Pathogenic variants in the human m6A reader YTHDC2 are associated with primary ovarian insufficiency

Primary ovarian insufficiency (POI) affects 1% of women and carries significant medical and psychosocial sequelae. Approximately 10% of POI has a defined genetic cause, with most implicated genes relating to biological processes involved in early fetal ovary development and function. Recently, Ythdc2, an RNA helicase and N6-methyladenosine reader, has emerged as a regulator of meiosis in mice. Here, we describe homozygous pathogenic variants in YTHDC2 in 3 women with early-onset POI from 2 families: c. 2567C>G, p.P856R in the helicase-associated (HA2) domain and c.1129G>T, p.E377*. We demonstrated that YTHDC2 is expressed in the developing human fetal ovary and is upregulated in meiotic germ cells, together with related meiosis-associated factors. The p.P856R variant resulted in a less flexible protein that likely disrupted downstream conformational kinetics of the HA2 domain, whereas the p.E377* variant truncated the helicase core. Taken together, our results reveal that YTHDC2 is a key regulator of meiosis in humans and pathogenic variants within this gene are associated with POI.


Introduction
Primary ovarian insufficiency (POI) affects 1% of women and is characterized by early loss of normal ovarian function and depletion of the ovarian reserve (1). POI is diagnosed in women who present before the age of 40 with estrogen deficiency, amenorrhea for at least 4 months, and evidence of raised follicle-stimulating hormone of more than 40 IU/L on 2 occasions (2). While most women with POI present with secondary amenorrhea, approximately 10% present with primary amenorrhea with or without absent puberty (3). POI is a condition that is associated with significant medical and psychosocial consequences (3). In a minority of women, the etiology is easily identified, with POI occurring secondary to ovarian damage arising from chemotherapy, radiotherapy, autoimmunity, or environmental agents (4). X chromosome abnormalities, such as Turner and fragile X syndromes, account for up to 5% of cases (5). Many of the remaining affected women may have an Primary ovarian insufficiency (POI) affects 1% of women and carries significant medical and psychosocial sequelae. Approximately 10% of POI has a defined genetic cause, with most implicated genes relating to biological processes involved in early fetal ovary development and function. Recently, Ythdc2, an RNA helicase and N6-methyladenosine reader, has emerged as a regulator of meiosis in mice. Here, we describe homozygous pathogenic variants in YTHDC2 in 3 women with early-onset POI from 2 families: c. 2567C>G, p.P856R in the helicase-associated (HA2) domain and c.1129G>T, p.E377*. We demonstrated that YTHDC2 is expressed in the developing human fetal ovary and is upregulated in meiotic germ cells, together with related meiosisassociated factors. The p.P856R variant resulted in a less flexible protein that likely disrupted downstream conformational kinetics of the HA2 domain, whereas the p.E377* variant truncated the helicase core. Taken together, our results reveal that YTHDC2 is a key regulator of meiosis in humans and pathogenic variants within this gene are associated with POI.

Results
YTHDC2 variants in women with early-onset POI. Exome sequencing of a cohort of women with unexplained early-onset POI revealed pathogenic variants in YTHDC2 in 3 women from 2 families. All 3 patients presented with a severe phenotype of absent puberty and primary amenorrhea and were diagnosed with POI at an early age (Table 1). Extended characterization for the etiology of ovarian insufficiency demonstrated a negative screen for adrenal and ovary autoantibodies, a normal 46,XX karyotype (excluding Turner syndrome), and negative FRAXA premutation analysis (excluding fragile X syndrome). Pelvic imaging revealed no ovaries or very-small-volume ovarian tissue (24). All young women progressed through puberty normally after hormone replacement therapy was commenced.
In one consanguineous family from eastern Gujarat in India, the 2 affected sisters (patients 1.1 and 1.2) had a homozygous single nucleotide substitution, c.2567C>G, p.P856R, in YTHDC2 at position chr5:g.113563983C>G (exon 20; RefSeq transcript NM_022828.5, GRCh38) ( Figure 1A). This change was the only biallelic nonsynonymous variant that cosegregated with the POI phenotype under an autosomal recessive mode of inheritance in the family, where parents are heterozygous and the affected siblings are homozygous ( Figure 1A and Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.154671DS1). The unaffected brother (WT homozygote) progressed through puberty and had a normal reproductive endocrine profile.
