Bidirectional disruption of GNAS transcripts causes broad methylation defects in pseudohypoparathyroidism type 1B

Significance

Pseudohypoparathyroidism type 1B (PHP1B) is a disorder caused by end-organ hormone resistance due to epigenetic alterations at GNAS, a complex locus encoding Gsα as well as other sense and antisense RNAs. The promoters of these transcripts are differentially methylated, allowing parent-specific expression. PHP1B patients show abnormal GNAS methylation patterns, but the underlying mechanisms are incompletely understood, partly because of the lack of mouse models. Here, we show that different PHP1B subtypes are caused by abnormal transcription, which disrupts parent-specific GNAS methylation. Furthermore, a retrotransposon insertion within GNAS, identified in a PHP1B kindred, impairs bidirectionally run-through transcription, thus perturbing GNAS methylation. Our findings support a model of transcription-mediated regulation of GNAS methylation, thus refining systematic searches for PHP1B-causing mutations.

Abstract

Pseudohypoparathyroidism type 1B (PHP1B) is a multihormone resistance disorder caused by aberrant GNAS methylation. Characteristic epigenetic changes at GNAS differentially methylated regions (DMRs), i.e., NESP, AS1, AS2, XL, and A/B, are associated with specific structural defects in different autosomal dominant PHP1B (AD-PHP1B) subtypes. However, mechanisms underlying abnormal GNAS methylation remain incompletely defined, largely because viable PHP1B mouse models are lacking. Using lymphoblastoid cells and induced pluripotent stem cells, we show that various GNAS methylation patterns in PHP1B reflect differential disruption of sense and antisense GNAS transcripts. In cases with broad GNAS methylation changes, loss of the maternal, sense-transcribed exon H/AS region impairs methylation of the AS1 DMR, which results in biallelic expression of an antisense transcript, GNAS-AS1, and NESP hypermethylation. In contrast, cases with normal AS1 methylation, including STX16 deletions, show monoallelic GNAS-AS1 expression and normal NESP methylation. The roles of these GNAS transcripts were confirmed by a retrotransposon in GNAS-AS1 intron 1, identified in an AD-PHP1B family. This insertion impaired exon H/AS transcription when located on the maternal allele, thus preventing the complete establishment of methylation at all maternal GNAS DMRs, leading to biallelic GNAS-AS1 transcription. However, maternal GNAS-AS1 transcription was profoundly attenuated, thus allowing only a small gain-of-methylation at NESP. Likewise, on the paternal allele, the retrotransposon attenuated GNAS-AS1 transcription, thus preventing complete NESP methylation. Our findings support a model of bidirectional transcription-mediated regulation of methylation at GNAS DMRs and will help to refine systematic approaches for establishing molecular defects underlying different PHP1B subtypes.

GNAS encodes the α-subunit of the heterotrimeric stimulatory G protein (Gsα), which mediates cAMP/PKA signaling at various hormone receptors, such as the parathyroid hormone (PTH) receptor (1). Due to epigenetic mechanisms regulating GNAS-derived transcriptions, Gsα is silenced in a tissue- and age-dependent manner (2). Maternal Gsα inactivating mutations cause PHP type 1A, characterized by hormone resistance, particularly PTH resistance, and physical abnormalities known as Albright’s hereditary osteodystrophy (AHO) (1, 3). In contrast, paternal Gsα inactivating mutations cause pseudopseudohypoparathyroidism, characterized by AHO features without hormone resistance (4). Epigenetic GNAS changes cause PHP type 1B (PHP1B) characterized primarily by PTH- and mild TSH-resistance, but only occasional AHO-like findings (1, 2).

Besides exons 1 to 13 encoding Gsα, GNAS comprises several alternative first exons, including NESP, H, XL, and A/B, that are transcribed in sense direction (1, 5) (Fig. 1A). In addition, an antisense transcript, GNAS-AS1, traverses exons H and NESP (1). These alternative exons are located within or adjacent to differentially methylated regions (DMRs), resulting in parental allele-specific expression. Exons H and NESP are only maternally expressed since their promoters are paternally methylated, while exons GNAS-AS1, XL, and A/B have maternally methylated promoters, thus allowing only paternal expression (1, 5). The Gsα promoter is not differentially methylated, but the unmethylated paternal A/B DMR, which gives rise to A/B transcription, likely silences in cis paternal Gsα expression in specific tissues (6, 7).

Fig. 1.

Biallelic GNAS-AS1 transcription is a specific transcriptional change in category 1 AD-PHP1B patients. (A) Schematic representation of GNAS and STX16 loci. Black and white boxes represent exons with associated numbers indicating exon numbers. Black arrows indicate transcriptions with directions, and the magenta curved arrow indicates the enhancer effect of STX16 intron 4 (5). Beige boxes correspond to CpG islands with red and white circles representing methylated and unmethylated CpGs, respectively, in DMRs. Locations of genetic causes in AD-PHP1B are also shown, with red, blue, and yellow boxes representing category 1, 2, and 3 cases, respectively. (B) RNAseq signals in the centromeric portion of GNAS (GRCh38, ch20:58,830,620-58,860,827). Publicly available data obtained from human GV oocytes, MII oocytes, hESCs (HUES62), and hiPSCs were visualized on IGV (8). Horizontal bars indicate microdeletions or single nucleotide variants identified in Category 1 (red) or Category 2 (blue) AD-PHP1B kindreds. (C–H) Epigenetic categorization of AD-PHP1B patients. (C, E, and G) MS-MLPA results of AD-PHP1B kindreds, including NESP- exon H/AS-region deletions (category 1, n = 2) (C), STX16 deletions (category 2, n = 7) (E), and a chromosomal inversion with a centromeric breakpoint between exons XL and A/B (category 3, 131/II-1, n = 1) (G) are shown. The horizontal axes show the locations of MLPA probes. Dashed horizontal lines indicate expected methylation levels in WT samples. (D, F, and H) Analysis of GNAS-AS1 transcripts in patient-derived LCLs. (D) RT-PCR result of GNAS-AS1 using LCLs from a category 1 patient with a maternal deletion affecting the exon H/AS region. LCLs from WT and a carrier in the same family who possesses the same deletion on the paternal allele were used as controls. (F and H) Sanger sequencing across heterozygous SNPs in the GNAS-AS1 transcript. (F) Sequencing across rs55995956 in genomic DNA (gDNA) or complementary DNA (cDNA) derived from LCLs of a category 2 patient (W3) with maternal STX16 deletions (9) showing monoallelic paternal GNAS-AS1 transcription. (H) Sequencing across rs3787497 using gDNA or cDNA derived from LCLs of a category 3 patient (131/II-1) with maternal chromosomal inversion with a centromeric breakpoint between the XL and A/B exons (10). See also SI Appendix, Fig. S2, which confirms SNP assignment to the nonmethylated maternal alleles.

