STIGMA: Single-cell tissue-specific gene prioritization using machine learning.

Clinical exome and genome sequencing have revolutionized the understanding of human disease genetics. Yet many genes remain functionally uncharacterized, complicating the establishment of causal disease links for genetic variants. While several scoring methods have been devised to prioritize these candidate genes, these methods fall short of capturing the expression heterogeneity across cell subpopulations within tissues. Here, we introduce single-cell tissue-specific gene prioritization using machine learning (STIGMA), an approach that leverages single-cell RNA-seq (scRNA-seq) data to prioritize candidate genes associated with rare congenital diseases. STIGMA prioritizes genes by learning the temporal dynamics of gene expression across cell types during healthy organogenesis. To assess the efficacy of our framework, we applied STIGMA to mouse limb and human fetal heart scRNA-seq datasets. In a cohort of individuals with congenital limb malformation, STIGMA prioritized 469 variants in 345 genes, with UBA2 as a notable example. For congenital heart defects, we detected 34 genes harboring nonsynonymous de novo variants (nsDNVs) in two or more individuals from a set of 7,958 individuals, including the ortholog of Prdm1, which is associated with hypoplastic left ventricle and hypoplastic aortic arch. Overall, our findings demonstrate that STIGMA effectively prioritizes tissue-specific candidate genes by utilizing single-cell transcriptome data. The ability to capture the heterogeneity of gene expression across cell populations makes STIGMA a powerful tool for the discovery of disease-associated genes and facilitates the identification of causal variants underlying human genetic disorders.

Mosaic divergent repeat interruptions in XDP influence repeat stability and disease onset.

While many genetic causes of movement disorders have been identified, modifiers of disease expression are largely unknown. X-linked dystonia-parkinsonism (XDP) is a neurodegenerative disease caused by a SINE-VNTR-Alu(AGAGGG)n retrotransposon insertion in TAF1, with a polymorphic (AGAGGG)n repeat. Repeat length and variants in MSH3 and PMS2 explain ∼65% of the variance in age at onset (AAO) in XDP. However, additional genetic modifiers are conceivably at play in XDP, such as repeat interruptions. Long-read nanopore sequencing of PCR amplicons from XDP patients (n = 202) was performed to assess potential repeat interruption and instability. Repeat-primed PCR and Cas9-mediated targeted enrichment confirmed the presence of identified divergent repeat motifs. In addition to the canonical pure SINE-VNTR-Alu-5'-(AGAGGG)n, we observed a mosaic of divergent repeat motifs that polarized at the beginning of the tract, where the divergent repeat interruptions varied in motif length by having one, two, or three nucleotides fewer than the hexameric motif, distinct from interruptions in other disease-associated repeats, which match the lengths of the canonical motifs. All divergent configurations occurred mosaically and in two investigated brain regions (basal ganglia, cerebellum) and in blood-derived DNA from the same patient. The most common divergent interruption was AGG [5'-SINE-VNTR-Alu(AGAGGG)2AGG(AGAGGG)n], similar to the pure tract, followed by AGGG [5'-SINE-VNTR-Alu(AGAGGG)2AGGG(AGAGGG)n], at median frequencies of 0.425 (IQR: 0.42-0.43) and 0.128 (IQR: 0.12-0.13), respectively. The mosaic AGG motif was not associated with repeat number (estimate = -3.8342, P = 0.869). The mosaic pure tract frequency was associated with repeat number (estimate = 45.32, P = 0.0441) but not AAO (estimate = -41.486, P = 0.378). Importantly, the mosaic frequency of the AGGG negatively correlated with repeat number after adjusting for age at sampling (estimate = -161.09, P = 3.44 × 10-5). When including the XDP-relevant MSH3/PMS2 modifier single nucleotide polymorphisms into the model, the mosaic AGGG frequency was associated with AAO (estimate = 155.1063, P = 0.047); however, the association dissipated after including the repeat number (estimate = -92.46430, P = 0.079). We reveal novel mosaic divergent repeat interruptions affecting both motif length and sequence (DRILS) of the canonical motif polarized within the SINE-VNTR-Alu(AGAGGG)n repeat. Our study illustrates: (i) the importance of somatic mosaic genotypes; (ii) the biological plausibility of multiple modifiers (both germline and somatic) that can have additive effects on repeat instability; and (iii) that these variations may remain undetected without assessment of single molecules.