The detected p.P856R variant replaces a basic proline with polar arginine in the helicase-associated (HA2) domain of the YTHDC2 protein ( Figure 1B). This results in a protein predicted to be damaging in all in silico models studied (Combined Annotation Dependent Depletion score 27.0) (Supplemental Table 1). The affected amino acid is located within the HA2 domain and is highly conserved in YTHDC2 JCI Insight 2022;7(5):e154671 https://doi.org/10.1172/jci.insight.154671 across different species ( Figure 1C), and the variant is not present in public databases (gnomAD v.2.1.1 and v.3.1.2; >400,000 control alleles and >35,000 South Asian alleles) (25).
In a second family, a homozygous single nucleotide substitution, c.1129G>T, resulting in a stop-gain variant (p.E377*) in YTHDC2 at position chr5:g.113539100G>T (exon 8), was identified in patient 2 ( Table 1, Supplemental Table 1, and Figure 1A). Her unaffected parents originate from northern Pakistan. Three brothers and 1 sister were clinically unaffected (not tested). This stop-gain mutation is within the helicase core of the YTHDC2 protein and likely results in a truncated protein ( Figure 1B). All YTHDC2 variants were confirmed by Sanger sequencing (Supplemental Figure 1).
YTHDC2 is expressed in the human ovary and testis during meiosis. YTHDC2, a 3′-5′ DExH-box RNA helicase, is believed to be a key player in mammalian gonadal development and germ cell function. Disruption of Ythdc2 in mice has recently been shown to result in impaired meiosis and an ovarian insufficiency phenotype (ketu mouse mutation, ref. 26; and other Ythdc2 -/knockout mice, refs. 27,28). YTHDC2 binds to partner proteins MEIOC and XRN1 ( Figure 1D) (27,29,30). YTHDC2 is a "reader" of the N6-methyladenosine methyltransferase (m 6 a) complex, an mRNA modification with roles in posttranscriptional gene expression control.
Detected YTHDC2 variants do not affect RNA stability and splicing in peripheral leukocytes. Peripheral leukocytes were used as a biological system to investigate the effect of the YTHDC2 variants on the stability and splicing of the YTHDC2 transcript, as these cells express YTHDC2 (qRT-PCR and Genotype-Tissue Expression data; ref. 33) (Supplemental Figure 2A). RNA sequencing of leukocytes from the fresh blood of the 3 patients with YTHDC2 changes (c.2567C>G, p.P856R; c.1129G>T, p.E377*) and 7 patients with no pathogenic changes in YTHDC2 revealed YTHDC2 transcript levels to be similar between the 2 groups (Supplemental Figure 2B). This was confirmed on qRT-PCR ( Figure 3A). Furthermore, differential transcript analysis using DEXSeq bioinformatic analysis showed no differences in exon usage or exon splicing between patients and controls (FDR < 10%) (Supplemental Figure 2C). Taken together, these findings suggest that YTHDC2 variants do not affect RNA stability or splicing of the YTHDC2 transcript within peripheral leukocytes.
Global transcriptomic differences between patients with YTHDC2 variants and controls. We analyzed the peripheral leukocyte RNA-sequencing data set for global transcriptomic differences between patients  . Gene expression analysis revealed 13 differentially expressed genes in YTH-DC2 variants compared with controls (absolute log 2 FC > 0.5, p.adj < 0.05) (Supplemental Figure 2F). Genes of interest were validated by qRT-PCR. TRIM9, an E3 ubiquitin-protein ligase, and SYCP2L, a paralog of meiosis-specific synaptonemal complex protein 2 (SYCP2), were upregulated in patients with YTHDC2 variants compared with controls ( Figure 3A). Using thresholds of log 2 FC > 0.5 and p.adj < 0.05, there were no differentially expressed transposable elements (TEs) when comparing patients with YTHDC2 variants and controls (TEtranscripts).
