Functional analysis of a novel de novo variant in PPP5C associated with microcephaly, seizures, and developmental delay

We describe a proband evaluated through the Undiagnosed Diseases Network (UDN) who presented with microcephaly, developmental delay, and refractory epilepsy with a de novo p.Ala47Thr missense variant in the protein phosphatase gene, PPP5C. This gene has not previously been associated with a Mendelian disease, and based on the population database, gnomAD, the gene has a low tolerance for loss-of-function variants (pLI=1, o/e=0.07). We functionally evaluated the PPP5C variant in C. elegans by knocking the variant into the orthologous gene, pph-5, at the corresponding residue, Ala48Thr. We employed assays in three different biological processes where pph-5 was known to function through opposing the activity of genes, mec-15 and sep-1. We demonstrated that, in contrast to control animals, the pph-5 Ala48Thr variant suppresses the neurite growth phenotype and the GABA signaling defects of mec-15 mutants, and the embryonic lethality of sep-1 mutants. The Ala48Thr variant did not display dominance and behaved similarly to the reference pph-5 null, indicating that the variant is likely a strong hypomorph or complete loss-of-function. We conclude that pph-5 Ala48Thr is damaging in C. elegans. By extension in the proband, PPP5C p.Ala47Thr is likely damaging, the de novo dominant presentation is consistent with haplo-insufficiency, and the PPP5C variant is likely responsible for one or more of the proband’s phenotypes.


INTRODUCTION
Protein phosphatase 5, encoded by the PPP5C gene, is a serine/threonine phosphatase that participates in the modulation of protein function by removing phosphate groups at serine and threonine residues. PPP5C interacts with a wide range of proteins, including nuclear receptors, kinases and transcription factors, and plays an important role in the regulation of numerous cellular processes including cell growth and differentiation, migration, apoptosis, DNA damage response and regulation of circadian rhythm [1][2][3][4][5][6][7][8] (reviewed in [9]). PPP5C is expressed ubiquitously but is present in high levels in the mammalian brain [10][11][12][13], and altered phosphatase activity has been associated with disease. For example, elevated PPP5C expression was detected in tumors, but not in normal tissues taken from patients with prostate cancer [14].
Additionally, knock down of PPP5C caused a decrease in the proliferation of cultured prostate, lung and urinary bladder cancer cells [5,14,15]. In contrast, overexpression of PPP5C in a breast cancer cell line (MCF7) enabled estrogen-dependent cells to grow independently of estrogen [16]. PPP5C has also been shown to regulate ciliogenesis by dephosphorylating DVL2, a key mediator of WNT signaling, in mammalian cells [17].
PPP5C is known to physically interact with and dephosphorylate HSP90 and glucocorticoid receptors suggesting a role in metabolism, stress response, immune function, skeletal growth, reproduction and cognition [18,19]. HSP90 is a protein chaperone responsible for stabilizing many different proteins. For example, in conjunction with the TEL2-TTI1-TTI2 (TTT) complex, HSP90 stabilizes phosphatidylinositol 3-kinase related kinases (PIKKs), which in turn are responsible for regulating DNA damage responses, cell growth, and transcriptome surveillance (reviewed in [20,21]).
Ppp5c homozygous knockout mice are viable and fertile, however, these mice have lower body weights as compared to their wild type littermates and do not gain weight when fed a high-fat diet [22][23][24]. In Drosophila, the PPP5C homolog (PpD3) is most highly expressed during embryonic development and has been implicated in centrosome separation during mitosis [25,26]. Ppt1p, the yeast homolog of PPP5C, shows peak expression during early log cell growth, suggesting that Ppt1p may be involved in cell growth similar to what has been reported in human cell culture [27].
Analogous to the mammalian PPP5C, the Caenorhabditis elegans (C. elegans) ortholog PPH-5 is known to associate with the HSP90 and the glucocorticoid receptor complex [2,28]. While pph-5 null animals are superficially wild-type, genetic studies in C. elegans have revealed a role for pph-5 in the regulation of GABA signaling and in neurite growth by opposing the activity of the F-box protein, MEC-15 [28]. Similarly, numerous pph-5 loss-of-function mutants were identified in a genetic screen for interactors with separase (sep-1), a cysteine protease that plays an important role in chromosome segregation during mitosis, cortical granule exocytosis, and eggshell formation [29,30].
In a sep-1 hypomorphic mutant that is unable to form an eggshell, loss-of-function mutations in pph-5 were found to rescue the embryonic lethality, suggesting a role for pph-5 in the regulation of exocytosis and eggshell formation.
Here we present a proband accepted to the Undiagnosed Diseases Network This gene has not previously been associated with a Mendelian disease. Therefore, we evaluated this gene-variant by performing in vivo functional analysis in the model organism C. elegans, as has been performed for other genes of uncertain significance [32][33][34]. We employed CRISPR-Cas9 to introduce the proband's missense variant into pph-5, the C. elegans ortholog of PPP5C, and assessed the impact of pph-5 Ala48Thr, corresponding to human PPP5C p.Ala47Thr. We provide data showing that the pph-5 Ala48Thr variant alters neurite growth, GABA signaling, and eggshell formation, by acting as a strong hypomorph or loss-of-function allele. By extension, our data suggest that the human PPP5C p.Ala47Thr is damaging and likely responsible for one or more of the neurodevelopmental phenotypes in the proband.