Hypomethylation at the A/B DMR is shared by all PHP1B subtypes, while additional characteristic methylation patterns at other GNAS DMRs depend on the underlying genetic defect (1, 11). Our previous work classified these disease-causing methylation patterns into three categories: 1) broad methylation changes, including loss-of-methylation at all maternal GNAS DMRs and gain-of-methylation at the NESP DMR; 2) hypomethylation restricted to the AS2 and A/B DMRs; 3) hypomethylation solely at the A/B DMR (11). Genetic mutations located downstream (telomeric) of AS2 lead to the category 3 methylation pattern, whereas those blunting NESP transcription postzygotically, mostly STX16 deletions, constitute category 2 (11). The pathogenic mechanism underlying category 1, the most frequent subtype, remains unresolved, except for few patients with paternal uniparental disomy (UPD) of chromosome 20q.

Human and mouse GNAS undergo indistinguishable parent-specific epigenetic changes (12, 13). However, due to interspecies differences in genomic structure, viable mouse models of PHP1B are lacking, which has made it difficult to define the mechanisms underlying abnormal GNAS methylation caused by different mutations. In fact, mechanistic insights into the disease continue to rely on studying familial PHP1B subtypes and cell lines derived from affected patients and unaffected carriers (1, 14). Here, by analyzing lymphoblastoid cells (LCLs) established from patients in each epigenetic category and a previously not reported AD-PHP1B kindred, we defined two temporally distinct transcriptional changes specifically underlying category 1 cases, i.e., blunted sense transcription from the centromeric part of GNAS that leads, in the affected patients, to loss of all maternal GNAS methylation imprints during oogenesis, thus allowing postzygotically biallelic GNAS-AS1 transcription. These epigenetic and transcriptional changes are caused by the insertion of a retrotransposon into the maternally derived allele of GNAS-AS1 intron 1, thus defining a previously unrecognized cause for AD-PHP1B. Analysis of LCLs and induced pluripotent stem cells (iPSCs) generated from this kindred revealed that the inserted retrotransposon attenuates transcriptions bidirectionally, leading to the category 1 epigenotype. Due to the inserted retrotransposon that blunts GNAS-AS1 transcription, the affected patients show, besides broad maternal LOM, an incomplete methylation gain at the NESP DMR. These findings not only support our model of transcription-mediated mechanisms underlying the category 1 epigenotype but will help guide the search for structural GNAS defects in genetically unresolved PHP1B patients.

Results

Variants in the Exon H/AS Region Are Accompanied by Loss of All Maternal GNAS Imprints in Category 1 PHP1B Patients.

Maternal methylation imprints are established during oogenesis, whereas the paternal NESP DMR is methylated after fertilization (15–17). Given these temporal differences in epigenetic GNAS changes, we hypothesized two distinct events causing the broad methylation GNAS changes in category 1 PHP1B, namely failure to establish methylation at maternal GNAS DMRs during oogenesis and aberrant postzygotic gain-of-methylation at maternal NESP.

Loss of maternal methylation imprints had been attributed to loss of cis-acting transcripts that originate from the most centromeric GNAS portion and traverse several exons (5, 15, 16, 18, 19). Maternal exon NESP deletions are predicted to cause A/B and AS2 hypomethylation (category 2) (20, 21), as shown for postzygotic human embryonic stem cells (hESCs) (5, 11). By contrast, a microdeletion comprising a region telomeric of NESP that includes exon H and GNAS-AS1 exons 3/4, hereafter referred to as “exon H/AS region,” leads to broad methylation changes (category 1) (14). Furthermore, microdeletions or nucleotide variants identified in category 1 AD-PHP1B kindreds invariably involve the exon H/AS region (Fig. 1B and SI Appendix, Fig. S1A) (14, 22–25), indicating that this GNAS portion is particularly important for establishing the maternal methylation imprints. Therefore, we analyzed publicly available RNAseq data of human oocytes and postzygotic hESCs/hiPSCs, which revealed that the exon H/AS region is actively transcribed in oocytes, while NESP transcription is limited to postzygotic hESCs/hiPSCs (Fig. 1B and SI Appendix, Fig. S1A). Although the exon H/AS region-derived RNA is readily detectable also in postzygotic hESCs/hiPSCs, these transcripts were less abundant than those derived from the NESP promoter. Furthermore, maternal deletion of this region in hESCs did not cause GNAS methylation changes (SI Appendix, Fig. S1 B and C). These findings support the conclusion that the exon H/AS region is of particular importance during oogenesis rather than the postzygotic period. Since methylation at maternal GNAS DMRs, including the AS1 DMR (GNAS-AS1 promoter), is introduced before fertilization (17), an exon H/AS region-derived transcript most likely mediates the establishment of methylation at these DMRs during oogenesis. Thus, it is plausible that lack of or blunted exon H/AS transcription causes loss of all maternal GNAS imprints.