The YTHDC2 missense change p.P856R does not affect the YTHDC2-MEIOC interaction. We assessed the cellular expression and localization of the pP856R missense variant in an in vitro cell system. Transfection of WT or p.P856R FLAG-tagged YTHDC2 with HA-tagged MEIOC into HeLa cells showed similar cytoplasmic localization ( Figure 3B). In most cells, WT or p.P856R YTHDC2 colocalized with MEIOC ( Figure 3B). Both WT and p.P856R YTHDC2 associated with MEIOC in co-immunoprecipitation assays (Supplemental Figure 3). These studies suggest that YTHDC2/MEIOC interactions are not significantly affected by the p.P856R variant.
The YTHDC2 p.P856R variant results in a conformational change of the native protein. Next, we characterized the p.P856R YTHDC2 missense change using in silico modeling ( Figure 4A). DExH-box helicases bind RNA (ATP-independent) and unwind this duplex RNA via progressive 3′-5′ translocation on single-stranded RNA (ATP-dependent) (34). The biological action of YTHDC2 is likely the product of coordinated RNA binding, RNA unwinding, and RNA helicase enzymatic activity mediated by its heterogeneous domains. The proline residue at 856 of the HA2 domain is not predicted to interact with RNA directly ( Figure 4A). However, the p.P856R change results in replacement of a nonpolar proline residue with a larger charged polar arginine, resulting in strong hydrogen bonds between residue R856 and residues Q905 and D855 (intrahelical R-D ion pair) and a higher number of protein-RNA hydrogen bonds ( Figure 4, B and C). These changes are predicted to lead a more stable, compact protein with a lower RoG that is less flexible and tolerant for the conformational changes needed for RNA processing (35,36) (Figure 4, C and D, and Supplemental Figure 5). Predicted changes in surface electrostatic potential could also change protein-protein interactions, although this change is considered of lower importance given the protein size ( Figure 4E and Supplemental Figure 5).

Discussion
Here, we provide evidence that the m 6 A reader YTHDC2 plays a role in the regulation of meiosis and human gonadal development and, for the first time to our knowledge, show that loss of function in YTHDC2 is associated with POI.
YTHDC2 is a DExH-box RNA helicase that binds to and unwinds RNA via 3′-5′ translocation on single-stranded RNA (26,27,34). YTHDC2 is a known reader of the m 6 a complex that, together with  (40), m 6 a-marked transcript degradation (41), and promotion of methylated mRNA expression and translation (41,42). Animal data suggest that the m 6 a complex plays an important role in the posttranscriptional regulation of mammalian gonadal development. Mice and zebrafish deficient in METTL3 and WTAP (writers) (43,44), ALKBH5 (eraser) (45), and YTHDC1 and YTHDF2 (readers) (46,47) demonstrate impaired gametogenesis. Similarly, mice deficient in YTHDC2 are infertile with small or atrophied gonads (26-28, 48, 49). Our data suggest that YTHDC2 is also important for human meiosis, with highest expression of YTHDC2 and MEIOC in the fetal ovary and adult testis at times of maximum sex-specific meiotic activity, localization of YTHDC2 to meiotic germ cells (oogonia) in the fetal ovary, and an increased expression of YTHDC2 with several known meiotic markers during stages of peak meiosis.