Proband evaluation and consent
Informed consent was obtained from the proband's parents as part of enrollment into the Undiagnosed Diseases Network study, which is overseen by the National Institutes of Health Institutional Review Board (protocol #15HG0130). Phenotypic evaluation was completed through review of medical records and in-person examination by physicians who are co-investigators on the study.

Exome sequencing and variant analysis
Clinical trio exome sequencing was performed clinically in a CLIA certified laboratory and the data was transferred for research analysis within the Undiagnosed Diseases Network. Codified Genomics (www.codifiedgenomics.com) and emedgene (www.emedgene.com) were used to prioritize novel, de novo variants that were not observed in gnomAD or in our database of over 1000 exomes and genomes (unless the phenotype of the probands matched) and biallelic variants in the coding region or near intron-exon boundaries with allele frequency of less than 1% and without homozygous entries in gnomAD [31] (Table S1). Variants of interest were confirmed using Sanger Sequencing within a CLIA certified laboratory.

C. elegans strain and culture information
C. elegans were grown on standard Nematode Growth Media (NGM) and fed OP50 E. coli [35]. As mec-15(u1042) phenotypes were previously reported to be more extreme at higher temperatures [28], strains used for aldicarb assay and PLM-anterior neurite length were grown at 25C for at least two generations before analysis. Strains used in this paper include (TU5237) mec-15(u1042); uIs115 (a generous gift from the Chalfie lab), pph-5(ok3498), sep-1(e2406) and VC2010. Strains with pph- 5(udn90-94) generated in this study are described below. A full list of strains used in this study is in Table S2. Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440).

CRISPR/Cas9-based gene editing
pph-5(udn90-94) alleles were generated by injecting VC2010 1 day adults with Cas9 (IDT 1081060), tracrRNA (IDT 1072533), crRNA with the targeting sequence TGCAGATCAAGTGTACGACG (IDT Alt-R® CRISPR-Cas9 sgRNA), pRF4 rol-6 plasmid as a positive injection marker, and single-stranded DNA oligonucleotide repair template (IDT) as previously described [32,33]. Repair templates were 100 bases in length, centering on the locus of the proband variant, and were used to destroy the PAM site to prevent further editing, add a silent AvaII (NEB R0153L) restriction site to aid in genotyping, and to introduce the proband variant. Control repair templates containing only the silent PAM site change and the silent AvaII sites were also used, to determine that any phenotypes observed in the proband variant lines were due to the proband variant alone and not due to any silent mutations. Complete sequences of the repair templates are shown in Table S3. Each edited line was Sanger sequence verified and backcrossed twice to VC2010 to remove any background mutations. A line was considered independent if obtained from a separate injected animal. Three variant Ala48Thr lines were initially generated, and as all three lines showed no statistical difference in a preliminary blinded aldicarb paralysis assay (Fig. S1), only two lines were used for further analysis. Additionally, two control Ala48Ala edit lines were initially generated, and as both showed no statistical difference from each other or to the corresponding background strains in a blinded aldicarb paralysis assay and in the sep-1 embryo lethality assay, only one control edit line was used for analysis.

