The most relevant contribution to date of microarray technology to the biology of CLL is certainly the identification of a homogeneous phenotype related to memory B cells  and its classification as a distinct type of B lymphoma . In particular, for a given CLL subset defined by disease aggressiveness, this method may provide new insights into mechanisms that have yet to be revealed. In addition to transcriptional changes defined by the microarray approach, it has become evident that epigenetic alterations that determine structural and functional chromatin organization should be integrated into these types of studies. Indeed, molecular profiling in CLL has allowed the identification of new genes for which the expression is dependent on CpG island methylation . The down-regulation of the death-associated protein kinase 1 (DAPK1) gene in CLL indicates that both genetic and epigenetic factors may define both the sporadic and inherited forms of this disease .
In our current study, we employed a microarray methodology to search for potential candidate genes that may discriminate aggressive cases of CLL according to the sex of the patient. In our previous studies, we reported that in aggressive forms of CLL, the cells display partially defined mechanisms enabling them to avoid apoptotic death following DNA damage [7, 22, 25]. Thus, comparing resistant and sensitive CLL samples according to patient sex, the gene expression profile of resistant cells in male patients was found unexpectedly to differ from females. This suggests that analyses based on cellular susceptibility to DNA damage-induced apoptosis combined with patient sex may reveal a new sub-class of CLL that has previously been unknown.
CLL cells that are resistant to DNA damage-induced apoptosis may harbor mutated as well as wild type IgVH genes, mutated or wild-type TP53, and any different (probably multiple) types of chromosomal aberrations. In contrast, sensitive cells contain wild-type TP53 only and any (but probably a unique) type of chromosomal aberration  (Table 1). These biological markers have previously been used as criteria to select and to class CLL samples for microarray analyses but are not necessarily linked to the susceptibility of CLL cells to DNA damage-induced apoptosis. We thus evaluated whether this last type of CLL classification may be useful in identifying new molecular markers which have not been elucidated using other characteristics of this disease.
Remarkably, when male and female CLL samples were compared in terms of their ability to undergo apoptosis, all but 2% of the genes in the male samples were under-expressed in resistant compared with sensitive CLL types. In contrast, 41% of the genes were under-expressed in resistant female CLL samples (Additional file 1, Figure S1). The significant difference between resistant male and female CLL is emphasized by the very low number of genes in common, (17 in total) between these subsets (Table 2). Thus, it was striking that in the resistant male subset, 98% of the genes were down-regulated. Based on this observation, we addressed whether global chromatin condensation could underlie this transcriptional repression. We used immunofluorescence in situ labeling of a repressive chromatin state to evaluate the global levels of DNA and histone methylation in the sample CLLs (Figure 3). Immunofluorescence staining with HP1α, 3metH3K9 and 5metC showed differences in the distribution of inactive chromatin markers between resistant and sensitive cells and is indicative of a global repressive state of chromatin in the resistant male CLL subset. Indeed, HP1α, 3metH3K9 and 5metC are associated with heterochromatin-driven transcriptional repression .
These observations are in agreement with recently described features of CLL that involve altered epigenetic control i.e. the over-expression of nucleolin and its cytoplasmic retention in CLL cells  and an altered telomere length in a subset of CLL with a poor prognosis . Indeed, the nucleolar structure involving nucleolin and telomere lengths has been shown to be controlled by epigenetic factors, such as H3K9 methylation [28, 29]. RELB was elucidated as one of three discriminatory genes since its relative expression, reflected by the ratio R/S, was opposite between male and female CLL. The level of histone H3K9 trimethylation significantly increased in three promoter regions of RELB, two of them contain potential DNA-binding sites for the STAT3 and PAX5 (Figure 1, see also additional file Table S2) transcriptional factors that may be blocked by this type of histone modification. Of note, this region contains the motifs for 20 TFs which may be potentially involved in the RELB transcriptional control. The observation that the level of acetyl-H3, used as the marker of positive gene expression, was extremely low in male samples while in resistant female samples it was significantly high (Figure 2), further highlighted the role of epigenetic differences between male and female CLL samples. Hence, in the same promoter regions, histone modifications were not found to be associated with DNA methylation (Figure 7). Effectively, DNA methylation was investigated in regions that were more proximal from TSS and first exon than those investigated for 3metH3K9. According to UCSC Genome Browser http://genome.ucsc.edu on Human Mar. 2006 (NCBI36/hg18), the assembly used for this study, RELB gene position was 50,196,552 - 50,233,292 and mRNA at 50,196,535. The most proximal CpG islet is located 50,196,420 - 50,197,139. This means that the most probable site for DNA methylation is located -115 bp upstream TSS. The region of the RELB locus that was bisulfite sequenced was a 428-bp region comprising the promoter (from -259 bp) and first exon (to +169 bp). Thus, we probably cover the region that is the most likely affected by DNA methylation.
