In mantle cell lymphoma (MCL) and some cases of multiple myeloma (MM), cyclin D1 expression is deregulated by chromosome translocations involving the immunoglobulin heavy chain (IgH) locus. To evaluate the mechanisms responsible, gene targeting was used to study long-distance gene regulation. Remarkably, these targeted cell lines lost the translocated chromosome (t(11;14)). In these MCL and MM cells, the nonrearranged cyclin D1 (CCND1) locus reverts from CpG hypomethylated to hypermethylated. Reintroduction of the translocated chromosome induced a loss of methylation at the unrearranged CCND1 locus, providing evidence of a transallelic regulatory effect. In these cell lines and primary MCL patient samples, the CCND1 loci are packaged in chromatin-containing CCCTC binding factor (CTCF) and nucleophosmin (NPM) at the nucleolus. We show that CTCF and NPM are bound at the IgH 3′ regulatory elements only in the t(11;14) MCL cell lines. Furthermore, NPM short hairpin RNA produces a specific growth arrest in these cells. Our data demonstrate transvection in human cancer and suggest a functional role for CTCF and NPM.
B cell malignancies such as non-Hodgkin's lymphoma and multiple myeloma (MM) are characterized by 14q32 translocations involving the IgH locus (1). These translocations serve to juxtapose IgH regulatory elements, such as the intronic enhancer (Eμ;1) or 3′ Cα locus control region (LCR) (2, 3), that deregulate transcription of target genes over several hundred kilobases of DNA. The mechanisms involved in long-distance deregulation of target genes by IgH regulatory elements are unknown; however, regulatory elements in the IgH locus are thought to derepress or increase the transcription of target genes such as CMYC (1, 2) and cyclin D1 (CCND1) (1, 4) that are involved in B cell malignancies. Epigenetic changes in chromatin structure involving DNA methylation (5) and/or histone modifications (6) have been implicated in deregulated gene expression (4).
The activation of the CCND1 gene that occurs in mantle cell lymphoma (MCL) and a subset of MM was used as a model system to investigate the mechanisms responsible for long-distance gene deregulation in B cell malignancies (see Fig. 1 A). Cyclin D1 is not expressed in normal lymphocytes, where the unlinked family members cyclin D2 and/or D3 are active (7). In B cell malignancies, cyclin D1 gene expression is activated by the insertion or translocation of IgH regulatory elements, such as the Eμ intronic or 3′ Cα enhancer/LCR, that can be as far as 100–300 kb away from the CCND1 gene (4, 8). The majority of the breakpoints in MCL map to the major translocation cluster (MTC) region located ∼120 kb upstream (centromeric) of the CCND1 gene (8). The nearest gene to CCND1, MyeOV, is located 360 kb centromeric to CCND1 and is expressed in a subset of t(11;14) MM, but not in MCL (9). Although translocations involving the t(11;14) are most common, the MM cell line U266 contains an insertion of IgH regulatory sequences ∼10 kb centromeric of the CCND1 promoter (10).
The CCND1 promoter contains a CpG island that can be potentially regulated by DNA methylation (4, 11). We have found that in normal B cells the CCND1 locus is organized into hypomethylated DNA bound by acetylated nucleosomes. Analysis of the CCND1 DNA methylation status in MM and MCL cell lines with t(11;14) indicates that the deregulated as well as the normal, silent CCND1 loci are CpG hypomethylated (4).
Control of DNA methylation in mammalian cells has been shown to involve a cis-acting mechanism (12). However, observations in higher plants (13–15), the fungus Ascobolus immerses (16), and more recently in mice (13, 17) demonstrated that DNA methylation can also be regulated in trans by interactions between homologous chromosomes.
