Identification of an episignature for CHD3-related Snijders Blok-Campeau syndrome reveals heterogeneity in the CHARGE syndrome episignature: towards a better characterisation of chromatinopathies.

Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Genome Med . 2026 Apr 8;18:39. doi: 10.1186/s13073-026-01639-5 Search in PMC Search in PubMed View in NLM Catalog Add to search Identification of an episignature for CHD3 -related Snijders Blok-Campeau syndrome reveals heterogeneity in the CHARGE syndrome episignature: towards a better characterisation of chromatinopathies Amandine Santini Amandine Santini 1 Department of Genetics and Reference Center for Developmental Abnormalities, Univ Rouen Normandie, Normandie Univ, Inserm U1245 and CHU Rouen, Rouen, France Find articles by Amandine Santini 1 , Angelo Tognon Angelo Tognon 2 Université Paris Cité, INSERM U1163, Imagine Institute, Paris, France Find articles by Angelo Tognon 2 , Anne-Claire Richard Anne-Claire Richard 3 Department of Genetics and Reference Center for Developmental Abnormalities, University Rouen Normandie, Normandie University, Inserm U1245 and CHU Rouen, 76000 Rouen, France Find articles by Anne-Claire Richard 3 , Guillaume Velasco Guillaume Velasco 4 Université Paris Cité, CNRS, Epigenetics and Cell Fate, UMR7216, Paris, France Find articles by Guillaume Velasco 4 , Gilles Phan Gilles Phan 5 UMR 8038, Laboratoire CiTCoM (Cibles Thérapeutiques Et Conception de Médicaments), Faculté de Pharmacie de Paris, Université Paris Cité, CNRS, Paris, France Find articles by Gilles Phan 5 , Pauline Marzin Pauline Marzin 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France 7 Service de Gynécologie Et d’obstétrique, UF de Génétique Médicale, Centre Hospitalier Universitaire de La Réunion, La Réunion, France Find articles by Pauline Marzin 6, 7 , Fabien Maury Fabien Maury 2 Université Paris Cité, INSERM U1163, Imagine Institute, Paris, France Find articles by Fabien Maury 2 , Angele May Angele May 3 Department of Genetics and Reference Center for Developmental Abnormalities, University Rouen Normandie, Normandie University, Inserm U1245 and CHU Rouen, 76000 Rouen, France Find articles by Angele May 3 , Caroline Michot Caroline Michot 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France Find articles by Caroline Michot 6 , Adela Chirita-Emandi Adela Chirita-Emandi 8 Department of Microscopic Morphology, Genetics Discipline, Center of Genomic Medicine, University of Medicine and Pharmacy “Victor Babes”, Timisoara, Romania 9 Regional Center of Medical Genetics Timis, Clinical Emergency Hospital for Children “Louis Turcanu”, Timisoara, Romania Find articles by Adela Chirita-Emandi 8, 9 , Jorge M Saraiva Jorge M Saraiva 10 Medical Genetics Department, Hospital Pediátrico de Coimbra, Unidade Local de Saúde de Coimbra, Coimbra, Portugal 11 Faculty of Medicine, University Clinic of Pediatrics, University of Coimbra, Coimbra, Portugal 12 Clinical Academic Center of Coimbra, Hospital Pediátrico de Coimbra, Unidade Local de Saúde de Coimbra, Coimbra, Portugal Find articles by Jorge M Saraiva 10, 11, 12 , Maria Juliana Ballesta-Martinez Maria Juliana Ballesta-Martinez 13 Sección Genética Médica. Servicio de Pediatría. Hospital Clinico Universitario Virgen de La Arrixaca, Murcia, Spain Find articles by Maria Juliana Ballesta-Martinez 13 , Stanislas Lyonnet Stanislas Lyonnet 2 Université Paris Cité, INSERM U1163, Imagine Institute, Paris, France 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France Find articles by Stanislas Lyonnet 2, 6 , Ivona Sansović Ivona Sansović 14 Department of Medical and Laboratory Genetics, Endocrinology and Diabetology, Children’s Hospital Zagreb, University of Zagreb School of Medicine, Zagreb, Croatia Find articles by Ivona Sansović 14 , Tahsin Stefan Barakat Tahsin Stefan Barakat 15 Department of Clinical Genetics, Erasmus MC University Medical Center, Rotterdam, The Netherlands Find articles by Tahsin Stefan Barakat 15 , Perrine Brunelle Perrine Brunelle 16 Institut de Génétique Médicale, University Lille, CHU Lille, Lille, France Find articles by Perrine Brunelle 16 , Jamal Ghoumid Jamal Ghoumid 17 ULR7364 – RADEME – Maladies RAres du DEveloppement embryonnaire et du Métabolisme, University Lille, Clinique de Génétique, Lille, France Find articles by Jamal Ghoumid 17 , Xavier Le Guillou Xavier Le Guillou 18 Service de Génétique Médicale, CHU de Poitiers, Poitiers, France Find articles by Xavier Le Guillou 18 , Pauline Le Tanno Pauline Le Tanno 19 Genetic, Genomic and Procreation Department, CHU Grenoble Alpes, Grenoble, France Find articles by Pauline Le Tanno 19 , Marjolaine Willems Marjolaine Willems 20 Département de Génétique Clinique, CHRU de Montpellier, Hôpital Arnaud de Villeneuve, Montpellier, France 21 Institute for Neurosciences of Montpellier, University Montpellier, INSERM, Montpellier, France Find articles by Marjolaine Willems 20, 21 , Martin Zenker Martin Zenker 22 Institute of Human Genetics, University Hospital Magdeburg, Magdeburg, Germany Find articles by Martin Zenker 22 , Ina Schanze Ina Schanze 22 Institute of Human Genetics, University Hospital Magdeburg, Magdeburg, Germany Find articles by Ina Schanze 22 , Stéphanie Moortgat Stéphanie Moortgat 23 Centre de Génétique Humaine, Institut de Pathologie et de Génétique, Gosselies, Belgium Find articles by Stéphanie Moortgat 23 , Bertrand Isidor Bertrand Isidor 24 Department of Genetics, Centre Hospitalier Universitaire de Nantes, Nantes, France Find articles by Bertrand Isidor 24 , Alix Paulet Alix Paulet 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France Find articles by Alix Paulet 6 , Alison Yeung Alison Yeung 25 Victorian Clinical Genetics Services, Murdoch Children’s Research Institute, Melbourne, Australia Find articles by Alison Yeung 25 , Jonathan Levy Jonathan Levy 26 Département de Génétique, Hôpital Robert Debré, Paris, France 27 Laboratoire de Médecine Génomique SeqOIA, Paris, France Find articles by Jonathan Levy 26, 27 , Federica Ruscitti Federica Ruscitti 26 Département de Génétique, Hôpital Robert Debré, Paris, France Find articles by Federica Ruscitti 26 , Leticia Pias-Peleteiro Leticia Pias-Peleteiro 28 Neurometabolic Disorders Unit, Department of Child Neurology/Department of Genetics and Molecular Medicine, Sant Joan de Déu Hospital, Barcelona, Spain Find articles by Leticia Pias-Peleteiro 28 , Marlène Rio Marlène Rio 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France Find articles by Marlène Rio 6 , Thomas Courtin Thomas Courtin 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France Find articles by Thomas Courtin 6 , Hamza Hadj Abdallah Hamza Hadj Abdallah 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France Find articles by Hamza Hadj Abdallah 6 , Stéphanie Ducreux Stéphanie Ducreux 27 Laboratoire de Médecine Génomique SeqOIA, Paris, France Find articles by Stéphanie Ducreux 27 , Jean-Sérène Laloy Jean-Sérène Laloy 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France Find articles by Jean-Sérène Laloy 6 , Paul Rollier Paul Rollier 29 Génétique Clinique - Centre de Référence Maladies Rares CLAD-Ouest, FHU GenOMedS, CHU de Rennes, Rennes, France Find articles by Paul Rollier 29 , Anne-Marie Guerrot Anne-Marie Guerrot 3 Department of Genetics and Reference Center for Developmental Abnormalities, University Rouen Normandie, Normandie University, Inserm U1245 and CHU Rouen, 76000 Rouen, France Find articles by Anne-Marie Guerrot 3 , Nicolas Chatron Nicolas Chatron 30 Genetics Department, Hospices Civils de Lyon, Lyon, France 31 Pathophysiology and Genetics of Neuron and Muscle (PNMG), UMR5261 - INSERM, UCBL, CNRS, U1315 Lyon, France Find articles by Nicolas Chatron 30, 31 , Florence Demurger Florence Demurger 32 Service de Génétique, CHBA, Vannes, France Find articles by Florence Demurger 32 , Alice Goldenberg Alice Goldenberg 3 Department of Genetics and Reference Center for Developmental Abnormalities, University Rouen Normandie, Normandie University, Inserm U1245 and CHU Rouen, 76000 Rouen, France Find articles by Alice Goldenberg 3 , Julian Delanne Julian Delanne 33 Inserm, CTM UMR1231, Équipe GAD, FHU TRANSLAD, Centre de Génétique, Centre de Référence Anomalies du Développement Et Syndromes Malformatifs, Centre de Référence Déficiences Intellectuelles de Causes Rares, Université Bourgogne Europe, CHU Dijon Bourgogne, Et Centre de Référence GénoPsy, Dijon, France Find articles by Julian Delanne 33 , Laurence Faivre Laurence Faivre 33 Inserm, CTM UMR1231, Équipe GAD, FHU TRANSLAD, Centre de Génétique, Centre de Référence Anomalies du Développement Et Syndromes Malformatifs, Centre de Référence Déficiences Intellectuelles de Causes Rares, Université Bourgogne Europe, CHU Dijon Bourgogne, Et Centre de Référence GénoPsy, Dijon, France Find articles by Laurence Faivre 33 , François Lecoquierre François Lecoquierre 3 Department of Genetics and Reference Center for Developmental Abnormalities, University Rouen Normandie, Normandie University, Inserm U1245 and CHU Rouen, 76000 Rouen, France Find articles by François Lecoquierre 3 , Gaël Nicolas Gaël Nicolas 3 Department of Genetics and Reference Center for Developmental Abnormalities, University Rouen Normandie, Normandie University, Inserm U1245 and CHU Rouen, 76000 Rouen, France Find articles by Gaël Nicolas 3 , Aurélie Coussement Aurélie Coussement 34 Fédération de Génétique Et Médecine Génomique, Service de Médecine Génomique Des Maladies de Système Et d’Organes, AP-HP, Hôpital Cochin, Paris, France Find articles by Aurélie Coussement 34 , Corinne Collet Corinne Collet 2 Université Paris Cité, INSERM U1163, Imagine Institute, Paris, France 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France Find articles by Corinne Collet 2, 6 , Yvan Herenger Yvan Herenger 35 Genetica AG, Zurich, Human Genetics and Genetic Counselling Unit, Zurich, Switzerland Find articles by Yvan Herenger 35 , Matthieu Defrance Matthieu Defrance 36 Interuniversity Institute of Bioinformatics in Brussels, Université Libre de Bruxelles, Brussels, Belgium Find articles by Matthieu Defrance 36 , Valérie Cormier-Daire Valérie Cormier-Daire 2 Université Paris Cité, INSERM U1163, Imagine Institute, Paris, France 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France Find articles by Valérie Cormier-Daire 2, 6 , Camille Charbonnier Camille Charbonnier 37 Department of Biostatistics and Reference Center for Developmental Abnormalities, Univ Rouen Normandie, Normandie Univ, Inserm U1245 and CHU Rouen, Rouen, France Find articles by Camille Charbonnier 37 , Maud de Dieuleveult Maud de Dieuleveult 2 Université Paris Cité, INSERM U1163, Imagine Institute, Paris, France Find articles by Maud de Dieuleveult 2, ✉ Author information Article notes Copyright and License information 1 Department of Genetics and Reference Center for Developmental Abnormalities, Univ Rouen Normandie, Normandie Univ, Inserm U1245 and CHU Rouen, Rouen, France 2 Université Paris Cité, INSERM U1163, Imagine Institute, Paris, France 3 Department of Genetics and Reference Center for Developmental Abnormalities, University Rouen Normandie, Normandie University, Inserm U1245 and CHU Rouen, 76000 Rouen, France 4 Université Paris Cité, CNRS, Epigenetics and Cell Fate, UMR7216, Paris, France 5 UMR 8038, Laboratoire CiTCoM (Cibles Thérapeutiques Et Conception de Médicaments), Faculté de Pharmacie de Paris, Université Paris Cité, CNRS, Paris, France 6 Service de Médecine Génomique Des Maladies Rares, Faculté de Médecine, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris, Université de Paris Cité, Paris, France 7 Service de Gynécologie Et d’obstétrique, UF de Génétique Médicale, Centre Hospitalier Universitaire de La Réunion, La Réunion, France 8 Department of Microscopic Morphology, Genetics Discipline, Center of Genomic Medicine, University of Medicine and Pharmacy “Victor Babes”, Timisoara, Romania 9 Regional Center of Medical Genetics Timis, Clinical Emergency Hospital for Children “Louis Turcanu”, Timisoara, Romania 10 Medical Genetics Department, Hospital Pediátrico de Coimbra, Unidade Local de Saúde de Coimbra, Coimbra, Portugal 11 Faculty of Medicine, University Clinic of Pediatrics, University of Coimbra, Coimbra, Portugal 12 Clinical Academic Center of Coimbra, Hospital Pediátrico de