Aldicarb paralysis assay
To make 0.25mM aldicarb plates, 22.5 L of 100mM aldicarb (Sigma #33386) stock in 70% ethanol was spread onto each NGM plate and allowed to dry at room temperature for one hour. A drop of OP50 suspended in LB was added to the center of each plate to help keep animals concentrated in the middle of the plate. Plates were then allowed to dry at room temperature for an additional hour. Ten one-day-old adult animals were added to each plate, with three replicates per strain. Paralyzed animals, defined as those not independently moving following two gentle taps with a platinum pick, were counted every twenty minutes. Animals were scored every thirty minutes from t=30 minutes to 130 minutes. Each experiment was performed three times, and the experimenter was blinded to genotypes. Plates that had two or more animals crawl off the agar were discarded from analysis.

Obtaining heterozygote animals for phenotyping
To ensure heterozygous animals were generated, mec-15; uIs115 larval stage L4 animals were raised on fem-1 RNAi plates at 25C, and females from the next generation were picked (i.e. 1-day adults observed with stacked oocytes and no embryos that did not lay embryos for 4 hours). These females were mated to pph-5(xxx); mec-15; uIs115 males, and resulting cross progeny were analyzed as described above. At least 29 PLM-AN lengths were recorded per line, from at least 15 animals.

Quantification of embryo lethality in sep-1 mutant animals
Embryo lethality in sep-1 separase and pph-5 mutant animals were quantified as previously described [30]. In brief, strains were maintained at 15ºC to obtain gravid mutant adult animals as sep-1(e2406) mutants have elevated sterility at higher temperatures. A single hermaphrodite animal staged between L4 and young adult was plated to individual 35mm NGM plates for 24 to 30 hours at 20ºC and allowed to lay embryos. Animals were removed from the plate and embryos were incubated at 20ºC for another 24 hours to allow hatching to occur. Dead embryos and live young larvae were counted to determine embryo lethality.

Clinical presentation
The proband was a 6-year-old girl with epilepsy, nystagmus, congenital microcephaly, and developmental delay (Fig. 1). Brain MRI initially at 11 months old showed slightly incomplete myelination that improved on an updated brain MRI at 4 years old but still with diminutive cerebral white matter volume Developmentally, she rolled over at [8][9] months with therapies, sat up after age 1 year, and currently crawls and pulls to a stand but is unable to walk independently.    [31], and the gene has not yet been associated with disease apart from a possible association with autism [36]. The gene level scores, pLI = 1 and o/e = 0.07, suggest a haploinsufficiency model for the genetic mechanism in humans.

PPP5C is conserved in C. elegans
To assess the effect of PPP5C p.Ala47Thr on in vivo function, we evaluated the impact of this gene variant in the model organism C. elegans. The ortholog of PPP5C in C. elegans is pph-5, which overall has an identity of 55.9% and a similarity of 72.3% to the human gene. In both organisms, PPP5C and pph-5 encode two large domains: a phosphatase domain, which shares 62.7% identity (77.8% similarity), and three tetratricopeptide repeats (TPR) that are responsible for autoinhibition of the phosphatase domain by physically blocking the active site [37,38]. The vast majority of mutations in C. elegans pph-5 were previously identified in a suppressor screen [30] and are conserved or moderately conserved in humans ( Table   1). The PPP5C p.Ala47Thr variant is located at the N-terminus of the protein in the TPR1 domain ( Fig. 2A). The TPRs are highly conserved between C. elegans and humans, with PPH-5 and PPP5C sharing ~54% identity (72.5% similarity) in this domain (Fig. 2B). We therefore modeled the PPP5C p.Ala47Thr proband variant in the corresponding C. elegans residue (Ala48) of pph-5 using CRISPR/Cas9-based gene editing. In addition to the variant residue, we introduced synonymous changes to block Cas9 re-cleavage following homology-directed repair and a restriction enzyme cleavage site to aid in allele genotyping. As controls, we generated lines that contain only the synonymous changes, but not the proband's variant. In total, we generated three independent variant (Ala48Thr) alleles and two independent control (Ala48Ala) alleles (Table S2). All the CRISPR-edited lines were superficially wild-type, similar to the pph-5(ok3498) null mutants (hereafter referred to as pph-5(null)) [29].