A ubiquitous, complex protein that has emerged as a critical mediator of multiple epigenetic processes is the zinc finger protein CCCTC binding factor (CTCF). CTCF is a highly conserved, 11 zinc finger protein first identified as a c-myc binding factor and subsequently shown to bind to metazoan regulatory elements known as insulators (18). Insulator elements, which act from an intervening position to prevent flanking cis-acting elements from interacting, mediate their function through CTCF (18, 19). Furthermore, the binding of CTCF to insulator elements is blocked by CpG methylation, allowing interactions between distal regulatory elements in the imprinted Igf2/H19 locus (19, 20). Recent evidence has identified the nucleolar protein nucleophosmin (NPM) as a major CTCF-interacting protein that functions to tether promoters and regulatory elements separated by large linear distances together at the nucleolar periphery (21).
To investigate the mechanisms of long-distance activation of the CCND1 gene by the IgH enhancer/LCR, homologous recombination was used to target regulatory regions potentially involved in cyclin D1 deregulation. Recombinants that lost the translocated chromosome also lost the ability to maintain hypomethylation at the normal CCND1 allele. These clones no longer expressed cyclin D1. These findings suggest that the translocated chromosome exerts a long-distance cis DNA hypomethylating effect on the linked CCND1 promoter as well as a transallelic effect on the unlinked CCND1 promoter on the homologous chromosome. Thus, in the absence of the translocated chromosome, the unrearranged CCND1 locus is densely DNA methylated. CTCF and NPM are associated with both the translocated and unrearranged CCND1 loci in cell lines with t(11;14). Interestingly, binding of these proteins at the unrearranged CCND1 locus is dependent on the presence of the translocated allele. These results provide further evidence that long-distance control of DNA methylation in mammalian cells can be communicated in trans, and demonstrate a novel mechanism for chromosome structure and nuclear organization to participate in deregulated gene expression in human malignancy.
B cell lines that have lost the t(11;14) translocation do not express cyclin D1 and have altered DNA methylation patterns at the unrearranged CCND1 locus
A gene targeting strategy was designed to study regulatory regions involved in cyclin D1 deregulation in human B cell malignancies. Our initial approach was to target the potential regulatory regions involved in long-distance cyclin D1 deregulation in B cell malignancies. By analogy, results in the β-globin locus demonstrated that insertion of a selectable marker gene into the β-globin LCR abrogated transcription of downstream genes (22). Two gene targeting vectors (TV1 and 2; Fig. 1 A) were constructed and electroporated into two MCL cell lines, Granta 519 (Granta [G]) and NCEB-1 (NCEB [N]), and the MM cell line U266 (U), all of which express cyclin D1. The initial Southern blot analysis was consistent with successful gene targeting (Fig. 1 B); however, subsequent analyses showed instability and loss of the targeted/translocated chromosomes. Additional clones were isolated using TV1 in NCEB (N1–3) cells, and TV2 in U266 (LU clones) and Granta (LG clones) cells. Southern blot analysis using cell line–specific translocation breakpoint probes (P519 for Granta and MTC for NCEB) demonstrated the absence of translocated chromosome breakpoint sequences in the mutant MCL cell lines (Fig. 1 C). Fluorescent in situ hybridization (FISH) analysis demonstrated that the translocated chromosomes had been lost in all of these clones (Fig. 1, D and E; and Table S1).
We assessed the expression of D-type cyclins in each of the parental and targeted MCL cells to determine whether loss of t(11;14) affected expression of the remaining cyclin D alleles. Parental Granta and NCEB MCL cells expressed cyclin D1 (4). However, the derivative MCL cell lines, which had lost t(11;14), did not express cyclin D1 (Fig. 2, A and B). Instead, up-regulation of cyclin D3 mRNA (not depicted) and protein (Fig. 2 C) levels was observed, presumably to compensate for the loss of cyclin D1 that has been observed in other experimental cell culture and animal systems (23, 24).
DNA methylation patterns within the CCND1 locus in the variant cyclin D1− MCL cells were analyzed. The CCND1 promoter region and sequences 3 kb upstream were extensively methylated, as shown by Southern blot analysis (Fig. 3 B) and bisulfite sequencing (Fig. 3 C). DNA methylation changes occurred in sequences 90 kb (P519) upstream of the CCND1 gene. Because the translocation breakpoints in the NCEB and Granta cell lines are located at the 3′ end and proximal to the MTC (Fig. 3 A), respectively, most of this region lies on the reciprocal translocation partner. At the MTC, CpG sequences such as at the Hpa II/Msp site remained hypomethylated in both cyclin D1-expressing and nonexpressing clones. Similar results were obtained with independently isolated Granta and NCEB clones (Fig. 3, B and C).