DNA methylation could be correlated (but not necessarily) with histone H3K9 modifications and vice versa. This discrepancy may be due to the fact that histone methylation is not always overlapping with CpG islands regions. The fact that we investigate upstream sequences for histone modifications might lead to under-estimate a putative 3metH3K9 modification in the sequences investigated for DNA methylation, and this results that RELB could be even more silenced than we proposed. Therefore, our findings support that H3K9 methylation rather than deoxycytosine methylation is responsible for RELB silencing.
As learned from the expression regulation of some autosomal genes, genes that are specifically maternally repressed may be CpG-hypomethylated to the same extent as paternally over-expressed genes. This suggests that DNA methylation is not necessarily modified in the transcriptionally "open" nucleosomal state at the promoter region . In addition, an impairment of DNA and histone modifications has been observed in a mouse leukemia cell line, L1210, where an inverse relationship in the regulation of the silencing of candidate genes was reported . This suggests that the methylation of DNA and histones is less cooperative in hematological cancers compared with other types of malignant cells. We recently reported that an epigenetic heterochromatinisation may be the mechanism involved in telomere shortening in both female- and male-derived resistant CLL cells [32, 33] resulting in an aberrant chromatin structure of telomeric regions. Together with current data, it appears evident that there is an altered epigenetic control involved in heterochromatinisation in CLL cells. While B cell lineage differentiation have been shown to depend on inherited epigenetic factors it could be speculated that malignant transformation of CLL cells results from an epigenetic defect resulting in a modified expression of RELB which is potentially involved in this process. Whether RELB inactivation in males occurs concomitantly with malignant B cell transformation, warrants further investigations.
The NF-κB family is a tightly coordinated group of five identified transcription factors with structural homologies enabling them to be involved in two major cellular responses mainly related to stress and inflammation as well as to cancer development and progression . With regards to the aggressiveness of CLL, the RelA member of the NF-κB family, involved in its heterodimeric form with p50 and through its temporal transactivation of "regulon" type genes in a canonical manner, has been recently reported to be associated with both in vitro survival and clinical disease progression in CLL. NF-κB pathway activation is achievable exogenously by stress (reactive oxygen species or DNA damage) or by death receptor stimulation. While this activation might results in a pro- or anti-apoptotic signaling we could speculate that the success of therapy would depend on its activity. In consequence, this finding suggests that RELB is a promising new therapeutic target for this disease . Moreover, as an NF-κB family member associated with the control of the adaptive immune response and the response to metabolic stress, RelB may be of particular interest in the emergence of progressive form of CLL since the alternative NF-κB pathway is potentially involved in both early and late differentiation of B cells [36, 37]. Together with our demonstrations of local heterochromatinization of telomeric regions [32, 33] and in agreement with the recently discovered role for the multifunctional transcriptional positive coactivator 4 (PC4) in targeting heterochromatinization in non-neuronal human cells , different levels of RelB expression between female and male CLL cells may involve a cell differentiation- and/or gender-specific factor that controls a dynamic state of chromatin region containing RELB.