A humanized Caenorhabditis elegans model of hereditary spastic paraplegia-associated variants in KLC4

Over 10,000 disorders are classified as rare diseases, each affecting fewer than 1/2000 people (Ferreira, 2019). Together, they are not rare; over 4% of the world's population is currently suffering from a rare disease (Nguengang Wakap et al., 2020). Diagnoses, let alone treatments, of rare diseases are difficult because underlying mutations are spread over 8000 genes (Ferreira, 2019). Even whole-genome sequencing leads to a definitive diagnosis only ∼25% of the time (Smedley et al., 2021). More often, a definitive diagnosis is not returned, and the clinician is left with a list of variants of uncertain significance and little idea as to which of these variants are pathogenic. Thus, one of the biggest challenges in genomic medicine is the validation of which identified variant is pathogenic. The bottleneck facing clinical geneticists is a need for functional data that can assess the pathogenicity of a variant of uncertain significance.

Hereditary spastic paraplegia (HSP) is a group of monogenetic diseases that are classified as rare diseases that present at various times throughout life. Individuals characteristically suffer from neurodegeneration in the longest motor neurons, leading to progressive spasticity and lower-limb weakness (Parodi et al., 2017; Gumeni et al., 2021; Shribman et al., 2019). Upwards of 79 genes have been linked to HSP, yet geneticists fail to obtain definitive genetic diagnoses in over half of suspected HSP individuals (Parodi et al., 2017; Gumeni et al., 2021; Shribman et al., 2019). This suggests that mutations in additional unknown genes lead to HSP. Moreover, once new candidate HSP genes are identified, we need an in vivo model to access the physiological significance of newly identified variants for a timely clinical diagnosis (Hopkins et al., 2022).

Functional studies in vivo are important for variant assessment. Caenorhabditis elegans is a model system that can relatively inexpensively test variants of uncertain significance at the speed needed for inclusion in a clinical report (Baldridge et al., 2021). C. elegans also allows examination of function in the context of a developing tissue and the use of a variety of biochemical, developmental and quantitative cellular assays needed to detect subtleties of variant biology. These advantages have led to many reports modeling human diseases in C. elegans (Kropp et al., 2021). Thus, humanized C. elegans models are likely to be useful in testing the in vivo consequences of variants of uncertain significance identified in the clinic. Here, we report the design and use of a humanized C. elegans model to test the clinical significance of variants of uncertain significance in the human kinesin light chain gene identified in individuals with HSP, KLC4.

Molecular motor-based transport along microtubules is essential for the function and survival of eukaryotic cells (Ross et al., 2008). Microtubule motors are especially important in transporting organelles and molecules down the length of long motor neuron axons (Hirokawa et al., 2010; Perlson et al., 2010; Saxton and Hollenbeck, 2012). Disrupting motors leads to a variety of neurodegenerative diseases (Mandelkow and Mandelkow, 2002; Kurd and Saxton, 1996; Giudice et al., 2014). Kinesin-1 is the founding member of the kinesin superfamily of microtubule motors (Vale et al., 1985). Kinesin-1 consists of a tetramer of two kinesin light chains that bind to the tails of two kinesin heavy chains. The heavy chains, called KIF5B in humans, bind microtubules and provide the ATPase motor activity, whereas the light chains serve as cargo adapters (Verhey et al., 1998). In the presence of a cargo bound to the light chains, kinesin-1 is activated to move towards the plus end of microtubules. In humans, there are four different kinesin-1 light chains: KLC1, KLC2, KLC3 and KLC4. Although KLC1, KLC2 and KLC3 are relatively well studied (Rahman et al., 1998; Junco et al., 2001; Zhang et al., 2012), and their functions and how they are involved in certain diseases are known, KLC4 is understudied. Yet, mutations in KLC4 are linked to diseases including lung cancer (Baek et al., 2018, 2020) and HSP (Bayrakli et al., 2015). The goal of this study was to further explore the link between KLC4 and HSP by developing a humanized C. elegans model as a clinical avatar to test the functions of variants of uncertain significance in the gene KLC4.