DNA methylation in the cyclin D1− genetic variants was also assayed by methylation-specific PCR (MSP) (25), which complements the Southern blotting and sodium bisulfite sequencing approaches (26). MSP analysis of the CCND1 promoter and the 3-kb upstream regions in Granta, NCEB, G4, and N1 cells revealed that both regions were heavily CpG methylated in the genetic variants (Fig. S1).
These data are consistent with a model in which the translocated chromosome exerts a transallelic hypomethylating effect on the nontranslocated chromosome in MCL parental cell lines. This effect disappeared in the targeted variants with the loss of the translocated chromosome, leading to extensive methylation of the CCND1 locus. The transallelic hypomethylating effect extended from the CCND1 promoter at least to the P519 region 90 kb upstream.
To determine if our observations in MCL cells were applicable to cyclin D1–expressing MM cells that contained an insertion of IgH sequences instead of the classical t(11;14), we analyzed similar variant clones from parental U266 cells that contain IgH regulatory sequences inserted ∼10 kb centromeric of the CCND1 promoter (12). Similar to the variant Granta and NCEB clones, U7, LU1, and LU2 clones were isolated where the active chromosome 11 was lost (Fig. 1 E; and Table S1), and no cyclin D1 protein was detected (Fig. 2 B). In addition, two other control MM cell lines that contain t(11;14) and express cyclin D1 (KMS12 and KMS21BM) (27) were analyzed.
Analysis of the DNA methylation status at the CCND1 locus in the parental MM cell lines and the U266 cyclin D1− derivatives revealed that the CCND1 promoter and P519 regions were hypomethylated in all of the cyclin D1+ MM cell lines, including U266, KMS12, and KMS21BM (Fig. 3 D), similar to the results obtained with the parental Granta and NCEB MCL cell lines. In contrast, DNA hypermethylation was observed at the CCND1 locus in the cyclin D1− U266 MM cell lines that had lost the active CCND1 locus with inserted IgH regulatory elements. Thus, transallelic effects at the CCND1 locus are observed with inserted as well as translocated IgH sequences.
Monoallelic histone acetylation patterns at the cyclin D1 loci
To determine if posttranslational histone modifications such as histone acetylation were coordinately involved with DNA methylation in transallelic effects at the cyclin D1 locus, histone acetylation patterns in the cyclin D1− genetic variants and parental cell lines were studied by chromatin immunoprecipitation (ChIP) assays. In contrast to the parental cell lines, the cyclin D1 coding and promoter regions and sequences extending 3 kb upstream were histone H3 and H4 deacetylated in the mutant cell lines, consistent with the DNA methylation patterns and the lack of transcription of the cyclin D1 gene. However, regions far upstream of the cyclin D1 gene, including the MTC, P519, and the IgH regulatory regions, remained histone H3 and H4 hyperacetylated in the mutant cell lines (Fig. 4 B) (4). Similar results were obtained with specific antibodies to acetyl H3 lysine 9 (AcK9; Fig. 4 C).
Because the changes in histone acetylation observed in the mutant cells were also consistent with the loss of the histone H3 and H4 acetylated, translocated allele, allele-specific histone acetylation assays were performed. Specifically, we wished to determine whether both the translocated and nontranslocated CCND1 alleles in the parental cells contained acetylated histones. Allele-specific patterns of histone acetylation were studied in the U266 MM cell line, whose cyclin D1 alleles can be distinguished by a polymorphism at the cyclin D1 intron/exon 4 junction (Fig. 4 D). Allelic ChIP assays performed on U266 cells indicated that acetylated histones were only present on the translocated allele (Fig. 4 E), suggesting that histone acetylation was not participating in transallelic effects at the CCND1 locus.