The mitosis-meiosis switch is a key and carefully timed transition during germ cell development. YTHDC2 is required for the mitosis-meiosis switch in animals. Ythdc2 and Meioc are expressed at the onset of and throughout the extended meiotic phase of prophase 1 in mice (30). Germ cells of Ythdc2 mutants enter early meiotic prophase but do not complete the mitotic to meiotic transition correctly, with persistent expression of mitotic markers (cyclin A2, cyclin D1), some expression of early meiotic markers (Stra8, Sycp3), and no expression of later meiotic markers (Dmc1, Spo11) (27,28). As early as preleptotene, germ cells appear to prematurely enter an aberrant metaphase-like state that results in germ cell apoptosis (26,30,50). The exact molecular mechanisms underlying the role of YTHDC2 in the mitosis-meiosis switch have not been fully elucidated, and we did not find impaired YTHDC2-ME-IOC interaction in patients with YTHDC2 variants. However, it is plausible that the YTHDC2-MEIOC complex binds to m 6 a-marked mRNAs to facilitate a correctly timed mitosis to meiosis progression in humans. This is supported by the finding in mice that YTHDC2 and MEIOC pull down mitosis-associated transcripts in WT germ cells, suggesting that the YTHDC2-MEIOC complex downregulates the mitotic cell cycle program (30). Indeed, given what is known about the function of other m 6 a readers, YTHDC2 may both promote the translation efficiency of required meiotic transcripts and degrade unwanted transcripts used earlier in the process -the latter possibly mediated via XRN1 (27) ( Figure  6). This may explain the higher expression of peripheral leucocyte expression of SYCP2L in patients with YTHDC2 variants compared with controls. SYCP2L, a synaptonemal complex protein, is an early meiotic marker, and defects in this gene have very recently been reported in association with POI (51). Ythdc2 mutant mice can express Sycp2l (leptotene) but not later meiotic markers (pachytene) (26). Abnormal YTHDC2 appears to prevent m 6 a-mediated degradation of mitotic and early meiotic markers such as SYCP2L, impeding normal meiotic progression and possibly explaining the higher expression of SYCP2L in patients with YTHDC2 variants. A faulty mitotic-meiotic switch represents a severe and early folliculogenesis block and may explain the small-volume, follicle-deplete ovaries seen in the 3 patients with early-onset POI and YTHDC2 variants.
The YTHDC2 variants identified in the 3 women with early-onset POI likely have different functional consequences. The biallelic stop-gain variant (c.1129G>T, p.E377*) might typically be predicted to result in nonsense-mediated decay; however, we found that this transcript appears to avoid nonsense-mediated decay in patient leukocyte studies and so is more likely to be translated into a protein that is truncated after the amino-terminal DExDc motif of the RNA helicase core. Notably, several Ythdc2 mouse models involve disruption within or near this highly conserved YTHDC2 helicase core. In addition, we identified a biallelic c.2567C>G variant in 2 sisters that is predicted to replace a basic proline with a polar arginine (p.P856R) in the HA2 domain of the YTHDC2 protein and that does not influence transcript splicing or stability. Investigation of this variant in vitro did not show clear disruption of YTHDC2 expression or YTHDC2/MEIOC colocalization. We hypothesized that the p.P856R variant may induce a conformational change affecting downstream function. The HA2 domain is found in many DExH-box RNA helicases; conserved motifs within the helicase core bind to RNA, promote ATP binding and hydrolysis, and -via their tertiary structure -result in the formation of an RNA-binding tunnel that facilitates efficient substrate loading (52)(53)(54). The RNA-binding domains of DExH-box RNA helicases may indirectly regulate helicase activity because their deletion reduces the rate of unwinding activity (54). Deletion of the HA2 domain results in undetectable RNA unwinding, although RNA binding is preserved (54). Confirmational kinetics may therefore play a particular role here, with the HA2 domain acting as a "wedge" to prevent reannealing of the newly separated single-stranded RNAs (54). In silico structural modeling of the WT and P856R mutant YTHDC2 protein predicted that substitution of proline for a charged arginine residue would result in hydrogen bonds that increase the stability of this region of the protein, making it more compact and altering the surface charge. This prediction was supported by simulation models, which showed that the p.P856R protein has reduced flexibility, as indicated by a lower RoG. These changes would impede the conformational flexibility required for functions such as RNA unwinding, as well as possibly changing electrostatic potential, which in turn could disrupt protein-protein interactions. Further approaches such as knock-in animal models and RNA-binding studies would be needed to study this in more detail.