A male individual with the clinical KLC4 variant presented to the Undiagnosed Diseases Network (UDN) with slowly progressive myelopathy and neuropathy since ∼50 years of age. Initial symptoms included numbness, followed by weakness and lower-extremity hypertonia, hyperreflexia and spasticity. The individual's symptoms worsened over the next 20 years, although he maintained normal cognitive ability. Ophthalmologic evaluation showed thinning of the ganglion cell layer and papillomacular bundle, but no visual changes. Of note, he also had celiac disease, which can cause myelopathy and neuropathy, but he had been compliant with treatment with no obvious symptoms. The individual worked as an agronomist and was frequently exposed to herbicides and pesticides, potentially complicating the diagnosis.

A heterozygous suspected pathogenic variant was identified in the individual with HSP described above, in which a GG pair of nucleotides in the open reading frame of KLC4 was deleted to cause a frame shift (NM_201523.2; c.1160-1161delGG; p.G369Afs*8) (Fig. S1). The predicted mutant KLC4 protein (G369fs) replaces the glycine at position 369 with an alanine, followed by eight additional novel residues and a premature stop codon that truncates more than a third of the protein. For KLC4, only 33% of the expected loss-of-function variants were observed (https://gnomad.broadinstitute.org/; GnomAD v2.1.1, assessed 5 June 2023), suggesting that it is under selection against loss-of-function variants (Karczewski et al., 2020). This suggests that, in a subset of genetic contexts, the gene variants can be associated with an autosomal-dominant disorder, and therefore it is considered a heterozygous variant of uncertain significance. There was no evidence of any duplication or deletion of the KLC4 gene. The individual has an unaffected sibling who did not harbor the KLC4 variant.

We turned to bioinformatic databases in attempts to identify four KLC4 variants of uncertain significance, two predicted pathogenic and two predicted benign variants, as reference alleles in addition to the clinical KLC4 G369fs variant (Fig. S1). Two missense KLC4 mutations, R72H and R358H, were chosen as predicted benign controls. R72H was observed at a very-high frequency (6803×) in unaffected populations using GnomAD (Karczewski et al., 2020). R358H was seen at 9× in GnomAD and was scored as possibly damaging in PolyPhen-2 (Adzhubei et al., 2010), tolerated in Sorting Intolerant From Tolerant (SIFT; Sim et al., 2012), and neutral in Combined Annotation-Dependent Depletion (CADD; Kircher et al., 2014) and Rare Exome Variant Ensemble Learner (REVEL; Ioannidis et al., 2016). Two other missense mutations were chosen because they were predicted to be pathogenic variants. Both T381I and A295P were identified as damaging by PolyPhen-2 (Adzhubei et al., 2010) and possibly damaging by SIFT (Sim et al., 2012). Thus, we obtained a collection of five KLC4 mutations for testing in an in vivo model.

