Neurodevelopmental Disorder Caused by Deletion of CHASERR, a lncRNA Gene

SUMMARY

CHASERR encodes a human long noncoding RNA (lncRNA) adjacent to CHD2, a coding gene in which de novo loss-of-function variants cause developmental and epileptic encephalopathy. Here, we report our findings in three unrelated children with a syndromic, early-onset neurodevelopmental disorder, each of whom had a de novo deletion in the CHASERR locus. The children had severe encephalopathy, shared facial dysmorphisms, cortical atrophy, and cerebral hypomyelination — a phenotype that is distinct from the phenotypes of patients with CHD2 haploinsufficiency. We found that the CHASERR deletion results in increased CHD2 protein abundance in patient-derived cell lines and increased expression of the CHD2 transcript in cis. These findings indicate that CHD2 has bidirectional dosage sensitivity in human disease, and we recommend that other lncRNA-encoding genes be evaluated, particularly those upstream of genes associated with mendelian disorders. (Funded by the National Human Genome Research Institute and others.)

Developmental and epileptic encephalopathies are genetically and phenotypically heterogeneous disorders that cause severe neurodevelopmental delay.¹ De novo loss-of-function variants in several genes, including chromodomain helicase DNA-binding protein 2 (CHD2), are known to cause early-onset epileptic encephalopathy (Mendelian Inheritance in Man number, 615369).² CHD2 is a DNA-binding protein that has been implicated in chromatin remodeling and plays a role in the neurogenesis of cortical neurons and interneurons.^3–6 The most common phenotypes among patients with CHD2-related neurodevelopmental disorders include seizures, developmental delay, and intellectual disability (occurring in more than 90% of patients), and findings on brain imaging are often normal.^7,8 To date, all pathogenic variants in CHD2 have been found to be de novo, with a suspected loss-of-function mechanism; the majority are deletions or truncating variants, and a few are missense variants in the highly conserved DNA-binding or helicase domains.⁷ CHD2 was curated by the Clinical Genome Resource (ClinGen) as a haploinsufficient gene (wherein loss of one copy results in disease), which is in line with the loss-of-function constraint seen for CHD2 in the Genome Aggregation Database (as shown by a pLI [probability of being loss-of-function intolerant] score of 1.0 [scores range from 0.0 to 1.0, with higher scores indicating a greater degree of negative selection on loss-of-function variants]).⁹ An analysis of rare copy-number variation in the human population supported the finding of intolerance to three gene copies of CHD2 (i.e., triplosensitivity),¹⁰ but increased CHD2 dosage has not yet been shown in human disease.

CHASERR is a highly conserved long noncoding RNA (lncRNA) located upstream of CHD2, separated by only 1.6 kb. Its mouse homologue (Chaserr) localizes in embryonic stem cells within a topologically associated domain containing the Chaserr RNA and Chd2 protein.¹¹ Chaserr mediates cis-acting transcriptional repression of Chd2 and is essential for postnatal survival in mice.¹² However, CHASERR has not been implicated in human disease. Only a few lncRNAs have been associated with mendelian disease. Homozygous deletions of LINC01956 were shown to cause a severe congenital limb and hindbrain malformation syndrome through loss-of-target gene expression of EN1.¹³ Variants in a locus producing a lncRNA were found to be associated with HELLP syndrome (characterized by hemolysis, elevated liver enzymes, and low platelet count), but these variants have not been shown to have an effect on lncRNA levels or downstream gene expression.¹⁴ An inherited balanced translocation disrupting lnc-NR2F1 was observed in a father and son with variable expressive language impairment, but similarly, the effect of the translocation on lncRNA or target gene expression is not known.¹⁵ Balanced translocations in the CISTR–ACT locus on chromosome 12 were found to segregate with autosomal dominant brachydactyly in two families and were associated with an increased expression of a lncRNA within the locus.¹⁶ However, haploinsufficiency of a lncRNA that results in increased target gene expression has not been shown to cause a mendelian condition.

Here, we report our findings in three unrelated children with a syndromic, early-onset neurodevelopmental disorder, each of whom had a heterozygous de novo deletion involving CHASERR that did not overlap the promoter or coding region of CHD2. These children had severe encephalopathy, shared facial dysmorphisms, cortical atrophy, and cerebral hypomyelination — a phenotype that is distinct from the known clinical spectrum of CHD2 haploinsufficiency. In two of these children, we found an allelic imbalance of CHD2 messenger RNA (mRNA) (i.e., the CHD2 allele in cis with the de novo CHASERR deletion was transcribed to a greater extent), with corresponding overexpression of CHD2 protein.

