ATP1A3 mosaicism in families with alternating hemiplegia of childhood

Alternating hemiplegia of childhood (AHC) is a rare and severe neurodevelopmental disorder characterized by recurrent hemiplegic episodes. Most AHC cases are sporadic and caused by de novo ATP1A3 pathogenic variants. In this study, the aim was to identify the origin of ATP1A3 pathogenic variants in a Chinese cohort. In 105 probands including 101 sporadic and 4 familial cases, 98 patients with ATP1A3 pathogenic variants were identified, and 96.8% were confirmed as de novo. Micro‐droplet digital polymerase chain reaction was applied for detecting ATP1A3 mosaicism in 80 available families. In blood samples, four asymptomatic parents, including two paternal and two maternal, and one proband with a milder phenotype were identified as mosaicism. Six (7.5%) parental mosaicisms were identified in multiple tissues, including four previously identified in blood and two additional cases identified from paternal sperms. Mosaicism was identified in multiple tissues with varied mutant allele fractions (MAFs, 0.03%‐33.03%). The results suggested that MAF of mosaicism may be related to phenotype severity. This is the first systematic report of ATP1A3 mosaicism in AHC and showed mosaicism as an unrecognized source of previously considered “de novo” AHC. Identifying ATP1A3 mosaicism provides more evidence for estimating recurrence risk and has implications in genetic counseling of AHC.


| INTRODUCTION
Alternating hemiplegia of childhood (AHC, MIM: 614820) is a rare and predominantly sporadic neurological disorder characterized by attacks of hemiplegia and other paroxysmal manifestations such as abnormal eye movement and dystonia. 1 Additional symptoms may appear later in the disease process. 2 Several study groups have identified de novo pathogenic variants in the ATP1A3 gene as the cause of AHC in 93.3% to 100% of patients. [3][4][5] De novo variants have been discovered to be a more prominent disease-causing mutation type than inherited variants in numerous neurological genetic disorders, such as Dravet syndrome and autism spectrum disorders. 6,7 De novo variant was an alteration in a gene that is present for the first time in one family member as a result of a mutation in a germ cell (egg or sperm) of one of the parents or in the fertilized egg itself. De novo variants are detectable in the diseaseaffected probands but undetectable in either of their parents in genomic DNA from peripheral blood sample. The consistency of geneassociated phenotypes and patients' manifestations were confirmed clinically. 8 Many variants were considered de novo by either direct Sanger sequencing or next-generation sequencing (NGS) results validated by Sanger sequencing. [3][4][5] De novo variants are considered to be either mainly prezygotic variants in the germ cells of parents, or postzygotic variants in the offspring. 9,10 They have been reported to have strong paternal origin bias in large-scale family data. [11][12][13] Some postzygotic mosaicisms in the parents may occur at the early stage and affect both the germ cells and somatic cells, known as gonosomal mosaicism. 14 Recent publications have further reported that some "de novo" pathogenic variants are actually inherited from underestimated parental mosaicism. 15,16 In AHC, de novo single nucleotide pathogenic variants are more common than inherited variants. [3][4][5] However, few families with autosomal dominant inheritance have been reported. A family with two AHC half-sisters born to an asymptomatic mother and different fathers, suggesting maternal mosaicism. 17 Hully et al reported two unrelated families with two full siblings with ATP1A3 variants presenting epilepsy and ataxia, and the authors suggested the potential occurrence of parental ATP1A3 mosaicism. 18 More sensitive genetic testing approaches are required for the detection and validation of mosaicism in AHC-affected families with ATP1A3 variants, especially in parents with multiple AHC-affected children. NGS-based deep sequencing methods, such as personal genome machine amplicon sequencing for mosaicism (PASM) and molecular inversion probes, can be used to detect mutant allele fraction (MAF) as low as 1%, and duplex sequencing can detect even lower percentages. 19 However, deep sequencing methods targeting the whole exome or genome require significantly increased sequencing depth to achieve higher detection limits. Nextgeneration micro-droplet digital polymerase chain reaction (mDDPCR) can be used to detect MAF as low as 0.01%, as is suitable for hotspot variants. 20 Here, we first report parental ATP1A3 mosaicism in AHC families, regarded as "de novo" via Sanger sequencing, by mDDPCR and PASM using blood, sperm, and other tissues. We also identified ATP1A3 mosaicism in one proband with milder AHC.  Figure 1A). Fifty-one patients had been described in our previous publication, 5 and the remaining patients were not reported.
The four multiplex families included one dominant family with two affected family members, two sets of monozygotic twins who were both affected with AHC, and one family with sibship of an affected brother and sister.
2.2 | DNA isolation, ATP1A3 variant screening, and functional prediction of variants Genomic DNA from peripheral blood was extracted using a saltingout procedure. Genomic DNA from buccal swab, saliva, hair follicles and urine was extracted using a QIAamp DNA Micro Kit (#56304, Qiagen, Germany). Skin biopsies were collected from a 5-mmdiameter skin punch, and DNA was extracted using an All Prep   Table S1. All 23 coding exons and their flanking regions were sequenced using primers as previously described. 4 Variants described in this article were based on reference cDNA NM_152296.4, reference amino acid sequence NP_689509.1, and reference genome assembly hg19. Deleterious variants detected in the probands were then sequenced in the blood samples of their parents or other available family members. Hundred unrelated healthy individuals were sequenced as controls. Functional predictions of the variants were carried out using conventional prediction software as well as the classifiers we developed for AHC or ATP1A3. 5,23 The variant and phenotype information have been deposited in https://databases.lovd.nl/shared/genes/ATP1A3.