Comparative single-cell analysis of the adult heart and coronary vasculature.

The structure and function of the circulatory system, including the heart, have undergone substantial changes with the vertebrate evolution. Although the basic function of the heart is to pump blood through the body, its size, shape, speed, regeneration capacity, etc. vary considerably across species. Here, we address the differences among vertebrate hearts using a single-cell transcriptomics approach. Published datasets of macaque (Macaca fascicularis), mouse, and zebrafish hearts were integrated and compared to the human heart as a reference. While the three mammalian hearts integrated well, the zebrafish heart showed very little overlap with the other species. Our analysis revealed a mouse-specific cell subpopulation of ventricular cardiomyocytes (CM), represented by strikingly different expression patterns of specific genes related to high-energy metabolism. Interestingly, the observed differences between mouse and human CM coincided with actual biological differences between the two species. Smooth muscle and endothelial cells (EC) exhibited species-specific differences in clustering and gene expression, respectively, which we attribute to the tissues selected for sequencing, given different focuses of the original studies. Finally, we compared human and zebrafish heart-specific fibroblasts (FB) and identified a distinctively high expression of genes associated with heart regeneration following injury in zebrafish. Together, our results show that integration of numerous datasets of different species and different sequencing technologies is feasible and that this approach can identify species-specific differences and similarities in the heart.

Transcriptional Alterations in X-Linked Dystonia-Parkinsonism Caused by the SVA Retrotransposon.

X-linked dystonia-parkinsonism (XDP) is a severe neurodegenerative disorder that manifests as adult-onset dystonia combined with parkinsonism. A SINE-VNTR-Alu (SVA) retrotransposon inserted in an intron of the TAF1 gene reduces its expression and alters splicing in XDP patient-derived cells. As a consequence, increased levels of the TAF1 intron retention transcript TAF1-32i can be found in XDP cells as compared to healthy controls. Here, we investigate the sequence of the deep intronic region included in this transcript and show that it is also present in cells from healthy individuals, albeit in lower amounts than in XDP cells, and that it undergoes degradation by nonsense-mediated mRNA decay. Furthermore, we investigate epigenetic marks (e.g., DNA methylation and histone modifications) present in this intronic region and the spanning sequence. Finally, we show that the SVA evinces regulatory potential, as demonstrated by its ability to repress the TAF1 promoter in vitro. Our results enable a better understanding of the disease mechanisms underlying XDP and transcriptional alterations caused by SVA retrotransposons.

The importance of genetic testing for dystonia patients and translational research.

Genetic testing through a variety of methods is a fundamental but underutilized approach for establishing the precise genetic diagnosis in patients with heritable forms of dystonia. Our knowledge of numerous dystonia-related genes, variants that they may contain, associated clinical presentations, and molecular disease mechanism may have significant translational potential for patients with genetically confirmed dystonia or their family members. Importantly, genetic testing permits the assembly of patient cohorts pertinent for dystonia-related research and developing therapeutics. Here we review the genetic testing approaches relevant to dystonia patients, and summarize and illustrate the multifold benefits of establishing an accurate molecular diagnosis for patients imminently or for translational research in the long run.

A hexanucleotide repeat modifies expressivity of X-linked dystonia parkinsonism.

OBJECTIVE: X-linked dystonia parkinsonism (XDP) is a neurodegenerative movement disorder caused by a single mutation: SINE-VNTR-Alu (SVA) retrotransposon insertion in TAF1. Recently, a (CCCTCT)n repeat within the SVA insertion has been reported as an age-at-onset (AAO) modifier in XDP. Here we investigate the role of this hexanucleotide repeat in modifying expressivity of XDP. METHODS: We genotyped the hexanucleotide repeat in 355 XDP patients and correlated the repeat number (RN) with AAO (n = 295), initial clinical manifestation (n = 294), site of dystonia onset (n = 238), disease severity (n = 28), and cognitive function (n = 15). Furthermore, we investigated i) repeat instability by segregation analysis and Southern blotting using postmortem brain samples from two affected individuals and ii) relative TAF1 expression in blood RNA from 31 XDP patients. RESULTS: RN showed significant inverse correlations with AAO and with TAF1 expression and a positive correlation with disease severity and cognitive dysfunction. Importantly, AAO (and not RN) was directly associated with whether dystonia or parkinsonism will manifest at onset. RN was lower in patients affected by mouth/tongue dystonia compared with blepharospasm. RN was unstable across germline transmissions with an overall tendency to increase in length and exhibited somatic mosaicism in brain. INTERPRETATION: The hexanucleotide repeat within the SVA insertion acts as a genetic modifier of disease expressivity in XDP. RN-dependent TAF1 repression and subsequent differences in TAF1 mRNA levels in patients may be potentiated in the brain through somatic variability leading to the neurological phenotype. ANN NEUROL 2019;85:812-822.