Coimbra, Unidade Local de Saúde de Coimbra, Coimbra, Portugal 13 Sección Genética Médica. Servicio de Pediatría. Hospital Clinico Universitario Virgen de La Arrixaca, Murcia, Spain 14 Department of Medical and Laboratory Genetics, Endocrinology and Diabetology, Children’s Hospital Zagreb, University of Zagreb School of Medicine, Zagreb, Croatia 15 Department of Clinical Genetics, Erasmus MC University Medical Center, Rotterdam, The Netherlands 16 Institut de Génétique Médicale, University Lille, CHU Lille, Lille, France 17 ULR7364 – RADEME – Maladies RAres du DEveloppement embryonnaire et du Métabolisme, University Lille, Clinique de Génétique, Lille, France 18 Service de Génétique Médicale, CHU de Poitiers, Poitiers, France 19 Genetic, Genomic and Procreation Department, CHU Grenoble Alpes, Grenoble, France 20 Département de Génétique Clinique, CHRU de Montpellier, Hôpital Arnaud de Villeneuve, Montpellier, France 21 Institute for Neurosciences of Montpellier, University Montpellier, INSERM, Montpellier, France 22 Institute of Human Genetics, University Hospital Magdeburg, Magdeburg, Germany 23 Centre de Génétique Humaine, Institut de Pathologie et de Génétique, Gosselies, Belgium 24 Department of Genetics, Centre Hospitalier Universitaire de Nantes, Nantes, France 25 Victorian Clinical Genetics Services, Murdoch Children’s Research Institute, Melbourne, Australia 26 Département de Génétique, Hôpital Robert Debré, Paris, France 27 Laboratoire de Médecine Génomique SeqOIA, Paris, France 28 Neurometabolic Disorders Unit, Department of Child Neurology/Department of Genetics and Molecular Medicine, Sant Joan de Déu Hospital, Barcelona, Spain 29 Génétique Clinique - Centre de Référence Maladies Rares CLAD-Ouest, FHU GenOMedS, CHU de Rennes, Rennes, France 30 Genetics Department, Hospices Civils de Lyon, Lyon, France 31 Pathophysiology and Genetics of Neuron and Muscle (PNMG), UMR5261 - INSERM, UCBL, CNRS, U1315 Lyon, France 32 Service de Génétique, CHBA, Vannes, France 33 Inserm, CTM UMR1231, Équipe GAD, FHU TRANSLAD, Centre de Génétique, Centre de Référence Anomalies du Développement Et Syndromes Malformatifs, Centre de Référence Déficiences Intellectuelles de Causes Rares, Université Bourgogne Europe, CHU Dijon Bourgogne, Et Centre de Référence GénoPsy, Dijon, France 34 Fédération de Génétique Et Médecine Génomique, Service de Médecine Génomique Des Maladies de Système Et d’Organes, AP-HP, Hôpital Cochin, Paris, France 35 Genetica AG, Zurich, Human Genetics and Genetic Counselling Unit, Zurich, Switzerland 36 Interuniversity Institute of Bioinformatics in Brussels, Université Libre de Bruxelles, Brussels, Belgium 37 Department of Biostatistics and Reference Center for Developmental Abnormalities, Univ Rouen Normandie, Normandie Univ, Inserm U1245 and CHU Rouen, Rouen, France ✉ Corresponding author. Received 2025 May 20; Accepted 2026 Mar 27; Collection date 2026. © The Author(s) 2026 Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ . PMC Copyright notice PMCID: PMC13067702  PMID: 41952182 Abstract Background Recent advances in sequencing technologies have enhanced patient diagnosis; however, causal pathogenic variants remain unidentified for a significant number of patients due to limited understanding of certain variants, regulatory sequences, or sequencing challenges, such as complex rearrangements. Investigating the epigenetic landscape has become essential to improve the diagnostic yield. Diseases caused by pathogenic variants in epigenetic regulators, often associated with growth abnormalities, intellectual disability, and facial dysmorphism, are prime models for studying episignatures. Among them, Snijders Blok-Campeau syndrome (ORPHA:599082), caused by pathogenic variants in the CHD3 gene, remains largely understudied. Methods A European cohort of 23 patients displaying typical Snijders Blok-Campeau syndrome traits and carrying pathogenic/likely pathogenic CHD3 variants was analysed using the Illumina EPIC array, identifying 270 differentially methylated positions distinguishing patients from 62 healthy matched controls. A subset of these regions serves as diagnostic tools for complex cases or variants of uncertain significance and helps uncover deregulated pathways linked to this syndrome. Four patients carrying pathogenic/likely pathogenic variants but with atypical clinical presentation, as well as 10 patients with variants of uncertain significance, were analysed as the testing set. Results Comparing methylomes of patients carrying pathogenic variants in CHD3 , CHD7 (CHARGE syndrome, ORPHA:138), and CHD8 (Intellectual developmental disorder with autism and macrocephaly, ORPHA:642675) genes allows us to identify distinct subgroups with unique methylation profiles. This CHD3 DNA methylation signature aids in reclassifying variants and diagnosing atypical cases. Conclusions Our findings advance the field of epigenetic signatures in rare diseases. We have opened new avenues for further investigation into subtypes defined by methylome assays (such as in the context of chromatinopathies), which could refine the phenotype spectrum and help predict patient outcomes. Supplementary Information The online version contains supplementary material available at 10.1186/s13073-026-01639-5. Keywords: Rare Diseases, CHD3 , Episignature, DNA methylation, Snijders Blok-Campeau syndrome Background Rare diseases (RDs) encompass more than 6,718 pathologies that altogether concern more than 300 million people worldwide [ 1 ] (Orphanet data in Dec 2024). In September 2021, 30% of RD patients were reported with a missing diagnosis [ 2 , 3 ]. Finding a diagnosis and shortening the clinical odyssey is thus one of the most important priorities. Currently, the profiling approaches for diagnostic purposes could be limited to a small part of the genome (e.g., CGH array, whole-exome sequencing or exome-focused analysis of whole-genome sequencing), making the genetic investigation sometimes incomplete. More and more frequently, access to the entire genome is possible thanks to high-throughput sequencing platforms. Still, there are many interpretation difficulties, notably due to variants of uncertain significance (VUS). Moreover, epigenomic information is still not or very rarely explored. Increasing attention has been spent on DNA methylation of RD patients, and specific patterns (also called episignatures) have been identified for several pathologies [ 4 – 7 ]. Mendelian disorders of the epigenetic machinery Among the rare disorders with developmental anomalies, the Mendelian disorders of the epigenetic machinery have raised attention in recent years. Indeed, they encompass all disorders caused by pathogenic variants in genes involved in the writing, erasing, maintenance, or reading of the epigenome (i.e. chromatinopathies). This group of disorders is characterised by developmental/growth anomalies and intellectual disabilities [ 8 – 10 ]. Pathogenic variants are often autosomal dominant, consistent with haploinsufficiency. Chromatinopathies represent a relevant model for studying epigenetic alterations. These changes can serve as episignatures for diagnostic purposes and as epigenomic markers that may reflect underlying molecular deregulations, thus contributing to a better understanding of the associated pathophysiological mechanisms. Chromatinopathies represent a relatively recently defined group of disorders. Their broad and frequently overlapping clinical manifestations, often shared with conditions such as autism spectrum disorders, make them particularly difficult to delineate with precision. The nomenclature and clinical definitions of chromatinopathies continue to evolve (Orphanet, OMIM) [ 3 , 11 ]. Strikingly, variants in the same causative gene can lead to disorders of varying severity (e.g., ATRX , SETD2 ), entirely distinct conditions (e.g., DNMT1 ), or even opposite phenotypes (e.g., DNMT3A ) [ 3 , 10 ]. A recent study reported that pathogenic variants in 26 epigenes can give rise to multiple distinct disorders or to a phenotypic continuum [ 11 ]. Since epigenes are expressed very early in embryonic development, pathogenic variants usually have multisystemic consequences. Furthermore, the protein domains in which these variants occur are affected in distinct ways, leading to differential impacts on the activity of these proteins and their downstream targets. For all these reasons, the precise characterisation of chromatinopathies is of crucial importance. Among SNF2 -domain chromatin remodelers, characterised by the presence of an ATPase domain, the CHD subfamily is defined by a tandem chromodomain capable of binding methylated histones [ 12 ]. This subfamily comprises nine genes encoding enzymes involved in various molecular processes, including cell differentiation and tissue-specific transcriptional regulation [ 13 – 16 ]. Initially, only pathogenic variants in CHD7 (CHARGE syndrome; ORPHA:138) and CHD8 (intellectual developmental disorder with autism and macrocephaly/AUTS18; ORPHA:642675) had been implicated in human disease [ 17 ]. More recently, pathogenic variants in CHD1 (Pilarowski-Bjornsson syndrome; ORPHA:529965), CHD2 (developmental and epileptic encephalopathy 94; ORPHA:2382 and more precisely MIM #615369), CHD3 (Snijders Blok-Campeau syndrome; ORPHA:599082), CHD4 (Sifrim-Hitz-Weiss syndrome; ORPHA:653712), CHD5 (Parenti-Mignot neurodevelopmental syndrome; MIM #619873), and CHD6 (Hallermann-Streiff syndrome; ORPHA:2108) have also been reported in neurodevelopmental disorders [ 18 – 20 ]. To complete the CHD subfamily, currently, no pathogenic variants have been associated with the CHD9 gene despite extensive research. Snijders Blok-Campeau syndrome The Snijders Blok-Campeau syndrome (SBCS, ORPHA:599082, MIM #618205, MONDO:0032600), first identified in 2018 [ 21 ], is characterised by developmental delay and mild to severe intellectual disability, autism spectrum disorder traits or attention deficit hyperactivity disorder, and sometimes hypersociability [ 22 – 26 ]. Facial features such as a wide nasal bridge, a thin upper lip, and downturned mouth corners [ 24 , 27 ] and skeletal anomalies, like scoliosis or joint hypermobility, are recurrently found. Heart defects and kidney abnormalities have been reported in some rare cases [ 28 ]. All these features are common and nonspecific among neurodevelopmental disorders, making clinical diagnosis challenging. This syndrome is caused by monoallelic pathogenic variants in the CHD3 gene, which plays a crucial role in chromatin remodelling, a process essential for DNA repair and gene expression [ 29 ]. Recurrent arginine mutations are described in previous cohorts [ 22 , 30 ]. To date, no specific treatment is available for this disorder. Management of SBCS requires a multidisciplinary approach, including speech therapy, educational support, and medical treatment for physical health issues. This project aims to identify a robust episignature useful to reclassify VUS and diagnose SBCS. To reach this objective, we established a cohort of SBCS patients by establishing new European collaborations. Indeed, to identify and validate a robust episignature, a minimum cohort of approximately 20 patients is recommended, ideally encompassing variants distributed across the entire gene and including diverse mutation types (e.g., missense, nonsense, frameshift variants) to accurately capture the full epigenetic heterogeneity associated with these genetic alterations [ 7 ]. Here, we report a specific episignature to patients with a typical phenotype of 270 differentially methylated positions (DMPs), which allows the reclassification of three out of ten VUS and differentiation from other closely related syndromes. Methods Cohort construction and clinical data In 2023, national and European calls for collaboration were released (FSMR AnDDi-Rares and ERN ITHACA), and 38 patients from 10 countries were identified. Clinical data were extracted from medical records by the medical doctors who previously examined the patients during routine medical examinations. A significant number of healthy matched donors (negative controls) is essential to accurately identify episignatures, i.e. sets of differentially methylated positions or regions in patients compared to