As well as a postulated role in the mitosis-meiosis switch, Ythdc2 has recently been suggested to have a later role at pachytene of meiosis 1, and this is also suggested by our data (49). A male Ythdc2 mouse model deficient for YTHDC2 at pachytene demonstrated aberrant microtubule network changes and telomere clustering with germ cell apoptosis (49). Here, we gained further insight into the role of YTHDC2 in the developing human ovary from cluster and pathway enrichment analyses of differentially expressed genes between early-and late-stage ovaries throughout human prophase 1. A distinct subset of genes within this cluster related to piRNA biogenesis, including PIWIL2-4, HENMT1, PLD6, and MAEL (37)(38)(39). At pachytene, MYBL1 initiates transcription of piRNA biogenesis genes and piRNA precursors (55). piR-NAs then accumulate in cytoplasmic germ cell granules, where they are suggested to protect the integrity of germ cells against TEs and to regulate gene expression by directing PIWI proteins to cleave target meiotic mRNAs at the appropriate times (56,57). Downstream targets of murine MYBL1 include MILI (PIWIL2), MAEL, TDRD9, MOV10L1, DDX39A, and VASA (DDX4) (55). Consistent with a role in piRNA metabolism, TDRD9-and MOV10L1-deficient mice and VASA-and DDX3-deficient Drosophila demonstrate derepressed TEs, abnormal piRNA profiles, and meiotic arrest within germ cells (39,(58)(59)(60). Similarly, Ythdc2-deficient mice show downregulation of 1 piRNA precursor, impaired interaction with MIWI (murine PIWIL1) and MSY2 (murine YBX2), and downregulation of Miwi within germ cell granules (28,49). In humans, TDRD9 pathogenic variants are associated with male factor infertility (61), DHX37 variants with 46,XY gonadal dysgenesis (62), and here, YTHDC2 variants with POI. Most of these above-described genes are either piRNA biogenesis or DExH/D helicase genes; furthermore, the 3 latter genes are paralogs and share both a helicase core and HA2 domain. We speculate that DExH/D helicases have a role at pachytene, supported by our observation of perinuclear YTHDC2 staining and upregulation of DExH/D helicases, PIWIL genes, and other piRNA biogenesis genes at meiosis.
We did not find differential TE expression in the peripheral leukocytes of patients with YTHDC2 variants compared with controls, suggesting that any potential role of YTHDC2 in piRNA-mediated TE regulation is meiosis specific. However, in addition to dysregulated TE expression, PIWI knockdown in a Drosophila ovarian follicle cell line was found to indirectly regulate a small subset of protein-coding genes -including TRIM9, which was 3-fold upregulated in the PIWI knockdown model (63). We show that TRIM9 was more highly expressed in the peripheral leukocytes of patients with YTHDC2 variants, intimating that piRNA-mediated TE disruption during meiosis may result in enduring downstream genomic changes. Furthermore, a functioning piRNA pathway protects against TEs, and this protection is required for normal microtubule organization during gonadal development in Drosophila (64,65). We noted dual clustering of piRNA and microtubule processes at human meiosis and additionally noted the abnormal microtubule function in murine germ cells deficient in Ythdc2 at pachytene (49). Taken together, we speculate that YTHDC2, both an m 6 A-reader and a DExH helicase, may have dual meiotic roles in humans: m 6 A-mediated processing of mRNA transcripts in the mitosis-meiosis switch of early prophase and a later role in piRNA processing within germ cell granules at pachytene (Figure 6) (59). We suggest that human YTHDC2 variants result in several meiotic abnormalities similar to Ythdc2 mutant mice, including persistence of the mitotic program, a dysregulated transcriptomic and epitranscriptomic profile, microtubule-driven aberrant telomere clustering at pachytene, and apoptosis of faulty germ cells.