| Detection of mutant allele origin and paternity
Allele-specific PCR (ASPCR) was carried out to determine the origins of the mutant alleles in the probands. Primers for amplifying the SNP fragments are listed in Table S2. For ASPCR, one or two pairs of allele-specific primers were designed based on the informative SNPs.
The patients' two alleles were analyzed separately using paternal or maternal-specific primers containing different bases of the target variants. The primers used for ASPCR are listed in Table S3.
Short tandem repeat (STR) analysis was carried out for all informative families to confirm paternity and gender using 11 highly polymorphic microsatellite markers. The forward primer of each pair was labeled with FAM fluorophores. Primers for STR analysis are provided in Table S4. Labeled amplicons were detected on an ABI 3730 automated sequencer (Applied Biosystems by Thermo Fisher) and analyzed using genotype software (Gene Marker 1.5; SoftGenetics).

| mDDPCR screening and PASM for mosaicism
Raindrop mDDPCR was used for the detection of low-fraction mosaicism in 80 available families ( Figure 1B). Genotyping assays were first tested using an endpoint-genotyping experiment. Ultraviolet treatments were carried out after each reaction to reduce contamination.
After droplet emulsion generation and PCR reactions, post-PCR emulsions were subjected to a droplet detector, and data were analyzed  15 MAFs in detectable ATP1A3 mosaic families were further quantified using PASM. The primers of PASM detection were provided in Table S5.

| Prenatal diagnosis
Prenatal diagnosis was performed upon further pregnancy of one proband's mother at 22  F I G U R E 3 mDDPCR and PASM detected ATP1A3 variants from blood samples of 80 AHC families. A, Four parental mosaic variants were detected by mDDPCR, and their pedigree charts are shown. Mosaic variants are clearly showed on the flow cytometry scatter plots of mDDPCR under the cluster name of "MU" at the bottom right corner. Peripheral blood from the proband was used as a positive control; the performance of the negative controls is not shown. B, mDDPCR considered AHC-affected families with "de novo" ATP1A3 variants. Probands were all detected with a mutant allelic fraction (MAF) of 50%. The lower bounds of the 95% binominal CIs of the measured MAFs were under 0.01% in the nonmosaic parents and negative controls. C, The MAF identified by mDDPCR and PASM confirmed that A01203 carried a pathogenic variant on c.2839G>C/p.(Gly947Arg) with the MAF and the 95% confidence intervals that met the detection criteria for a mosaic variant. The proband A06603, who was detected with a heterozygous variant at the same base c.2839G>C/p.(Gly947Arg), was chosen as a positive control [Colour figure can be viewed at wileyonlinelibrary.com] standard and guidelines, 24 all variants were classified as likely pathogenic.
For 93 out of 98 probands with ATP1A3 variants, Sanger sequencing was performed in the blood DNA of their parents. We identified 96.8% (90/93) as de novo pathogenic variants and 3.2% (3/93) as inherited. For the three patients with inherited variants, one patient inherited the heterozygous variant from her AHC-affected mother, 5 and the other two patients inherited the variant from the asymptomatic mother (A05202) or father (A11201), suggesting the mosaicism due to the lower signal ( Figure 3A). ASPCR amplification were performed in 33 families with available SNPs. Twenty-nine (87.9%) were paternal and four (12.1%) were maternal, indicating significant parentof-origin sex bias. One index case of paternal origin is shown in Figure S3, and the variants and parental origins are listed in Table S7.
In the family with two affected siblings (Family A065, Figure S4

