The locus for ICF was localized to 20q11-q13 by homozygosity mapping 10. This led to the discovery that the DNMT3B gene is often the site of ICF mutations 111213. In the 35 ICF patients analyzed to date for DNMT3B mutations, 57% had mutations in both DNMT3B alleles (OMIM #602900) within the coding portion 5141516. These mutations are frequently found in the C-terminal portion of the protein that contains the catalytic domain 17 and are often missense mutations 1113141819. Although DNMT3B has repressor activity that is independent of its DNA methyltransferase activity, repression involves the central portion of the protein, which does not overlap with the methyltransferase domain 20. DNMT3B also forms a complex with DNMT1 and with small ubiquitin-like modifier 1, but these interactions involve the N-terminus of DNMT3B 2122. In contrast, many ICF patients have a missense mutation in one or both of their mutant alleles to give an amino acid substitution in one of ten motifs conserved among all cytosine-C5 methyltransferases 17. These findings suggest that it is the loss of DNA methyltransferase activity, and not some other function of the protein, that is responsible for the syndrome. The involvement of DNA hypomethylation in the phenotype of ICF is supported at the cytogenetic level because the ICF-specific rearrangements in mitogen-treated lymphocytes are the same in frequency, spectrum, and chromosomal specificity as those found in a normal pro-B lymphoblastoid cell line treated with the DNA methylation inhibitors 5-azacytidine or 5-azadeoxycytidine 2324. Also, biochemical analyses of recombinant proteins with known ICF mutations 17 and the invariant hypomethylation of certain portions of the genome from ICF patients 2526 are consistent with ICF being due to a deficiency in DNA methylation.

For the ~40% of ICF patients with no mutations discovered in exonic DNA from DNMT3B, there might be mutations in the promoter or other transcription control elements or mutations affecting splicing or polyadenylation. However, for several of these ICF patients without detected DNMT3B mutations, the most common isoform of DNMT3B RNA was still observed 15, and for the one patient examined in two putative promoter regions of DNMT3B, no mutations were found 16. The patients without DNMT3B coding region mutations seem to be derived from a different subtype of ICF. Lymphocytes or fibroblasts from 9 out of 9 patients in this category displayed hypomethylation in satellite α (centromeric) DNA while 5 out of 5 patients who had mutations in DNMT3B did not 15. ICF patients, including those without demonstrated mutations in DNMT3B, invariantly exhibit hypomethylation in the major DNA component of 1qh and 16qh (classical satellite 2 DNA) and of 9qh (classical satellite 3 DNA) in all examined tissues and in B-cell lines. In addition, they display chromosomal anomalies at 1qh and 16qh in mitogen-stimulated cells 16252728. As described below, it is unclear how the limited amount of DNA demethylation associated with ICF (a 7% decrease in the genomic 5-methylcytosine level in ICF tissue compared to that in normal tissue 28) may cause the immunodeficiency and other variable symptoms associated with the disease.

One other genetic disease, the X-linked α-thalassemia/mental retardation syndrome (ATR-X linked syndrome), is associated with an apparently Mendelian inheritance of DNA methylation abnormalities in a small part of the genome 2930. This syndrome is accompanied by either decreases (satellite 1 in Yqh but not in satellites 2 or 3) or increases (ribosomal RNA genes) in DNA methylation, apparently as a result of mutations in a putative ATP-dependent DNA helicase. ICF is the only human genetic disease currently known to involve mutations in a DNA methyltransferase gene. In mice, insertional inactivation of Dnmt3b or of the other two main DNA methyltransferase genes, Dnmt1 and Dnmt3a, results in prenatal death or (for Dnmt3a) death several weeks after birth 18. It is likely that ICF-causing mutations in DNMT3B leave residual activity, otherwise embryonic lethality would probably result. This is consistent with analyses of the biochemical effects of ICF-associated mutations on DNMT3B activity in vitro 17. Therefore, homozygous null DNMT3B (or DNMT1 or DNMT3A) mutations might make a small contribution to spontaneous abortions.

The immunodeficiency, despite the presence of B cells almost always 31, results in severe recurrent infections, often seen in early childhood and usually as the presenting finding 3233. A review of the literature shows that almost all ICF patients have severe respiratory infections and more than half have recurrent gastrointestinal infections. Pericarditis, ear infections, septicemia, and oral Candida infections have also been observed 193435. The immunodeficiency in ICF patients ranges from agammaglobulinemia to a mild reduction in immune response 36. Most patients have a poor immune response with low amounts or undetectable levels of IgA, IgG and/or IgM 14, but the exact nature of the immunodeficiency is variable. For example, in some patients only certain subclasses of IgG display deficiencies 1437, while in others, there are very low levels of IgG, IgA, and IgM or just of two of these three classes of immunoglobulins 26. Low levels of T-cells are present in about half of the patients and levels of B cells are also sometimes low, although in some patients only one or the other type of lymphocyte shows reduced levels 343538.