DNA Methylation as a Potential Molecular Mechanism in X-linked Dystonia-Parkinsonism.

BACKGROUND: X-linked dystonia-parkinsonism is a neurodegenerative movement disorder. The underlying molecular basis has still not been completely elucidated, but likely involves dysregulation of TAF1 expression. In X-linked dystonia-parkinsonism, 3 disease-specific single-nucleotide changes (DSCs) introduce (DSC12) or abolish (DSC2 and DSC3) CpG dinucleotides and consequently sites of putative DNA methylation. Because transcriptional regulation tightly correlates with specific epigenetic marks, we investigated the role of DNA methylation in the pathogenesis of X-linked dystonia-parkinsonism. METHODS: DNA methylation at DSC12, DSC3, and DSC2 was quantified by bisulfite pyrosequencing in DNA from peripheral blood leukocytes, fibroblasts, induced pluripotent stem cell-derived cortical neurons and brain tissue from X-linked dystonia-parkinsonism patients and age- and sex-matched healthy Filipino controls in a prospective study. RESULTS: Compared with controls, X-linked dystonia-parkinsonism patients showed striking differences in DNA methylation at the 3 investigated CpG sites. Using methylation-sensitive luciferase reporter gene assays and immunoprecipitation, we demonstrated (1) that lack of DNA methylation because of DSC2 and DSC3 affects gene promoter activity and (2) that methylation at all 3 investigated CpG sites alters DNA-protein interaction. Interestingly, DSC3 decreased promoter activity per se compared with wild type, and promoter activity further decreased when methylation was present. Moreover, we identified specific binding of proteins to the investigated DSCs that are associated with splicing and RNA and DNA binding. CONCLUSIONS: We identified altered DNA methylation in X-linked dystonia-parkinsonism patients as a possible additional mechanism modulating TAF1 expression and putative novel targets for future therapies using DNA methylation-modifying agents. © 2020 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Regulation of the cohesin-loading factor NIPBL: Role of the lncRNA NIPBL-AS1 and identification of a distal enhancer element.

Cohesin is crucial for genome stability, cell division, transcription and chromatin organization. Its functions critically depend on NIPBL, the cohesin-loader protein that is found to be mutated in >60% of the cases of Cornelia de Lange syndrome (CdLS). Other mutations are described in the cohesin subunits SMC1A, RAD21, SMC3 and the HDAC8 protein. In 25-30% of CdLS cases no mutation in the known CdLS genes is detected. Until now, functional elements in the noncoding genome were not characterized in the molecular etiology of CdLS and therefore are excluded from mutation screening, although the impact of such mutations has now been recognized for a wide range of diseases. We have identified different elements of the noncoding genome involved in regulation of the NIPBL gene. NIPBL-AS1 is a long non-coding RNA transcribed upstream and antisense to NIPBL. By knockdown and transcription blocking experiments, we could show that not the NIPBL-AS1 gene product, but its actual transcription is important to regulate NIPBL expression levels. This reveals a possibility to boost the transcriptional activity of the NIPBL gene by interfering with the NIPBL-AS1 lncRNA. Further, we have identified a novel distal enhancer regulating both NIPBL and NIPBL-AS1. Deletion of the enhancer using CRISPR genome editing in HEK293T cells reduces expression of NIPBL, NIPBL-AS1 as well as genes found to be dysregulated in CdLS.

Mutations in chromatin regulators functionally link Cornelia de Lange syndrome and clinically overlapping phenotypes.