METHODS

CLINICAL DATA AND SPECIMEN COLLECTION

Patient 1 was enrolled in the Undiagnosed Diseases Network. Written informed consent was obtained from the patient and parents in accordance with the protocol of the institutional review board at the National Institutes of Health. Patient 1, a sibling, and the parents were also enrolled in the Rare Genomes Project at the Broad Institute of Massachusetts Institute of Technology and Harvard, with approval by the institutional review board at Mass General Brigham. As part of a study approved by the ethics committee at Centre Hospitalier Universitaire de Nantes (Research Programme “Génétique Médicale”), written informed consent for Patients 2 and 3 was obtained from their legal guardians according to the French law on bioethics and the Declaration of Helsinki. The MatchMaker Exchange¹⁷ network facilitated the connection between the investigators for Patients 1 and 2 through the GeneMatcher node.¹⁸

SEQUENCING AND GENOMIC ANALYSIS

Trio genome sequencing (mother, father, and proband) was performed on genomic DNA isolated from peripheral blood samples obtained from the family members of Patients 1, 2, and 3. Long-read genome sequencing was also performed on the family members of Patient 1 according to the circular consensus sequencing protocol of Pacific Biosciences. Additional details regarding genomic analyses are provided in the Supplementary Appendix, available with the full text of this article at NEJM.org.

EXPRESSION ANALYSIS

RNA sequencing was performed on whole-blood samples and cultured fibroblasts obtained from Patients 1 and 2, and the findings were compared with those for corresponding control tissue samples from the Genotype-Tissue Expression (GTEx) project.¹⁹ In addition, RNA sequencing was performed on induced pluripotent stem cells (Patients 1 and 2) and neural precursor cells (Patient 1). Quantification of CHD2 protein abundance was performed on induced pluripotent stem cells from Patients 1 and 2. Details regarding expression analyses are provided in the Supplementary Appendix.

RESULTS

CLINICAL EVALUATIONS

The children were born after uncomplicated pregnancies, with birth weight, length, and head circumference in the normal range and a decreased head circumference percentile at 2 years of age (Table S1 in the Supplementary Appendix). Facial dysmorphisms were noted at birth, including widely spaced eyes, anteverted nares, low-set ears, and a long philtrum (Fig. 1). Developmental delay and encephalopathy were noted within 2 months of age. By 4 years of age, there was evidence of generalized background slowing on electroencephalography (EEG) in all three patients, with infantile spasms and generalized polyspike-and-wave discharges seen in Patients 1 and 3. No photosensitive or generalized tonic–clonic seizures have been observed to date. All three patients showed cortical atrophy, optic nerve atrophy, and cerebral hypomyelination, with a thin corpus callosum, on T1- or T2-weighted magnetic resonance imaging (MRI) of the brain, performed by 4 years of age (Fig. 1), and all three had severe global developmental delay and could not communicate with words, ambulate independently, or perform tasks involving fine-motor skills. In terms of nonneurologic features, Patients 1 and 3 had severe vesicoureteral reflux that was noted at birth and later led to vesicostomy. The findings from full clinical evaluations are described in the Supplementary Appendix.

Figure 1. Facial Features and Findings on Magnetic Resonance Imaging of the Brain in Three Children with CHASERR Haploinsufficiency.

Shared facial dysmorphisms among the three children include wide-set eyes, anteverted nares, low-set ears, and a long philtrum. T1 and T2-weighted magnetic resonance imaging of the brain showed frontal predominant cortical atrophy with reduced volume of the brain stem (asterisks) and generalized hypomyelination of the corpus callosum and subcortical white matter (arrows) at 4 years or 8 months of age (middle row), which were all less apparent at 1 month of age (bottom row).

The phenotypes of the three children were distinct from those of persons with CHD2 haploinsufficiency, in whom generalized convulsive epilepsy and interictal generalized spike-and-wave discharges on EEG are common. Also, among persons with CHD2 haploinsufficiency, background slowing on EEG is uncommon, growth retardation is not seen, and findings on brain MRI are typically normal, showing no evidence of cerebral hypomyelination and only rare evidence of cortical atrophy.⁷ Global developmental delay and intellectual disability are common among persons with CHD2 haploinsufficiency, but the conditions are much less severe than those in the children with CHASERR haploinsufficiency described herein.

DETECTION AND PHASING OF DELETIONS IN CHASERR

In Patients 1, 2 and 3, short-read genome sequencing identified 22-kb, 8.4-kb, and 25-kb heterozygous de novo deletions, respectively. These deletions encompassed the promoter and the first three exons of CHASERR (Fig. 2A) and did not overlap CHD2 or its promoter.