Biallelic GNAS-AS1 Transcription Underlies NESP Hypermethylation in Category 1 PHP1B Cases.

To explore the postzygotic gain-of-methylation at the NESP DMR, we focused on the GNAS-AS1 transcript, which is required, based on limited mouse studies, for establishing methylation at the paternal Nesp DMR (15, 19). Consistent with these findings, AD-PHP1B patients with a maternal deletion involving NESP and the exon H/AS region showed biallelic GNAS-AS1 expression, resulting in category 1 disease (22). To further confirm the role of GNAS-AS1 transcription in NESP hypermethylation, we employed LCLs derived from AD-PHP1B patients of each epigenetic category. Biallelic GNAS-AS1 expression was observed in LCLs from a category 1 AD-PHP1B patient with a maternal deletion of the exon H/AS region that does not include exon NESP (Fig. 1 C and D). In contrast, LCLs from category 2 cases with two different STX16 deletions, one category 3 case caused by a chromosomal inversion with a centromeric breakpoint between XL and A/B (9, 10, 26), and a WT control showed only paternal GNAS-AS1 expression, as revealed by sequencing RT-PCR products comprising informative SNPs (Fig. 1 E–H and SI Appendix, Fig. S2 A–D). These results supported the conclusion that relaxation of GNAS-AS1 imprinting, which results in biallelic GNAS-AS1 transcription, underlies in cis NESP hypermethylation.

Affected Patients in a New AD-PHP1B Kindred as Well as Their Unaffected Mother and Maternal Grandfather Show Atypical GNAS Methylation Changes.

Important confirmation for our model of transcriptional changes was obtained when encountering AD-PHP1B patients (kindred #182) with broad but unusual GNAS methylation changes. The female index patient (182/II-1) was first diagnosed with hypothyroidism when being evaluated at age 8 y for short stature, a rounded face, and mild central obesity; at this age, her fingers as well as left third and fourth metatarsals were short (SI Appendix, Fig. S3 A–C). By age 17 y, the shortening of these two metatarsals had become more prominent, and shortening of the right fourth metatarsal had become obvious (SI Appendix, Fig. S3 D–F). A head CT at age 16 y showed intracranial calcifications. Her PTH level was initially only slightly above the normal range but frankly elevated by age 10 y when treatment with calcium and calcitriol was started.

Assuming that the index patient (182/II-1) might be a sporadic case of PHP1B, we analyzed GNAS methylation by MS-MLPA, which showed a complete (A/B and XL) or almost complete (AS) loss-of-methylation at the maternal GNAS DMRs (Fig. 2A); AS2 methylation, measured by MSRE-qPCR (11), was completely lost (0.14%). These broad epigenetic changes at the maternal DMRs were similar to the methylation pattern observed in sporadic PHP1B cases (11). However, gain-of-methylation at the NESP DMR was only partial, a finding distinct from typical category 1 cases. This result prompted analysis of available family members, revealing that the younger brother (182/II-2) had GNAS methylation changes that were indistinguishable from those of 182/II-1 (Fig. 2A). 182/II-2 had normal calcium, PTH, and TSH levels at age 8 y and 4 mo; PTH was subsequently well above the normal range but thus far required only treatment with vitamin D (SI Appendix, Table S1). At the age of 12 y and 6 mo, he showed no evidence of short metacarpals and metatarsals. Notably, the mother of the affected children (182/I-2) and the maternal grandfather (182/0-2), who are both biochemically unaffected, showed incomplete loss-of-methylation at the paternal NESP DMR with normal methylation at all maternal DMRs (Fig. 2A and SI Appendix, Fig. S4). Microsatellite analysis of chromosome 20 showed that the two affected children shared the same GNAS allele with their unaffected mother and maternal grandfather, while this allele was absent in two siblings with normal GNAS methylation (Fig. 2B), making it likely that a maternal GNAS mutation led to AD-PHP1B in this kindred.

Fig. 2.

A novel category 1 AD-PHP1B kindred caused by a retrotransposon insertion in GNAS-AS1 intron 1. (A) MS-MLPA results of members in the previously not reported AD-PHP1B kindred 182. Two patients (182/II-1 and 182/II-2, purple) and two unaffected carriers (182/0-2 and 182/I-2, green italics) were analyzed. Note that NESP methylation levels in unaffected carriers (182/0-2 and 182/I-2) are significantly lower than noncarrier siblings (SI Appendix, Fig. S4). (B) Pedigree of kindred 182 and the results of microsatellite marker analyses for seven family members as well as one SNP for the two affected patients and their carrier mother. Fully informative markers are shown in bold. Small open circle, not available. (C) Identification of partially aligned reads from WGS. (Top) UCSC browser track showing the position of insertion (red triangle) identified in gDNA of patient 182/II-1. (Bottom) Screenshot from the Integrative Genomics Viewer (IGV) showing chimeric reads comprising the gray wild-type sequence and insert-specific colored sequences that are unrelated to GNAS (upper reads) and entirely gray wild-type sequences (lower reads). The colored reads indicate homology with multiple chromosomal regions. (D) Locations of PCR amplicons used in panels (E) and (F). (E) Inserted allele-specific PCR using genomic DNA of the members of kindred 182. Forward primer (a) is specific for the WT genomic sequence, while the reverse primer (b) is specific for the inserted retrotransposon. Code for each family member is shown underneath the pedigree; control, WT; H₂0, no DNA; DNA ladder shown on the Left. (F) PCR using genomic DNA of patient 182/II-1 with GNAS-specific primers (c and d) flanking the inserted sequence. The arrowhead indicates a PCR product larger than the expected size of 1,829 bp amplified only from gDNA of patient 182/II-1 that underwent amplicon sequencing.