pph-5 Ala48Thr variant is defective in function and suppresses the touch receptor neuron neurite growth phenotype in mec-15 F-box protein mutants
Since the pph-5 Ala48Thr proband variant and the pph-5 null mutants were superficially wild-type, we employed sensitized genetic backgrounds, the mec-15 null or the sep-1 hypomorph, where it is known that pph-5 acts to oppose the function of these genes. In C. elegans, pph-5 has previously been reported to have genetic interactions with mec-15 in regulating neurite growth [28]. MEC-15 is an F-box protein (orthologous to human FBXW9) that promotes touch receptor neurite growth by aiding degradation of DLK-1 (orthologous to human MAP3K12), an inhibitor of microtubule stability [28,39]. As a consequence, in mec-15(u1042) mutants, the microtubules are less stable, and the PLM anterior neurites (PLM-AN) are considerably shorter than those of wild-type animals.
PPH-5 is part of the HSP90 chaperone complex, which aids in protein folding [40,41], including that of DLK-1 [28]. Thus, PPH-5 negatively regulates neurite growth by opposing the function of MEC-15. Loss-of-function alleles of pph-5, and other components of the HSP90 complex, therefore restore microtubule stability in mec-15(u1042) mutants, and suppress the short PLM-AN phenotype [28]. To examine the role of the PPH-5 Ala48Thr proband variant in neurite growth, we crossed the pph-5 variants into the mec- 15(u1042) mutant background and measured PLM-AN neurite length. As previously reported, mec-15 mutants have significantly shorter PLM-AN neurites compared to wild-type animals ( Fig. 3A and B). However, when combined with the proband variant, the short neurite phenotype was completely suppressed and neurite length was restored to that of the wildtype ( Fig. 3A and B). In contrast, the pph-5 Ala48Ala control edited lines did not suppress the short neurite phenotype of mec-15(u1042) animals. These results indicate that pph-5 Ala48Thr is defective in function as a negative regulator of PLM-AN neurite length.
Moreover, the extent of the suppression by the proband's variant was similar to that of the reference null allele of pph-5, indicating that the pph-5 Ala48Thr variant is likely a strong hypomorph or a complete loss-of-function.

pph-5 Ala48Thr variant suppresses aldicarb-induced paralysis of mec-15 mutants
In the C. elegans neuromuscular junction, opposing neurotransmitter signals regulate muscle contraction. Acetylcholine (ACh) provides the stimulatory signal for muscle contraction while gamma-aminobutyric acid (GABA) provides the inhibitory signal.
ACh activity is regulated by acetylcholinesterase, an enzyme that degrades ACh in the neuromuscular junction. Inhibition of acetylcholinesterase by treatment with the drug aldicarb leads to ACh accumulation in the extracellular space and subsequent hypercontraction of the muscle and eventual paralysis (reviewed in [42]). MEC-15 promotes synaptic transmission in GABAergic motor neurons to inhibit muscle contraction [43,44].
Consequently, when exposed to aldicarb, mec-15(u1042) mutants release less GABA and reach paralysis at a faster rate than wild-type animals (Fig. 4A)   obtained with the pph-5(null); mec-15(u1042) animals (Fig. 4B). Taken together, our aldicarb-sensitivity data provide additional evidence for pph-5 Ala48Thr variant being damaging to gene function and behaving similarly to the null allele.