healthy individuals, and to minimise the risk of false positives, including those arising from technical artefacts. Those regions are called differentially methylated positions (DMPs) or differentially methylated regions (DMRs) when several DMPs are identified and can be used to develop accurate machine-learning predictors to later classify candidate variants and patients with evocative symptoms in a clinical setting. All patients or legal representatives provided informed written consent for exome/genome analyses in a medical setting that contains a query on the use of residual samples for research. DNA collection DNA from the patient’s blood was collected in the hospital DNA bank and anonymised. DNA was purified using the standard procedures. The QiaAmp kit #51194 from Qiagen was used. Briefly, DNA was previously extracted for each patient for exome or whole genome sequencing. The aliquots were collected with the agreement of each patient. DNA quantity was determined with the Qubit fluorometer and integrity was checked by agarose gel profiling. DNAs showing degradation profile were discarded. All samples were stored within the Imagine/Inserm U1163 biological collection (DC-2019–3504). Samples and healthy controls analysed on the ASGARD platform were also stored within the biological collection of the CRBi, Rouen (DC 2008–711, access authorisation n°MCRBi/2024/02). Structural analysis of CHD3 mutations 3D structures were analysed based on human nucleosome CHD1 and CHD4 complex structures, PDB codes 5O9G and 6RYR respectively, superposed with AlphaFold model of human CHD3 (AF- Q12873 -F1-v4) [ 31 ]. Structures superposition and images were generated using PyMOL Molecular Graphics System, version 2.5 Schrödinger, LLC [ 32 ]. EPIC arrays DNA methylation arrays were then generated. Four patients (ST5, ST15, ST46, ST47) and 15 healthy controls were analysed at Diagenode S.A. (Liege, Belgium) using Illumina EPIC v1.0 arrays, and 15 patients (ST49, ST56, ST64, ST66, ST70, ST71, ST73, ST74, ST75, ST76, ST77, ST155, ST146, ST147, ST148) and 14 healthy controls using Illumina EPIC v2.0 arrays. Twenty patients (ST79, ST80, ST98, ST90B, ST78, ST81.1, ST82, ST90A, ST81.2, ST83, ST69, ST53, ST67, ST68, ST72, ST121, ST136, ST139, ST135, ST134) and 33 healthy controls were analysed at the ASGARD platform (Advanced sequencing solutions and genomic analysis for research and diagnosis, Rouen, France) using Illumina EPIC v2.0 arrays. Such analysis methylation profiles was approved by the CERDE ethics committee (notification n° E2023-13) from the Rouen University Hospital. IDATs were retrieved and analysed with the Meffil package [ 33 ]. Quality control of DNA & sample preparation Diagenode For samples ST15, ST5, ST46, ST47, ST49, ST56, ST64, ST66, ST70, ST71, ST73, ST74, ST75, ST76, ST77, ST155, ST146, ST147, ST148 genomic DNA from whole blood was deaminated with the EZ-96 DNA Methylation Kit (#D5004, Zymo Research) according to Illumina’s recommended deamination protocol and then hybridized on EPIC v1.0 or v2.0 arrays (Additional file 4 ). ASGARD platform analysis For samples ST79, ST80, ST98, ST90B, ST78, ST81.1, ST82, ST90A, ST81.2, ST83, ST69, ST53, ST67, ST68, ST72, ST121, ST136, ST139, ST135, ST134 genomic DNA from whole blood was bisulfite-converted using the EZ DNA Methylation Lightning Kit (Zymo Research). In accordance with the manufacturer’s protocol, DNA methylation profile was then derived using Illumina’s Infinium EPIC array v2.0. Patients and negative controls were balanced across 20 arrays and within each array rows to minimise technical biases, allowing a reliable estimation of methylation signal variability within and between arrays. DNA methylation arrays were generated at the ASGARD-Rouen genomic platform (University of Rouen and Rouen university hospital, Rouen, France) on an Illumina NextSeq550 scanner. Healthy controls consisted of individuals without NDD who underwent pre-symptomatic testing for other conditions and were found to be non-carriers or unaffected relatives of patients with a genetic disease, among non-carriers of pathogenic variants. Raw IDAT data were processed and normalised using the default Meffil R package. This package efficiently handles large DNA methylation datasets. Briefly, probes that failed methylation detection (detection p -value > 0.01) in more than 5% of samples were removed. Samples with > 1% of failed probes or an outlier methylation distribution (methylation/unmethylation ≥ 3 S.D. from the mean) were flagged. DNA methylation analysis & Statistics Samples from both platforms were imported and analysed separately. Once the IDAT files were retrieved, a quality control (QC) process was conducted using the meffil.qc function from the Meffil R package. Several predictions were obtained from methylation values to apply additional QC and normalisation steps. Sex predictions were extracted from the standard Meffil normalised object. Blood cell counts were estimated with the meffil.cell.count.estimates function. DNA methylation age was predicted with the DNAmAge function from the methylclock R package [ 34 ]. The skinHorvath clock, which was trained on 450 K skin and blood samples displayed a very strong correlation with actual age at blood sample on our dataset (Pearson correlation r = 0.96). Following extensive QC checks over sex mismatches, control probes and overall poor-quality flags, 100 samples were kept for further statistical analyses, 38 patients and 62 healthy controls. After performing quality controls, we apply two successive normalisation steps using the functions meffil.normalize.quantiles and meffil.normalize.samples, respectively. As advocated in the Meffil documentation, random effect adjustment was performed on the array and sentrix rows, as well as fixed effect adjustment on the first two PCs, before computing β-values. Next, we conducted a linear regression to account for confounding factors, including age at sampling, sex, and inferred blood cellular composition. Finally, we performed a differential analysis using the meffil.ewas function. To analyse the two platform-specific datasets together while minimising batch effects, two separate differential analyses were performed. One analysis included only Diagenode samples, and the other included only Rouen samples. The set of DMPs found for each platform dataset was identified with the meffil.ewas function on the subset of controls and pathogenic or likely pathogenic CHD3 typical variant carriers. To correct for well-known confounders of methylation levels, the differential analysis accounted for age at sampling, sex and inferred blood cell composition. Namely, a baseline methylation level model adjusting for age at sampling, sex and inferred blood cell composition was fitted on negative control samples for each probe. Adjusted methylation levels were computed for each sample from this model by correcting each β-value for the expected baseline level according to this model. When comparing predicted age using the skinHorvath clock to chronological age at sampling, a systematic bias was observed in samples processed with the EPIC v2.0 array, resulting in a curved prediction trend. To avoid confounding effects, chronological age at sampling was used in all subsequent analyses. Meta analysis The results of these analyses were then combined using a fixed-effect meta-analysis approach, which included the observed mean methylation differences (Δβ) and their corresponding standard deviations. Three embedded sets of selected CpG positions were used in the following analyses, depending on the level of required stringency: i) The set of differentially methylated positions (DMPs) was defined by a false-discovery rate (FDR, using Benjamin-Hochberg method) threshold of 5% (Additional file 5 ), ii) Signature positions were defined by a double constraint: p -value < 10 –5 and |Δβ|> 0.05, where Δβ represents the average methylation difference between positive and negative controls at 0.05 (Additional file 5 ). iii) The DMP set for between-gene comparison (multi-gene comparison positions) was defined by double constraint: FDR < 10% and |Δβ|> 0.05 to balance the number of positions between genes (Additional file 5 ). Adjusted methylation levels were visualised on the first two principal components as well as on a heatmap of DNA methylation residuals of episignature positions (pheatmap package with Euclidean distance and Ward aggregation method) [ 35 ]. Testing sensitivity and specificity of CHD3 DNAm signature The robustness of the signature was challenged through a leave-one-out approach to negative controls and typical CHD3 samples. An SVM model was trained on signature positions on each cross-validation training set and applied to the corresponding test set to obtain unbiased estimates of sensitivity and specificity, both overall and by status class, along with 95% binomial confidence intervals. Two additional cross-validations were conducted to validate the reproducibility of the signature. First, a cross-platform validation was implemented. Each platform was used in turn as a training set for the SVM while the other served as a validation set. As an extended cross-platform validation (including both EWAS and SVM training) was rendered impossible by the sample size of platform-specific datasets, we performed a global leave-one-out cross-validation. To reduce the computational burden, all CpG positions reaching FDR < 50% in the original main meta-analysis were included in this extended cross-validation (including EWAS filtering and SVM training). For CHD3 VUS, atypical CHD3 , CHD7 pathogenic variants and CHD8 pahogenic variants, the pathogenicity score was derived from a prediction model based on the full training set which consisted of 23 patients with P/LP CHD3 variants and 62 healthy age-matched controls. To balance the number of positions associated with each episignature, and for coherence with the following CHD3/CHD7/CHD8 joint analysis, the positions used in this test correspond to a more lenient definition of DMPs, namely the set of multi-gene comparison positions. We used the default parameters of the e1071 package to build all SVM models [ 36 ]. Comparison of CHD3 / CHD7 / CHD8 episignatures Combined analysis of CHD3 , CHD7 and CHD8 signatures was done similarly to the main CHD3 analysis, using multi-gene comparison positions for CHD3 . Raw EPIC v1.0 data from CHD7 , CHD8 and negative controls were collected from Husson et al. original dataset [ 7 ]. EPIC v1.0 and EPIC v2.0 samples were imported and normalised separately with standard meffil functions. Baseline methylation models were fitted separately on EPIC v1.0 and EPIC v2.0 positive and negative control samples. Adjusted methylation profiles of CpG positions belonging to the union of published CHD7 and CHD8 signatures as well as the CHD3 signature. For the CHD7 and CHD8 signatures, two separate probe lists were curated for each signature. The probes selected were those with the highest performance as reported in the study by Husson et al. [ 7 ]. Specifically, the final probe list for CHD7 was derived from the study of Aref-Eshghi et al. [ 4 ], while the probe list for CHD8 was derived from the work of Siu et al. [ 37 ]. These combined profiles were visualised through PCA and heatmap generated using the pheatmap package, with Euclidean distance and the Ward aggregation method applied to both rows and columns. For easier visualisation of the heatmap results, a raincloud plot was created. To achieve this, the heatmap was divided into seven blocks, each representing a specific signature ( CHD3 , CHD7 , or CHD8 ). Additionally, all probes within each block are homogeneous and predict the same result, either all are hypomethylated or all are hypermethylated. For each block, we calculate the average methylation level per individual. This is then plotted as a density curve, resulting in a density plot for each category (Control, CHD3 , CHD7 , and CHD8 ). As the controls are always at zero, we can see how each group behaves within each block: a shift to the left of the distribution indicates hypomethylation, whereas a shift to the right indicates hypermethylation. Below each density distribution, a boxplot visualises the distribution of individual data