The coordinated tissue-specific regulation of gene expression is essential for the proper development of all organisms. Mutations in multiple transcriptional regulators cause a group of neurodevelopmental disorders termed "transcriptomopathies" that share core phenotypical features including growth retardation, developmental delay, intellectual disability and facial dysmorphism. Cornelia de Lange syndrome (CdLS) belongs to this class of disorders and is caused by mutations in different subunits or regulators of the cohesin complex. Herein, we report on the clinical and molecular characterization of seven patients with features overlapping with CdLS who were found to carry mutations in chromatin regulators previously associated to other neurodevelopmental disorders that are frequently considered in the differential diagnosis of CdLS. The identified mutations affect the methyltransferase-encoding genes KMT2A and SETD5 and different subunits of the SWI/SNF chromatin-remodeling complex. Complementary to this, a patient with Coffin-Siris syndrome was found to carry a missense substitution in NIPBL. Our findings indicate that mutations in a variety of chromatin-associated factors result in overlapping clinical phenotypes, underscoring the genetic heterogeneity that should be considered when assessing the clinical and molecular diagnosis of neurodevelopmental syndromes. It is clear that emerging molecular mechanisms of chromatin dysregulation are central to understanding the pathogenesis of these clinically overlapping genetic disorders.

Heterozygous truncation mutations of the SMC1A gene cause a severe early onset epilepsy with cluster seizures in females: Detailed phenotyping of 10 new cases.

OBJECTIVE: The phenotype of seizure clustering with febrile illnesses in infancy/early childhood is well recognized. To date the only genetic epilepsy consistently associated with this phenotype is PCDH19, an X-linked disorder restricted to females, and males with mosaicism. The SMC1A gene, which encodes a structural component of the cohesin complex is also located on the X chromosome. Missense variants and small in-frame deletions of SMC1A cause approximately 5% of Cornelia de Lange Syndrome (CdLS). Recently, protein truncating mutations in SMC1A have been reported in five females, all of whom have been affected by a drug-resistant epilepsy, and severe developmental impairment. Our objective was to further delineate the phenotype of SMC1A truncation. METHOD: Female cases with de novo truncation mutations in SMC1A were identified from the Deciphering Developmental Disorders (DDD) study (n = 8), from postmortem testing of an affected twin (n = 1), and from clinical testing with an epilepsy gene panel (n = 1). Detailed information on the phenotype in each case was obtained. RESULTS: Ten cases with heterozygous de novo mutations in the SMC1A gene are presented. All 10 mutations identified are predicted to result in premature truncation of the SMC1A protein. All cases are female, and none had a clinical diagnosis of CdLS. They presented with onset of epileptic seizures between <4 weeks and 28 months of age. In the majority of cases, a marked preponderance for seizures to occur in clusters was noted. Seizure clusters were associated with developmental regression. Moderate or severe developmental impairment was apparent in all cases. SIGNIFICANCE: Truncation mutations in SMC1A cause a severe epilepsy phenotype with cluster seizures in females. These mutations are likely to be nonviable in males.

Hidden mutations in Cornelia de Lange syndrome limitations of sanger sequencing in molecular diagnostics.

Cornelia de Lange syndrome (CdLS) is a well-characterized developmental disorder. The genetic cause of CdLS is a mutation in one of five associated genes (NIPBL, SMC1A, SMC3, RAD21, and HDAC8) accounting for about 70% of cases. To improve our current molecular diagnostic and to analyze some of CdLS candidate genes, we developed and established a gene panel approach. Because recent data indicate a high frequency of mosaic NIPBL mutations that were not detected by conventional sequencing approaches of blood DNA, we started to collect buccal mucosa (BM) samples of our patients that were negative for mutations in the known CdLS genes. Here, we report the identification of three mosaic NIPBL mutations by our high-coverage gene panel sequencing approach that were undetected by classical Sanger sequencing analysis of BM DNA. All mutations were confirmed by the use of highly sensitive SNaPshot fragment analysis using DNA from BM, urine, and fibroblast samples. In blood samples, we could not detect the respective mutation. Finally, in fibroblast samples from all three patients, Sanger sequencing could identify all the mutations. Thus, our study highlights the need for highly sensitive technologies in molecular diagnostic of CdLS to improve genetic diagnosis and counseling of patients and their families.

De novo heterozygous mutations in SMC3 cause a range of Cornelia de Lange syndrome-overlapping phenotypes.

