Pathogenic Variants in RNPC3 are Associated with Hypopituitarism and Primary Ovarian Insufficiency

Purpose We aimed to investigate the molecular basis underlying a novel phenotype including hypopituitarism associated with POI. Methods We used NGS to identify variants in all pedigrees. Expression of Rnpc3/RNPC3 was analysed by in situ hybridization on murine/human embryonic sections. Crispr/Cas9 was used to generate mice carrying the p.L483F pathogenic variant in the conserved murine Rnpc3 RRM2 domain. Results We described 15 patients from nine pedigrees with biallelic mutations in RNPC3, encoding a specific protein component of the minor spliceosome, associated with a hypopituitary phenotype including severe GHD, hypoprolactinaemia, variable TSH deficiency and anterior pituitary hypoplasia. POI was diagnosed in eight of nine affected females whilst males had normal gonadal function. Additionally, two affected males displayed normal growth off GH treatment, despite severe biochemical GHD. In both mouse and human embryos, Rnpc3/RNPC3 was expressed in the developing forebrain including the hypothalamus and Rathke’s pouch. Female rnpc3 mutant mice displayed a reduction in pituitary GH content, but with no reproductive impairment in young mice. Male mice exhibited no obvious phenotype. Conclusion Our findings suggest novel insights into the role of RNPC3 in female-specific gonadal function and emphasize a critical role of the minor spliceosome in pituitary and ovarian development and function.

As whole exome variants were processed at different times and at several institutions/countries, the filtering parameters applied varied. In Pedigrees 1-5, the following variants were kept: quality score ≥20 and read depth ≥10, population frequency ≤1% (1000 genomes, NHLBI Exomes, ExAC), exonic changes other than synonymous, intronic splice sites (affecting first 7 bases of an intron), synonymous changes that are predicted to affect splicing as determined by MaxEntScan. In Pedigree 7, the following variants were kept: quality score ≥20 and coverage ≥4X, located outside of segmental duplications and simple repeats, variants with allele frequency ≤1% in the 1000 Genomes or NHLBI Exomes, non-synonymous variants with Polyphen2 4 scores >0.8 and SIFT 5 scores ≤ 0.05, splice-site (±15 nt), nonsense, non-stop, and small in-frame or frame-shift indels. In Pedigree 8, quality score ≥20 and coverage ≥10X, exonic variants, intronic variants within 10 bases from exon-intron boundaries, minor allele frequency ≤0.1% (gnomAD) were kept. In Pedigree 9, non-synonymous exonic changes and invariant splicing variants with a minor allele frequency ≤0.1% in the Genome Aggregation Database (gnomAD) were kept.
In all pedigrees, variants were assessed for segregations under the following inheritance models: Xlinked, de novo and autosomal recessive (homozygous and compound heterozygous).
Methodology of the whole exome sequencing performed in Pedigree 6 was reported previously. 6

Sanger sequencing confirmation
The variants identified by WES in each pedigree were confirmed by Sanger sequencing. Genomic DNA was isolated from leukocytes of peripheral blood by using standard protocols in each individual. The regions of interest were amplified by PCR using primers depicted in Table S1 and sequenced using BigDye Terminator V.3.1. Reference sequences were obtained from NCBI (reference sequence: NM_017619). The PCR amplification conditions for each reaction are available upon request.