points. P-values were calculated using the Wilcoxon test. Each patient category was compared to controls, and the two CHD7 subgroups were also compared to each other. To assess whether probe clusters were homogeneous based on their signature, hierarchical clustering was performed using Ward’s method. This approach groups probes by minimising the intra-cluster variance at each aggregation step. The resulting dendrogram was then divided into seven clusters, corresponding to the expected number of probe groups. Differentially methylated regions (DMRs) episignatures To identify differentially methylated regions (DMRs), we used the DMRcate package in the R software [ 38 ]. The model was adjusted for covariates, including platform, age at sampling, sex, and inferred blood cell composition. We used the default parameters of the package, with lambda = 1000 and C = 2. The lambda parameter defines the window size (in base pairs) used to group adjacent probes into potential DMRs. The C parameter controls the weight given to the test statistic when smoothing probe values within a region. GO term analysis We used the ClusterProfiler package in R [ 39 ] to identify GO terms associated with the genes corresponding to the identified probe list, for DMRs. We kept the default parameter settings but restricted the analysis to GO categories related to biological processes. In addition, a minimum number of genes was required for each category. For visualisation, we used the functions provided by the package, such as barplot and emapplot. In Gene Ontology analysis, we categorized gene targets into three biological domains: specific molecular activities carried out by gene products (Molecular Function, MF), larger processes accomplished by multiple molecular functions (Biological Process, BP), and cellular structures or locations where these genes products were found (Cellular Component, CC). In this study, we considered only Biological Process. Venn diagram We retrieved the BED files corresponding to the probe lists for differentially methylated positions, differentially methylated regions, and the probes present in the CHD3 ChIP-Seq data. A Venn diagram was created using the venn.diagram function [ 40 ]. ChIP-Seq analysis In 2011, Ram et al . made a bed file containing ChIP-Seq peaks of CHD3 available (hg19) [ 41 ]. ChIP-Seq peaks were downloaded from the Cistome project [ 42 ]. In 2016, Yet et al . reported a study about the role of the FOXA1 pioneer factor on chromatin in LNCaP cells [ 43 ]. They performed ChIP-Seq of eight chromatin remodeler factors including CHD3 ( GSM1868500 ). We converted that bigwig file to bedgraph format using the UCSC Genome Browser bigwigToBedGraph tool. We then used MACS3 [ 44 ] to identify ChIP-Seq peaks from the bedgraph file and took the top 2.5% and top 1% peaks. All processed data are available on the Zenodo repository (10.5281/zenodo.17566757). Results CHD3 cohort of 30 patients and characterisation Thirty-eight SBCS patients were identified (Additional file 4 and Fig. 1 ), among them 28 patients carried pathogenic or likely pathogenic variants, and 10 patients carried VUS (Fig. 1 A). Clinical sum-up, phenotypic and molecular data are reported in Additional file 4 . Thirteen patients were reported as severe based on birth length, birth occipital frontal circumference (OFC), neonatal complications, height and OFC at last examination, early developmental delay, intellectual disability, speech at last examination, seizures and facial dysmorphism (Fig. 1 A, B, C, Additional file 1 : Fig S1A). Ten patients were reported as moderate, thirteen patients were reported as mild, and two patients could not be classified due to a lack of clinical information. The heterogeneity of patients' clinical presentation was important, so an additional patient classification was needed to fine-tune the classification. The clinical patient's profile was considered and determined for each patient whether they presented the typical symptoms of SBCS based on the previous publications, as we defined and reported in Additional file 4 [ 21 , 22 , 24 , 29 ]. This classification enabled us to determine two classes: typical and atypical (Fig. 1 A). Among the 28 patients (LP + P, Additional file 1 : Fig S1A) carrying a pathogenic variant in the CHD3 gene, 4 patients were reported as atypical. A detailed description of the different types of mutations, along with a comparison with previously published cohort data, is presented in Additional file 2 : Table S1. Briefly, 63% of patients in our cohort carry a missense mutation, which is consistent with prior reports. Fig. 1. Open in a new tab A Description of the cohort of individuals carrying pathogenic variants in the CHD3 gene. B & C Schematic diagram of the CHD3 protein depending on the pathogenicity of the variant (likely pathogenic & pathogenic versus VUS) with the functional domains design with IBS 2.0 [ 45 ]. Missense mutations are in yellow and truncated mutations are in black. D Mapping of CHD3 mutations. The mutated residues are in red sphere on the structural model of CHD3 (AlphaFold code AF- Q12873 -F1-v4: Chromo domain, Helicase domain, Helicase C-term domain and CHDCT2 domain are in orange, dark green, green and cyan respectively), superposed with the structure of the nucleosome complex with CHD1 in blue (PDB code 5O9G). ATP analogue AMP-PNP (Adenylyl-imidodiphosphate) is shown in yellow sphere. Structural details of highly conserved residues R966, R1169 and R1172 show hydrogen bonds between R1169 and R1172 with ATP analogue in the active site of Helicase domain Patients with VUS 10 patients had a VUS in CHD3 . Patients ST53, ST68 and ST147 had a VUS inherited from a likely asymptomatic parent. For patients ST67, ST69, ST72, and ST148, no segregation data were available. Atypical patients Among the 28 patients carrying a likely pathogenic/pathogenic variant, 4 individuals present atypical clinical signs (see Supplemental Information). While the individuals presented clinical signs compatible with SBCS, such as intellectual disability, tall stature and macrocephaly, other features, such as craniosynostosis in ST78, small stature and malabsorption disorder in ST80, microcephalia and very short stature in ST135 (sharing the same pathogenic variant as ST80) and coarctation of the aorta in ST56, are not generally reported. Altogether, those 4 clinically atypical SBCS patients carrying a pathogenic/likely pathogenic (P/LP) CHD3 variant underline the need for a CHD3 episignature to refine the diagnosis. Localisation of the variants in the CHD3 protein Most of the variants associated with severe phenotype are localised in the highly conserved helicase domain; in particular, Thr768 (T768), Arg1169 (R1169) and Arg1172 (R1172) interact directly with ATP substrate in the active site based on the nucleosome complex structure of CHD1 and CHD4 (Fig. 1 D, Additional file 1 : Fig S1B) [ 46 , 47 ]. Snijders Blok-Campeau syndrome episignature To search for an episignature, we compared the genome-wide methylation profiles of 23 patients with P/LP CHD3 variants with typical phenotype with those of 62 healthy age-matched and platform-matched controls. Adjusting for age at sampling, sex and inferred blood cell composition, we identified 270 DMPs (differentially methylated positions) FDR < 0.05, after meta-analysis of platform-specific analyses (Additional file 5 , Additional file 1 : Fig S2A,B,C). The genomic distribution of these DMPs across promoters (≤ 1 kb, 1–2 kb, 2–3 kb), introns, exons, untranslated regions (UTRs), distal intergenic, and downstream regions. Most DMPs were found within promoter regions (≤ 1 kb; n ≈ 80; ~ 29%), distal intergenic regions ( n ≈ 68; ~ 25%), and intronic regions ( n ≈ 73; ~ 27%). To enhance the specificity and accuracy of the selected DMPs used to train the machine-learning predictor, additional DMP selection criteria have been established, defining signature positions: an unadjusted p-value ≤ 10e-5 was required to apply Bonferroni family-wise error rate correction for multiple testing and an average |Δꞵ|≥ 0.05 between patients and healthy controls was requested. Accordingly, we identified 36 DMPs (Additional file 5 ). These DMPs were used to construct PCA and heatmap representations as well as to train the episignature classifier (respectively Fig. 2 A, B and C). For consistency with the differential analysis and to reduce artefactual variations, both representations do not display raw methylation levels but differential methylation levels after correction for expected methylation levels according to platform, age at sampling, sex and inferred blood cell composition in the control population. The PCA well separated the SBCS typical patients from the healthy controls on the first axis, which captured 25.8% of the overall methylation dispersion. The heatmap clearly shows a separation between SBCS typical patients and controls. Importantly, no clustering is observed based on array technology (EPIC v1.0 vs EPIC v2.0) or on platform, as both technologies are well intertwined within each group (Fig. 2 B). Of note, prior to CpG differential analysis, patients and controls appeared well matched either on PCA of normalised beta values or residual methylation levels (Additional file 1 : Fig S3A,B). Fig. 2. Open in a new tab A Principal component analysis of adjusted methylation levels, after correcting for expected methylation based on age, sex, and estimated blood cell counts. Status is indicated by colour: green for controls, red for CHD3 typical, purple for CHD3 atypical, blue for CHD3 VUS, while phenotype is determined based on point size, with small points representing controls, mild phenotypes, and not available; medium-sized points representing moderate phenotypes; and large points representing severe phenotypes. The percentage of explained variance is provided for each axis. B Heatmap of adjusted methylation levels displays hierarchical clustering of controls and patients with CHD3 LP/P variants and CHD3 VUS. Blue indicates hypo-methylated positions while red indicates hyper-methylated positions with respect to expected methylation levels at equivalent age, sex and inferred blood cell composition. Status is indicated by colour: green for controls, red for CHD3 typical, purple for CHD3 atypical, and blue for CHD3 VUS. The technology used for each sample is also indicated: yellow for EPIC v1.0 and orange for EPIC v2.0. Also the origin is indicated: grey for Diagenode platform and brown for Rouen. C Pathogenicity scores for each patient with CHD3 LP/P variants were obtained by leave-one-out, and pathogenicity scores for CHD3 atypical and CHD3 VUS were derived from a prediction model based on the complete training set, using a support vector machine predictor. Colours follow the same rules as in panels A and B. D Barplot representing the Gene Ontology (GO) enrichment analysis of differentially methylated genes for DMRs of the CHD3 signature. The x-axis represents the number of genes associated with each Biological Process category. The colour gradient indicates the adjusted p-value, with lower p-values (higher significance) shown in pink and higher p-values in blue VUS patients, most of whom exhibit a mild phenotype, are classified between the patients and controls. The severity of the clinical signs does not correlate with the PCA representation (wide versus small circles). The 4 atypical individuals also demonstrate an atypical methylation pattern. ST80 and ST135 clusters with the controls. Like their clinical phenotype, those patients harbour a different methylation profile than the other SBCS patients. ST56 exhibits a mild phenotype with mainly cardiac anomalies at birth (coarctation of the aorta, bicuspid aortic valve, left ventricular asymmetry), intellectual disability and limited facial dysmorphism. Altogether, she does not recapitulate a typical SBCS, and the methylation analysis corroborates this finding as she clusters with the controls. ST78, a girl, display a very atypical methylation pattern and appears far from the controls and the SBCS cohort (both typical and VUS cases). This girl is the child of a consanguineous couple, exhibits a severe phenotype with a strong neurodevelopment delay, feeding difficulties, Attention-Deficit/Hyperactivity