Figure 2. De Novo CHASERR Deletions and Characterization of CHD2 Expression.

Panel A shows human chromosome 15q26.1, with a 500-kb inset showing a gene-sparse region upstream of the long noncoding RNA (lncRNA) CHASERR (teal) and its tandem coding gene CHD2 (light green). Reads were aligned to human genome assembly GRCh38. Arrows indicate the direction of transcription. De novo deletions (bracketed pink lines) in Patient 1 (22 kb), Patient 2 (8.4 kb), and Patient 3 (25 kb) overlap the promoter and first three exons of CHASERR. Long-read DNA sequencing in Patient 1 identified three de novo single-nucleotide variants (SNVs, pink dots) in cis with the CHASERR deletion. SINE denotes short interspersed element. The signs “(+)” and “(−)” denote the strand orientation of the Alu element. Panel B shows increased CHD2 protein abundance in induced pluripotent stem cells (iPSCs) from Patients 1 and 2, as compared with iPSCs from sex-matched wild-type controls, an unrelated female with CHD2 haploinsufficiency (CHD2^+/−), and CHD2^−/− CRISPR (clustered regularly interspaced short palindromic repeats)–Cas9 knockout iPSCs. CHD2 protein abundance was normalized to histone deacetylase 1. I bars indicate the standard deviation. P values were calculated with a two-tailed, heteroscedastic Student’s t test. Panel C shows RNA sequencing in whole-blood samples, cultured fibroblasts, and iPSCs obtained from Patients 1 and 2 and cultured neural precursor cells obtained from Patient 1. All showed allelic imbalance toward expression of the CHD2 allele in cis with the CHASERR deletion, across all phasable variants. The control whole-blood samples and cultured fibroblasts from the Genotype-Tissue Expression (GTEx) project showed no allelic imbalance in CHD2. White and black bars represent the read fraction for each of the two CHD2 alleles. I bars indicate the 95% confidence interval (with the assumption of binomial distribution). P values were calculated with a chi-square test, with an allele balance of 1:1 as the null hypothesis.

Long-read DNA sequencing in Patient 1 confirmed the de novo 22-kb deletion and mapped the breakpoint to two negative strand–oriented SINE (short interspersed element)–Alu elements (AluSg and AluYb8), with 40 bp of microhomology at the breakpoint junction (Fig. S1). These findings suggest that the deletion was generated by recombination between two Alu repeats. In addition, long-read sequencing confirmed three non-coding de novo single-nucleotide variants (SNVs) in cis with the deletion (Fig. 2A), which were probably generated by error-prone polymerase during repair of double-stranded DNA breaks.²⁰ The CHASERR deletion was most likely mediating the phenotype, because the incidentally identified SNVs were not in highly conserved regions of the genome, and no de novo SNVs were identified in Patients 2 or 3 in the CHASERR–CHD2 locus.

Short-read genome sequencing data from Patients 2 and 3 showed a similar deletion mechanism: breakpoints between two negative strand–oriented SINE–Alu elements (AluSx and AluSz), with 20 bp of microhomology at the junction, were identified in Patient 3, and a deletion in close proximity to two positive strand–oriented SINE–Alu elements (AluJr and AluSx3), with 8 bp of microhomology at the junction, was identified in Patient 2. The microhomology sequences at the breakpoint junction in Patients 1 and 3 map to a known recombination hotspot within SINE–Alu elements.²¹

In Patient 1, phasing of SNVs by long-read sequencing established that the de novo deletion in CHASERR occurred on the paternally inherited chromosome. In Patient 2, the de novo deletion also occurred on the paternally inherited chromosome, with phasing ascertained by the detection of a mendelian violation at an SNV within the deletion interval (i.e., Patient 2 was hemizygous for the maternally inherited variant g.92882694C→T, which suggests that the deletion was on the paternally inherited chromosome). The parental origin of the de novo deletion in CHASERR in Patient 3 could not be determined from short-read genome sequencing.

A genomewide search for other de novo variants in the three trios identified a likely pathogenic variant in CITED2 (NM_006079.5, c.701A→C, p.Glu234Ala) in Patient 1, which was suspected to account for the atrial septal defect but not the neurologic phenotype. In Patients 1, 2, and 3, standard variant analysis with the use of both dominant and recessive models did not identify any other pathogenic or likely pathogenic SNVs, small insertions or deletions, short-tandem-repeat expansions, or structural variants in the nuclear or mitochondrial genomes.