Identification of a Retrotransposon Insertion in GNAS-AS1 Intron 1.

Given that maternal GNAS DMR methylation relies on transcription from the exon H/AS region, we predicted that a genetic defect might prevent this transcript in kindred 182 patients from reaching the maternal DMRs, thereby preventing methylation during oogenesis. Analysis of WGS data of 182/II-1 revealed several reads in the first intron of GNAS-AS1 that mapped partially to GNAS and partially to multiple other genomic regions (Fig. 2C) (27). These findings were reminiscent of those in another AD-PHP1B kindred in which a retrotransposon insertion was recently identified between GNAS exons XL and A/B (27). Therefore, we assumed that the portion of inserted DNA that was partly aligned to the GNAS region has multiple homologous sequences elsewhere in the genome. Using an amplicon specific for the sequence unrelated to GNAS (Fig. 2D, primers a and b), we showed that the intronic insertion was present only in the affected patients and unaffected carriers (Fig. 2 D and E).

By performing long-read sequence analyses of genomic DNA from the affected female (182/II-1) and her unaffected carrier mother (182/I-2), we were able to confirm the presence of an approximately 3 kb insertion located on haplotype 2 that 182/II-1 had inherited from 182/I-2 (SI Appendix, Fig. S5). The long-read sequences furthermore confirmed the GNAS methylation changes identified by MS-MLPA for patient 182/II-1, namely a complete loss-of-methylation at all the maternal DMRs and only a partial gain-of-methylation at the NESP DMR on the maternal allele, while 182/I-2 revealed normal methylation at the maternal DMRs and an incomplete loss-of-methylation at the NESP DMR on the paternal allele.

With primers flanking the predicted insertion (Fig. 2D, c and d), we were able to amplify the product of 1,829 bp derived from the wild-type allele, as well as a longer product that was expected from long-read sequencing of the patient 182/II-1 (Fig. 2F). Amplicon sequencing of the insert-specific PCR product revealed that the inserted sequence is an SVA retrotransposon of ~2.8 kb containing an internal repeat of GC-rich sequences and a 3’ terminal SINE-R domain homologous to the long terminal repeat of the human endogenous retrovirus, HERV-K10 (28). The SINE-R domain, which includes a polyadenylation signal close to the 3′-end, is highly homologous to the inserted SVA retrotransposons previously identified in two AD-PHP1B kindreds (27, 29). Based on the orientation of the retrotransposon in kindred 182, the polyadenylation signal is located at the telomeric end of the insertion (Fig. 3A and SI Appendix, Fig. S6 A and B).

Fig. 3.

Impaired maternal methylation caused by retrotransposon insertion in GNAS-AS1 intron 1. (A) Schematic showing the positions of retrotransposon insertions in the current (red bracket and arrow) and two previously reported AD-PHP1B kindreds (yellow arrows) (27, 29). Black and white boxes represent GNAS-AS1 and GNAS exons, respectively, with black arrows indicating transcriptional directions on each parental allele. Locations of an informative SNP (rs55995056) and qPCR amplicons are also shown. See SI Appendix, Fig. S7A for detailed locations. (B) Quantitative analysis of the transcript level traversing from the centromeric end toward the telomeric end of GNAS. DNase-treated RNA from LCLs of AD-PHP1B patients 182/II-1, 182/II-2, and 208/II-2 was reverse-transcribed with GNAS intron-specific primers. Transcript levels at the amplicons upstream (UP) and downstream (DOWN) of the retrotransposon were measured by qPCR, and the transcript levels at (DOWN) were normalized by those at (UP). (C) qPCR to assess NESP and exon H transcription in iPSCs from 182/II-2 and 182/I-2. For (B and C), each data point represents an independent experimental result. Intergroup comparisons were performed by one-way ANOVA with Tukey’s post hoc test. ns, not significant; **P < 0.01; ***P < 0.001. (D) Schematics showing the positions where the exon H/AS region-derived transcription was presumably blunted. (E and F) Allelic determination of the GNAS-AS1 transcript using a heterozygous SNP (rs55995056) in the two patients of kindred 182. DNase-treated RNA obtained from LCLs was reverse-transcribed using primer #1, and PCR was performed with primers #2 and #3. Sanger sequencing results of heterozygous SNP (rs55995056) are shown for 182/II-1 (D) and 182/II-2 (E). See also SI Appendix, Fig. S7 D–F for the determination of parental alleles.

Insertion of a Retrotransposon Into the Maternally Derived Allele Attenuates Transcription Derived from the Exon H/AS Region and Prevents Establishment of Maternal GNAS Imprints.

The 3′ terminal portion of the retrotransposon surrounding the polyadenylation signal was highly homologous to insertions previously identified in two unrelated AD-PHP1B kindreds (SI Appendix, Fig. S6A) (27, 29), which blunt an exon H/AS region-derived transcript that traverses all maternally methylated DMRs, i.e., AS, XL, and A/B DMRs, as we have recently shown (11). However, in those previously reported kindreds, retrotransposons were inserted between XL and A/B DMRs (Fig. 3A and SI Appendix, Fig. S7A), and hypomethylation in genomic DNA of affected patients was restricted to the A/B DMR (category 3) (27, 29). To test whether distinct GNAS methylation patterns reflect the position at which the transcription from the exon H/AS region is impeded, we analyzed LCL-derived RNA obtained from patients 182/II-1 and 182/II-2, and from patient 208/II-2 with the previously described more telomeric transposon insertion (27). By normalizing transcript levels of a telomeric (DOWN) amplicon relative to a centromeric (UP) amplicon, we showed that the DOWN/UP ratio was decreased in LCLs from both patients in kindred 182 but similar to WT in 208/II-2-derived LCLs (Fig. 3B). To assess the levels of transcripts derived from NESP and exon H, we reprogrammed LCLs from 182/I-2 and 182/II-2 to generate iPSCs because those exons are actively transcribed in pluripotent stem cells (Fig. 1B and SI Appendix, Fig. S1A). Patient (182/II-2)-derived iPSCs showed significantly lower expression levels of NESP and exon H (Fig. 3C) compared to WT hESCs and a carrier (182/I-2)-derived iPSCs. These findings were consistent with a transcription-blunting effect of the retrotransposon when inserted into GNAS-AS1 intron 1, thus supporting a model in which a prematurely truncated transcript from the exon H/AS region underlies loss-of-methylation at maternal GNAS DMRs (Fig. 3D).