Disorder (ADHD). Patients ST64, ST75, and ST78 all carry the same pathogenic variant, R1172Q. Patient ST78 is considered atypical for the reasons mentioned above (see Additional file 1 : Fig S1). Both patients ST64 and ST75 present with macrocephaly and neurodevelopmental delay associated with poor language development. Patient ST64 additionally exhibits abnormal behaviour characterised by aggressive outbursts and epilepsy treated with Depakine. The dysmorphic features are more pronounced in patient ST64, including brachycephaly, a prominent forehead, a poorly lobulated tragus and helix, and wide spacing between irregularly shaped teeth. In contrast, patient ST75 shows a milder phenotype, leading to their classification as moderate, while patient ST64 is considered severe. Patients ST66 and ST98 carry pathogenic variants affecting the same amino acid (R966) residue but resulting in different substitutions: tryptophan in ST66 and glutamine in ST98. Both patients lack macrocephaly and exhibit more severe intellectual disability than other individuals in the cohort, along with facial dysmorphism, particularly affecting the ears. However, patient ST66 exhibits attention and concentration difficulties, whereas patient ST98 displays autism spectrum-related behavioural disturbances, including self- and hetero-aggression, as well as sleep disorders. In summary, numerous factors contribute to shaping the epigenetic landscape, and even patients carrying the same pathogenic variant may not exhibit strictly overlapping epigenetic profiles. To complement the analysis, instead of focusing on differentially methylated positions (DMPs), we examined differentially methylated regions (DMRs). Again, adjustments for age at sampling, sex, and inferred blood cell composition were accounted for in the analysis. In total, we identified 32 differentially methylated regions (DMRs) at an FDR threshold of 10%, encompassing 218 individual CpG sites (Additional file 5 , Fig. 2 D). When comparing the CpG sites identified in the DMP and DMR analyses, we found that only 13 DMPs (out of the 270) overlapped with DMRs. Sensitivity and specificity of the SBCS episignature We trained a machine-learning classification model, called a support vector machine (SVM), to robustly classify variants as ‘typical CHD3 ’ or ‘negative controls’ based on methylation levels corrected for platform, age at sampling, sex and inferred blood sample composition at signature positions following a leave-one-out cross-validation framework. The SVM model generated a probability ranging from 0 'negative controls' to 1 'typical CHD3 ' for each sample. We started by evaluating the predictive performance of the model on the 23 patients with P/LP CHD3 variants and 62 healthy age-matched controls. Within the cross-validation framework, each positive and negative sample was iteratively left out of the training set. For each iteration, a new SVM classifier was then trained on all but one samples, allowing the excluded sample to serve as internal independent validation. All typical CHD3 cases had a probability of pathogenicity greater than 0.80, except for ST49, ST64, and ST136. In contrast, all negative controls had a probability between 0 and 0.30. Adopting a 50% cut-off, the model achieved an incomplete but high sensitivity of 0.88 (95% CI [0.68–0.97]) and a specificity of 100% (95% CI [0.95–1]) (Fig. 2 C). We further challenged the episignature, first by conducting cross-platform SVM validation, then by extending the leave-one out scheme to the complete EWAS plus SVM process, to validate the overall reproducibility of the episignature. Despite the degradation of the separation between patients and controls, all patients except ST49, ST64, and ST136, maintained SVM scores above 50% and all controls maintained SVM scores below 50% in both scenarios (Additional file 1 : Fig S3C, D). We then classified a test cohort of four patients with P/LP variants but atypical clinical presentation and ten patients with VUS. For atypical CHD3 cases and CHD3 VUS, we trained the model using the 36 DMPs identified during the CHD3 episignature search, selected based on an unadjusted p-value ≤ 10e-5 and an average |Δꞵ|≥ 0.05 between patients and healthy controls. The training was performed on all 23 typical patients and 62 negative controls (Fig. 2 D). The predictor classified all atypical patients as controls, in accordance with their position in the PCA. ST80 and ST135 cluster with the negative controls, while ST78 appears at the bottom of the heatmap, showing no resemblance to the methylation profile of pathogenic cases. The predictor attributed a score of 36% to ST56, which was higher than any of the controls in our training set but did not reach the traditional 50% cut-off. This intermediate score was consistent with its position on the PCA: on the first axis, ST56 lies between the controls and the pathogenic cases, while on the second axis, it appears above the pathogenic cases. Interpretation of the CHD3 VUS remained challenging. The PCA and SVM scores classified three VUS profiles as close to P/LP variant profiles, namely ST53, ST139 and ST69, all harbouring a mild phenotype. All remaining VUS had low SVM scores and found themselves at intermediate positions between patients and controls on the PCA. The two atypical VUS, ST134 and ST46, could not be classified due to a low score. These results confirm that the SVM classifier reliably distinguishes typical CHD3 cases from controls. The intermediate epigenetic profiles observed among the CHD3 VUS samples, in both PCA and SVM probability plots, may reflect biological heterogeneity or misclassification of these variants, rather than a limitation of the episignature itself. This highlights the potential of episignatures as an additional line of evidence for variant interpretation, especially in uncertain cases. CHD3 episignature is specific and differs from the CHD8 and CHD7 episignatures Pathologies associated with pathogenic variants in CHD family genes are associated with distinct pathologies, which may share certain clinical features (Additional file 6 , Fig. 3 A, Additional file 1 : Fig S4A). The question of whether their episignatures also partially overlap or correlate thus arises in the case of mutations in CHD s genes. DMPs associated with each pathology were grouped for a comparative study (Additional file 7 ). For AUTS18 ( CHD8 ) and CHARGE ( CHD7 ) syndromes, multiple episignatures already existed in the literature. The DMP lists published by Siu et al . and Aref-Eshghi et al . respectively were selected as they achieved the best sensitivity in an independent evaluation [ 4 , 7 , 37 ]. No strict overlap was detected between the three lists of DMPs (Additional file 2 : Table S2). The lack of nominal overlap could result from the trimming of DMPs prior to episignature training. However, the degree of similarity between the probes from the three DMP lists revealed that, aside from a small subset of correlated probes, most probe pairs from distinct signatures were largely uncorrelated (Additional file 1 : Fig S4B). Fig. 3. Open in a new tab A Breakdown histogram showing the number of clinical phenotypes associated with each disease based on data from the MONDO database ( https://monarchinitiative.org/ ). B Principal component analysis of adjusted methylation levels, after correcting for expected methylation based on age, sex, and inferred blood cell composition. The probes used correspond to the union of the CHD7 , CHD8 (previously published), and CHD3 (newly identified DMPs) signatures (Additional file 5). Status is indicated by colour: green for controls, red for CHD3 typical patients, purple for CHD8 patients, and orange for CHD7 patients, with a further separation into two subgroups among CHD7 patients, one of which is outlined in black. C Heatmap of adjusted methylation levels, displaying hierarchical clustering of controls and CHD3 typical, CHD8 , and CHD7 patients. Hypomethylated regions are shown in blue, while hypermethylated regions appear in red, relative to the expected methylation levels for individuals of the same age, sex, and inferred blood cell composition. Colours follow the same scheme as in the principal component analysis above. D Raincloud plot illustrating the distribution of adjusted methylation levels. For each category, a mean methylation level is computed per block and per individual, generating a density plot for each group. Below each density distribution, a boxplot visualises the spread of individual data points. Colours follow the same scheme as in the plots on the left. P-values were determined using the Wilcoxon test. Each patient category was compared to the controls, and the two CHD7 subgroups were also compared to each other On the comparative CHD3 - CHD7 - CHD8 PCA, the first axis opposed CHARGE patients to SBCS and AUTS18 patients, while the second axis separated SBCS from AUTS18 patients (Fig. 3 B). The CHD3 - CHD7 - CHD8 heatmap provided the best clustering of samples (columns) and DMPs (rows) into homogeneous groups at seven DMP sets and five sample clusters (Fig. 3 C). While patients and negative controls were perfectly separated between syndromes, CHARGE patients appeared to be split into two very distinct clusters: one cluster (epi-strong) which corresponded to the first and more substantial split on the dendrogram, and a second cluster which was closer to CHD3 patients, in coherence with their position on the first axes of the comparative PCA (epi-mild). To ease the interpretation of the heatmap, Fig. 3 D provides a closer look at the distribution of the average methylation profile on each of these DMP sets by patient. In coherence with the absence of correlation between probes from distinct signatures, most DMP sets were mostly specific to a single signature. Namely, in five out of seven DMP sets, a single signature accounted for more than 70% of the DMPs, suggesting that, beyond the lack of technical overlap between the original DMP lists, the statistical correlation between the three signatures was also very low (probes sets number 2,4,5,6,7). On the DMP axis, DMP set 2, which gathered a majority of CHD3 signature probes (53.6% CHD3 , 34% CHD8 ; Fig. 3 C), unveiled the dichotomy between CHARGE sub-signatures. Compared to average control methylation levels, CHD3 patients presented significant hypomethylated levels, while CHD7 epi-strong patients displayed a significant hypermethylation (Fig. 3 D). In DMP set 4, CHD3 patients presented significant hypermethylation along with CHD7 epi-strong patients. DMPs from CHARGE original episignature were split into two DMP sets (1 and 7) consisting of 31.8% and 53% CHD7 probes, respectively. CHD7 epi-strong and epi-mild patients shared similar profiles hypermethylated on set 7 but significantly differed on set 1, with epi-strong patients displaying stronger hypomethylation levels (Fig. 3 D). PCA analysis of the CHD7 signature confirmed that this gradient was already present within the original signature itself (Additional file 1 : Fig S4B). Four DMP sets (1, 3, 5, 6) included most CHD8 probes. Set 5 was specific to CHD8 strong hypomethylation. On set 1, CHD7 patients resembled AUTS18 patients with hypomethylated levels, epi-strong patients displaying stronger hypomethylation. Set 3 showed CHD3 and CHD8 hypomethylated profiles opposed to CHD7 epi-strong highly hypermethylated profiles. On set 6, CHD7 epi-strong patients mirrored CHD8 profiles: while CHD8 patients were hypomethylated, CHD7 epi-strong patients were hypermethylated (Fig. 3 D). Interestingly, the addition of atypical patients to the PCA analysis (Additional file 1 : Fig S4C) shows that patient ST78, the most severely affected, clusters with the most severe CHARGE epi-strong patients, while patient ST56 clusters with the AUST18 patients. SVM analysis revealed that inter-syndrome specificity remained excellent. However, for certain CHARGE or AUST18 patients, the probability of pathogenicity could reach up to 40%, highlighting the importance of cross-over studies to better understand the strengths and limitations of methylation analyses (Additional file 1 : Fig S4D). New insights on CHARGE episignature The atypical pattern of CHD7 patients in Fig. 3 B and D (probe set 1, 2, 3, 4, 6) raises questions about the accuracy of the CHARGE syndrome episignature. In particular, within CHARGE patients, epi-strong cluster patients show distinctive patterns of methylation within a subset of the original CHARGE signature DMPs, but also extremely altered methylation levels on both CHD3 and CHD8 signatures. SkinHorvath analysis showed that there was no bias in the biological age between the two patient groups, epi-strong and epi-mild (Additional file 1 : Fig S5A). The two DMP lists given by Aref-Eshghi et al . and Butcher et al . were combined, and a PCA analysis revealed that the two groups of CHARGE patients were well distinguished (Additional file 1 : Fig S5B,C, Additional file 2 : Table S2) [ 4 , 48 ]. When the same analysis with separate DMPs lists is performed, the two different profiles are also observed, although the distinction is less clear (Additional file 1 : Fig S5D,E). Back in the literature, Aref-Eshghi et al. 