The dysmorphic facial features are variable 3134 and usually mild; moreover, several patients did not display them (unpub. data). The typical facial features are a broad flat nasal bridge, hypertelorism (very widely spaced eyes), and epicanthic folds. Less frequent, but still often associated with the syndrome, are micrognathia (small jaw), low-set ears, and macroglossia (protrusion or enlargement of the tongue) 1519343940.

Mental retardation and neurologic defects have been seen in about one-third and one-fifth of the patients, respectively 313437, and include slow cognitive and motor development and psychomotor impairment (ataxic gait and muscle hypotonia) 383141. In one case, delayed psychomotor development changed into age-appropriate development at 36 months 35.

Other congenital abnormalities in ICF are highly variable. Intrauterine growth retardation has been observed in ICF patients 343940. Several patients have been described with protruding abdomens 3440 and thin arms and legs 34. At least two patients have been found to have bipartite nipples 37. Skin pigment changes have been seen in some patients, with café au lait spots or irregularly outlined mildly hyperpigmented spots reported 3637. Scleral telangiectasias were found in at least one patient 37.

Mitogen-stimulated blood samples from unaffected individuals rarely display abnormalities in the vicinity of the centromeres of chromosomes 1, 16, and 9, in contrast to analogous cultures from ICF patients. However, 2% of metaphases from random chorionic villus (CV) cultures at day 8 displayed 1qh or 16qh decondensation and, less frequently, pericentromeric breaks, whole-arm deletions, quadriradials, or triradials 56. Furthermore, yet higher frequencies of these types of chromosomal aberrations have been reported by others in CV cultures 57. One exceptional CV sample taken during routine screening exhibited four cells with pericentromeric breaks in chromosome 1 and seven cells with decondensation of 1qh out of a total of 20 examined metaphases, even though the amniotic fluid-derived culture was normal and the patient, at almost 2 years of age, appeared normal 56. Furthermore, high-passage cells in CV or amniotic fluid-derived (AF) cultures from normal individuals routinely display high frequencies of ICF-like karyotypic abnormalities, although AF cultures require more passages to display these anomalies 58. The chromosomal metaphase anomalies in late-passage AF cultures are largely restricted to 1qh and 16qh decondensation, while those from mid- or late-passage CV cultures exhibit 1qh and 16qh decondensation and, in addition, whole-arm deletions, chromosome breaks, and multiradials involving the 1qh or 16qh region. Also, some control B-cell lines after long-term passage develop ICF-like chromosomal abnormalities, which appear to increase in frequency if satellite 2 becomes spontaneously demethylated upon prolonged cell culture 5960, but 1qh decondensation has been observed even in a control cell line that did not exhibit ICF-like hypomethylation of satellite 2 (M. Ehrlich, L. Qi, R. Nishiyama, and C. Tuck-Muller, unpub. data).

Fasth et al. 6 reported on AF cells taken at the time of a cesarean section of a 34-week sibling of an ICF patient. Five out of 15 metaphases in the AF culture showed decondensation (despiralization) of the heterochromatic region of chromosome 1. This was also reported in the bone marrow cells from the affected sibling. At birth, the younger sibling had 90% of lymphocytes showing despiralization or a break in the juxtacentromeric heterochromatin of chromosome 1. From the above-mentioned studies, short-term AF cultures are less likely to have culture-associated chromosomal artifacts than analogous CV cultures. Fetal blood sampling (cordocentesis) has been successful in at least one family 61.

Linkage analysis via CV sampling at 12 weeks of gestation was reported 62. Linkage markers from a 9-cM region of chromosome 20, at which the ICF locus had been mapped at that time, were used. A marker, D20S850, was informative, indicating that the fetus was heterozygous for the gene. The couple was given a greater than 90% probability that the fetus was not affected with ICF. Cordocentesis was declined, and postnatal blood chromosome analysis revealed a normal male karyotype, without juxtacentromeric heterochromatin instability.

DNA analysis of the DNMT3B gene is also being done on a research basis in some laboratories, and could be used for diagnosis. However, because of the variety of mutations in DNMT3B that cause this syndrome 515, it is feasible only for analyzing close relatives in ICF families with known DNMT3B mutations. Therefore, this prenatal diagnosis requires a first-degree affected relative who had sequenced mutations in both alleles.