We aimed to make a humanized C. elegans model to test the physiological significance of KLC4 mutations (Fig. 1). Kinesin-1 plays similar important roles in C. elegans as it does in humans, including moving synaptic vesicles in motor neurons and nuclei in hypodermal precursors (Meyerzon et al., 2009; Sakamoto et al., 2005). The C. elegans genome encodes two kinesin light chains, KLC-1 and KLC-2 (Koushika and Nonet, 2000). Although they act redundantly in the C. elegans embryo during the pre-anaphase translocation of the meiotic spindle (Yang et al., 2005), KLC-2 is likely to be the primary light chain for kinesin-1 in most of development. Compared with KLC-1, KLC-2 has a higher level of sequence homology to other metazoan kinesin light chains. KLC-2 also directly interacts with the kinesin heavy chain (UNC-116) to form a kinesin-1 complex, and null alleles of klc-2 cause severe larval lethality (Sakamoto et al., 2005). Human KLC4 and C. elegans KLC-2 proteins both have a predicted coiled-coil domain that binds to the kinesin heavy chain and six tetratricopeptide (TPR) repeats that function together to bind cargo (Fig. 2A-B′). The coiled-coil regions of KLC4 and KLC-2 are 46% identical and the TPR domains are 78% identical (Fig. 2A; Fig. S1). Using AlphaFold (Jumper et al., 2021), we were able to model the predicted structures of KLC-2 and KLC4 as well as their interactions with UNC-116 and the C. elegans protein UNC-83 that acts as a binding adaptor for kinesin-1 (Meyerzon et al., 2009; Taiber et al., 2022). In addition to the evolutionarily conserved sequences the two proteins share, we saw that KLC-2 and KLC4 share a high level of similarity in their structural features and how they interact with the C. elegans proteins UNC-116 and UNC-83. Thus, we chose the klc-2 locus to engineer in human KLC4 to make a humanized model. Human KLC4 has a major isoform encoding a 619 residue protein that is expressed more broadly and at higher levels than the 637 residue isoform, according to the Transcript Support Level (TSL) method (Yates et al., 2016). Therefore, we used the shorter 619 residue protein isoform to make a humanized C. elegans line. A humanized C. elegans model was generated in which the endogenous klc-2 gene was replaced by human KLC4 using CRISPR/Cas9 genome editing (Fig. 1). The coding region for human KLC4 was codon optimized, and placed under control of the endogenous klc-2 promoter and the 5′ and 3′ untranslated regions (UTRs) of the klc-2 gene. In addition, three synthetic C. elegans introns were inserted into KLC4 to maximize its expression in C. elegans (Fig. 1).

After generating the humanized C. elegans line in which human KLC4 replaced klc-2 (hereafter referred to as the hKLC4 line), we compared the fitness of the new model to wild type. We quantified the viability of the hKLC4 line by measuring its brood size and embryonic lethality in comparison to wild type (Fig. 2C,D). The hKLC4 strain was viable as a homozygous strain, while klc-2 null animals are 100% embryonic or L1 larval lethal, suggesting that hKLC4 rescued many of the essential klc-2 functions. However, the hKLC4 line had significant levels of embryonic lethality (11.2±5.4% compared to 4.1±1.8% in wild type; mean±95% confidence intervals) and a slightly lower brood size (142.6±30.2 compared to 191.1±31.9 in wild type) (Fig. 2C,D), suggesting that hKLC4 animals were not as fit as wild type, perhaps because C. elegans were maintained at ∼22°C instead of the 37°C that the human protein has evolved to have optimal activity at. Nonetheless, most animals with only human KLC4 in place of endogenous KLC-2 were quite viable, fertile and appeared relatively normal, suggesting that the hKLC4 model would be of use as a clinical avatar.

The hKLC4 animals had no obvious phenotypes affecting their ability to crawl on the surface of an agar plate. However, swimming and crawling are two different forms of movement in terms of their kinematics and muscle activity (Pierce-Shimomura et al., 2008). Therefore, to conduct a more comprehensive movement analysis of the humanized animals, we used a swimming assay in which we observed head thrashing in liquid. We scored swimming by measuring the number of body bends per second (BBPS) and average speed per animal to assess the motility of hKLC4 animals (Pierce-Shimomura et al., 2008; Mattout et al., 2011). We observed that hKLC4 animals had a significantly lower number of BBPS and average speed than wild-type worms (0.72±0.21 BBPS compared to 1.48±0.09 BBPS in wild type, and 11.16±2.24 pixels/s compared to 29.43±2.64 pixels/s in wild type; mean±95% confidence intervals) (Fig. 3A,B; Movies 1 and 2). The twofold effect in BBPS and almost threefold effect in the average speed suggest that the humanized line has a significant motility defect.