's (Fig. 4 B in the original paper) clearly shows two groups of CHARGE patients, without this dichotomy being commented on [ 4 ]. Fig. 4. Open in a new tab A , B , C , D , E IGV visualisation examples showing DMPs, DMRs, and CHD3 ChIP-Seq results (public data, see Mat & Meth), which often colocalise at gene promoters. F Venn diagram representing the common positions between DMPs and DMRs of the CHD3 signature and CHD3 ChIP-Seq We therefore sought to better understand why two profiles appeared with the same episignature and whether this methylation pattern correlated with phenotype profiles. To do this, we collected clinical data on CHARGE patients (Additional file 7). We observed that patients with an epi-strong profile were the most severely affected and had also had their blood drawn very early in life, 6/8 being less than one month old at blood sampling, whereas those in the epi-mild group were less severely affected and had their blood drawn later in life. CHD3 DMPs and DMRs correlate with ChIP-Seq data at some genomic regions To better understand the location of the DMRs in the patients, we compared them with the regions where the CHD3 protein is located. To do this, we used ChIP-Seq data from two previous studies in the literature, obtained from an erythroblast cell line (K562) and a cell line derived from a lymph node from prostate adenocarcinoma (LNCaP) [ 41 , 43 ]. The Venn diagram shows that 1 DMP is simultaneously found in the CHD3 ChIP-Seq peaks, the set of DMPs, and within DMRs, as illustrated in Fig. 4 . Although the small number of events does not allow statistical enrichment analyses, the recurrence of regions of interest (DMPs and/or DMRs) is observed at promoters (e.g. METTL8 ) and alternative transcription start sites (e.g. TCEA2 ), with or without detectable CHD3 binding in ChIP-Seq data, represent promising avenues for further investigation. Discussion CHD3 cohort Herein, we reported a new SBCS cohort of 38 patients across Europe. Among them, 23 individuals harbour typical SBCS phenotype. Only truncated or frameshift variants were located in the N-terminal or C-terminal domains, whereas variants found in the central chromo and helicase domains of the protein are exclusively missense mutations. Ten individuals carrying VUS were classified separately; they are all missense variants (except one) distributed across the CHD3 gene. Interestingly, three out of the ten VUS involve arginine substitutions (mutated to tryptophan in one individual, ST67, and to cysteine in two, ST72 and ST68), with a probable strong impact on the CHD3 function [ 49 ]. We focused on the 23 individuals showing typical SBCS with pathogenic variants to identify 270 DMPs in the genome mostly located in gene promoters. A more stringent filtering approach identified 36 DMPs highly specific to the SBCS typical patients. To date, our data do not allow correlation of clinical severity with DNA methylation profiles. Patients' classification The identified DMPs place CHD3 in the realm of mild episignatures, easily affected by technical artefacts and requiring methodological precautions. The episignature reaches high but incomplete sensitivity on patients with typical SBCS phenotype along with high specificity. Patients with atypical phenotype do not share the signature. The separation between typical patients and controls weakens following cross-platform or extended cross-validation but preserves prediction performances on our dataset. All in all, the episignature allowed reclassification of three VUSs out of ten (ST69 NM_001005273.2 :c.5198C > T, ST53 c.1840G > A and ST139 c.1072_1074del) as likely pathogenic variants with at least good confidence (ST72 c.1903C > T with an intermediate status requires a deeper analysis). Despite a good separation of positive and negative controls, seven VUSs showed intermediate episignature profiles that preclude definitive classification. A larger training set, including an increased number of positive and negative controls and a more refined distinction between typical and atypical categories, could help clarify patient classification and enhance the reproducibility of the episignature. Individuals with atypical SBCS clinical presentation also displayed an atypical methylation pattern (far from the control and typical CHD3 group), various scenarios come into play. The patient with the most severe phenotype has related parents (ST78), so it is very likely that the CHD3 variant is not the only one contributing to the observed phenotype. In the case of patient ST56, the cardiac malformation unreported in the other patients in our cohort correlates with a distinct methylation profile. Finally, the ST80 individual ( NM_001005273.3 ( CHD3 ):c.5754G > A, silent mutation possibly affecting the splicing site, patient’s RNA not available), atypical for his short stature and microcephaly, preserved intelligence and digestive disorder, presented a methylation profile similar to the controls. However, RNA analysis done by Australian collaborators in ST135 (short stature, small OFC, moderate or mild attention deficit) confirms the pathogenicity of the mutation (not published yet). One hypothesis is that this variant may indeed be responsible for the clinical symptoms observed in this patient, although the resulting pathology could differ from typical SBCS. Further investigations are required to better understand the pathophysiological mechanisms of those variants. Interpreting episignatures Numerous previous studies have explored episignatures by distinguishing well-characterised positive cases used to define the episignature from less definitive cases, where the diagnosis was presumed based on a VUS requiring further evaluation. Our analyses show that it is important to consider the DNA methylation profile as a biomarker connected to the clinical profile. The combination of both the clinical and the DNA methylation profiles is crucial to ensure the correct interpretation of the variants. As an example, the four atypical cases pointed out based on the clinical questionnaire also harbour atypical DNA methylation profiles. DNA methylation profile could then be considered as a molecular phenotype (phenotype at the cellular level) of the disease. Specific methodology A significant methodological strength of this study is the identification of an episignature by combining data from two different methylation platforms as well as two methylation array technologies. To our knowledge, there have been no studies that have established a reliable episignature integrating methylation profiles generated from both platforms simultaneously. This represents a significant technical challenge, as differences in probe design, coverage, and array performance can introduce substantial biases. To surmount these obstacles, a rigorous and tailored meta-analysis approach was implemented. Differential methylation analyses were performed separately for each platform, with careful adjustment for confounding factors such as age at sampling, sex, and inferred blood cell composition. Subsequent to this, a meta-analysis was conducted to select only those robust and reproducible DMPs. This strategy was developed to minimise batch effects and ensure that the identified signatures truly reflected biological differences rather than technical artefacts associated with platform changes. Consequently, the methodological approach adopted was pivotal to the success of the CHD3 episignature discovery, underscoring the critical importance of adapting analytical strategies in response to rapidly evolving epigenomic profiling technologies. Extension of episignatures Studying the episignature of SBCS enabled more refined molecular and clinical characterisation of affected individuals. Comparative analyses with other chromatinopathies allowed us to further characterise methylation signatures associated with AUST18 and CHARGE syndromes. Notably, two distinct sub-profiles emerged within CHARGE syndrome. While previously unrecognised, these subgroups became evident with the integration of multiple episignatures and focus on the CHD sub-family [ 4 ]. A closer examination of the clinical data suggests that epi-strong patients exhibit more severe phenotypes and undergo earlier molecular diagnostic investigations. At present, we cannot discard some age effect that linear adjustment cannot sufficiently account for. Additional blood resampling at an older age for the most severely affected patients could help confirm whether the distinction between these two groups truly reflects severity more than an artefact of age at sampling. The rise of new long-read sequencing technologies, which can detect not only canonical bases but also modified DNA bases such as methylation or hydroxymethylation, among others, is a real time-saver for patient diagnosis, since in a single experiment the genetic and epigenetic sequence is available [ 50 ]. Bioinformatics pipelines are rapidly evolving to integrate and analyse patient methylation profiles by including both types of analysis, microarrays and long-read sequencing (and potentially other technologies which are more costly and less common, such as WGBS, RRBS or MeDIP-Seq) [ 51 , 52 ]. At present, microarray data is analysed by various bioinformatics pipelines lacking precise conventions, even though this valuable patient data warrants meticulous analysis with the highest standards of rigour and transparency [ 53 ]. Analysis tools need to be shared and harmonised to enable fast and accurate diagnoses for patients. Finally, since epigenetic dysregulation in chromatinopathies extends beyond DNA methylation, future work should also consider other epigenetic modifications, potentially employing single-cell resolution technologies. Conclusions Based on the detailed clinical characterisation of a cohort of 38 patients, including 23 individuals carrying likely pathogenic or pathogenic CHD3 variants and presenting with typical features of Snijders Blok-Campeau syndrome (SBCS), we identified a CHD3-associated DNA methylation episignature that supports molecular diagnosis and variant interpretation. This methylation signal is moderate, with partial loss of sensitivity observed during cross-platform validation, reflecting real-world variability in clinical testing; however, it remains specific to individuals exhibiting the canonical SBCS phenotype. The currently limited sensitivity across the full phenotypic spectrum of CHD3 -associated conditions (typical and atypical) underscores that DNA methylation profiling should complement, rather than replace, genomic analyses in diagnostic workflows. Expanding reference cohorts, refining clinically defined subgroups, and harmonising analytical strategies across technologies will be essential to improve robustness and enable reliable clinical implementation of episignatures in chromatinopathies. Supplementary Information 13073_2026_1639_MOESM1_ESM.docx (2.1MB, docx) Additional file 1. Fig S1. Clinical and structural characterization of CHD3 variants. (A) Representative patient photographs illustrating severity spectrum and distinction between typical and atypical forms. (B) Structural representation of CHD3 protein, shown from an alternative angle focusing on the ATP-binding site. Fig S2. Quality control and differential methylation meta-analysis. (A) Correlation between SkinHorvath epigenetic age prediction and chronological age at blood sampling (EPIC v1 in yellow; EPIC v2 in orange). (B) Barplot showing the number of probes per category. (C) Volcano plot of the meta-analysis including 62 controls and 23 CHD3 pathogenic variant carriers. Fig S3. Principal component analyses and cross-validation procedures. (A) PCA based on normalized beta values. (B) PCA based on residual methylation values. (C) Cross-platform SVM validation using reciprocal training and validation sets. (D) Global leave-one-out cross-validation integrating EWAS filtering (FDR < 50%) and SVM training. Fig S4. Comparative analyses of CHD-family episignatures. (A) Phylogenetic analysis of the human CHD ATPase remodeler subfamily using Neighbor-Joining (JTT model; 433 conserved amino acids; MAFFT). (B) Correlation matrix of probes from the union of CHD7, CHD8, and CHD3 signatures. (C) PCA of adjusted methylation levels using the combined probe set. (D) Extended PCA and pathogenicity score prediction including CHARGE and AUTS18 patients. Fig S5. Analysis of CHARGE episignature subtypes. (A) Correlation between epigenetic and chronological age in CHD7 epi-strong and epi-mild groups. (B) Overlap between Aref-Eshghi and Butcher CHARGE episignatures. (C,D,E) PCA analyses using previously published CHD7/CHARGE episignatures. 