CHD2 EXPRESSION

RNA sequencing of whole-blood samples obtained from Patient 1 and an unaffected sibling showed haploinsufficiency of CHASERR in Patient 1, with 6.95 transcripts per million, as compared with 23.92 transcripts per million in the unaffected sibling. No significant transcriptome-wide gene expression outliers were seen in whole-blood samples from Patient 1, as compared with the control samples from the GTEx project (Fig. S2). The CHD2 protein abundance in induced pluripotent stem cells obtained from Patients 1 and 2 was significantly greater than that in sex-matched wild-type controls by a factor of 1.8 and 1.7, respectively (Fig. 2B and Figs S3 and S4). Together, these observations support the hypothesis that CHASERR haploinsufficiency increases CHD2 protein abundance, as seen in the Chaserr^+/− mouse model.¹² No evidence of haploinsufficiency or triplosensitivity was shown for other nearby genes within a 1-Mb window centered on CHASERR.¹⁰

On closer inspection of CHD2 expression from RNA sequencing data, consistent allelic imbalance across all phasable variants in CHD2 mRNA was noted in the whole-blood samples, cultured fibroblasts, induced pluripotent cells, and neural progenitor cells obtained from Patients 1 and 2 (Fig. 2C and Table S2). No phasable coding variants in CHD2 were present in Patient 3. Moreover, the imbalance significantly skewed toward the allele in cis with the CHASERR deletion — a finding that suggests cis-regulatory derepression of CHD2 expression, which was expected from the Chaserr^+/− mouse model.¹² Collectively, these results show that CHASERR is a negative cis regulator of CHD2 and that CHASERR deletion results in increased CHD2 abundance.

DISCUSSION

We report our findings in three children in whom de novo heterozygous deletions of a lncRNA (CHASERR) and the resulting increased expression of a target coding gene (CHD2) caused a neurodevelopmental disorder: a syndrome of severe encephalopathy with cortical atrophy and cerebral hypomyelination and dysmorphic facies. Each deletion overlapped CHASERR but not its downstream gene, CHD2. The deletion of CHASERR resulted in a corresponding increase in CHD2 transcription and protein abundance that were specifically due to the up-regulation of the CHD2 allele in cis with the CHASERR deletion. This functional effect is consistent with observations made in mice, in which Chaserr haploinsufficiency derepresses Chd2 expression in cis, resulting in growth impairment, facial dysmorphisms, and neonatal lethality.¹²

The three children with CHASERR deletions had greater impairment than even the most severely affected persons with CHD2 haploinsufficiency. Whereas most persons with CHD2 haploinsufficiency are able to speak and walk, none of the children described here have reached those developmental milestones. All three children had an electrographic phenotype (none or few clinical seizures and no photosensitivity) and radiographic phenotype (cortical atrophy and cerebral hypomyelination) distinct from the phenotypes seen in persons with CHD2-related disorders (Fig. 1), which suggests that CHASERR haploinsufficiency represents a disease that is clinically distinguishable from CHD2 haploinsufficiency–related disorder. Because the de novo deletions of CHASERR reported here implicate recombination mediated by Alu–Alu elements as the causal mechanism, we anticipate reanalysis of genome sequencing from other large cohorts of neurodevelopmental disorders (or array data when there is sufficient probe coverage of CHASERR) to uncover additional cases of CHASERR haploinsufficiency.

Our results extend the model of dosage sensitivity of CHD2 beyond haploinsufficiency to show that CHD2 overexpression also perturbs human brain development. CHD2 triplosensitivity is inferred from gene dosage–sensitivity metrics imputed from a large study of rare copy-number variation.¹⁰ However, we are not aware of any CHD2-only duplications (i.e., duplications dissociating CHD2 from CHASERR) in affected patients or unaffected controls.

In a previous characterization of Chaserr-haploinsufficient mice, it was suggested that CHASERR may be an appropriate RNA therapeutic target to rescue CHD2 haploinsufficiency.¹² An analogous treatment strategy for the Angelman syndrome is currently being evaluated in trials involving humans. In one study, antisense oligonucleotide therapy with a cis-acting lncRNA (UBE3A-ATS) aimed at derepressing its target coding gene, UBE3A.²² However, our study shows that excessive CHD2 abundance causes a potentially worse neurologic outcome than that caused by CHD2 haploinsufficiency. This “Goldilocks problem” will need to be addressed, potentially through preclinical models of partial inhibition of CHASERR or CHD2 or both to ameliorate haploinsufficiency of either gene.

