Figure 1.
The critical region at 13q14.3 displays relaxed chromatin at the transcriptional start site (TSS) of the long non-coding RNA genes DLEU1 and DLEU2 in CLL cells.
The critical region on 13q14.3 was analyzed for DNA-methylation (lanes 1–3), distribution of histone modifications (lanes 4, 5) and gene expression (lanes 6, 7). DNA-hypomethylation was detected in CLL cells at the transcriptional start sites of the lncRNA genes DLEU1 and DLEU2 variant Alt1. This finding coincides with enrichment of H3K4me2 and depletion of macroH2A, corroborating relaxation of 13q14.3 in CLL. While expression of the protein coding genes and DLEU2 is decreased in CLL cells, lncRNA genes DLEU1 and DLEU2/Alt1 show upregulation, suggesting a direct regulation by DNA-methylation. Base pair positions on top refer to genome GRCh37 build hg18. The conserved CpG islands (A–E) are shown in green, less stringent CpG islands shown in light green. Lane 1: Methyl-CpG Immunoprecipitation (MCIp) allowed semi-quantitative DNA-methylation analysis in non-malignant B-cells sorted from healthy donors (n = 7) and CLL samples (n = 25). Precipitated DNA was hybridized onto a custom tiling microarray covering the 13q14.3 critical region. Depicted is the median log2 fold enrichment of CLL samples from which the median of log2 fold enrichment of healthy donor B-cell samples has been subtracted. Lane 2: Restriction digest-based analysis of DNA-methylation (aPRIMES) was performed in CpG islands C–E at 13q14.3 at 1 kbp resolution. Shown is the median log2 signal intensity of CLL patients from which the median of non-malignant B cell samples has been subtracted. Lane 3: Hypomethylation at D6 and E6 was validated by BioCOBRA and MassARRAY/Sequenome analyses (for details see Figure 2). Shown is the difference of the median percentage of methylation in CLL cells and non-malignant B cells. Lanes 4,5: ChIP was performed for macroH2A (lane 4) and H3K4me2 (lane 5). Precipitated DNA was quantified using qPCR. Enrichment was normalized to non-specific IgG and two control promoters (CDH12, HK2) that displayed similar enrichment for the two histone marks in CLL samples (n = 7) and peripheral blood mononuclear cells (PBMC) from healthy probands (n = 5). Enrichment was calculated as median log2 fold enrichment in precipitate vs. input for CLL samples after subtraction of enrichment from non-malignant B-cells. Lane 6: Gene expression profiling (GEP) was performed using bead chip arrays (Illumina) in CLL patients (n = 25) and sorted B-cells from healthy donors (n = 9). Plotted is the difference of log fold changes of CLL samples and non-malignant B cells. Lane 7: QRT-PCR for gene expression analysis of 13q14.3 candidate genes (for details see Figure 3).
Figure 2.
Hypomethylation at the TSS of the large ncRNAs affects the majority of CLL patients independent of 13q gene dosage and IGHV mutation status.
(A) For validation of DNA-hypomethylation at different CpG islands in 13q14.3 (x-axis; for localization see figure 1), quantitative measurement with BioCOBRA and MassARRAY (“Rb”) technology was performed in B-cells of healthy donors (“h”, n = 15–19, white diamonds) and PBMCs of CLL patients (“p”, n = 47–82, black diamonds; “miR” = 308 bp fragment spanning miR-15a; “Rb” = intragenic CpG island). (B, C) CLL patients with different genomic aberrations show different degree of hypomethylation that is distinct from healthy probands younger (“young”) or older (“old”) than 45 years. DNA-methylation was quantified with BioCOBRA (D6) and MassARRAY (E6). Analyzed were patients with normal karyotype (“nk”), deletion of 13q14.3 (“13q”), deletion of 11q22-q23 (“11q”), deletion of 17p (“17p”) or combinations thereof. Statistics were performed using Wilcoxon rank sum test (*** p<0.001, ** p 0.001 to 0.01, * 0.01 to 0.05 and ns not significant p>0.05). (D) DNA-methylation of D6 or E6 is aberrant in 58/61 CLL patients. Black diamonds: Patient samples with two copies of the critical region in 13q, grey diamonds: samples with deletion of one copy of 13q14.3, white diamonds: healthy donor samples. Dashed lines represent lowest levels of methylation of non-malignant B-cells.