POI is a condition with high genetic heterogeneity that can pose challenges for clinicians tasked with providing tailored counseling to patients with this diagnosis. A greater understanding of the genetic mechanisms underlying POI has permitted more individualized diagnoses and advice on fertility preservation options and familial risk. Female meiosis and gonadal development require a series of carefully timed biological processes, such as meiosis initiation, synaptonemal complex formation, and homologous recombination. For the most part, POI-associated genetic variants have mapped to these processes in a relatively linear manner. Our association of YTHDC2 with human meiosis and POI highlights the emerging importance of epitranscriptomic regulation to these processes. It also requires that we widen the network of POI candidate genes to include those that function within posttranscriptional networks, such as m 6 A readers, RNA helicases, and piRNA biogenesis genes. Exploring the posttranscriptional regulation of gonadal development presents challenges as standard mRNA-sequencing techniques are not designed to detect all important signals from non-mRNA species, which is a limitation of our data. This issueis compounded by the inherent difficulty of interrogating a biological process occurring during fetal life within an inaccessible tissue, such as the ovary that atrophies by adult life. Using patient-derived material as surrogate experimental systems, such as used here, addresses these limitations to a degree. However, further studies using epitranscriptomic techniques such as m 6 A-Seq, ATAC-Seq, and small RNA sequencing will be required to fully interrogate the complex transcriptional and posttranscriptional landscape governing female meiosis and to better understand the genetic and epigenetic mechanisms underlying a clinical diagnosis of POI.

Methods
Participants. The affected women in this study were recruited as part of the Reproductive Life Course Study at University College London Hospital, London. Those with a known cause of ovarian dysfunction (e.g., abnormal karyotype, iatrogenic POI) were excluded.
Genetic analysis. Exome sequencing was performed using a customized Exome CG enrichment panel (Nonacus) followed by paired-end sequencing on a NovaSeq 6000 (S4, 2 × 151 bp, Illumina) or the BGIseq-500 platform (Beijing Genomics Institute). Reads were aligned against the human reference genome sequence (National Center for Biotechnology Information [NCBI] GRCh38) using the Burrows-Wheeler Aligner (66). Platypus software (v0.8.1) was used for variant calling using standard parameters (67). Multiple nucleotide polymorphisms were normalized and decomposed using vcflib (v1.0). Variant filtering was performed using Ingenuity Variant Analysis software (Qiagen) (Supplemental Figure 1). Variants were kept if they fulfilled the following criteria: read depth ≥ 5, maximum population frequency ≤ 1% (minor allele frequency < 0.01) (gnomAD), and segregation with an autosomal recessive model in the family, where parents are heterozygous for a variant and affected individuals are homozygous. Synonymous changes were excluded, unless they were associated with splice site loss or up to 7 bases into an intron or predicted to affect splicing using MaxEntScan. Potential changes in YTHDC2 were confirmed using Sanger sequencing (Supplemental Table 2) and compared to gnomAD (v2.1.1 and v3.1.2, https:// gnomad.broadinstitute.org, accessed December 2021) (68).
Single-cell RNA sequencing and spatial transcriptomics. Expression of YTHDC2 during human fetal ovary development was retrieved from single-cell RNA sequencing and spatial transcriptomic data sets from a previous study (31).
Immunohistochemistry. A human fetal ovary (16 wpc) was obtained from the HDBR and fixed in 4% paraformaldehyde before being processed, embedded, and sectioned for histology analysis including immunohistochemistry (IHC). The sections for IHC were produced at 4 μm thickness, and staining occurred on Leica Bond-max automated platform (Leica Biosystems). On the Bond-max, the sections underwent antigen retrieval to unmask the epitope (Heat Induced Epitope Retrieval protocol 1 for 20 minutes, pH 6, Bond-max protocol F). Endogenous activity was blocked with peroxidase with bond polymer refine kit (catalog DS9800) followed by incubation with rabbit polyclonal YTHDC2 primary antibody (MilliporeSigma, Merck KGaA, HPA037364, 1:200) for 1 hour. Then, the postprimary was applied to the sections (Bond polymer refine kit; catalog DS9800) and HRP-labeled polymer (Bond polymer refine kit; catalog DS9800). DAB (Bond polymer refine kit; catalog DS9800) chromogen solution was added to the sections to precipitate the locus of antigen-antibody interactions. Next, the sections were counterstained with hematoxylin (Bond polymer refine kit; catalog DS9800), washed in deionized water, dehydrated in graded alcohols, cleared in 2 xylene changes, and mounted for light microscopy. The stained slides were imaged on the Aperio CS2 Scanner (Leica Biosystems, catalog 5872) at 40× objective, and analysis was supported by QuPath (v.0.2.3) and Leica ImageScope (Leica Biosystems) software.