Reintroduction of the translocated t(11;14) alters DNA methylation in trans
Somatic cell fusion experiments have been used to study the contribution of individual alleles to both the transcription and DNA methylation status of several genetic loci in human malignancy (28, 29). Thus, cell fusion was used to determine if reintroduction of the translocated chromosome into the mutant cell lines where the CCND1 locus was methylated was capable of effecting transhypomethylation of the methylated allele. The Granta cell line (CCND1 hypomethylated) was fused with the N1 cell line (CCND1 methylated) to enable use of a DNA polymorphism within the CCND1 gene (30) to distinguish the translocated from the normal CCND1 loci (Fig. 5, A and B). All of the hybrid (G/N1) clones isolated contained both the untranslocated CCND1 locus from N1 cells and the translocated CCND1 locus from Granta cells, and expressed cyclin Dl (Fig. 5, C–F). No hybrid clones could be isolated that contained only the nontranslocated CCND1 loci (see Discussion). RFLP analysis demonstrated that cyclin D1 gene expression in these G/N1 hybrid clones originated from the Granta cell–derived translocated CCND1 locus (Fig. 5 E).
DNA methylation in the hybrid G/N1 cells was analyzed using Southern blotting (Fig. 5 G) and sodium bisulfite sequencing (Fig. 6) to determine whether introduction of the translocated CCND1 locus was capable of effecting transhypomethylation of the methylated allele. A 1:1 mix of the two parental DNAs and DNAs from nonhomologous clones were used as controls (Fig. S2). Analysis of DNA methylation patterns by bisulfite sequencing and MSP (Fig. 6) demonstrated that the translocated CCND1 locus was able to effect DNA hypomethylation at the promoter and 3 kb upstream of the promoter from the unrearranged, methylated CCND1 allele in G/N1 hybrid cells containing both the translocated and nonrearranged CCND1 loci. Extensive hypomethylation of the upstream 3-kb region was observed, but the transhypomethylation at the CCND1 promoter was incomplete in G/N1 hybrids. DNA methylation was reevaluated in these hybrid cells after multiple rounds of DNA replication to determine if the transhypomethylating effect at the promoter could be magnified (Fig. 6 B). In fact, DNA replication was associated with a small increase in CCND1 promoter methylation.
Treatment of these hybrid cells with a DNA hypomethylating agent (5-azacytidine) and/or a histone deacetylase inhibitor (trichostatin A) was performed (Fig. 6 C). Treatment with 5-azacytidine alone or in combination with trichostatin A led to almost total DNA hypomethylation at the CCND1 promoter in the hybrid cells, whereas treatment with trichostatin A alone did not change DNA methylation patterns.
Because hybrid clones containing only the nontranslocated CCND1 locus could not be isolated (see Discussion), control fusions were performed using N1 and lymphocyte cell line (LCL) cells to investigate the effects of introduction of a nontranslocated, hypomethylated CCND1 locus (4) on the densely methylated CCND1 locus in N1 cells. MSP analysis of the CCND1 3-kb upstream region demonstrated that in contrast to G/N1 cells, LCL/N1 hybrid cells did not demonstrate hypomethylation in trans at the CCND1 locus (Fig. 6 D).
To further demonstrate that transallelic effects in MCL were specific to t(11;14), malignant lymphoid cells that had deleted one CCND1 allele without the presence of t(11;14) were studied. A variant MCL patient (31) with marked peripheral blood lymphocytosis was identified (expressing cyclin D3 and no cyclin D1 mRNA; not depicted). FISH analysis demonstrated that this patient's malignant B cells contained a deletion of one allele of chromosome 11q, including the cyclin D1 locus (Fig. S3), and no t(11;14). MSP analysis of this 11q− patient's DNA demonstrated a hypomethylated CCND1 allele (Fig. 6 D). Thus, transallelic effects were specific to the presence of the translocated CCND1 locus.