13073_2026_1639_MOESM2_ESM.xlsx (21.2KB, xlsx) Additional file 2. Table S1: Summary of CHD3 variants reported in the literature. Compilation of previously published CHD3 cases, including variant types (frameshift, missense, nonsense, splice-site), zygosity, number of affected individuals per study, proportion of missense variants, and proportion of recurrent missense variants across cohorts. Table S2: List of differentially methylated probes (DMPs) from published CHARGE and AUTS18 episignatures. Comprehensive list of CpG probes previously reported in CHARGE and AUTS18 episignatures, including probe identifiers grouped by condition and reference study. 13073_2026_1639_MOESM3_ESM.docx (15.6KB, docx) Additional file 3. Supplementary data for CHD3 patients 13073_2026_1639_MOESM4_ESM.xlsx (19.6KB, xlsx) Additional file 4. CHD3 patient phenotypes Additional file 5. CHD3 DMPs and DMRs (295.2KB, xlsx) 13073_2026_1639_MOESM6_ESM.xlsx (9.8KB, xlsx) Additional file 6. CHD3 CHD7 CHD8 phenotypes 13073_2026_1639_MOESM7_ESM.xlsx (88.5KB, xlsx) Additional file 7. CHD7 patient clinical table Acknowledgements We thank Sophie Rondeau, Véronique Pingault, Angelique Gaba, Elise Pisan, Zaina Ait Arkoub (APHP Necker, France), Anne Dieux, Marie Balerdi (CHU Lille, France), Catherine Vincent-Delorme (CHU Arras, France), Océane Coudrieu (CHU Montpellier, France), Katarina Vulin and Ljubica Odak (Children's Hospital Zagreb, Croatia), Marie Bournez, Lucie Dauvier and Jade Heitz (CHU Dijon, France), Nicolas Garcelon and Céline Huber (Institut Imagine, France), Godeliève Morel (CHU la Réunion, France), Gema Escribano Serrano (Hospital Sant Joan de Déu, Spain) for their help in collecting clinical data and patient samples and helpful discussions. We thank Nicolas Doldi and Emilia Puig Lombardi (Institut Imagine, France) for their help in bioinformatics. We thank Catherine Schramm (Inserm U1245 Rouen, France) for her helpful discussions on statistical analyses. We thank Caterina Lucano and Marta Fructuoso (Orphanet, INSERM, France) for their valuable input on the ontology of chromatinopathies. We thank Matthieu Gérard (CEA Saclay, France) for initial discussion and previous work on the CHD family [15]. We acknowledge the use of the bioresources of the Necker Imagine DNA biobank (BB-033-00065). We thank our colleagues for their helpful discussions and constant support. We thank the European Reference Network for Rare Malformation Syndromes, Intellectual and Other Neurodevelopmental Disorders (ERN ITHACA) and the French rare diseases Healthcare Network: rare developmental defect and rare intellectual disability (FSMR AnDDI-rares) networks for helping to collect CHD3 patients. Some authors of this publication are members of the European Reference Network on Rare Congenital Malformations and Rare Intellectual Disability ERN-ITHACA. ERN-ITHACA is funded by the European Union, under the grant agreement N°101156387. Authors’ contributions AS, CaC performed the bioinformatical analyses; AT, MD, Md, GV performed preliminary bioinformatical analyses. ACR, AM, FL and GN contributed to the in-house DNA methylation analyses on the ASGARD platform. GP performed CHD3 structure analysis. FM helped with the CHD3 ChIP-seq analysis. PM, Md, VCD, CaC performed patient classification. CM, ACE, JMS, MJBM, SL, IS, TSB, PB, JG, XLG, PLT, YH, MW, MZ, IS, SM, BI, AP, AY, JL, FR, LPP, MR, TC, HHA, SD, JSL, VCD recruited SBCS patients. PM, AM, PR, AMG, NC, FD, AG, JD, LOF, AC, CoC, BI recruited healthy donors and CHARGE patients. VCD and Md had the original idea. CaC and Md received funding for this study. AS, CaC and Md co-wrote the manuscript. AS, CaC, PM, GV, MD, VCD and Md revised the manuscript. The manuscript has been proofread and accepted by all the authors before submission. Funding This work has been supported by the “Dotation nouveaux recruté INSERM 2021” and the Data Intelligence Institute of Paris (diiP), IdEx Université Paris Cité (ANR-18-IDEX-0001) and Fondation Maladies Rares (Omics2024-15089) to Md and 2022 MESSIDORE program (Inserm-MESSIDORE N°19) to CaC. AS received a grant from Region Normandie and GIRCI Nord Ouest (FHU-A2M2P). This work was supported by the French Ministry of Health within the framework of the National Plan for Rare Diseases, through a funding of the French Rare Disease Network for Malformations of the Head, Neck and Teeth (FSMR TETECOU). The Barakat lab was supported by the Netherlands Organisation for Scientific Research (ZonMw Vidi, grant 09150172110002). Data availability Methylation data generated at the Rouen ASGARD platform have been deposited in the European Genome-phenome Archive (EGA, hosted by EMBL-EBI) under study accession number EGAS00001008070 ([ https://ega-archive.org/studies/EGAS00001008070 ]). Similarly, methylation data generated at the Diagenode platform are available under the study accession number EGAS00001008414. These datasets are subject to a Data Processing Agreement, and access requests will be reviewed by a Data Access Committee to ensure compliance with ethical and legal requirements. Declarations Ethics approval and consent to participate This research was conducted as part of the research programmes of two biobanks: Imagine/INSERM U1163 (DC-2019–3504) and the Biological Resource Centre (BRC) in Rouen (DC 2008–711). The analysis of methylation profiles using previously stored DNA under these conditions was approved by the Ethics Committee for Research on Existing Data and/or outside the Jardé Law (CERDE) of Rouen University Hospital (approval no. E2023-13). The Necker Biological Resource Centre (BRC) received approval from the Ethics Committee for Research of AP-HP Centre (CERAPHP Centre). In accordance with data protection regulations, informed consent for the use of DNA in analyses complementary to sequencing was obtained from each patient by their attending physician. All samples were anonymised at the clinical site prior to their transfer to the laboratory, where they were used for downstream analyses. This study was conducted in accordance with the Declaration of Helsinki. Consent for publication Written informed consent for the use of biological samples and, where applicable, for the publication of photographs was obtained by the treating physicians at their respective institutions. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References 1. Nguengang Wakap S, Lambert DM, Olry A, Rodwell C, Gueydan C, Lanneau V, et al. Estimating cumulative point prevalence of rare diseases: analysis of the Orphanet database. Eur J Hum Genet. Nature Publishing Group; 2020;28:165–73. 10.1038/s41431-019-0508-0 [ DOI ] [ PMC free article ] [ PubMed ] 2. Pichon T, Messiaen C, Soussand L, Angin C, Sandrin A, Elarouci N, et al. Overview of patients’ cohorts in the French National rare disease registry. Orphanet J Rare Dis. 2023;18:176. 10.1186/s13023-023-02725-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. de Dieuleveult M, Velasco G. Les maladies génétiques de la machinerie épigénétique. médecine/sciences. 2024;40:914–24. 10.1051/medsci/2024181. (EDP Sciences). [ DOI ] [ PubMed ] [ Google Scholar ] 4. Aref-Eshghi E, Kerkhof J, Pedro VP, Groupe DI France, Barat-Houari M, Ruiz-Pallares N, et al. Evaluation of DNA Methylation Episignatures for Diagnosis and Phenotype Correlations in 42 Mendelian Neurodevelopmental Disorders. Am J Hum Genet. 2020;106:356–70. 10.1016/j.ajhg.2020.01.019. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Levy MA, McConkey H, Kerkhof J, Barat-Houari M, Bargiacchi S, Biamino E, et al. Novel diagnostic DNA methylation episignatures expand and refine the epigenetic landscapes of Mendelian disorders. Hum Genet Genomics Adv. 2022;3:100075. 10.1016/j.xhgg.2021.100075. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Awamleh Z, Goodman S, Kallurkar P, Wu W, Lu K, Choufani S, et al. Generation of DNA methylation signatures and classification of variants in rare neurodevelopmental disorders using EpigenCentral. Curr Protoc. 2022;2:e597. 10.1002/cpz1.597. [ DOI ] [ PubMed ] [ Google Scholar ] 7. Husson T, Lecoquierre F, Nicolas G, Richard A-C, Afenjar A, Audebert-Bellanger S, et al. Episignatures in practice: independent evaluation of published episignatures for the molecular diagnostics of ten neurodevelopmental disorders. Eur J Hum Genet. 2023. 10.1038/s41431-023-01474-x. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Fu MP, Merrill SM, Sharma M, Gibson WT, Turvey SE, Kobor MS. Rare diseases of epigenetic origin: Challenges and opportunities. Front Genet. 2023. 10.3389/fgene.2023.1113086. (Frontiers). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Fahrner JA, Bjornsson HT. Mendelian disorders of the epigenetic machinery: postnatal malleability and therapeutic prospects. Hum Mol Genet. 2019;28:R254–64. 10.1093/hmg/ddz174. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Harris JR, Gao CW, Britton JF, Applegate CD, Bjornsson HT, Fahrner JA. Five years of experience in the Epigenetics and Chromatin Clinic: what have we learned and where do we go from here? Hum Genet. 2024;143:607–24. 10.1007/s00439-023-02537-1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Bukowska-Olech E, Majchrzak-Celińska A, Przyborska M, Jamsheer A. Chromatinopathies: insight in clinical aspects and underlying epigenetic changes. J Appl Genet. 2024. 10.1007/s13353-023-00824-1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Clapier CR, Cairns BR. The biology of chromatin remodeling complexes. Annu Rev Biochem. 2009;78:273–304. 10.1146/annurev.biochem.77.062706.153223. [ DOI ] [ PubMed ] [ Google Scholar ] 13. Liu C, Kang N, Guo Y, Gong P. Advances in Chromodomain Helicase DNA-Binding (CHD) Proteins Regulating Stem Cell Differentiation and Human Diseases. Front Cell Dev Biol [Internet]. Frontiers; 2021;9. 10.3389/fcell.2021.710203 [ DOI ] [ PMC free article ] [ PubMed ] 14. Alendar A, Berns A. Sentinels of chromatin: chromodomain helicase DNA-binding proteins in development and disease. Genes Dev. 2021;35:1403–30. 10.1101/gad.348897.121. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. de Dieuleveult M, Yen K, Hmitou I, Depaux A, Boussouar F, Dargham DB, et al. Genome-wide nucleosome specificity and function of chromatin remodellers in ES cells. Nature. 2016;530:113–6. 10.1038/nature16505. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Trujillo JT, Long J, Aboelnour E, Ogas J, Wisecaver JH. CHD Chromatin Remodeling Protein Diversification Yields Novel Clades and Domains Absent in Classic Model Organisms. Genome Biol Evol. 2022;14:evac066. 10.1093/gbe/evac066. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Muhammad T, Pastore SF, Good K, Ausió J, Vincent JB. Chromatin gatekeeper and modifier CHD proteins in development, and in autism and other neurological disorders. Psychiatr Genet. 2023;33:213. 