Plasmid construction. Plasmids containing WT or mutated YTHDC2 (P856R) were constructed. Primer sequences are provided in Supplemental Table 2 (all MilliporeSigma). All PCRs were performed using Phusion DNA polymerase (New England Biolabs). All constructs were Sanger sequenced to verify correct insertion of variants and tag epitopes. To build the pcDNA3.1-HsYTHDC2-1-1430-Flag plasmid (SC0827 plasmid), the WT HsYTHDC2-1-1430 gene sequence was first amplified from a plasmid provided by H. Sahara (Azabu University School of Veterinary Medicine, Fuchinobe, Japan), then subcloned into a pcD-NA3.1(-) plasmid using NheI and XhoI restriction sites, with a Flag-tag-encoding sequence added to the C-terminus by PCR. The P856R mutation was introduced by rolling circle amplification to produce pcD-NA3.1-HsYTHDC2-1-1430-P856R-Flag (SC0828 plasmid). The human cDNA sequence of MEIOC was obtained using nested PCR with RNA extracted from 12 and 13 wpc human fetal ovaries following reverse transcription. BamH1 and EcoR1 restriction sites were introduced at the 5′ and 3′ ends of the MEIOC sequence for ligation into a pkH3 plasmid containing an HA-tag.
Subcellular localization of YTHDC2 and MEIOC. HeLa cells (ATCC) were cultivated in Dulbecco's modified Eagle medium supplemented with 15% FBS (Gibco) in a 5% CO 2 humidified incubator at 37°C. Plasmid transfection was performed using Lipofectamine 2000 (Life Technologies) according to the manufacturer's instructions. HeLa cells were seeded on 170 μm coverslips (Marienfeld) 24 hours before transfection. After 48 hours of transfection, cells were washed twice with PBS with 500 μM MgCl 2 and 500 μM CaCl 2 (PBS-S) and fixed using 4% paraformaldehyde. Cells were permeabilized in PBS-S, placed in 0.5% Triton for 10 minutes. and incubated in blocking buffer (PBS-S, BSA 5%, Tween 0.1%) for 30 minutes at room temperature. Coverslips were incubated with primary antibodies (mouse, anti-Flag [clone 9A3], 1:200, Cell Signaling Technology; and rabbit, anti-HA [catalog ab9110], 1:500, Abcam) overnight in blocking solution at 4°C. Cells were then incubated with appropriate fluorescent dye-coupled secondary antibodies (Alexa Fluor AB_2340854 and AB_2313584, Jackson ImmunoResearch) for 1 hour at 37°C in blocking solution, stained for 5 minutes with DAPI, and washed once with PBS-S. Coverslips were mounted on glass slides using ProLong Gold antifade reagent (Thermo Fisher Scientific). Imaging was performed using an AX70 epi-fluorescence microscope (Olympus) equipped with a CoolSNAP MYO charge-coupled device camera (Teledyne Photometrics).