Consistent with hypomethylation at all maternal GNAS DMRs, transcription from XL, A/B, as well as GNAS-AS1 promoters was biallelic in LCLs derived from both 182 patients (Fig. 3 E and F and SI Appendix, Fig. S7 B–F). To further confirm maternal derepression of GNAS-AS1 transcription, we reverse-transcribed DNase-treated RNA using a primer within the inserted retrotransposon (SI Appendix, Fig. S7A), which demonstrated biallelic expression in 182/II-1 and 182/II-2 (SI Appendix, Fig. S7G). These results supported the conclusion that maternal NESP methylation results from derepressed maternal GNAS-AS1 transcription.

GNAS-AS1 Transcription Was Attenuated Due to the Retrotransposon Insertion.

Last, we explored the mechanism underlying the incomplete gain-of-methylation at the NESP DMR in both patients of kindred 182, an epigenetic pattern distinct from the complete gain-of-methylation at this DMR in all other category 1 cases (1, 11, 14). In addition, NESP methylation levels were partially reduced in the unaffected carriers 182/I-2 and 182/0-2, in whom the retrotransposon insertion had been inherited paternally (Fig. 2B). Thus, we assumed that a specific underlying mechanism in 182 kindred hampers GNAS-AS1 transcription-mediated NESP methylation, irrespective of whether the retrotransposon insertion is located on the maternal or the paternal allele.

There was no evidence for aberrant splicing of GNAS-AS1 transcripts caused by the retrotransposon insertion (SI Appendix, Fig. S8). However, in LCLs from 182/II-1 and 182/II-2, maternal GNAS-AS1 transcription (nucleotide A at rs55995056) was weaker than that from the paternal allele (nucleotide G) (Fig. 3 E and F). In addition, LCLs and iPSCs from the unaffected carrier (182/I-2) with the paternal retrotransposon insertion showed a lower GNAS-AS1 transcript level than WT (Fig. 4 A and B). Furthermore, in a luciferase reporter assay using human embryonic stem cells, retrotransposon insertion in the same location and orientation as in kindred 182 revealed considerably attenuated reporter activity (Fig. 4C), indicating that the insert blunts GNAS-AS1 transcription even though the polyadenylation signal was inverted. Therefore, “partial” NESP methylation changes in kindred 182 further support the conclusion that intact GNAS-AS1 transcription is required for establishing NESP methylation (Fig. 4D).

Fig. 4.

Blunted GNAS-AS1 transcription due to the retrotransposon insertion results in partial NESP methylation changes. (A and B) qPCR measurement of GNAS-AS1 transcript levels in LCLs (A) and hESCs/hiPSCs (B). WT LCLs (A) and WT hESCs (B) were used as controls. RNA was reverse-transcribed using oligo-dT primers, and GNAS-AS1 expression levels were quantified by qPCR. Each data point represents an independent experimental result. (C) Luciferase assay using transfected hESCs. A promoterless vector (Top) was used as a negative control. WT construct (Middle) contains the GNAS-AS1 promoter (prom) exon 1 and a portion of intron 1 surrounding the inserted retrotransposon. The retrotransposon-inserted construct (Bottom) includes the entire retrotransposon sequence identified in kindred 182 in the same orientation and location as in the patient. Firefly counts were normalized using Renilla counts. (D) Schematic representation of altered GNAS and GNAS-AS1 transcripts on the retrotransposon-inserted alleles in kindred 182. Blue vertical bars with a red arrowhead show the position of the inserted retrotransposon; red crosses indicate where transcription is blunted. Black arrows indicate transcripts and their directions, and dotted arrows mean attenuated transcriptions. Intergroup comparisons were performed by one-way ANOVA with Tukey’s post hoc test. *P < 0.05; **P < 0.01; ****P < 0.0001. cent, centromeric; tel, telomeric.

Discussion

Current findings collectively support a model of two temporally distinct but linked mechanisms underlying category 1 PHP1B cases (Fig. 5A). The first mechanism hampers the establishment of methylation at all maternal GNAS DMRs, i.e., AS1, AS2, XL, and A/B, that normally requires transcription from the exon H/AS region during oogenesis. AD-PHP1B patients with deletions comprising the exon H/AS region or with the retrotransposon insertion herein (kindred 182) thus truncate this transcript before reaching all maternal DMRs (14, 22, 23); the fact that two unrelated AD-PHP1B kindreds with retrotransposon insertions between the XL and A/B DMRs show isolated A/B hypomethylation (category 3) further supports this model (27, 29). Although the exon H/AS region is not well-conserved between humans and mice, murine PHP1B-like models with a deletion of the second Nesp exon or insertion of a polyadenylation cassette into the second Nesp intron recapitulated loss-of-methylation at all maternal DMRs (15, 18, 19, 30), suggesting that an evolutionarily conserved transcription-dependent mechanism dictates establishment of methylation at the maternal GNAS DMRs during oogenesis.

Fig. 5.