Overall, our findings identify an essential, highly conserved lncRNA in human brain development and disease. Chaserr-mediated regulation of Chd2 has been suggested to be part of an autoregulatory feedback loop,²³ a hypothesis that highlights that Chaserr and related lncRNAs are adapted to regulate target genes that require exquisite spatial and temporal expression. Genetic lesions of lncRNAs that regulate coding genes in cis will lead to imbalances in the expression of these tandem genes, potentially resulting in disease. Homozygous deletions of the lncRNA LINC01956 were shown to cause a severe congenital limb and brain malformation syndrome through the disruption of cis-acting regulation of the nearby gene EN1¹³; the tandem gene structure and cis-regulatory model of these two genes are similar to those of CHASERR–CHD2. Yet lncRNAs are largely overlooked in clinical genomic analyses. Altogether, our results support the disruption of lncRNAs as a cause of human mendelian conditions, and we recommend that de novo structural variants that encompass highly conserved lncRNAs be examined closely in persons with undiagnosed disorders.

APPENDIX

The authors’ full names and academic degrees are as follows: Vijay S. Ganesh, M.D., Ph.D., Kevin Riquin, Ph.D., Nicolas Chatron, M.D., Ph.D., Esther Yoon, B.S., Kay-Marie Lamar, Ph.D., Miriam C. Aziz, B.S., Pauline Monin, M.D., Melanie C. O’Leary, M.Sc., Julia K. Goodrich, Ph.D., Kiran V. Garimella, D.Phil., Eleina England, M.S., Ben Weisburd, B.S., François Aguet, Ph.D., Carlos A. Bacino, M.D., David R. Murdock, M.D., Hongzheng Dai, Ph.D., Jill A. Rosenfeld, M.Sc., Lisa T. Emrick, M.D., Shamika Ketkar, Ph.D., Yael Sarusi, M.Sc., Damien Sanlaville, M.D., Ph.D., Saima Kayani, M.D., Brian Broadbent, M.B.A., Alisée Pengam, M.D., Bertrand Isidor, M.D., Ph.D., Stéphane Bezieau, Pharm.D., Ph.D., Benjamin Cogné, Pharm.D., Ph.D., Daniel G. MacArthur, Ph.D., Igor Ulitsky, Ph.D., Gemma L. Carvill, Ph.D., and Anne O’Donnell-Luria, M.D., Ph.D.

The authors’ affiliations are as follows: the Broad Center for Mendelian Genomics, Program in Medical and Population Genetics, Broad Institute of Massachusetts Institute of Technology and Harvard, Cambridge (V.S.G., M.C.O., J.K.G., K.V.G., E.E., B.W., F.A., D.G.M., A.O.-L.), and the Department of Neurology, Brigham and Women’s Hospital (V.S.G.), the Division of Genetics and Genomics, Boston Children’s Hospital (V.S.G., A.O.-L.), and Harvard Medical School (V.S.G., A.O.-L.), Boston — all in Massachusetts; L’institut du Thorax (K.R., B.I., S.B., B.C.), Service de Radiopediatrie (A.P.), and Service de Génétique Médicale (B.I., S.B., B.C.), Nantes Université, Centre Hospitalier Universitaire (CHU) de Nantes, Centre National de la Recherche Scientifique (CNRS), INSERM, Nantes, and Institut Neuromyogène, Laboratoire Physiopathologie et Génétique du Neurone et du Muscle, CNRS, INSERM (N.C., D.S.), and Service de Génétique, Hospices Civils de Lyon (N.C., P.M., D.S.), Lyon — all in France; the Departments of Neurology (E.Y., K.-M.L., M.C.A., G.L.C.) and Pharmacology (G.L.C.), Northwestern University Feinberg School of Medicine, Chicago; the Undiagnosed Diseases Network and the Department of Molecular and Human Genetics, Baylor College of Medicine, Houston (C.A.B., D.R.M., H.D., J.A.R., L.T.E., S. Ketkar), and the Department of Pediatrics, University of Texas Southwestern Medical Center (S. Kayani), and Coalition to Cure CHD2 (B.B.), Dallas; the Departments of Immunology and Regenerative Biology and Molecular Neuroscience, Weizmann Institute of Science, Rehovot, Israel (Y.S., I.U.); and the Centre for Population Genomics, Garvan Institute of Medical Research and University of New South Wales Sydney, Sydney (D.G.M.), and the Centre for Population Genomics, Murdoch Children’s Research Institute, Melbourne, VIC (D.G.M.) — both in Australia.