DNA Methylation Episignature as a Novel Diagnostic Tool for Diamond-Blackfan Anemia Syndrome

Abstract

Diamond-Blackfan Anemia Syndrome (DBAS) is a rare inherited bone marrow failure syndrome (IBMFS) characterized by impaired erythropoiesis and significant genetic heterogeneity. Diagnosis can be challenging due to clinical variability and the lack of sensitive and specific biomarkers. We investigated the evidence for a DNA methylation (DNAm) episignature in a cohort of 80 DBAS patients with causative variants in various ribosomal protein genes: DBA1 (RPS19, n = 30), DBA4 (RPS17, n = 6), DBA5 (RPL35A, n = 8), DBA6 (RPL5, n = 15), DBA7 (RPL11, n = 13), DBA10 (RPS26, n = 8). We identified a distinct and highly accurate episignature biomarker for DBAS, clearly differentiating it from both Fanconi anemia and a broad spectrum of other episignature-positive disorders. Furthermore, we developed a specific DNAm classifier for the clinically similar DBA6 and DBA7 subtypes. Applying the DBAS episignature analysis to six molecularly uncharacterized cases, three exhibited the DBAS pattern. Subsequent genome sequencing identified causative genetic variants in two (RPL5: c.325-380A>G:p.?; RPL26: c.-6 + 3_-6 + 25del:p.?), validating the test robustness. Methylation profiles from two revertant cases (RPS19:P47L and RPS17 full gene deletion) exhibited the DBAS episignature, suggesting it to be a stable epigenetic mark associated with the underlying genetic mutation, likely established early in development. In conclusion, we propose DNAm profiling as a robust diagnostic tool for DBAS, providing a biomarker applicable to all patients with clinical suspicion of the disease and critically aiding in the resolution of variants of uncertain significance and molecularly uncharacterized cases.

Keywords: DBAS; episignature; methylation.

FIGURE 1

Genetic characterization of the entire DBAS patient cohort. Representation of DBAS subgroups associated with mutations in minor ribosomal subunit genes, those associated with alterations in major ribosomal subunit genes, and the subgroup of patients with suspected DBAS but unknown molecular alterations. [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 2

Discovery of an episignature in DBAS cohort. (A) Heatmap of DNA methylation patterns: each column is one individual (either with DBAS or a healthy control). Each row is one of the 206 differentially methylated CpG probes identified by comparing the DBAS cases versus controls. The red columns are DBAS cases, and the blue columns are controls. A clear separation was observed between DBAS cases and controls. (B) Plot of samples similarity: the multidimensional scaling (MDS) plot shows how the DBAS cases (red) cluster apart from the controls (blue) based on their DNA methylation patterns. [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 3

DNA methylation profiling of DBAS, FA (FANC), and control samples. (A) Heatmap showing the separation of methylation profiles between DBAS cases (red), FA cases (green), and control samples (blue). (B) MDS plot illustrating the distinct clustering of DBAS cases (red), FA cases (green), and control populations (blue) based on their methylation profiles. [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 4

Discovery of a sub episignature for DBA6 and DBA7 cohorts. (A) Heatmap that shows a shared DNA methylation pattern in DBA6 (orange) and DBA7 (red) samples clearly distinct from the controls (blue) and other DBAS subtypes (purple = DBA1, yellow = DBA4, pink = DBA5, black = DBA10). (B) MDS plots that demonstrated the presence of a distinct cluster corresponding to DBA6 (orange) and DBA7 (red) differentiating them from controls (blue) and other DBAS cases (purple, yellow, pink, and black). (C) SVM classifier, a type of supervised machine learning algorithm, built to distinguish between DBA6/DBA7 cases and other samples. The algorithm was developed in two steps. First, during the training phase, it learned to recognize the pattern of DBA6/DBA7 cases (n = 28) compared with their matched controls, two‐thirds of other controls, and other disorders with detectable DNA methylation patterns (shown as blue circles). In the second, testing phase, the remaining samples (one‐third of controls and other disorders) were used to evaluate the model's performance (gray circles). Additional DBAS subtypes (DBA1, DBA4, DBA5, DBA10) were also tested (gray circles). The model assigned scores close to 1 for DBA6/DBA7 samples and near 0 for other DBAS subtypes and unrelated disorders, indicating the specificity of the model in detecting the methylation signature of DBA6/DBA7. (D) MVP score, a metric derived from the SVM classifier, was calculated for each DBA6 and DBA7 case using a leave‐one‐out approach (each case tested individually while the others trained the model) and compared with the average scores of controls and other episignature disorders across all rounds. DBA6/DBA7 cases (red dots) consistently scored high (near 1), while controls and other disorders stayed low (near 0), confirming the distinct and reproducible episignature. The abbreviations for the disorders are listed in Table S5. [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 5

Testing the methylation profile of DBAS cases without molecular diagnosis. (A) Heatmap: Each column represents an individual DBAS case (red), control (blue), or unsolved DBAS case (purple). (B) MDS plot showing the clustering of DBAS cases (red), unsolved DBAS molecular cases (purple), and controls (blue). (C) SVM classifier: the model highlights three unsolved DBAS samples with high MVP scores (near 1), while the remaining three unsolved DBAS samples have MVP scores below 0.25 (gray circles). The abbreviations for the listed epigenetic profiles are provided in Table S5. [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 6

Testing the DBAS cases with molecular reversion. Heatmaps (left) and MDS plots (right) illustrate the DNA methylation profiles of (A) the entire DBAS cohort, (B) the DBA1 subtype, and (C) the DBA4 subtype. In each analysis, cases with molecular reversion (yellow) were used as the testing set while other DBAS cases (red) and matched controls (blue) were used to train the model. No differences were observed between the reverted case patient #25 (yellow) and patient #26 (green), which carries the same variant, in the DBA1 cohort. Similarly, patient #35 (yellow) clustered alongside the training DBA4 cases (red), indicating no significant deviation in methylation patterns. [Color figure can be viewed at wileyonlinelibrary.com]