Blood samples of six healthy volunteers, aged 35, 24, 26, 27, 29 and 41 years, respectively, were collected and were numbered from 1 to 6. Donors 3 and 6 are women.

Our cohort of patients included thirteen ICF1 patients with mutations in DNMT3B, six ICF2 patients with mutations in ZBTB24 and four ICFX patients with as of yet unknown mutations (Additional file 2). Most patients were described earlier 11 12 25 26 27 except for five newly enrolled patients (Table  1). The ICF B-lymphoblastoid and fibroblasts named here pCor were obtained from the Coriell Cell Repositories (USA) ( http://ccr.coriell.org/). Patients pG, pR, pI, pH, pC, pD, pN, pP, pS were recruited by the ICF Consortium and described together with patients pG, pR, pI, pH, pC, pD, pN, pP, pS in 12 . Patients pW, pT and P5 were described earlier 11 25 26 . Patients P7 and P8 were recently classified as ICF2 patients 27 . Patients pC, pS, pU and pN were classified as ICFX patients since sequence analysis of DNMT3B and ZBTB24 genes performed as previously described 12 27 did not reveal any mutation in their coding sequences.

M: male; F: female; *siblings.

New ICF syndrome patients P1, P2, P3, P4 and pY were diagnosed according to cytogenetic analysis that revealed typical multiradial chromosome configurations with multiple arms from chromosomes 1 and 16, and clinical features, mainly primary immunodeficiency. For these patients, primary immunodeficiency and hypomethylation at satellite repeats are detailed in Additional files 3 and 4. These patients were classified as ICF1 patients (Necker Hospital, Paris, France) based on mutations found in DNMT3B using previously described sequencing methods 12 27 . Written informed consent was obtained from the parents of the patients.

Primary fibroblasts (passage 7 for pG, pW and pT; passage 5 for pR; passage 6 for pI; passage 14 for pC and pP; passage 9 for pCor and pS) and Lymphoblastoid cell lines (LCLs) from healthy donors and ICF patients were cultured in DMEM and RPMI 1640 respectively, supplemented with 15% FCS, glutamine and antibiotics (Invitrogen).

Genomic DNA (200 ng) was digested at 37°C for 4 h with 10 U of the methylation-sensitive enzyme AciI, or NcoI (New England Biolabs) which does not have cutting sites in our regions of interest and served to normalize the data. The endonucleases were subsequently inactivated by incubation at 65°C for 20 min. Real-time PCR was carried out using the light cycler-DNA MasterPLUS SYBR Green I mix (Roche) supplemented with 0.5 μM specific primer pairs and with 2 μL of digested DNA. Real-time quantification PCR were run on a light cycler rapid thermal system (LightCycler®480 2.0 Real time PCR system, Roche) with 20 sec of denaturation at 95°C, 20 sec of annealing at 65°C and 20 sec of extension at 72°C for all primers, and analyzed by the comparative CT (∆CT) method according to the formula: methylation (%) = E^(∆CT) × 100 where E represents PCR efficiency and ∆CT = CT_sample (AciI digest) - CT_sample (NcoI digest). Sequences of primers within CpG islands at germline gene promoters are shown in Additional file 1. Each data shown on histograms is the mean result of qPCR analysis on at least three independent experiments performed on at least three independent genomic extractions.

Peripheral blood samples were obtained from healthy donors and ICF patients. We assessed profiles of gene expression focusing on the germline genes transcriptional signature that we previously established in a mouse model for ICF1 23 24 .