10.1097/YPG.0000000000000353. [ DOI ] [ PubMed ] [ Google Scholar ] 18. Sadler B, Wilborn J, Antunes L, Kuensting T, Hale AT, Gannon SR, et al. Rare and de novo coding variants in chromodomain genes in Chiari I malformation. Am J Hum Genet. 2021;108:100–14. 10.1016/j.ajhg.2020.12.001. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Mossink B, Negwer M, Schubert D, Nadif Kasri N. The emerging role of chromatin remodelers in neurodevelopmental disorders: a developmental perspective. Cell Mol Life Sci. 2021;78:2517–63. 10.1007/s00018-020-03714-5. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Kasah S, Oddy C, Basson MA. Autism-linked CHD gene expression patterns during development predict multi-organ disease phenotypes. J Anat. 2018;233:755–69. 10.1111/joa.12889. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Snijders Blok L, Rousseau J, Twist J, Ehresmann S, Takaku M, Venselaar H, et al. CHD3 helicase domain mutations cause a neurodevelopmental syndrome with macrocephaly and impaired speech and language. Nat Commun. 2018;9:4619. 10.1038/s41467-018-06014-6. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Pascual P, Tenorio-Castano J, Mignot C, Afenjar A, Arias P, Gallego-Zazo N, et al. Snijders blok-campeau syndrome: description of 20 additional individuals with variants in CHD3 and literature review. Genes. 2023;14:1664. 10.3390/genes14091664. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Coursimault J, Lecoquierre F, Saugier-Veber P, Drouin-Garraud V, Lechevallier J, Boland A, et al. Hypersociability associated with developmental delay, macrocephaly and facial dysmorphism points to CHD3 mutations. Eur J Med Genet. 2021;64:104166. 10.1016/j.ejmg.2021.104166. [ DOI ] [ PubMed ] [ Google Scholar ] 24. Drivas TG, Li D, Nair D, Alaimo JT, Alders M, Altmüller J, et al. A second cohort of CHD3 patients expands the molecular mechanisms known to cause Snijders Blok-Campeau syndrome. Eur J Hum Genet. 2020;28:1422–31. 10.1038/s41431-020-0654-4. (Nature Publishing Group). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Mizukami M, Ishikawa A, Miyazaki S, Tsuzuki A, Saito S, Niihori T, et al. A de novo CHD3 variant in a child with intellectual disability, autism, joint laxity, and dysmorphisms. Brain Dev. 2021;43:563–5. 10.1016/j.braindev.2020.12.004. [ DOI ] [ PubMed ] [ Google Scholar ] 26. Goldfarb Yaacobi R, Sukenik Halevy R. A severe neurocognitive phenotype caused by biallelic CHD3 variants in two siblings. Am J Med Genet A. 2024;194:e63503. 10.1002/ajmg.a.63503. [ DOI ] [ PubMed ] [ Google Scholar ] 27. Gao Y, Wang P, Chen M, Pang K, Sun Y, Zheng B, et al. Novel genotypes and phenotypes in Snijders blok-campeau syndrome caused by CHD3 mutations. Front Genet. 2024;15:1347933. 10.3389/fgene.2024.1347933. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. LeBreton L, Allain EP, Parscan RC, Crapoulet N, Almaghraby A, Ben Amor M. A novel CHD3 variant in a patient with central precocious puberty: expanded phenotype of Snijders Blok-Campeau syndrome? Am J Med Genet A. 2023;191:1065–9. 10.1002/ajmg.a.63096. [ DOI ] [ PubMed ] [ Google Scholar ] 29. van der Spek J, den Hoed J, Snijders Blok L, Dingemans AJM, Schijven D, Nellaker C, et al. Inherited variants in CHD3 show variable expressivity in Snijders Blok-Campeau syndrome. Genet Med Off J Am Coll Med Genet. 2022;24:1283–96. 10.1016/j.gim.2022.02.014. [ DOI ] [ PubMed ] [ Google Scholar ] 30. Tie X, Che F, Liu S, Mo L, Zhang L, Li B, et al. Insights From a Novel Splicing Variant and Recurrent Arginine Variants in the CHD3 Gene Causing Snijders Blok-Campeau Syndrome. Am J Med Genet A. 2024. 10.1002/ajmg.a.63930. [ DOI ] [ PubMed ] [ Google Scholar ] 31. Jumper J, Evans R, Pritzel A, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583–9. 10.1038/s41586-021-03819-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Schrödinger LLC. The PyMOL Molecular Graphics System, Version 2.5. 2022. https://pymol.org/2/ 33. Min JL, Hemani G, Davey Smith G, Relton C, Suderman M. Meffil: efficient normalization and analysis of very large DNA methylation datasets. Bioinformatics. 2018;34:3983–9. 10.1093/bioinformatics/bty476. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. Pelegí-Sisó D, de Prado P, Ronkainen J, Bustamante M, González JR. Methylclock: a bioconductor package to estimate DNA methylation age. Bioinformatics. 2021;37:1759–60. 10.1093/bioinformatics/btaa825. [ DOI ] [ PubMed ] [ Google Scholar ] 35. Kolde R. pheatmap: Pretty heatmaps. R package version 1.0.12. https://CRAN.R-project.org/package=pheatmap 36. Meyer D, et al. e1071: Misc functions for probability, confidence and ROC. R package version 1.7–14. https://CRAN.R-project.org/package=e1071 37. Siu MT, Butcher DT, Turinsky AL, Cytrynbaum C, Stavropoulos DJ, Walker S, et al. Functional DNA methylation signatures for autism spectrum disorder genomic risk loci: 16p11.2 deletions and CHD8 variants. Clin Epigenetics. 2019;11:103. 10.1186/s13148-019-0684-3. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. Peters TJ, et al. DMRcate: differential methylation analysis for BS-seq data. Bioinformatics. 2015;31(13):1992–2000. https://bioconductor.org/packages/DMRcate 39. Yu G, et al. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS. 2012;16(5):284–287. https://bioconductor.org/packages/clusterProfiler [ DOI ] [ PMC free article ] [ PubMed ] 40. Chen H, Boutros PC. VennDiagram: Generate high-resolution Venn and Euler plots. R package version 1.7.3. https://CRAN.R-project.org/package=VennDiagram 41. Ram O, Goren A, Amit I, Shoresh N, Yosef N, Ernst J, et al. Combinatorial patterning of chromatin regulators uncovered by genome-wide location analysis in human cells. Cell. 2011;147:1628–39. 10.1016/j.cell.2011.09.057. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 42. Zheng R, et al. Cistrome data browser: expanded datasets and new tools for gene regulatory analysis. Nucleic Acids Res. 2019;47(D1):D729–35. 10.1093/nar/gky1094. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Ye Z, Chen Z, Sunkel B, Frietze S, Huang TH-M, Wang Q, et al. Genome-wide analysis reveals positional-nucleosome-oriented binding pattern of pioneer factor FOXA1. Nucleic Acids Res. 2016;44:7540–54. 10.1093/nar/gkw659. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Yu G, et al. Model-based Analysis of ChIP-Seq (MACS3). https://github.com/macs3-project/MACS (2023) 45. Xie Y, Li H, Luo X, Li H, Gao Q, Zhang L, et al. IBS 2.0: an upgraded illustrator for the visualization of biological sequences. Nucleic Acids Res. 2022;50:W420–6. 10.1093/nar/gkac373. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 46. Farnung L, Ochmann M, Cramer P. Nucleosome-CHD4 chromatin remodeler structure maps human disease mutations. Elife. 2020;9:e56178. 10.7554/eLife.56178. (eLife Sciences Publications, Ltd). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 47. Zhong Y, Moghaddas Sani H, Paudel BP, Low JKK, Silva APG, Mueller S, et al. The role of auxiliary domains in modulating CHD4 activity suggests mechanistic commonality between enzyme families. Nat Commun. 2022;13:7524. 10.1038/s41467-022-35002-0. (Nature Publishing Group). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 48. Butcher DT, Cytrynbaum C, Turinsky AL, Siu MT, Inbar-Feigenberg M, Mendoza-Londono R, et al. CHARGE and Kabuki syndromes: gene-specific DNA methylation signatures identify epigenetic mechanisms linking these clinically overlapping conditions. Am J Hum Genet. 2017;100:773–88. 10.1016/j.ajhg.2017.04.004. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 49. Tsuber V, Kadamov Y, Brautigam L, Warpman Berglund U, Helleday T. Mutations in cancer cause gain of cysteine, histidine, and tryptophan at the expense of a net loss of arginine on the proteome level. Biomolecules. 2017;7:49. 10.3390/biom7030049. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 50. Lucas MC, Novoa EM. Long-read sequencing in the era of epigenomics and epitranscriptomics. Nat Methods. 2023;20:25–9. 10.1038/s41592-022-01724-8. [ DOI ] [ PubMed ] [ Google Scholar ] 51. Rauluseviciute I, Drabløs F, Rye MB. DNA methylation data by sequencing: experimental approaches and recommendations for tools and pipelines for data analysis. Clin Epigenetics. 2019;11:193. 10.1186/s13148-019-0795-x. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 52. Arora I, Tollefsbol TO. Computational methods and next-generation sequencing approaches to analyze epigenetics data: profiling of methods and applications. Methods. 2021;187:92–103. 10.1016/j.ymeth.2020.09.008. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 53. Sahoo K, Sundararajan V. Methods in DNA methylation array dataset analysis: a review. Comput Struct Biotechnol J. 2024;23:2304–25. 10.1016/j.csbj.2024.05.015. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials 13073_2026_1639_MOESM1_ESM.docx (2.1MB, docx) Additional file 1. Fig S1. Clinical and structural characterization of CHD3 variants. (A) Representative patient photographs illustrating severity spectrum and distinction between typical and atypical forms. (B) Structural representation of CHD3 protein, shown from an alternative angle focusing on the ATP-binding site. Fig S2. Quality control and differential methylation meta-analysis. (A) Correlation between SkinHorvath epigenetic age prediction and chronological age at blood sampling (EPIC v1 in yellow; EPIC v2 in orange). (B) Barplot showing the number of probes per category. (C) Volcano plot of the meta-analysis including 62 controls and 23 CHD3 pathogenic variant carriers. Fig S3. Principal component analyses and cross-validation procedures. (A) PCA based on normalized beta values. (B) PCA based on residual methylation values. (C) Cross-platform SVM validation using reciprocal training and validation sets. (D) Global leave-one-out cross-validation integrating EWAS filtering (FDR < 50%) and SVM training. Fig S4. Comparative analyses of CHD-family episignatures. (A) Phylogenetic analysis of the human CHD ATPase remodeler subfamily using Neighbor-Joining (JTT model; 433 conserved amino acids; MAFFT). (B) Correlation matrix of probes from the union of CHD7, CHD8, and CHD3 signatures. (C) PCA of adjusted methylation levels using the combined probe set. (D) Extended PCA and pathogenicity score prediction including CHARGE and AUTS18 patients. Fig S5. Analysis of CHARGE episignature subtypes. (A) Correlation between epigenetic and chronological age in CHD7 epi-strong and epi-mild groups. (B) Overlap between Aref-Eshghi and Butcher CHARGE episignatures. (C,D,E) PCA analyses using previously published CHD7/CHARGE episignatures. 13073_2026_1639_MOESM2_ESM.xlsx (21.2KB, xlsx) Additional file 2. Table S1: Summary of CHD3 variants reported in the literature. Compilation of previously published CHD3 cases, including variant types (frameshift, missense, nonsense, splice-site), zygosity, number of affected individuals per study, proportion of missense variants, and proportion of recurrent missense variants across cohorts. Table S2: List of differentially methylated probes (DMPs) from published CHARGE and AUTS18 episignatures. Comprehensive list of CpG probes previously reported in CHARGE and AUTS18 episignatures, including probe identifiers grouped by condition and reference study. 13073_2026_1639_MOESM3_ESM.docx (15.6KB, docx) Additional file 3. Supplementary data for CHD3 patients 13073_2026_1639_MOESM4_ESM.xlsx (19.6KB, xlsx) Additional file 4. CHD3 patient phenotypes Additional file 5. CHD3 DMPs and DMRs (295.2KB, xlsx) 13073_2026_1639_MOESM6_ESM.xlsx (9.8KB, xlsx) Additional file 6. CHD3 CHD7 CHD8 phenotypes 13073_2026_1639_MOESM7_ESM.xlsx (88.5KB, xlsx) Additional file 7. CHD7 patient clinical table Data Availability Statement Methylation data generated at the Rouen ASGARD platform have been deposited in the European Genome-phenome Archive (EGA, hosted by EMBL-EBI) under study accession number EGAS00001008070 ([ https://ega-archive.org/studies/EGAS00001008070 ]). Similarly, methylation data generated at the Diagenode platform are available under the study accession number EGAS00001008414. These datasets are subject to a Data Processing Agreement, and access requests will be reviewed by a Data Access Committee to ensure compliance with ethical and legal requirements. Articles from Genome Medicine are provided here courtesy of BMC ACTIONS View on publisher site PDF (2.6 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top