pph-5 Ala48Thr variant suppresses embryonic lethality of sep-1 mutants
As a final test of the proband's variant function, we examined the role of PPH-5 Ala48Thr in embryogenesis. C. elegans sep-1 (orthologous to human ESPL1) encodes a cysteine protease required for the cleavage of cohesin complex subunits during mitosis, exocytosis of cortical granules and eggshell formation [46,47]. Consequently, sep-1(e2406) hypomorphic mutants do not form an eggshell, fail to hatch, and are embryonic lethal [47]. However, reduction of function mutations in pph-5, including mutations within the TPR domain, are known to suppress the embryonic lethality of sep-1 mutants, indicating that PPH-5 phosphatase activity has a role in negatively regulating both SEP-1 and eggshell formation [30]. Therefore, we used the suppression of embryonic lethality as a readout to assess the functionality of the proband's variant. Consistent with previous results, we found sep-1 animals to be 100% embryonic lethal (Fig. 4C). When the control edit pph-5 Ala48Ala allele was combined with the sep-1 allele, the resulting animals were also 100% embryonic lethal as indicated by the lack of hatched embryos (Fig. 4C). In contrast, embryonic lethality was strongly suppressed by the pph-5 Ala48Thr variant and 89-96% of the embryos successfully hatched (Fig. 4C). The percent hatched values are similar to the previous data obtained for the pph-5(tm2979) null; sep-1(e2406) double mutant [30]. Combined, our results indicate that Ala48Thr variant leads to defective PPH-5 function with a severity that is similar to that of the pph-5 null alleles.

pph-5 null and Ala48Thr variants behave recessively in C. elegans
The underrepresentation of loss-of-function variants in the control human population (gnomAD v2.1.1, pLI=1, o/e = 0.07) suggests that PPP5C is a haploinsufficient gene. However, missense variants in haploinsufficient genes may act in a gain-of-function manner, for example, via a dominant negative or a hyperactive mechanism. We therefore used the PLM-AN neurite growth assay to compare heterozygous and homozygous phenotypes for the pph-5 null and Ala48Thr alleles.