A second klc-2-dependent assay examined nuclear migration (Fridolfsson et al., 2018). There is no known link between nuclear migration and HSP, but because interkinetic nuclear migration is important for neurodevelopment and defects lead to disease (Bone and Starr, 2016), and we had an assay that is dependent on klc-2 (Meyerzon et al., 2009), we reasoned that this assay could be used to study the function of the hKLC4 variants. In mid-embryogenesis, two rows of hyp7 precursors on the dorsal surface of embryos intercalate to form a single row spanning the dorsal midline. Next, nuclei migrate contralaterally toward the plus ends of microtubules across the dorsal midline to the opposite side of the embryo (Sulston et al., 1983) (Fig. 4A). Successful completion of nuclear migration in embryonic hyp7 hypodermal cells requires kinesin-1 heavy chain and KLC-2 (Meyerzon et al., 2009). The linker of nucleoskeleton and cytoskeleton (LINC) complex, consisting of the Klarsicht/ANC-1/SYNE homology (KASH) protein UNC-83 at the outer nuclear membrane and the Sad1/UNC-84 (SUN) protein UNC-84 at the inner nuclear membrane, recruits kinesin-1 to the surface of nuclei and transmits the forces to inside the nucleus (Starr, 2019) (Fig. 4A). Null mutant klc-2(km28) larvae that barely escape embryonic lethality had an average of 10.4±1.4 hyp7 nuclei abnormally located in the dorsal cord, compared to 0.07±0.08 in wild type (Fig. 4B). We have previously shown that this represents a nearly penetrant nuclear migration defect (Fridolfsson and Starr, 2010) and that klc-2(km28) is a likely null allele resulting from a 0.6 kb deletion of parts of exons 2-4 (Sakamoto et al., 2005). To analyze whether KLC4 was able to retain KLC-2 function in the humanized worms, we counted the number of hyp7 nuclei abnormally present in the dorsal cord. The hKLC4 animals had no significant nuclear migration defects compared to wild type (Fig. 4B,C). Together, these data suggest that hKLC4 can substitute for most of the function of klc-2 in C. elegans.

After showing that the hKLC4 animals were healthy, our goal was to introduce missense variants into the hKLC4 avatar to test their possible effects on C. elegans development as an indicator of clinical interest. Four hKLC4 variants were introduced into our hKLC4 worm line using CRISPR/Cas9 gene editing, chosen as discussed above. Variants hKLC4 R72H and hKLC4 R358H were predicted to be benign, and hKLC4 A295P and hKLC4 T381I were predicted to be pathogenic (Fig. S1). We quantified the brood size and embryonic lethality of the missense variants and found that neither the predicted benign nor the predicted pathogenic hKLC4 mutations had a significantly deleterious effect on the percentage embryonic lethality observed in the parental hKLC4 line (Fig. 2D). However, one of the predicted pathogenic mutations, hKLC4 T381I, led to a significant decrease in the brood size (Fig. 2C).

We had similar results when we observed the swimming behavior of the missense mutants. The two predicted benign variants and the A295P variant had no effect on the swimming rate of the hKLC4 parental line. However, the predicted disease allele hKLC4 T381I caused a severe motility defect, with only 0.085±0.038 BBPS and 3.07±0.38 pixels/s average speed (Movie 3), compared to the parental hKLC4 worms that had 0.72±0.21 BBPS and 11.16±2.24 pixels/s average speed (Fig. 3). After seeing such severe motility defects in the hKLC4 T381I homozygous animals, to further analyze the characteristics of this particular point mutation, we generated hKLC4 T381I heterozygous (hKLC4 T381I/+) animals by mating hKLC4 T381I homozygous hermaphrodites with hKLC4 homozygous males. Although there was a statistically significant improvement in terms of motility compared to the hKLC4 T381I homozygous animals, hKLC4 T381I/+ animals still had a severe motility defect (Movie 4), with 0.19±0.038 BBPS and 5.04±0.59 pixels/s average speed (Fig. 3). We also observed no statistically significant change in the total brood size and embryonic lethality of the heterozygous animals compared to those of the homozygous animals or the hKLC4 starting strain (Fig. 2C,D).