Figure 3.
Downregulation of 13q14.3 candidate tumor suppressor genes and upregulation of lncRNA genes DLEU1 and DLEU2/Alt1 correlates with DNA-methylation.
(A–C) Expression of the protein-coding genes KPNA3, C13ORF1 and RFP2, the miRNA genes miR-15a and miR-16-1 and the lncRNA transcripts from genes DLEU1, DLEU2 and alternative transcript Alt1 were quantified with qRT-PCR in CLL cells from patients with retention of both copies of 13q14.3 (“p”, n = 34; see Table S1) and compared to B-cells from healthy donors (“h”, n = 20). Mean expression is indicated by a black horizontal line. “HKG” = housekeeping genes, average of PGK, LMNB1 and PPIA. (D, E) DNA-methylation levels correlate with transcript levels of genes in 13q14.3: While expression (y-axis) of candidate tumor suppressor genes is directly correlated to DNA-methylation of regions D6 and E6 (x-axis; left panel), lncRNA genes are anti-correlated (right panel). Pearson correlation coefficients are color-coded, blue = negative, yellow = positive correlation. (F) Pearson coefficients of correlation of gene expression (rows) and DNA-methylation at D6 and E6 (columns), colour coded (see legend). DLEU1a = exon1 to exon4; DLEU1b = exon 1 to exon 2; ALT1a/b = exon1, two different amplicons (G) Significance values of Pearson correlation coefficients (t-distribution), values p<0.05 are coded green. (H) The bidirectional promoter of DLEU1 and DLEU2 and the flanking CpG island was cloned into pCpGL luciferase vector, either including D6 (orange) or excluding D6 (green). (I) Constructs depicted in (H) were either methylated in-vitro using SSsI methylase (m, dark bars) or left unmethylated (u, light bars) and subsequently transfected into Mec1, Granta519 and HeLa cells and luciferase activity quantified. Grey: Luciferase CMV expression vector (V) not containing CpGs was used as negative control, green: excluding D6 region, orange: including D6 region. Error bars signify standard deviation of 2 independent experiments, each performed in duplicate. No changes were detected for E6 (Figure S4G–S4J). (J) Binding of the DNA-methylation sensitive chromatin reader CTCF was enriched in CLL samples both in the D6 and E6 regions compared to non-malignant B-cells. CTCF-bound chromatin was immunoprecipitated and quantitated with qPCR. Specific binding was shown by quantification of a sequence not bound by CTCF 2 kb upstream of the DM1 insulator (“ctr”). Statistics were performed using Wilcoxon rank sum test (*** p<0.001, ** p 0.001 to 0.01, * 0.01 to 0.05 and ns not significant p>0.05).
Figure 4.
MiR-15/miR-16 family is the strongest inducer of NF-kB.
(A) NF-kB induction by miRNA families (see Table S5 for complete list) was measured using NF-kB luciferase reporter assay. Median z-scores are indicated by black horizontal lines. Each dot represents one miRNA-mimic. (B) Transfection of miRNA-mimics miR-15a and miR-16 induces NF-kB activity as compared to unspecific control siRNA. Error bars represent standard deviation of triplicate measurements. (C–E) Upon transfection of miRNA mimic miR-15a and miR-16, expression of NF-kB target genes measured by qRT-PCR is induced in HEK293FT, MCF7 and MDA-MB-231 cells. (F) 3′UTRs of genes involved in the modulation of NF-kB activity and shown to be miR-15a and/or miR-16 targets were cloned into luciferase reporter plasmids. After co-transfection with miR-mimics into HEK cells, firefly luciferase activity was measured and normalized to renilla luciferase activity and empty luciferase plasmid as transfection and background control, respectively, and to no-template control (“NTC”). Data is depicted on a logarithmic scale to balance lower and higher activity compared to control, error bars show standard error of the mean of three independent experiments.
Figure 5.
13q14.3 candidate genes are a functionally related gene cluster that modulates NF-kB signalling.