It is not clear how DNA hypomethylation in ICF patients may result in the immunodeficiency. However, it seems likely, as described under "Etiology," that it is the DNA methylation deficiency, and not some other aspect of impaired DNMT3B activity, that is responsible for the disease. Hypomethylation in some region(s) of the genome probably changes transcription of one or more genes that are primary targets for the deficiency in DNMT3B activity.

The immunodeficiency in ICF is manifested as low serum immunoglobulin levels, although there can be normal B- and T-cell numbers in peripheral blood. The latter finding indicates that the early stages of lymphocyte differentiation are not abnormal. High percentages of cells in ICF B-cell lines that have membrane-bound IgM and/or IgD have been observed even in patients with extremely low levels of IgM. Therefore, class switching was normal for the precursors of these cells, but there seems to be a defect in lymphocyte maturation or activation 14. In a comparison of ICF lymphoblastoid cell lines (B-cell lines with known DNMT3B mutations) and analogous control cell lines by a microarray expression analysis, only a small number of genes, including IgG- and IgA-encoding sequences, were found to have ICF-specific differences in RNA levels 14. A larger microarray expression analysis, which was followed by quantitative real-time reverse transcription-polymerase chain reaction (RT-PCR) for selected genes, gave similar findings (M. Ehrlich, C. Sanchez, C. Shao, R. Kuick, S. Hanash, unpub. data). In the first study, no ICF-specific differences in promoter methylation were seen, even for genes with elevated RNA levels in ICF cell lines 14 and similar methylation analyses are underway for selected genes from the second study. A number of reports of promoter-region hypomethylation in ICF cells, including in the inactive X chromosome, have revealed inconsistent hypomethylation at given gene regions among different patients 32366465. This finding matches results from a genome-wide search for consistently hypomethylated DNA sequences in ICF vs. control B-cells lines by 2-dimensional gel electrophoresis of DNA digested with two restriction endonucleases, one of which was CpG methylation-sensitive 66. Only repeated DNA sequences (a subtelomeric repeat and a repeat from the acrocentric chromosomes) were identified as hypomethylated specifically in all the ICF samples by this analysis. It has been hypothesized that in ICF lymphoid cells, hypomethylation of regions of the genome that are normally constitutively heterochromatic (e.g., 1qh and 16qh) could affect regulation of expression of genes elsewhere in the genome by altering the postulated normal sequestration of DNA sequence-specific proteins at this heterochromatin 1426. There are precedents for such binding of transcription factors to centromeric heterochromatin 6768. The hypothesis that constitutive heterochromatin itself can act in trans to modulate gene expression remains to be tested for ICF cells. Alternatively, there may be only a small number of currently unidentified gene regions with consistent hypomethylation specific to ICF lymphoid cells that are responsible for ICF-type immune dysfunction.

As explained above, the effect of the ICF DNMT3B mutations on immune functions is likely to be the result of DNA hypomethylation, probably through one or more genes that initiate the abnormalities in late maturation and activation of lymphoid cells. The above-mentioned microarray expression analyses [14, M. Ehrlich, C. Sanchez, C. Shao, R. Kuick, and S. Hanash, unpub. data] indicate that there are a small number of candidate genes for ICF-specific alterations in gene expression that might determine the phenotype. These include genes that are involved in cell signaling, transcription control, or chromatin remodeling. It was suggested that altered RNA levels in ICF B-cells compared to control cells might simply be a reflection of an abnormally prevalent immature state of these cells in vivo 2669. However, the genes that displayed ICF-specific differences in RNA levels, other than the immunoglobulin sequences, were not those predicted to be differentially expressed just because the ICF B-cell lines may have been derived from less mature cells than is normally the case. More research is needed to test which of these microarray candidates might be the proximal gene(s) involved in the lymphogenesis dysregulation in ICF patients as a result of DNMT3B mutations.

No obvious candidate genes for the ICF chromosome instability have been found from the above-mentioned microarray studies on ICF B-cell lines that exhibit high frequencies of 1qh or 16qh anomalies vs. control cell lines. It is possible that the hypomethylation of the satellite DNA in these regions in certain types of cells is responsible by itself for these chromosomal aberrations. However, most early-passage cultures from normal chorionic villi do not display appreciable numbers of abnormalities in these regions, despite the hypomethylation of 1qh and 16qh DNA in these cells due to the cell's extraembryonic mesodermal origin 5658. Therefore, there must be a cell-type specificity to this chromosome instability, which is in accord with the lower frequency of chromosomal abnormalities in bone marrow cells and fibroblasts from ICF patients than that found in stimulated lymphocytes 26. Moreover, the 1qh satellite DNA hypomethylation is not required for decondensation in these regions because normal amniotic fluid-derived cultures at late passage (essentially only embryonic fibroblasts) show high frequencies of 1qh decondensation despite a very high level of satellite DNA methylation at 1qh 58. It is likely that there is a DNA methylation-independent pathway (probably involving epigenetic chromatin changes) and a DNA methylation-stimulated pathway for decondensation and rearrangements targeted to the 1qh and 16qh regions. These mechanisms need to be elucidated. Further studies are also necessary to elucidate why there is a much lower frequency of these abnormalities in the 9qh region, despite the 9qh region usually being almost as long as the 1qh region and much longer than the 16qh region. Moreover, 9qh is predominantly composed of a similar DNA sequence (classical satellite 3; 2) to that of classical satellite 2 in 1qh and 16qh and, like 1qh, displays ICF-specific DNA hypomethylation of its satellite DNA.