In our nuclear migration assay, the benign variants hKLC4 R72H and R358H did not cause any hyp7 nuclear migration defects (Fig. 4). In contrast, both predicted pathogenic variants, hKLC4 A295P and T381I caused mild, but significant, nuclear migration defects, with 0.43±0.34 and 0.62±0.24 hyp7 nuclei observed in the dorsal cord of an average worm (Fig. 4B,C). The hKLC4 T381I/+ animals did not have any hyp7 nuclear migration defects (0.22±0.16 hyp7 nuclei located abnormally in the dorsal cord).

Finally, we introduced the clinical variant of uncertain significance hKLC4 G369fs using CRISPR/Cas9 genome editing. Attempts to generate a homozygote line failed so we suspected this variant to be lethal. We introduced G369fs into a strain humanized for hKLC4 that also contained a translational fusion klc-2::gfp rescue array. The rescuing array is expressed from an extrachromosomal array that in C. elegans is lost in a high percentage of animals during early embryonic cell divisions (Sakamoto et al., 2005). Thus, this strain produces hKLC4 G369fs animals both with and without the rescuing array. All hKLC4 G369fs animals that survived to adulthood maintained the klc-2::gfp rescuing array, suggesting that all the animals that lost the rescuing array died as embryos or early larvae. Therefore, animals homozygous for hKLC4 G369fs have a total brood size of zero and 100% embryonic lethality (Fig. 2C,D). We were therefore unable to measure the swimming ability of the hKLC4 G369fs animals. However, we were able to observe nuclear migration defects in the rare hKLC4 G369fs animals that escaped embryonic lethality and could be scored as young larvae before dying. The hKLC4 G369fs animals had severe nuclear migration defects with a mean of 5.9±1.3 hyp7 nuclei in the dorsal cord of an animal, compared to nearly zero nuclei in the dorsal cord of the hKLC4 animals (Fig. 4B). Taken together, these data show that the G369fs mutation is very severe, making the hKLC4 animals very sick.

To test the extent to which the KLC4 frameshift variant at residue 369 acts in a haplo-insufficient manner or whether it is the sole contributor to the clinical features, we crossed the homozygous truncation mutant to the hKLC4 strain to assay heterozygotes. Heterozygous hKLC4 G369fs/+ animals were healthy and viable, with no significantly deleterious effect on the total brood size and embryonic lethality observed in the parental hKLC4 line (Fig. 2C,D). The heterozygous hKLC4 G369fs/+ animals also did not have a significant swimming defect (Movie 5) compared to hKLC4 animals (Fig. 3A). However, the hKLC4 G369fs/+ heterozygotes did have a weak, but significant, nuclear migration defect in hyp7 precursors (Fig. 4B,C). This phenotype was similar to the one observed in predicted pathogenic variants hKLC4 A295P and hKLC4 T381I. Thus, although heterozygous hKLC4 G369fs/+ animals are healthier than the homozygous truncation animals, they still have a significant hyp7 nuclear migration defect.

In this study, we characterized a heterozygous KLC4 variant of uncertain significance detected in an individual with HSP. We engineered and used a humanized C. elegans model to test the physiological relevance of the variant in a heterologous in vivo system. We demonstrated that C. elegans can be used to model disease-associated variants of human KLC4, that we can use the hKLC4 C. elegans strain generated here to test the physiological impact of other KLC4 variants, and that this strategy could be used to model neuromuscular diseases in other genes with clear orthologs in C. elegans, including LINC complexes that target kinesin light chains to the nucleus.

There are thousands of rare diseases, each of which affect fewer than 1/2000 people. However, combined together, ∼300-400 million people suffer from rare diseases worldwide (Nguengang Wakap et al., 2020). Whole-genome sequencing has aided in the diagnosis of rare diseases, but the identification of a genetic underpinning of a disease still fails up to 75% of the time (Smedley et al., 2021). We therefore need to develop animal models to test variants of uncertain significance to make more efficient and better clinical diagnoses. One goal of the UDN is to bring together clinicians and basic scientists to test variants of uncertain significance in animal models (Baldridge et al., 2021). Here, we report clinical findings implicating a novel variant in the kinesin light chain gene KLC4 from a proband with HSP and the subsequent generation of a humanized C. elegans model to test the significance of the variant.