Unified mechanistic model of aberrant methylation at GNAS DMRs and AD-PHP1B categories. (A) During oogenesis, methylation at maternal DMRs depends on transcription from the exon H/AS region (squared H). Disrupted exon H/AS region-derived transcription causes categories 1 or 3 methylation patterns, depending on the location where the transcription is blunted, except for category 1 cases with paternal UPD20. After fertilization, the maternal AS2 and A/B DMRs undergo a second wave of demethylation and remethylation, in which a NESP-derived transcript is required that is activated by an enhancer at STX16. Disrupted NESP transcription selectively impairs methylation at the AS2 and A/B DMRs, thus leading to category 2 methylation pattern. (B) Locations of genetic defects underlying AD-PHP1B. Red, blue, and yellow colors indicate causes in category 1, 2, and 3 cases, respectively.

The second mechanism derepresses GNAS-AS1 transcription from the maternal allele, thus explaining NESP gain-of-methylation in category 1 cases. The NESP DMR is unmethylated on both alleles in the zygote and undergoes methylation on the paternal allele postzygotically in unaffected humans and mice because of paternal GNAS-AS1 transcription (15–17, 31). The resulting monoallelic NESP transcription after fertilization reestablishes maternal methylation at AS2 and A/B (5, 11). On the other hand, the maternal AS1 DMR is methylated during oogenesis, and GNAS-AS1 thus normally shows monoallelic paternal transcription. In the current study, we provided compelling evidence for the conclusion that GNAS-AS1 transcription mediates postzygotic NESP methylation in humans. First, biallelic GNAS-AS1 transcription is specific for category 1 cases, resulting in biallelic methylation at the NESP DMR. Second, attenuated GNAS-AS1 transcription is associated in kindred 182 with incomplete NESP methylation changes in affected patients (partial maternal gain-of-methylation) and unaffected carriers (partial paternal loss-of-methylation). In addition, patients in another category 1 kindred with a deletion limited to the exon H/AS region showed maternal derepression of GNAS-AS1 transcription resulting in biallelic NESP methylation (Fig. 1 C and D) (14). Consistent with these findings, patients with even smaller deletions or a single nucleotide variant in the exon H/AS region showed the category 1 epigenotype (24, 25).

The NESP and AS1 (GNAS-AS1 promoter) DMRs are likely to be unmethylated on both GNAS alleles in the zygotes of category 1 patients. In fact, NESP hypermethylation in these cases indicates that postzygotic GNAS-AS1 transcription-mediated methylation of the NESP DMR probably precedes NESP promoter activation. Once methylation at the NESP DMR is established, it prevents activation of the NESP promoter by the embryonic enhancer at STX16 (32). Consequently, the GNAS-AS1 transcript normally prevents an effect of the STX16 enhancer, which is biallelically active, on the paternal NESP DMR, thus limiting NESP expression to the maternal allele. However, the molecular mechanisms underlying methylation at the NESP DMR through the GNAS-AS1 transcript remain unknown. Although the Nespas (the murine counterpart of GNAS-AS1) transcript is reported to interact with polycomb repressive complex 2 in mouse ES cells, further studies are required to elucidate the human molecular machinery that mediates NESP methylation. Nonetheless, the current findings from AD-PHP1B patients collectively support a model in which the temporal regulation of methylation at the different maternal GNAS DMRs underlies broad epigenetic changes in category 1 cases (Fig. 5A).

Combined with our previous work defining the mechanism underlying category 2 AD-PHP1B cases, i.e., defective postzygotic activation of NESP promoter by its enhancer at STX16 (5, 11), we propose that each GNAS methylation category is caused by a specific pathogenic mechanism (Fig. 5 A and B). Thus, genetic alterations that truncate exon H/AS region transcripts at different positions determine the extent of hypomethylated maternal DMRs, i.e., loss of all maternal imprints (category 1) or isolated A/B hypomethylation (category 3). In contrast, category 2 cases are likely to result from postzygotic disruption of NESP transcription that is driven by the STX16 enhancer (5, 32), which is required for remethylation at the two maternal DMRs after undergoing a second wave of postzygotic demethylation, i.e., at the A/B and AS2 DMRs, but it is dispensable for methylation at AS1 and XL (5, 11). Therefore, category 2 cases show hypomethylation limited to the A/B and AS2 DMRs.

Taken together, a mechanism-based approach focusing on the specific genomic locations that cause, when modified, the different GNAS methylation patterns can help establish the molecular diagnosis in PHP1B patients. Category 1 cases should be screened for inherited or de novo mutations in the region between exon H and the AS1 DMR, and paternal UPDs involving the GNAS locus (1), but most are currently unresolved genetically. The remaining cases may have heritable defects involving either STX16 intron 4 (5) or are restricted to the NESP promoter/exon (20, 21) (category 2), or occur between AS2 and A/B DMRs (category 3) (10, 11, 27, 29). Because of the improved definition of genotype-epigenotype correlations, it is now feasible to readily search for genetic mutations underlying PHP1B subtypes, an approach potentially applicable to other imprinting disorders.

Methods

DNA Methylation Analyses.

Genomic DNA (gDNA) was extracted from peripheral blood leukocytes or buccal mucosa by proteinase K digestion followed by phenol-chloroform extraction or by the QIAamp DNA Mini kit (QIAGEN). Methylation-sensitive multiplex ligation-dependent probe amplification (MS-MLPA) was performed using the SALSA MS-MLPA Probemix ME031 GNAS (MRC Holland) following the manufacturer’s protocol. Fragment analysis was conducted in the MGH DNA Core using ABI3730xl Genetic Analyzer, and the data were analyzed using GeneMapper v6.0 software.