Molecular modeling and simulations of WT and p.P856R variant YTHDC2. The human YTHDC2 protein sequence was retrieved from the UniProt database (UniProt Q9H6S0) (48). Sequence searches were conducted to identify conserved domain signatures using PFAM (37). Blastp against the PDB database (82) identified the structures of the related RNA helicase proteins MLE (PDB 5AOR) and DHX36 (PDB 5VHE) as potential templates for modeling YTHDC2 and variants (Supplemental Figure 4) (52,83). Residues 181-421 to 575-1171 of YTHDC2 were finally considered for model generation. We first constructed a model of YTHDC2 based on DHX36, including the nucleic acid and cofactors (ADP + 2 Mg 2+ atoms). A second model based on MLE without cofactors was also constructed. Finally, we averaged the previously obtained models, by generating a set of 100 models using Modeller v9.25 (84). The best model according to Modeller's PDF and DOPE scores was selected and verified by Structural Analysis and Verification Server to evaluate its stereo-chemical quality (85,86). With this approach, structural representations of WT and p.P856R variant YTHDC2 in complex with nucleic acid and cofactors were obtained. For molecular dynamics (MD) simulations, WT and P856R mutant were solvated and ionized using CHARMM-GUI web server interface (87). The proteins were embedded in a water box of 120 × 120 × 120 Å and the system charges equilibrated with Na + and Clatoms up to a concentration of 0.15 M NaCl. The final systems contained approximately 155,000 atoms. All MD simulations were performed using the NAMD v2.14 package (88). The simulation systems were first relaxed with 10,000 steps of minimization followed by gradual heating from 0 to 310°K by running short MD simulations of 500 steps each cycle using the NVT ensemble. The simulation was switched to NPT conditions and was further equilibrated for 2.5 ns while constraining the protein backbone with an initial force constant of 10 kcal/(mol·Å 2 ) and gradually decreasing to 8, 6, 4, 2, 1, 0.5, 0.05 kcal/(mol·Å 2 ) every 250 ps of MD simulation. Finally, the systems were run without any constraints for 100 ns. Protein, cofactor, and ion atom types and parameters were described by the CHARMM36 force field (89). The van der Waals interactions were calculated applying a cutoff distance of 12 Å and switching the potential from 10 Å. A time step of 2 fs was used in the production phase, and Particle Mesh Ewald was employed for the treatment of long-range electrostatic interactions. The temperature was maintained at 310°K using Langevin dynamics. The Nose-Hoover Langevin piston method was used to control the target pressure (1 atm) with the LangevinPistonPeriod set to 200 fs and LangevinPistonDecay set to 50 fs. The last 100 ns of each simulation was extracted for trajectory analysis using Visual Molecular Dynamics v1.93 software (90). The effect of the arginine substitution at residue 856 was explored by measuring the RMSD of Ca atoms and root-mean square-fluctuation of individual residues. We also measured the RoG, solvent accessible surface area, and hydrogen bond interactions established between protein and nucleic acid as well as by residue R856 through simulation. The differences in electrostatic potential of surface electrostatic potential were calculated using the APBS plugin within PyMOL v2.5 (Schröedinger Inc.) and represented as color spectrum within 5 KT/e units.
Data availability. We do not have patient consent to release raw next-generation sequencing data. The data that support the findings of this study are available on request from the corresponding author (SMMB). RNA-sequencing data that support the findings of this study have been deposited in BioStudies under the accession number S-BSST693 at https://www.ebi.ac.uk/biostudies/.
Statistics. Statistical analyses were performed using GraphPad Prism v9.1.1 (GraphPad Software) and are described in the relevant sections. Student's t tests (2 tailed) were used for statistical testing. Correction for multiple comparisons was performed using the Holm-Šídák method. A P value of less than 0.05 was considered significant.
Study approval. The affected women recruited to this study provided written informed consent as part of the Reproductive Life Course Study at University College London Hospitals (ethical approval: NRES Committee London-Chelsea [16LO0682]). Human fetal samples were obtained with ethical approval (REC reference 18/LO/0822; 18/NE/0290) and written informed consent from the HDBR (http://www.hdbr.org).

Author contributions
SMMB, IDV, and PLQS wrote and edited the manuscript. SMMB, IDV, PN, and TB performed RNA-sequencing experiments and analyzed RNA-sequencing data. SMMB, PLQS, TB, DK, LAO, JPS, AG, and MI undertook DNA extraction and exome capture and obtained and/or analyzed exome-sequencing data. CFL performed protein modeling and analysis. FB, MTD, CFL, and DK contributed to discussions and/or data analysis. SMMB, CEB, MTD, and GSC provided clinical data. LB and GL performed cell line expression, localization, and co-immunoprecipitation experiments and provided expert input.
LGA and RVT contributed with analysis of single-cell RNA-sequencing and spatial transcriptomic data. BC and NM were involved with tissue selection and dissection (HDBR). OKO and TM performed histopathology experiments. GL, JCA, and GSC supervised the studies and contributed to the writing and editing of the manuscript.