CTCF, NPM, and the nucleolus are associated with cyclin D1 activation and transvection
ChIP assays were used to screen parental versus variant cyclin D1− G4 cells for DNA binding proteins that would demonstrate selective binding to the translocated, transcriptionally active allele (Fig. 7). Proteins were assayed that have been reported to bind to the CCND1 promoter in vitro or whose potential consensus binding sites are located in the CCND1 promoter and IgH regulatory regions (32–35). Our results demonstrated that CTCF, Oct2, and Sp1 exhibited selective in vivo binding at the CCND1 promoter region in G versus G4 cells, whereas OCA-B (Bob-1), NF-κB, or a subunit of NF-κB (Rel-B) did not. Additional ChIP assays were performed to further study in vivo DNA–protein interactions of CTCF, Oct2, and Sp1 at the CCND1 promoter, and IgH regulatory regions in cyclin D1–expressing and control cell lines. Oct2, Sp1, and CTCF binding to the CCND1 promoter region correlated with cyclin D1 gene expression in MCL cells (Fig. 7 B). Primers from the IgH Eμ enhancer 3′ Cα LCR regions showed CTCF, Oct2, and Sp1 binding to the Eμ enhancer in both MCL and LCL cells. CTCF but not Oct2 or Sp1 binding to 3′ Cα LCR HS3 and HS4 correlated with cyclin D1 expression (Fig. 7 B).
The binding of CTCF at the CCND1 locus and IgH regulatory regions was further examined using ChIP assays (Fig. 7 C). Although binding at IgH 3′ Cα LCR HS3 and HS4 only occurred in cell lines that expressed CCND1, primers for 3′ Cα HS12 demonstrated binding of CTCF in all cell lines examined (Fig. 7 C). Collectively, these results suggested that 3′ Cα HS3 and HS4 binding by CTCF could be important in deregulated cyclin D1 expression in MCL.
Western blot analyses were performed to assay the level of these proteins in the cell lines used (Fig. S4). cyclin D1 protein was detected in the MCL cell lines Granta and NCEB, and the breast cancer cell line MCF7, consistent with RT-PCR assays (4). The transcription factors CTCF, Oct2, and Sp1 were detected in all cell lines examined. Thus, the specific binding of CTCF observed was not secondary to variations in the levels of this protein in MCL cells.
CTCF has been shown to interact with the nucleolar protein NPM to tether an insulator sequence to subnuclear sites at the nucleolar periphery (21). ChIP assays demonstrated binding of NPM at the CCND1 promoter and IgH regulatory regions in MCL cells, similar to results obtained with CTCF (Fig. 7 D). These results suggested a potential role for CTCF in juxtaposing 3′ Cα IgH regulatory elements (HS3 and HS4) and the CCND1 promoter by tethering them to the nucleolar periphery with NPM. CTCF and NPM could also play a role in mediating pairing of the translocated and normal CCND1 loci, facilitating transallelic effects (Fig. 7 E).
If CTCF and NPM were involved in tethering the CCND1 loci together, evidence of binding of these proteins to both CCND1 alleles would be expected. Allelic ChIP assays were performed with antibodies to CTCF, NPM, and RNA polymerase II (Pol II). MCF7 cells, a cyclin D1–expressing breast cancer line without the t(11;14) translocation, demonstrated biallelic transcription of cyclin D1 (30) as well as biallelic binding of Pol II, NPM, and CTCF (Fig. 8 B). An EBV LCL (LCL1) that does not express cyclin D1 demonstrated no binding of Pol II or CTCF at either allele, whereas the CCND1 loci in this cell line are CpG hypomethylated and histone H3 and H4 are hyperacetylated (4). U266 MM cells only transcribed the CCND1 allele containing the IgH insertion, as shown by RT-PCR analysis and ChIP analysis using Pol II antibodies (Fig. 8, A and B). However, CTCF and NPM ChIP DNA from U266 but not LCL cells contained sequences from both the transcribed and nontranscribed CCND1 alleles (Fig. 8 B). Thus, both the transcribed and nontranscribed CCND1 loci bind CTCF and NPM, consistent with our hypothesis that CTCF and NPM are involved in tethering these loci together.