We identified an individual with late-onset HSP with a heterozygous variant in KLC4 predicted to cause a frame shift at residue 369, closely followed by a premature stop codon. An additional family was previously reported where a premature stop codon after residue 277 of KLC4 caused HSP in an autosomal-recessive manner; heterozygous family members did not have any symptoms (Bayrakli et al., 2015). Truncations in KLC4 after either 277 or 369 residues are predicted to disrupt the TPR domain, which mediates the interaction between kinesin and the cargo adaptor (Pernigo et al., 2013; Zhu et al., 2012), suggesting that both KLC4 variants should produce similar pathologies. The truncated KLC4 protein variant could conceivably be acting in a dominant-negative manner in the proband, which would not be recapitulated in C. elegans owing to likely nonsense-mediated decay of the transcript. Alternatively, there could be other protein variants in the proband acting synergistically with the truncated KLC4 protein to produce a phenotype. Variants in other genes were identified by whole-genome sequencing of the individual, and studies are underway to determine their contributions to pathogenicity.

We aimed to make a humanized C. elegans model to test clinical KLC4 variants. The open reading frame of the C. elegans ortholog klc-2 was successfully replaced by the human KLC4 coding sequence. Although the klc-2 null alleles are not viable (Sakamoto et al., 2005), the hKLC4 model had only low levels of embryonic lethality and a slightly reduced brood size. However, the hKLC4 animals had significant swimming defects. These could be due to defects in vesicle transport in motor neuron axons, which is a process that klc-2 is known to be required for (Sakamoto et al., 2005). Thus, the humanized model expressing KLC4 under control of the endogenous klc-2 locus retained much, but not all, of the function of klc-2. We used these phenotypes as a baseline to compare with the effects caused by the introduction of variants of uncertain significance into the hKLC4 line, including the clinical truncation variant. One of the four variants of uncertain significance with single amino acid changes, hKLC4 T381I, which was predicted to be possibly pathogenic, disrupted the function of hKLC4 in C. elegans. The T381I variant had phenotypes including a reduced brood size, slower thrashing in our swimming assay and an increased number of hyp7 precursor nuclei that failed to migrate. The heterozygous hKLC4 T381I/+ animals still had a severe motility defect, but no nuclear migration defects and no statistically significant change in their brood size or embryonic lethality compared to the homozygous T381I animals. These results indicate that while even a single copy of hKLC4 T381I is sufficient to cause a severe motility defect, a single copy of wild-type hKLC4 is sufficient to rescue the mild hyp7 nuclear migration defect. These data suggest that the T381I variant is also likely to disrupt KLC4 function in human cells and is, therefore, likely to be a pathogenic variant as predicted by PolyPhen-2 (AdzhubeIvan et al., 2010) and SIFT (Sim et al., 2012). However, the other three tested variants are unlikely to be pathogenic. Thus, these results support the bioinformatic predictions of significance made for the KLC4 mutations, R72H and R358H, which were predicted benign by SIFT (Sim et al., 2012), CADD (Kircher et al., 2014) and REVEL (Ioannidis et al., 2016), while not for the mutation A295P, which was identified as damaging by PolyPhen-2 (AdzhubeIvan et al., 2010) and possibly damaging by SIFT (Sim et al., 2012).

The clinical variant was homozygous lethal, and escaper larvae had severe nuclear migration defects like those with klc-2 null alleles (Fridolfsson et al., 2010), suggesting that the clinical variant would lead to severe pathologies when homozygous. We next examined whether the KLC4 frameshift variant at residue 369 acts in a haplo-insufficient manner. We observed a mild nuclear migration defect in heterozygous animals, consistent with other disease-associated variants. This mild phenotype in C. elegans could therefore be useful in predicting whether a variant of uncertain significance might cause clinical symptoms of HSP. Further work is needed to determine whether a heterozygous loss-of-function KLC4 variant can cause HSP; this includes finding more affected individuals who harbor disease-causing KLC4 variants.