As to the relationship in metaphase between 1qh and 16qh decondensation and 1qh and 16qh rearrangements, there is evidence that ICF B-cell lines compared to controls show decondensation in these juxtacentromeric heterochromatin regions even in interphase and that 1qh and 16qh exhibit a significantly increased colocalization 70. In addition, these regions colocalize with an aberrantly concentrated focus of heterochromatin proteins 1 (HP1) in G2 phase and with other proteins from promyelocytic leukemia nuclear bodies 71. Moreover, these ICF B-cell lines display abnormal looping of pericentromeric sequences at metaphase, formation of chromosome bridges at anaphase, chromosome 1 and 16 fragmentation at the telophase-interphase transition, increased apoptosis, and micronuclei with overrepresentation of chromosome 1 and 16 material 70. Another source of anaphase bridging in ICF B-cell lines is a significantly increased frequency of random telomeric associations between chromosomes 2870. These have not been described in stimulated ICF lymphocytes. Extensive karyotype analysis of ICF B-cell lines suggests that decondensation of 1qh and 16qh often leads to unresolved Holliday junctions, chromosome breakage, arm missegregation, and the formation of more stable translocations 28. The findings that ICF B-cell lines and lymphocytes in vivo are prone to micronucleus formation and apoptosis 9707172 are consistent with the apparently normal cell cycle checkpoints in ICF B-cell lines 55. These findings suggest that in stimulated lymphocytes and B-cell lines from ICF patients there is continuous generation of 1qh and 16qh chromosome rearrangements followed by removal of most of the cells with these chromosomal abnormalities, which gives rise to a great variety of rearrangements at 1qh and 16qh and a lack of clonal abnormalities.

The diverse collection of mutations in the exons of DNMT3B and the as-yet unmapped other mutations that cause ICF may explain the variety of symptoms. In addition, it has been proposed that the polymorphic nature of the length of the 1qh and 16qh region may contribute to the variety of phenotypes 71. However, in non-ICF patients, no phenotype has been associated with the natural length-polymorphism of these constitutive heterochromatic regions.

Curiously, one of the types of DNA repeats found to be strongly hypomethylated in most ICF B-cell lines compared to controls is the 4q35 and 10q26 D4Z4 repeat 6673. Facioscapulohumeral muscular dystrophy (FSHD) syndrome, a dominant disease, is almost always caused by a contraction of these copy-number-polymorphic tandem repeats at 4q35 (from 11-100 copies per locus in unaffected individuals to 1-10 copies in FSHD patients) although the mechanism for how the array contraction causes the disease is unknown 73. Evidence has been provided for a significant reduction in methylation at D4Z4 arrays that have a contracted size 75. Muscular dystrophy is not associated with ICF, and so there may not be a causal connection between D4Z4 hypomethylation and FSHD.

There is frequent hypomethylation of satellite 2 at 1qh and 16qh as well as satellite α in the centromeres of all the chromosomes in various cancers 7677. It has been proposed that this hypomethylation leads to altered gene expression in trans, as has been hypothesized for ICF lymphoid cells 267677. While the consequences of satellite 2 hypomethylation in ICF and cancer require much more study, it is already clear that the causes of satellite 2 hypomethylation in cancer and in ICF patients with DNMT3B-linked disease probably differ. No significant association between decreased DNMT RNA levels or increased number of mutations in the DNMT1 gene has been observed in cancers 7879. Furthermore, cancers typically have hypomethylation in some portions of the genome and hypermethylation in others 7678, while no hypermethylated component has been found in the ICF genome (unpub. data). Nonetheless, the presently unknown cause of the hypomethylation of both satellite 2 and satellite α in the ICF cases where DNMT3B mutations have not been found 15 might be related to the cause of hypomethylation of classical satellite 2 and centromeric satellite α in cancer.