pph-5 legacy mutations as functional tests of yet to be identified PPP5C missense variants
A collection of 25 pph-5 missense mutations has been generated in C. elegans forward genetic screens [29,30]. The pph-5 mutations were isolated following chemical mutagenesis as recessive suppressors of the sep-1 separase embryonic lethality phenotype, analogous to the assay shown in Figure 4C. These pph-5 legacy mutations may have already been observed or have yet to be observed in the human ortholog and provide pre-existing experimental functional information about conserved residues and variants in PPP5C.
We mapped the 25 legacy missense variants to the corresponding amino acid residues in PPP5C ( Table 1)  indicating that the screen is very sensitive [30]. We therefore subdivided the pph-5 mutations into three categories based on the extent of suppression of sep-1 mutant embryonic lethality [29,30]: strong hypomorphs that are similar to null alleles (greater than 70% suppression); weak hypomorphs (between 30 -70% suppression); and very weak hypomorphs (less than 30% suppression). Assuming haploinsufficiney for the The Ala47Thr variant is located in the TPR domain of PPP5C, which is thought to play an important role in the autoinhibition of the phosphatase activity through intramolecular interactions with the C-terminal domain that hinder substrate access to the phosphatase catalytic pocket [41]. The binding of HSP90 to the TPR domain of PPP5C relieves this autoinhibition and stimulates phosphatase activity. The Ala47Thr variant might disrupt binding with HSP90. The active phosphatase can subsequently activate one of several hundred known HSP90 client proteins [48]. One such client is ataxia and telangiectasia and Rad3 related (ATR), a kinase that activates DNA damage signaling response [48,49]. Mutations in ATR, and other ATR-mediated checkpoint response genes, MCPH1 and ATRIP, have been linked to microcephaly and specifically Seckel syndrome [50] (reviewed in [51]). This is thought to be because ATR has an important role in supporting the rapidly proliferating zones in the brains of mice [52,53]. ATR, as well as other PIKK genes, are stabilized by the Hsp90 complex, through interaction with the Tel2-Tti1-Tti2 (TTT) complex [49,54,55]. Compound heterozygous variants in TELO2 (the human homolog of yeast Tel2) have been identified in six probands, including three siblings and three unrelated individuals, with You-Hoover-Fong syndrome, which is characterized by microcephaly, movement disorder, and severely delayed development [56]. Recently, a proband with microcephaly was identified to have a homozygous missense variant in TTI2, further implicating ATR, HSP90 and TTT complex in microcephaly [57]. Given that PPP5C is highly expressed in the mammalian brain, and overexpression and knockdown studies showed hyper-and hypo-cellular proliferation, respectively, it is tempting to hypothesize that a reduction of PPP5C function as seen with the variant Ala47Thr could lead to microcephaly and other neurodevelopmental abnormalities observed in the proband [5,14]. Moreover, it is also intriguing to note the parallel between the proband's failure to thrive and the low body weight of Ppp5c -/mice [23]. However, the Ppp5c knockout mouse is not a perfect model of human phenotypes as it does not have reported microcephaly or intellectual disabilities. It has been suggested that mouse models of microcephaly can have limits, as the human brain is much larger when compared to the body, and mice do not have structures such as gyri and encephalization [51]. Additionally, a recent study looked at the correlation of phenotypes between known human disease gene phenotypes and knockout mouse model phenotypes, of which 49.53% do not match [58]. A recent study also demonstrated that there are differences in gene expression during the developmental timeline of humans as compared to mice and other model organisms (with approximately 15% of genes expressed in the brain during development having different expression patterns between mice and humans), which could explain differences in phenotype between human patients and mouse models [59].
Numerous chemical mutagenesis screens have been performed in C. elegans and other model organisms generating large collections of historical or legacy missense mutations. These missense mutations, as they disrupt protein function, provide functional information for the corresponding variant in the human ortholog or homolog. For example, the myosin heavy chain mutation unc-54(st132) Glu524Lys, isolated as a dominant extragenic suppressor of the unc-22 (human titan ortholog) twitching phenotype [60,61] is at the same position and the same change in the myosin motor domain as MYH7 Glu525Lys, a pathogenic variant in dilated cardiomyopathy [62]. This unc-54 legacy mutation provides experimental support for the MYH7 variant being damaging, where the data were available more than 10 years before identification of the variant from nextgeneration sequencing of a patient cohort.
We have examined a collection of 25 pph-5 legacy missense mutations identified from forward genetic screens [29,30]. For 10 of these the residue is identical or a conservative substitution in PPP5C, there are no instances of variants at these residues in gnomAD, and all the pph-5 mutations are strong hypomorphs or complete loss of function, and thus likely to contribute to human phenotype if observed in PPP5C, under a model of dominant presentation due to haploinsufficiency ( Table 1). The 10 variants have yet to be reported in the human population in ClinVar (search date 12/27/2021).
These 10 legacy mutations provide functional data that the missense changes are damaging in C. elegans and, by extension, when observed in humans likely to contribute to disease. Given the potential utility of model organism legacy missense mutation findings to human gene-variant analysis, model organism databases (WormBase, FlyBase, Alliance of Genome Resources, etc.) should aggregate and display such data similar to that shown in Table 1 and make that information available to databases such as ClinVar and MARRVEL.
In summary, PPP5C p.Ala47Thr, a de novo heterozygous missense gene-variant identified in a UDN proband, was modeled in C. elegans. In vivo functional analyses indicated that the pph-5 Ala48Thr variant is damaging, similar in extent to the complete pph-5 loss of function. By extension, our data suggest that PPP5C p.Ala47Thr is also damaging, likely resulting in a strong hypomorph or complete loss of function. Given that PPP5C, HSP90 and client proteins are important for the development of the brain, it is likely that the PPP5C p.Ala47Thr loss-of-function variant is responsible for one or more of the proband's phenotypes, such as microcephaly and/or the other neurodevelopmental disorders.