As expected from their expression primarily in germ cells (Additional file 5; and RNAseq and exon arrays datasets on normal tissues publically available at UCSC genome browser) and function in meiosis, expression of the germline genes tested was undetectable in blood cells from healthy controls, although that of SYCE1 showed some degree of variation between healthy donors that did not seem to correlate with age or sex (Figure  1 and Additional file 6). In contrast, strong expression was detected for two of them in blood cells from ICF patients; among all the germline genes tested, only that of MAELSTROM (MAEL), involved in repression of transposable elements during meiosis, and Synaptonemal Complex Central Element Protein 1 (SYCE1), involved in assembly of the synaptonemal complex during meiosis in germ cells, were transcribed in blood cells from ICF patients (Figure  1 and Additional files 6 and 7). This illegitimate expression in whole blood was very strong and remarkably specific for ICF1 patients while at significantly lower levels in ICF2 blood cells (raw data and p values in Additional file 6). Interestingly, patient P5 had only mild ICF phenotype 25 although he showed strong expression of MAEL and SYCE1. Therefore, given their repressed state in normal somatic tissues, detection of MAEL and SYCE1 expression in blood cells provides a statistically significant signature for the most prevalent sub-type of the disease, regardless of the severity of the symptoms.

Since the germline genes tested in this study are known to be repressed through DNA methylation in mice 23 24 28 we analyzed their methylation status on genomic DNA isolated from fresh whole blood samples (Figure  2A) or epithelial cells obtained from buccal swabs (Figure  2B), from ICF patients and healthy donors. Consistent with their expression only in blood cells from ICF1 patients (Figure  1), analysis of methylation at MAEL and SYCE1 revealed their hypomethylated state specifically in ICF1 patients since ICF2 and ICFX patients (except ICFX patient pN; discussed below) were methylated at values comparable to healthy controls (raw data and p values in Additional file 8). Similarly, Solute Carrier family 25 member 31 (SLC25A31 also known as ANT4), encoding a mitochondrial ADP/ATP carrier essential during spermatogenesis, was significantly hypomethylated only in ICF1 patients (Figure  2 and Additional file 8) although this hypomethylation did not correlate with transcriptional activation (Additional file 6).

Importantly, and as mentioned above, this transcriptional signature could not have been unraveled using classically available cultured patient cells like Epstein-Barr virus-transformed lymphoblastoid cell lines (LCLs) and primary fibroblasts. Indeed, although we could easily detect MAEL and SYCE1 transcripts in these cultured cells, this expression was independent of patient genotype (Additional file 7). We also noticed that the illicit expression of MAEL in fibroblasts from ICF1 patients became undetectable beyond passage 9 (Additional file 9), although DNA methylation levels at its promoter remained relatively unchanged, suggesting that extensive culture may limit findings of illegitimately expressed genes. Hypomethylation of DNMT3B-target genes in ICF2 and ICFX cells (Additional file 7), in which DNMT3B is not mutated, could result from well-known, direct or indirect, aberrant culture-induced defects of DNA methylation 29 . Furthermore, EBV transformation and cell culture perturb DNA methylation patterns in human LCLs, in particular at satellite repeats of juxtacentromeric regions 30 that are used to diagnose ICF patients. Therefore, like in previously reported transcriptomic studies examining differential gene expression profiles between ICF patients and healthy controls 15 16 , our data obtained in LCLs or fibroblasts probably reflected culture-associated defects in DNA methylation and gene expression, which have been superimposed on the intrinsic defects that affect patients, and do not allow the reliable identification of biomarkers with diagnostic value.

In contrast, the analysis that we conducted in uncultured cells from patients provided the first blood-based gene expression test to identify ICF1 syndrome, and the first discriminating transcriptional differences amongst ICF patients. The molecular functions of ZBTB24 remain to be identified, but the finding that the germline genes tested herein are not similarly affected by mutations in DNMT3B or ZBTB24 contradicts earlier suggestion that ZBTB24 is a mere adapter of DNMT3B function 12 .