MSRE-qPCR for AS2 DMR was performed as we described previously (5, 33). Briefly, twenty ng of gDNA was incubated with 5 U of HpaII (New England Biolabs) at 37 °C for 2 h, and qPCR was run using KOD SYBR (TOYOBO) on Quantstudio3.0 (Thermofisher). The relative amount of HpaII-digested genome was calculated based on a calibration line generated from the Ct values of serially diluted undigested samples.

Bisulfite PCR and amplicon sequencing were performed as we previously described (5). Genomic DNA was bisulfite-converted using the EZ DNA Methylation-Gold Kit (ZYMO RESEARCH), and the A/B DMR was PCR-amplified using KOD One (TOYOBO). Purified PCR products were subjected to NGS analysis at the MGH DNA Core. Alignment of FASTQ files and calculation of methylation levels were performed using bwameth in a Galaxy Project environment.

Analysis of RNAseq Data.

The raw FASTQ files (GV oocytes, SRA SRR18175278; MII oocytes, SRA SRR18175251; hESCs, SRA SRR2038476; hiPSCs, SRR26674265) were obtained from the SRA database (https://www.ncbi.nlm.nih.gov/sra). After removing the adaptor sequences on the Galaxy server 24.2.rc1, using Cutadapt, sequences were aligned to the reference genome (GRCh38) using RNA STAR and visualized on Integrative Genomics Viewer (IGV, Ver2.19.1) (8).

HESCs Culture and Genome Editing.

HESCs (HUES62) were obtained from the Harvard Stem Cell Research Institute and were maintained in mTeSR plus medium (STEMCELL Technologies). Genome editing of hESCs was performed as we previously described (5). Briefly, the hESC culture medium was changed to mTeSR1 plus containing 1× CloneR2 (StemCell Technologies) 2 h before electroporation. The ribonucleoprotein (RNP) complex, including Alt-R SpCas9-GFP V3 (IDT), one pair of Alt-R sgRNAs targeting the exon H/AS region (IDT), Alt-R Cas9 Electroporation Enhancer (IDT), and P3 solution with supplement (Lonza), was prepared, and hESCs were mixed with the RNP complex. Electroporation was performed using the program CA-137 of a 4D Nucleofector (Lonza). On the following day, the top 10% GFP-positive hESC population was single-cell sorted into 96-well plates using FACS AriaII (BD Biosciences). To determine the parental origin of the deleted allele, we used a heterozygous SNP (rs3787497; A/G) within exon H, in which A was maternal based on sequencing of the exon H transcript (5).

Analysis of GNAS Transcripts in Patient-Derived Cells.

Patient-derived LCLs were established at the MGH Molecular Neurogenetics Unit Tissue Culture Core using a previously published protocol (34) and maintained in RPMI1640 (Thermofisher) with 10% fetal bovine serum with antibiotics solution (CORNING). RNA was extracted using the RNeasy Mini Kit (QIAGEN). Extracted RNA was treated with DNase I (ThermoFisher) following the manufacturer’s protocol. DNase-treated RNA was reverse-transcribed using ProtoScript II Reverse Transcriptase (New England Biolabs).

To detect nascent GNAS-AS1 transcripts, we reverse-transcribed DNase-treated RNA using an intronic primer located between exons NESP and H, thus allowing detection of the unspliced GNAS-AS1 transcripts in the exon H/AS region (Figs. 1 F and H and 3 E and F and SI Appendix, Fig. S2 A–D). We amplified this cDNA using intronic primers so that the resulting amplicon comprises at least one informative SNP, rs55995056 or rs3787497, in the exon H/AS region. PCR products were purified using the QIAquick PCR Purification Kit (QIAGEN), and Sanger sequencing was performed at the MGH DNA Core. Parental origin of the transcript was identified either by sequencing of rs55995056 or rs3787497 using CpG methylation-sensitive HpaII-digested gDNA, in which unmethylated maternal allele of the exon H/AS region was digested (for SI Appendix, Fig. S2 A–D), or by sequencing of rs55995056 of the mother’s gDNA (for Fig. 3 E and F and SI Appendix, Fig. S7 D–F).

For the detection of spliced GNAS-AS1 transcripts, RNA was reverse-transcribed using oligo dT primers. For exon H/AS region-derived transcript detection, intron-specific primers were used for reverse transcription (UP and DOWN; see Fig. 3A and SI Appendix, Fig. S7A), and cDNA was amplified by qPCR at centromeric (UP amplicon) and telomeric (DOWN amplicon) of the inserted retrotransposon. To estimate the transcription-blunting effect of the inserted retrotransposon, we calculated the relative transcript levels at the DOWN amplicon using the levels at the UP amplicon for normalization. qPCR analyses were done using Quantstudio3.0 (Thermofisher).

Description of Kindred 182.

The pedigree and the clinical findings of the index patient are described in Fig. 2B and SI Appendix, Tables S1 and S2, respectively. The female index patient (182/II-1) was 8 y old when she was evaluated for short stature, a rounded face, and mild central obesity. She attended regular school but showed some delay in cognitive processing and had a history of gross motor delay. She had no history of seizures, muscle twitching, or cramping. Physical examination revealed short third and fourth left metatarsals, and all fingers were relatively short, which was confirmed by X-ray (SI Appendix, Fig. S3 A–F). A year earlier, she had been diagnosed with primary hypothyroidism, for which levothyroxine had been prescribed. PTH levels were slightly above the normal range when first evaluated, which gradually increased until calcitriol was started at age 12 (SI Appendix, Table S1). Total calcium and phosphate levels were within the normal range. Growth hormone stimulation tests showed blunted responses to clonidine and arginine (SI Appendix, Table S2). At the age of 16 y, when evaluated for persistent headaches with normal blood pressure, a head CT showed some intracranial calcifications, particularly surrounding the superior sagittal sinus. Consistent with his GNAS methylation defects, PTH levels increased in patient 182/II-2 after age 10 y (SI Appendix, Table S1); his head CT has not shown evidence for intracranial calcifications.