Cornelia de Lange syndrome (CdLS) is characterized by facial dysmorphism, growth failure, intellectual disability, limb malformations, and multiple organ involvement. Mutations in five genes, encoding subunits of the cohesin complex (SMC1A, SMC3, RAD21) and its regulators (NIPBL, HDAC8), account for at least 70% of patients with CdLS or CdLS-like phenotypes. To date, only the clinical features from a single CdLS patient with SMC3 mutation has been published. Here, we report the efforts of an international research and clinical collaboration to provide clinical comparison of 16 patients with CdLS-like features caused by mutations in SMC3. Modeling of the mutation effects on protein structure suggests a dominant-negative effect on the multimeric cohesin complex. When compared with typical CdLS, many SMC3-associated phenotypes are also characterized by postnatal microcephaly but with a less distinctive craniofacial appearance, a milder prenatal growth retardation that worsens in childhood, few congenital heart defects, and an absence of limb deficiencies. While most mutations are unique, two unrelated affected individuals shared the same mutation but presented with different phenotypes. This work confirms that de novo SMC3 mutations account for ∼ 1%-2% of CdLS-like phenotypes.

Exome sequencing unravels unexpected differential diagnoses in individuals with the tentative diagnosis of Coffin-Siris and Nicolaides-Baraitser syndromes.

Coffin-Siris syndrome (CSS) and Nicolaides-Baraitser syndrome (NCBRS) are rare intellectual disability/congenital malformation syndromes that represent distinct entities but show considerable clinical overlap. They are caused by mutations in genes encoding members of the BRG1- and BRM-associated factor (BAF) complex. However, there are a number of patients with the clinical diagnosis of CSS or NCBRS in whom the causative mutation has not been identified. In this study, we performed trio-based whole-exome sequencing (WES) in ten previously described but unsolved individuals with the tentative diagnosis of CSS or NCBRS and found causative mutations in nine out of ten individuals. Interestingly, our WES analysis disclosed overlapping differential diagnoses including Wiedemann-Steiner, Kabuki, and Adams-Oliver syndromes. In addition, most likely causative de novo mutations were identified in GRIN2A and SHANK3. Moreover, trio-based WES detected SMARCA2 and SMARCA4 deletions, which had not been annotated in a previous Haloplex target enrichment and next-generation sequencing of known CSS/NCBRS genes emphasizing the advantages of WES as a diagnostic tool. In summary, we discuss the phenotypic and diagnostic challenges in clinical genetics, establish important differential diagnoses, and emphasize the cardinal features and the broad clinical spectrum of BAF complex disorders and other disorders caused by mutations in epigenetic landscapers.

LINE1-mediated epigenetic repression of androgen receptor transcription causes androgen insensitivity syndrome.

Androgen insensitivity syndrome (AIS) is a difference of sex development (DSD) characterized by different degrees of undervirilization in individuals with a 46,XY karyotype despite normal to high gonadal testosterone production. Classically, AIS is explained by hemizygous mutations in the X-chromosomal androgen receptor (AR) gene. Nevertheless, the majority of individuals with clinically diagnosed AIS do not carry an AR gene mutation. Here, we present a patient with a 46,XY karyotype, born with undervirilized genitalia, age-appropriate testosterone levels and no uterus, characteristic for AIS. Diagnostic whole exome sequencing (WES) showed a maternally inherited LINE1 (L1) retrotransposon insertion in the 5' untranslated region (5'UTR) of the AR gene. Long-read nanopore sequencing confirmed this as an insertion of a truncated L1 element of ≈ 2.7 kb and showed an increased DNA methylation at the L1 insertion site in patient-derived genital skin fibroblasts (GSFs) compared to healthy controls. The insertion coincided with reduced AR transcript and protein levels in patient-derived GSFs confirming the clinical diagnosis AIS. Our results underline the relevance of retrotransposons in human disease, and expand the growing list of human diseases associated with them.

A GGC-repeat expansion in ZFHX3 encoding polyglycine causes spinocerebellar ataxia type 4 and impairs autophagy.