DNA methylation at the promoters of DEAD (Asp-Glu-Ala-Asp) Box Polypeptide 4 (DDX4), the homolog of VASA proteins in Drosophila with key roles in germ cells development, and Testis Expressed 12 (TEX12), was similar between patients and controls, suggesting that, in contrast to the mouse, methylation of these genes was not affected in human DNMT3B- or ZBTB24-deficient cells (Additional file 10). Interestingly, the Synaptonemal Complex Protein 1 (SYCP1) gene showed significant hypomethylation in all but one ICFX patient with unknown mutation. Similarly, we observed hypomethylation of the X-linked testis-expressed gene 11 (TEX11) irrespective of ICF subtype (Figure  2; black bars), but restricted to female ICF cells (raw data and p values in Additional file 8). Whether hypomethylation at SYCP1 and TEX11 promoters was a direct consequence of ZBTB24 loss of integrity remains to be seen. However, these data identifies SYCP1 and TEX11 as the first unique gene loci affected by ZBTB24 mutations and contribute to a significant progress in our understanding of the etiology of ICF syndrome. In the case of the X-linked gene TEX11, it is likely related to the global hypomethylation of the inactive X-chromosome reported in female ICF1 patients 31 32 . Therefore, the hypomethylation of SYCP1 and TEX11 (Figure  2 and Additional file 8) regardless of the ICF sub-type represents the first unifying molecular signature for ICF syndrome and suggests that hypomethylation of the inactive X-chromosome in female patients may represent an invariant feature of ICF syndrome. In addition, these data point to a putative role of ZBTB24 in establishment or maintenance of DNA methylation at the inactive X-chromosome.

The perturbations that we reported here represent a reliable index of DNMT3B dysfunction in ICF1 patients, with the notable exception of patient pN. Based on our predictions of a molecular signature that varies according to ICF genotype, the germline methylation (Figure  2) and expression (Figure  1) signatures that we found in ICFX patient pN would place this patient in the ICF1 subtype. However, no mutation was previously found in the coding regions of DNMT3B for this patient 12 or in a new sequencing of DNMT3B and ZBTB24 exons. Intriguingly, the ICFX patient pN differs from other ICFX patients in that this patient had some degree of hypomethylation at α-Sat repeats, intermediate between hypomethylation found in ICF2 cells and almost full methylation characteristic of normal and ICF1 cells (Additional file 4). In addition, this profile resembles that of patient P1, first classified as ICF1 because of mutations in DNMT3B maternal allele but for whom no mutation could be found in the coding regions of the paternal allele. The combination of hypomethylation at MAEL and SYCE1 germline genes, indicative of DNMT3B impaired activity, with modest but reproducible hypomethylation at α-Sat repeats, which is never observed in ICF1 patients and was instrumental in suggesting genetic heterogeneity among ICF patients, implies an even more complex genetic etiology for these two patients. In the absence of an identified “culprit gene” for these patients, it remains unclear whether a third locus is implicated in ICF syndrome. Alternatively, these patients could be “compound” heterozygotes for mutations in both DNMT3B and ZBTB24, though it would have to be outside coding regions since exome sequencing did not reveal any mutation in the exons of any of the two genes. The genetic defects in these patients will be probably highly informative from a mechanistic perspective regarding the establishment and maintenance of DNA methylation patterns. These observations also emphasize that additional molecular markers besides the methylation of α-Sat sequences are required to distinguish various classes of ICF patients.

We propose that profiling of DNA methylation at MAEL and SYCE1 germline genes, combined to the detection of their illicit expression using non-invasive diagnostic techniques, could serve as a powerful, quick, and simple procedure to diagnose patients with suspected ICF1 syndrome. Because these genes are normally repressed in somatic tissues, any diagnostic test based on their expression will give an unambiguous result. To test the power of these genes for diagnostic means, we assessed methylation and expression profiles of germline genes in a child born to a consanguineous family but whose genotype at the time of testing was not known. This child, who is a sibling of patient P3, was born during the course of this study, and we were able to sample buccal epithelial cells and peripheral blood when she was 3 months old. Our analysis revealed germline gene methylation and transcription profiles typical of that of ICF1 patients tested in this study (Figure  3 and p values in Additional file 8). These observations were highly suggestive of the ICF1 syndrome in this child, who we named patient P4 (Table  1). In line with our predictions, diagnosis of the ICF1 syndrome was subsequently confirmed by sequencing of DNMT3B exons and characterization of the same mutations identified in her older sister (Table  1).