Predicting the pathogenicity of novel variants in RNPC3
We predict that p.Pro474Leufs*10 is pathogenic as it at the same codon as a known pathogenic variant. The proline amino acid has a large side chain and therefore both, the known pathogenic change to Threonine (p.Pro474Thr), and our Leucine change will have a fundamental deleterious effect on the protein. Based on Deciphering Developmental Disorders (DDD) study (Decipher), any truncating changes beyond amino acid 481 in RNPC3 are predicted to be subjected to NMD (nonsense-mediated decay). Therefore, the variant p.Pro474Leufs*10 is predicted to be subjected to NMD or truncated protein missing 44 amino acids including an RNA-binding domain.
The figure below has pathogenic variants from the paper in red, exons in grey, coding sequence in yellow and protein domains in blue (made using geneious prime 2021.1.1): With respect to the p.Arg205* variant, there is precedence for nonsense variants to be pathogenic in this gene and the gene has a pLoF score of 0.19 which means that it is highly constrained and does not tolerate loss of function variants. The variant at position c.613C>T in intron 6/14 creates a premature stop codon at amino acid 205; the RNA transcript is likely to be subjected to nonsense-mediated decay; any transcripts escaping nonsense-mediated decay would encode for a severely truncated protein missing 312 amino acids and several functional domains including part of the PDCD7 binding domain and the entire m(7)G-capped U12 snRNA binding domain. Total length of the RNPC3 protein is 517 amino acids.
As for the splice site variant c.624+1G>T, it affects the highly conserved residue in the canonical splice site and is thus predicted to cause splicing defects with high confidence. The four in silico algorithms (SSF, MaxEnt, NNSplice and HSF) unambiguously agree that variant RNPC3 c.624+1G>T is predicted to cause loss of the canonical donor site and is likely to result in an in-frame exon skipping event of at least exon 6 of the NM_017619.3 transcript or likely more (Alamut Visual version 2.9.0, Interactive Biosoftware, Rouen, France). We used the Tayoun ACMG PVS1 guidelines to predict the effect of this on the reading frame 7 . RNA studies are needed to confirm the exact effect of c.624+1 on splicing.

In situ hybridization studies
Murine expression was analysed in C57Bl6 samples as previously described 8 using probes synthesized from cloned PCR fragments spanning nucleotides 17 to 820 for Rnpc3 and 524 to 1479 for Prmt6 of the corresponding coding sequences. Human brain and pelvic sections at different stages of embryonic development were provided by the Joint MRC/Wellcome Trust (grant# MR/R006237/1) Human Developmental Biology Resource (HDBR) (http://hdbr.org). A purified pCMV-SPORT6 vector containing full length human wild-type RNPC3 cDNA (IMAGE ID: 3873751) (Source Bioscience) was cut to make antisense and control sense digoxigenin-labelled RNA probes respectively. The in situ hybridization used to analyse human embryonic gene expression was performed as previously described. 9

Mouse studies
We used the Crispr/Cas9 technology to generate mice carrying the p.Leu483Phe variant in the conserved RRM2 domain of murine Rnpc3. We also generated mice harbouring the p.Pro350Arg variant in Prmt6. A plasmid encoding the Cas9 protein, a single stranded oligonucleotide containing the variant of interest, and sgRNA were injected in mouse C57Bl6 zygotes, as described previously. 10 The single stranded template also contained two synonymous variants in the guide region to avoid re-cutting once the targeting event had occurred. Two sgRNA were designed and independently injected to produce two different mouse strains for each of the genes, to discriminate between the effects of the desired variant, which would be common to both guides, and potential off-target effects of the guides, which would be different between the two lines ( Figure 2, Figure S2). Animal samples were grouped by litters which comprised both wild-type and mutant animals, allowing for randomization and appropriate littermate controls. To prevent subjective bias, samples were not identified by genotype when hormone level measurements were performed.

Growth curves in mouse studies
Weight measurements were taken blindly by animal technicians who were not involved in the study. Mouse weights were regressed against time (in weeks), segmented by genotype, within a linear mixed effect model where the random effect is set at the mice's individual variations using the following formula: Week is log-transformed to account for the fact that weight gain is rapid at first, but then plateaus as mice finish maturing. Mutant genotype is handled as a binary variable.

Mouse fertility studies
Eight-week-old Rnpc3 p.Leu483Phe/ p.Leu483Phe and wild-type littermate females were bred to C57Bl6 wild-type males. Each pair was housed in an individual cage for ten weeks. Litters were recorded at birth and removed before weaning stage.