The kinesin light chain is part of a network of proteins conserved from C. elegans to humans including the LINC complex-forming KASH and SUN proteins (Starr, 2019; Fridolfsson and Starr, 2010; Roux et al., 2009). Variants in the genes encoding components of human LINC complexes, including Nesprins, have been implicated in a wide variety of diseases, including neurological disorders, muscular dystrophies and various cancers (Janin et al., 2017). Humanized C. elegans strains for LINC complex components would provide additional reagents to test clinical variants in this important complex. The success of the humanized KLC4 C. elegans line described here suggests that this approach could be feasible for modeling other neuromuscular diseases associated with LINC complex dysfunction.

C. elegans strains were maintained on nematode growth medium plates seeded with OP50 Escherichia coli at room temperature; the N2 strain was used as wild type (Brenner, 1974). Some strains were obtained from the Caenorhabditis Genetics Center, funded by the National Institutes of Health Office of Research Infrastructure Programs (P40 OD010440). All strains were maintained, and phenotypic assays were performed at room temperature (∼22°C). The ycIs9 I strain, which was used to mark hypodermal nuclei with GFP, was generated along with ycIs10 V as previously described (Bone et al., 2014). The strains used in this study are listed in Table S1.

The hKLC4 strain was made utilizing the transgenesis services of InVivo Biosystems. To generate the hKLC4 strain, the coding region of the human KLC4 open reading frame was synthesized from the ATG to the stop codon of the most supported (Transcription Support Level 1) KLC4 isoform in Ensembl (ENST00000347162.10) and placed into a plasmid, pUC57 (Guo et al., 2014). The resulting gene coding for 619 amino acids was codon optimized for C. elegans (Mitreva et al., 2006), and synthetic introns, which have been shown to be essential for normal expression in C. elegans (Blumenthal, 2012), were inserted. The sequence of this gene block is shown in Fig. S2. The hKLC4 coding sequence was then flanked by 500 bp endogenous C. elegans klc-2 5′ and 3′ sequences from genomic DNA by PCR and Gibson assembly to create pNU2756. A hygromycin-resistant gene with a tbb-2 3′ UTR selection cassette flanked by loxP sites was included in pNU2756 to aid in identifying transgenic animals. We also included the 3′ UTR sequence of eft-3 (also known as eef-1A.1) after the hKLC4 stop codon because sometimes a strong 3′ UTR is needed for optimal expression (Chen et al., 2013). pNU2756 was used as the repair template for CRISPR/Cas9 genome editing (Fig. 1). Two sgRNAs targeting each end of the klc-2 open reading frame were used to guide CRISPR/Cas9 to cut out the coding region of klc-2 (Table S2). The sgRNAs that were preassembled with crRNA, Cas9 protein and the assembled ribonucleoprotein complex, and the repair template were then injected into the gonads of C. elegans young adults (Paix et al., 2015; Farboud et al., 2019). Progeny were screened for incorporation of hKLC4 into the worm genome by selecting for the animals that could survive upon hygromycin treatment. Two strains with hKLC4::eft-3 3′ UTR and the hygromycin selection cassette were obtained and backcrossed to N2 wild type to minimize off-target effects. Expression of hKLC4 was confirmed by reverse transcription quantitative PCR. Finally, using new sgRNAs (Table S2), the hygromycin::tbb-2 3′ UTR cassette and the eft-3 3′ UTR were removed, restoring the native klc-2 3′ UTR and generating the humanized klc-2(knu1031[hkLC4]) strain, referred to as hKLC4 (Fig. 1).

Point mutations were introduced into the hKLC4 line using CRISPR/Cas9 gene editing (Farboud et al., 2019). The sgRNA and ssDNA repair template sequences used to introduce missense mutations studied in this work are listed in Table S2. To identify successfully injected animals, co-CRISPR with templates to create dpy-10(gof) alleles was performed (Arribere et al., 2014). Point mutants were then screened for with genomic PCR followed by restriction digest analysis. Newly generated alleles were backcrossed to N2.