Despite linkage to chromosome 16q in 1996, the mutation causing spinocerebellar ataxia type 4 (SCA4), a late-onset sensory and cerebellar ataxia, remained unknown. Here, using long-read single-strand whole-genome sequencing (LR-GS), we identified a heterozygous GGC-repeat expansion in a large Utah pedigree encoding polyglycine (polyG) in zinc finger homeobox protein 3 (ZFHX3), also known as AT-binding transcription factor 1 (ATBF1). We queried 6,495 genome sequencing datasets and identified the repeat expansion in seven additional pedigrees. Ultrarare DNA variants near the repeat expansion indicate a common distant founder event in Sweden. Intranuclear ZFHX3-p62-ubiquitin aggregates were abundant in SCA4 basis pontis neurons. In fibroblasts and induced pluripotent stem cells, the GGC expansion led to increased ZFHX3 protein levels and abnormal autophagy, which were normalized with small interfering RNA-mediated ZFHX3 knockdown in both cell types. Improving autophagy points to a therapeutic avenue for this novel polyG disease. The coding GGC-repeat expansion in an extremely G+C-rich region was not detectable by short-read whole-exome sequencing, which demonstrates the power of LR-GS for variant discovery.

Single-Cell Sequencing in Neurodegenerative Disorders.

Neurodegenerative disorders are typically characterized by late onset progressive damage to specific (sub)populations of cells of the nervous system that are essential for mobility, coordination, strength, sensation, and cognition. Addressing this selective cellular vulnerability has become feasible with the emergence of single-cell-omics technologies, which now represent the state-of-the-art approach to profile heterogeneity of complex tissues including human post-mortem brain at unprecedented resolution. In this review, we briefly recapitulate the experimental workflow of single-cell RNA sequencing and summarize the recent knowledge acquired with it in the most common neurodegenerative diseases: Parkinson's, Alzheimer's, Huntington's disease, and multiple sclerosis. We also discuss the possibility of applying single-cell approaches in the diagnostics and therapy of neurodegenerative disorders, as well as the limitations. While we are currently at the point of deeply exploring the transcriptomic changes in the affected cells, further technological developments hold a promise of manipulating the affected pathways once we understand them better.

Factors influencing reduced penetrance and variable expressivity in X-linked dystonia-parkinsonism.

X-linked dystonia-parkinsonism (XDP) is a neurodegenerative movement disorder that primarily affects adult Filipino men. It is caused by a founder retrotransposon insertion in TAF1 that contains a hexanucleotide repeat, the number of which differs among the patients and correlates with the age at disease onset (AAO) and other clinical parameters. A recent work has identified additional genetic modifiers of age-associated penetrance in XDP, bringing to light the DNA mismatch repair genes MSH3 and PMS2. Despite X-linked recessive inheritance, a minor subset of patients are female, manifesting the disease via various mechanisms such as homozygosity, imbalanced X-chromosome inactivation, or aneuploidy. Here, we summarize and discuss clinical and genetic aspects of XDP, with a focus on variable disease expressivity as a consequence of subtle genetic differences within a seemingly homogenous population of patients.

Elucidating Hexanucleotide Repeat Number and Methylation within the X-Linked Dystonia-Parkinsonism (XDP)-Related SVA Retrotransposon in TAF1 with Nanopore Sequencing.

BACKGROUND: X-linked dystonia-parkinsonism (XDP) is an adult-onset neurodegenerative disorder characterized by progressive dystonia and parkinsonism. It is caused by a SINE-VNTR-Alu (SVA) retrotransposon insertion in the TAF1 gene with a polymorphic (CCCTCT)n domain that acts as a genetic modifier of disease onset and expressivity. METHODS: Herein, we used Nanopore sequencing to investigate SVA genetic variability and methylation. We used blood-derived DNA from 96 XDP patients for amplicon-based deep Nanopore sequencing and validated it with fragment analysis which was performed using fluorescence-based PCR. To detect methylation from blood- and brain-derived DNA, we used a Cas9-targeted approach. RESULTS: High concordance was observed for hexanucleotide repeat numbers detected with Nanopore sequencing and fragment analysis. Within the SVA locus, there was no difference in genetic variability other than variations of the repeat motif between patients. We detected high CpG methylation frequency (MF) of the SVA and flanking regions (mean MF = 0.94, SD = ±0.12). Our preliminary results suggest only subtle differences between the XDP patient and the control in predicted enhancer sites directly flanking the SVA locus. CONCLUSIONS: Nanopore sequencing can reliably detect SVA hexanucleotide repeat numbers, methylation and, lastly, variation in the repeat motif.