| Mosaic variants were detected by mDDPCR in the blood of four parents and one atypical proband
Among AHC families that were considered to carry "de novo" or "parental mosaic" ATP1A3 pathogenic variants by Sanger sequencing, 82 families agreed to participate to further genetic tests of mosaicism screening, including Family A052, A112 and A065. Twenty-four genotyping assays were available for 80 families according to an endpoint genotyping qPCR analysis ( Figure S6 and S7).
Four of 80 (5%) parental mosaic families were detected from the peripheral blood samples by mDDPCR. The MAFs were quantified and validated as 18.82% in the mother of Family A052, 17.05% in the father of Family A112, 7.65% in the mother of Family A067 and 7.53% in the father of Family A066. The MAFs of PASM were similar to those obtained by mDDPCR ( Figure 3A and Table 1).
Mosaicism was also confirmed by mDDPCR and PASM in an atypical AHC proband (A01203) carrying the hotspot variant c.2839G>C/p.(Gly947Arg). MAF measured using the blood sample of proband A01203 was 44.19% by mDDPCR and 32.5% by PASM, which was significantly lower than all the other probands ( Figure 3B, C). The onset age and the age of first hemiplegia of A01203 were distinctly later. The clinical phenotype of the A01203 was milder, compared with the four probands carrying the same heterozygous variant (   (Table S8), including parental mosaic cases, and healthy sperm was purified and subjected to mDDPCR. Apart from the mosaicisms that were detectable in paternal blood, mosaicisms were detected in sperms of two additional cases, A01501 (MAF of 0.03%) and A06501 (MAF of 12.42%; Figure 4A,B).
Taken together, six (7.5%) parental mosaic cases were identified from the ATP1A3 mutated AHC affected families.

| Mosaic variants were detected in multiple peripheral samples
Multiple samples from parents, including buccal swab, saliva, hair follicles, skin punch and urine, were collected from 13 individuals from 11 AHC families. Among the 13 individuals, 6 parents were from five mosaic families (A052, A065, A066, A067, and A112) and the other 7 parents were from six non-mosaic families (A012, A023, A050, A061, A082 and A111). Mosaic variants were also detected in multiple samples from blood-positive mosaic parents ( Figure 4B and Table 3), but with different MAF values. Heterogeneity of the MAFs was observed among the different tissues, and the shared mosaicism suggested that the variants might have occurred at the early stages of embryonic development. High correlation was observed between MAFs measured by PASM and mDDPCR (R 2 = 0.9722; Figure 4C), showing that the differences of MAF we observed are reliable.
To analyze the MAF differences among multiple samples, we performed hierarchical clustering for square-root-transformed MAFs detected by mDDPCR from the parental samples ( Figure 4D). Euclidean distances were used for comparisons. From the clustering result, the probands showed the lowest similarity to the other samples. Samples from the blood of normal controls and non-mosaic parents had higher similarities. Most samples from mosaic parents showed a similar pattern. MAFs in blood and saliva were the most similar because of their mesodermal origin. Buccal swabs containing ectoderm-derived oral epithelium as well as mesoderm-derived white blood cells showed a similar MAF to the skin punch. These findings showed that the cell type might be an important factor influencing MAF. Additional cell divisions from the common progenitor cells might also contribute to this phenomenon.
The mDDPCR raw flow cytometry scatter plots are presented in Figure S8, and MAF differences are compared in Figure S9.

| Prenatal diagnosis in families with parental ATP1A3 mosaicism
Prenatal tests were then carried out for the mother from Family Abnormal eye movement  We observed that the differences in MAF between tissues showed similarities between mesoderm and ectoderm-derived samples, similar to our previous observations for samples from Dravet syndrome families. 15 We also found that mosaic variants that were detectable in paternal blood could be detected in their sperm, carrying deleterious alleles at the time of donation. Our results emphasize the urgency of including sperm detection as well as multiple parental samples for genetic testing and counseling for parents who have already had a child with AHC.
Based on our published results in Dravet syndrome cohorts, probands carrying mosaicism with a higher MAF suggest a diseaserelated phenotype. 15 Six parents with lower level MAF of ATP1A3 mosaicism (<18.82%) were asymptomatic. One mosaic proband with higher level MAF of ATP1A3 mosaicism (44.19%) had milder phenotype compared with other heterozygous ATP1A3 AHC patients. We speculated that MAF of mosaicism was related to the severity of clinical manifestations. Parents carrying lower MAF mosaicism who were missed by panel NGS sequencing, whole-exome sequencing, or Sanger sequencing tended to be asymptomatic. 15,26 For the seven probands without ATP1A3 variant, the more likely hypothesis is that additional pathogenic genes are responsible for AHC. We are planning to carry out whole-exome sequencing to identify new pathogenic genes.
In conclusion, this is the first report to identify parental mosaicisms of ATP1A3 in AHC families. In this study, we showed that 7.5% (6/80) of presumed "de novo" ATP1A3 variants with AHC were discovered to be mosaic, and MAF of mosaicism was speculated to be related to severity of clinical manifestations. These findings have significant implications in genetic counseling for AHC patients.