To generate the humanized allele with the clinical variant of uncertain significance, klc-2(knu1102[hKLC4(G369fs)]), a wild-type klc-2::gfp translational fusion expressed from an extrachromosomal array (Sakamoto et al., 2005) was crossed into the hKLC4 strain. Then, CRISPR/Cas9 gene editing was used as described above to introduce the mutation into hKLC4. By having a translational klc-2::gfp rescuing array, the worms were able to carry homozygous variants in hKLC4, even if they were not functional. To further analyze the clinical variant for haploinsufficiency, heterozygous hKLC4 G369fs/+ animals were generated by mating hKLC4 males with hKLC4 G369fs, klc-2::gfp hermaphrodites and selecting for animals without the klc-2::gfp extrachromosomal array.

The atomic models for the protein complex were generated using ′AlphaFold2-multimer-v2′ via ColabFold v1.3.0 (Mirdita et al., 2022) with default settings without using structure templates. Five models were generated and subsequently relaxed using amber force fields. The top-ranking model was used for Fig. 2B,B′. For visualization, some residues within the low-confidence linker region in KLC4 and KLC-2 were straightened by altering the torsion angles (phi and psi) to approximately −135° and 135°, respectively, using the ‘torsion’ command in ChimeraX (Pettersen et al., 2021).

Informed consent was obtained from the participant to participate in the National Institutes of Health (NIH)-Undiagnosed Diseases Network (UDN) protocol (15-HG-0130). Proband-only genome sequencing was performed through the UDN. Clinical investigation was conducted according to the principles expressed in the Declaration of Helsinki.

Brood size assays were conducted to quantify the brood size and embryonic lethality of each viable homozygous strain used in this study. For brood size assays, starting at the L4 stage, 10-15 single animals of each genotype were transferred onto fresh OP50 E. coli plates, labeled as Day 1, and kept at room temperature (∼22°C) for 42 h so that they became adults and had ∼24 h to lay eggs. At 42 h, adult worms were moved to new plates, labeled as Day 2, and kept at room temperature for 24 h. At the 24-h mark, adult worms were moved to new plates, labeled as Day 3, and dead eggs and young worms from the Day 1 plates were counted. At the next 24-h mark, the adult worms on the Day 3 plates were killed, and dead eggs and young worms from Day 2 plates were counted. At the next and final 24-h mark, dead eggs and young worms from Day 3 plates were counted. The total number of viable worms and dead eggs counted over the 3-day time course were plotted as the total brood size for each worm (Fig. 2C). Simultaneously, the total number of dead eggs counted over the 3-day time course was divided by the total brood size to calculate the embryonic lethality percentage for each worm and plotted (Fig. 2D).

Motilities of the strains used in this study were quantified by performing swimming assays. For the swimming assays, eight to ten L4 stage animals were put into a plate and flooded with M9 buffer. We observed and filmed worms swimming in buffer for 30 s. Using the Fiji wrMTrck plugin (Nussbaum-Krammer et al., 2015), we measured the average speed and BBPS to quantify the overall motility of the animals. Only worms that were consecutively tracked for 30 s were included in the analysis.

To quantify nuclear migration, humanized hKLC4 strains were crossed into a GFP nuclear marker expressed in larval hypodermal nuclei (ycIs9 I; Table S1). Nuclear migration assays were performed as described (Fridolfsson et al., 2018). Briefly, larval worms with GFP-marked hypodermal nuclei were picked and mounted on 2% agarose pads in ∼5 µl of 1 mM tetramisole in M9 buffer. Syncytial hyp7 nuclei were scored as abnormally located if they were in the dorsal cord.

GraphPad Prism version 9.0 software was used for the statistical analyses. All the data from nuclear migration, brood size and embryonic lethality, and swimming assays were displayed as scatter plots with means and 95% confidence intervals as error bars. Sample sizes and the statistical tests are indicated in the figure legends. Unpaired two-tailed Student’s t-tests were performed on the indicated comparisons.

We thank members of the Starr–Luxton laboratory for their helpful comments, especially Daniel Elnatan for help with the AlphaFold figure preparation.