(A) Knockdown of KPNA3 and KPNA4 (x-axis) resulted in reduced stimulation of NF-kB by TNFa in U2OS cells. NF-kB activity was measured after 24 hrs with a reporter driving luciferase expression under the control of 5 canonical NF-kB recognition sites. SiRNA directed against EGFP was used as negative control. Standard deviation of duplicate measurements is shown. (B) KPNA3, KCNRG, RFP2 and C13ORF1 were knocked down in HEK293T cells and activity of NF-kB was measured. As negative control, siRNA without a physiological target was used (siCONP). Error bars signify SEM of 3 independent experiments. (C) RFP2 induces NF-kB activity in HEK293T and HEK293 (Figure S6B) cells. HEK293T cells were transiently transfected with CMV RFP2 expression plasmids or empty vector and NF-kB activity measured after 4, 7, 10, 12 and 24 hrs (bottom panel). The top panel is a representative Western blot of two experiments, error bars in bottom panel represent standard deviation of triplicate measurements. (D) RFP2 induces NF-kB activity in primary CLL cells (n = 6 patients; CLL3 is beneath CLL5). Experimental setup as in (C). (E) Induction of NF-kB by RFP2 can be inhibited with dominant-negative IkB kinase (dnIKK) and dominant-negative IkB (dnIKB). HEK293T cells were transfected with expression vector encoding wt-RFP2 or mutant RFP2 in combination with dnIKK or dnIkB or empty vector, and activity of NF-kB was measured with a luciferase reporter. Inhibition of NF-kB activity by dnIKK and dnIkB was controlled by stimulation with TNFa 6 hrs after transfection and quantification of NF-kB activity after 24 hrs. (F) RFP2 induces NF-kB activity via p65 and p105. p105, RELA (p65), cREL, RELB and p100 were knocked down in HEK293 and HEK293T cells that were transfected with RFP2 expression plasmid. NF-kB activity was measured with luciferase reporter assay. *p<0.05 with students t-test. (G) RFP2 induces DNA-binding of RELA. HEK293T cells were transfected with empty vector (“empty”), expression plasmid containing wildtype RFP2 (“RFP2”), RFP2 with mutated ubiquitin ligase activity (“RFP2mut”) or were stimulated with TNFalpha (“TNFa”). After 24 hrs, DNA-binding capacity of RELA was quantified by co-ELISA. Error bars represent standard deviation of 2 independent experiments. (H) Knockdown of VCP but not SQSTM1 modulates activation of NF-KB by RFP2. SQSTM1 and VCP were knocked down and plasmids for overexpression of RFP2 and luciferase reporters detecting NF-kB activity were transfected after 24 hrs into HEK-293. While only a minor change could be observed after knockdown of SQSTM1, knockdown of VCP led to a substantial increase in the activation of NF-kB after cotransfection of RFP2. (I) Cotransfection of RFP2 leads to stabilization of C13ORF1 protein. HEK-293 and HEK-293T cells were transfected with wildtype RFP2 or mutated RFP2 lacking ubiquitin ligase activity alone or in combination with C13ORF1. Protein levels of RFP2 and C13ORF1 were monitored by Western blot. β-actin served as loading control. Experiment is representative of two independent experiments.
Figure 6.
Model of the tumor suppressor mechanism localized in 13q14.3.
(A) Regions in CpG islands D and E that are DNA-methylated in non-malignant B-cells (left) become demethylated in the vast majority of CLL patients (right). This coincides with relaxed chromatin characterized by absence of macroH2A and enrichment of H3K4me2 at the promoters of the lncRNA genes DLEU1 and DLEU2/Alt1. (B) DNA-hypomethylation is correlated with transcriptional upregulation of splicing variants of the two lncRNA genes DLEU1 and DLEU2 and inversely correlated with the protein-coding genes in 13q14.3. No correlation could be found with levels of mature miR-15a and miR-16, probably because these transcripts are also deregulated by a posttranscriptional processing defect in CLL cells (Allegra, manuscript submitted). (C) Candidate genes localized in the critical region in 13q14.3 are functionally related and all modulate NF-kB signalling, albeit with different impact. Nuclear transporter KPNA3 is binding NF-kB components p65 and p50 and is therefore likely a positive regulator like e.g. RFP2. RFP2 binds to C13ORF1 and induces NF-kB activity via the canonical pathway components p50 and p65, for which its ubiquitin ligase actvity is required. The miRNA genes miR-15a and miR-16-1 were identified together with other members of this miR family to be among the strongest activators of NF-kB activity. Previously they have been shown to both modulate cell cycle regulators and inhibit NF-kB via TAB3/TAK1. DLEU7 has recently also been reported to inhibit NF-kB by binding and inactivating the TACI/BCMA receptors.