgene which encodes Pmec-7::gfp and is expressed in all mechanosensory neurons.
For analysis of the PLM chemical synapses, a Pmec-7::gfp::rab-3 transgene that expresses the RAB-3 synaptic vesicle marker in the touch receptor neurons was used to visualize synaptic vesicle clusters [44]. The size of each synaptic cluster was measured as previously described [77]. For analysis of PLM electrical synapses, a Pmec-4::gfp::unc-9 transgene was used to express the UNC-9 innexin fused to GFP in the touch receptor neurons [46].
For analysis of mechanosensation, we adopted an eyelash touch assay [78]. We assayed gentle touch responses by touching the lateral side of animals with an eyebrow hair. Each animal was subjected to five touches alternating between the anterior and posterior ends and scored by the number of responses elicited. Assays were performed blind to genotype. Three independent samples of 20 animals each were collected by three independent observers and reported as mean percentage scores.

Ethics statement
Because C. elegans are invertebrate animals they do not require review by an ethics committee.

S1 Fig. Model for function of VGCCs in relation to the RPM-1 pathway.
This hypothetical model could explain the genetic interactions observed between VGCC genes and genes that encode members of the RPM-1 pathway. In this model, VGCCs function with WDFY-3 to negatively regulate an unknown protein (labeled as X). Protein X functions with RPM-1 to promote signaling events downstream of FSN-1 that promote axon termination. Loss of VGCC function causes an increase in protein X function. The additional protein X function works with RPM-1 to promote axon termination mechanism B, thereby compensating for loss of FSN-1 function. Loss of VGCCs does not suppress loss of RPM-1 function because the function of protein X requires RPM-1. (TIF) S2 Fig. PPM-1 is a protein that functions with RPM-1 to promote signaling events downstream of FSN-1 that promote axon termination. It is possible that PPM-1, or another protein with a similar role, could be protein X (see S1 Fig). Both FSN-1 and PPM-1 promote axon termination by functioning with RPM-1 to negatively regulate the DLK-1 MAP kinase pathway. Thus, it is possible that VGCCs and WDFY-3 could negatively regulate PPM-1, or a protein with a similar role. The extra PPM-1 function could enhance negative regulation of the DLK-1 pathway, thereby compensating for loss of FSN-1 function. (TIF) S1 Data. Excel file containing the numerical data for Figs 1-5. (XLSX)