Genetic Analysis of Kindred 182.

For linkage analysis, microsatellite markers were selected based on our previous studies (23, 27) and analyzed at the Mission-Driven Service Core of the Massachusetts General Hospital. MLPA analysis was performed using the SALSA MS-MLPA Probemix ME031 GNAS (MRC Holland), which showed no evidence for deletions or duplications in the GNAS or STX16 regions. Whole-genome sequencing (WGS) was performed at the BGI, Beijing, China. FASTQ files were aligned to GRCh38 on Integrative Genome Viewer. Analysis of WGS data of 182/II-1 revealed several reads for which portions mapped of the first intron of GNAS-AS1, while the remaining nucleotide sequence mapped to several different chromosomes (Fig. 2C). To amplify specifically the inserted sequence in kindred 182, we used a primer located in GNAS-AS1 intron 1 (a) and an insertion-specific reverse primer (b) (Fig. 2D). Subsequently, the entire inserted region was amplified using primers flanking the predicted insertion (c and d) (Fig. 2D). Amplicon sequencing of the PCR product containing the entire insert was performed at the MGH DNA core. Primer sequences are provided in SI Appendix, Table. S3.

Targeted Long-Read Sequencing (T-LRS).

T-LRS was performed using Adaptive Sampling on an Oxford Nanopore Technologies (ONT) PromethION 24, as previously described (27). Briefly, approximately 1.5 μg of genomic DNA from individuals 182/II-1 and 182/I-2 (Fig. 2) was used to make sequencing libraries using the ONT SQK-LSK114 kit (ONT) following the manufacturer’s instructions. Target regions for sequencing were defined as chr20:53,800,000–63,950,000 (GRCh38 coordinates) for the GNAS region. Approximately 600 ng of prepared libraries were loaded onto a R10.4.1 flow cell and run for 24 h on a PromethION running MinKNOW software version 24.06.10. After 24 h, the flow cell was stopped. FASTQ files were generated from the raw sequencing data using Dorado 0.9.0 (ONT) using the superior model with 5mCG and 5hmCG methylation model 5.0.0 (ONT). Reads were aligned to GRCh38 using minimap2 (35), variants were called using Clair3 (36), followed by phasing using LongPhase (37). Aligned reads were visualized using IGV (8).

Generation of iPSCs.

LCLs from a patient (182/II-2), a carrier (182/I-2), and a patient from another AD-PHP1B kindred (208/II-1) were pooled, as recently described (38), before reprogramming was performed at the Harvard iPS Core Facility (Dr. Laurence Marie Daheron) using the Epi5TM Episomal iPSC reprogramming kit (Thermo Fisher) (SI Appendix, Fig. S9A). Briefly, 500,000 cells were transfected with 1 µL of Epi5 reprogramming vectors and 1 µL of Epi5 p53 and EBNA vectors using the Neon transfection. The setting for the Neon transfection was 1,350v, 30 ms, 1 pulse. Two days later, the cells were plated on irradiated MEFs and hESC medium (DMEM/F12, 20%KOSR). iPS colonies were picked three weeks later based on morphology. The expression of pluripotency markers (OCT4, NANOG, SSEA4, and Tra-1-60) was confirmed by immunocytochemistry (ICC), as previously described (39) (SI Appendix, Fig. S9B). Briefly, cells were fixed in 4% paraformaldehyde, permeabilized with PBS/0.1% Triton X-100, and blocked with 4% donkey serum/PBS. Primary antibodies, OCT4 (Abcam, ab19857, 1:100), NANOG (Abcam, ab21624, 1:50), SSEA4 (Millipore, MAB 4304, 1:200), and Tra-1-60 (Millipore, MAB 4360, 1:200), diluted in 4% donkey serum/PBS, were added to the cells and incubated for 1 h at room temperature. Secondary antibodies, anti-rabbit IgG-488 (for OCT4 and NANOG, Thermo Fisher, A21208), anti-mouse IgG1-488 (for SSEA4, Thermo Fisher, A31571), and anti-mouse IgM-555 (for Tra-1-60, BD Biosciences, 560850), all diluted in PBS (1:1,000), were added and incubated for 1 h at room temperature. After staining with DAPI, images were obtained using a fluorescent microscope. Pools of iPSCs were single-cell sorted into 96-well plates on Bigfoot cell sorter (Invitrogen) and individual clones were genotyped using primers that specifically detect the retrotransposon in kindred 182 (Fig. 2E). Primers specific for the Y chromosome (DYS14) were used to distinguish cells from 182/II-2 (male) and 182/I-2 (female) (SI Appendix, Fig. S9A).

Luciferase Assay in hESCs.

Luciferase reporter plasmids were constructed by inserting PCR-amplified promoter and the first exon of GNAS-AS1 with or without the retrotransposon in kindred 182 before the firefly coding sequence. Firefly and Renilla plasmids were cotransfected into hESCs using lipofectamine 3000 (ThermoFisher). Forty-eight hours after transfection, luciferase counts were measured using a Dual-Glo luciferase assay kit (Promega) and ENVISION (PerkinElmer). Firefly counts were normalized by Renilla counts.

Sex as a Biological Variable.

The sex of the patients and unaffected controls was not considered as a biological variable in this study.

Statistical Analysis.

Statistical analyses were performed using GraphPad Prism 9 software. One-way ANOVA with Tukey’s multiple comparison test was used for multiple-group comparisons. P values < 0.05 were considered statistically significant.

Study Approval.

Written informed consent was obtained from each patient and unaffected control; IRB Protocol No.: 2001P000648. All experiments were approved by the Institutional Biosafety Committee of Mass General Brigham (No. 2019B000050).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.