Hormonal evaluation in patients
Hormonal assays were performed in the local laboratories and evaluated according to the normal ranges for each assay.
In pedigrees 1-5, The serum oestradiol and testosterone concentrations were measured using the ELECSYS Estradiol III and Testosterone II assay (Roche Diagnostics GmbH, Mannheim, Germany) by Modular analytics E170 autoanalyser. These assays employ the electrochemiluminescence immunoassay (ECLIA) method and have been standardized against CRM 6004a via isotope dilution-gas chromatography/mass spectrometry (ID-GC/MS). The coefficient of variation (CV) of oestradiol and testosterone were 8.5 % and 4.7 %, respectively.
Measuring range for oestradiol is 5-3000 pg/mL (defined by the Limit of Detection and the maximum of the master curve). Values below the Limit of Detection are reported as < 5 pg/mL.
Measuring range for total testosterone is 0.025-15.0 ng/mL (defined by the Limit of Detection and the maximum of the master curve). Values below the Limit of Detection are reported as < 0.025 ng/mL. Values above the measuring range are reported as > 15.0 ng/mL. In pedigree 6, serum oestradiol concentrations were measured by The ARCHITECT Estradiol assay (Abbott; Abbott Park, Chicago; IL, USA), a delayed one-step immunoassay using chemiluminescent microparticle immunoassay technology. The analytical sensitivity of the ARCHITECT Estradiol assay is around 10 pg/ml. Calibration range: 0 to 1000 pg/ml. The ARCHITECT estradiol assay was compared to isotope dilution-gas chromatography/mass spectrometry and gave a correlation coefficient of 0.99, with a mean slope of 1.09.
In pedigree 7, serum total testosterone concentrations were measured by chemiluminescent assay (Access Testosterone assay in Access 2 Immunoassay Systems). Dynamic measurement range is 10-1600 ng/dl. Inter-assay variation is 4.22-7.08%.
In pedigree 9, Serum testosterone was measured by using the Immulite 2000 which is an automated, random-access immunoassay analyzer with a solid-phase washing process and a chemiluminescence detection system. The interassay precision (CV%) is 13.7 with a calibration range: 20-1600 ng/dL. Testosterone measured by most automated and manual methods are reported within 20% of those measured by LC-MSMS and capable of distinguishing eugonadal from hypogonadal males 11 .
The diagnosis of central hypothyroidism is based on low serum thyroid hormone concentrations in the face of low/normal or slightly elevated TSH concentrations.
Cortisol deficiency was ruled out with normal early morning serum cortisol and ACTH concentrations, which are checked regularly in the patients during follow-up visits to the clinic.
Primary hypogonadism was diagnosed based on medical history, physical examination along with increased serum gonadotropin concentrations for age and pubertal stage in the face of low/undetectable serum oestradiol concentrations and supporting ultrasonographic findings 12 . In older patients, POI was diagnosed based on ESHRE guidelines and established diagnostic criteria 13 which are raised gonadotropins (FSH >40IU/L) taken three months apart on two occasions.

Pelvic imaging protocol
The pelvic ultrasonography (US) examinations were performed by dedicated paediatric sonographers using a transabdominal approach in all patients. Transabdominal images were obtained after night fasting with repleted bladder and the patient in a supine position. At the time of examination, the sonographer measured all adequately visualized pelvic organs, including left and right ovaries and uterus, in three dimensions (length, width and height in centimetres), endometrial thickness (double wall) in millimetres and additional measurements of any pertinent findings if present. Uterine and ovarian volumes were subsequently calculated using the prolate ellipsoid formula (volume=lengthxwidthxheightx0.523). All images were reviewed by a paediatric radiologist at the completion of imaging and a clinical interpretation, including verified pelvic organ measurements, was provided based on the images.
In pedigree 1-5, Siemens Acuson Antares Ultrasound system (Siemens AG, Muenchen, Germany) with convex US probes with a bandwidth 1-4 MHz were used depending on sonographer preference and patient body habitus. In pedigree 6, pelvic ultrasonography studies were performed with a high-end Samsung RS80A ultrasound device with Prestige technology (high-quality S-vision) using a CA1-7ASvue transducer (Samsung Electronics 2014). In pedigree 8, the Toshiba Aplio 500 TUS-A500 device was used.
Pelvic organ measurements have been evaluated according to the normative data for children and adolescents. 12 Table S1: Primers used for PCR amplification and Sanger sequencing of variants.   B) Two single guide RNAs were designed where the proline codon (ccc) was changed to arginine (cgc).  Females were weighted once/week from 1 to 10-week-old. Regression of the weight against time was performed to assess the effect of the Rnpc3 alteration. While the mean curve of mutant female weight is below that of control, the difference did reach not statistical significance (intercept difference, ns). Furthermore, the animal weight gain pattern during growth is similar in both genotypes (shape difference, ns). This is in agreement with the limited reduction in GH observed by RIA in mutant females (Fig 2).