To provide further evidence that CTCF and NPM were directly cobinding at the CCND1 locus, sequential ChIP (36) assays were performed on the EBV-negative MCL cell line HBL-2 (37) using sequential IP assays with anti-CTCF and NPM antibodies. These results demonstrated that CTCF and NPM were binding together at the CCND1 locus in MCL cells (Fig. 8 C).
FISH analysis combined with NPM immunofluorescent FISH (immunoFISH) staining was used to visualize the position of the CCND1 loci in relationship to the nucleolus. Fixation protocols using methanol/acetic acid or paraformaldehyde alone, or in combination, yielded similar results. The CCND1 locus colocalized at the nucleolar periphery in CCND1-translocated cell lines (Fig. 9, A and G) and in neoplastic cyclin D1–expressing lymphocytes from MCL patients (Fig. 9 B; and Fig. S5). Pairing of the CCND1 loci at the nucleolus was also observed, as indicated by two juxtaposed FISH signals (Fig. 9, B and C; and G and H). The paired CCND1 loci observed were unlikely to be caused by replicated but unseparated sister chromatids, because the same frequency of pairing was observed in flow-sorted Granta nuclei in the G1 cell cycle phase (Fig. 9 H). Paired CCND1 loci were not observed in cyclin D1–expressing MCF7 cells and LCL cells where the CCND1 locus was hypomethylated and histone H3 and H4 were acetylated (4).
In MCL cells that lost the translocated CCND1 locus (G4), the nontranslocated chromosome 11 remained associated with the nucleolus, providing further evidence that pairing of the cyclin D1 loci at the nucleolus was occurring in the parental MCL cell lines (Fig. 9, D and H). These observations demonstrated the nucleolar environment is capable of both activation and repression functions at the CCND1 loci in MCL cells (see Discussion). In cyclin D1–expressing MCF7 cells, B lymphocytes (LCLs), and the variant MCL patient (Fig. S4) that do not contain the t(11;14) translocation, colocalization of the CCND1 loci and the nucleolus was not observed (Fig. 9, E and F; and not depicted).
To further investigate the role of the nucleolus and NPM in deregulated cyclin D1 expression in t(11;14) B cell malignancies, we used NPM short hairpin RNA (shRNA) to knock down levels of NPM. Cell growth was specifically inhibited by NPM shRNA in cell lines containing a t(11;14) but not in B lymphocytes without this translocation and MCF7 cells (Fig. 10 A). Western blotting demonstrated significant knockdown of NPM protein levels but no change in cyclin D1 protein levels in either MCF7 or Granta cells (Fig. 10 B). Western analysis showed a marked decrease in NPM levels in cells containing the NPM shRNA (Fig. 10 C). The NCEB MCL cell line demonstrated marked morphological changes that were consistent with apoptosis and were confirmed by Tdt-mediated dUTP-biotin nick-end labeling (TUNEL) assays. Lower levels of apoptosis were observed in Granta cells (Fig. 10 D), and minimal apoptosis was seen in G4 and N1 cells (Fig. S6). MCF7 and K562 erythroleukemia cells treated with the NPM but not control shRNA vectors showed loss of nucleolar NPM staining, and CTCF, cyclin D1, and residual NPM staining in the cytoplasm (Fig. 10 C). These results are consistent with the role of NPM as a shuttle protein (38) and suggest that CTCF and cyclin D1 are among the proteins regulated by NPM. immunoFISH assays revealed no significant change in nucleolar colocalization of the CCND1 loci in NPM knockdown cells (Fig. 10 E), suggesting that additional layers of complexity are involved in the tethering of the CCND1 loci in MCL cells. NPM knockdown cells also did not demonstrate any change in cyclin D1 protein levels (Fig. 10 B), suggesting that NPM binding at the CCND1 locus was not directly involved in regulating expression of the cyclin D1 gene but could be serving a structural role in tethering the CCND1 loci to the nucleolus.
Previously, the DNA methylation patterns of the 11q13 region surrounding the CCND1 gene in MCL and MM cells were examined (4). The activation of the cyclin D1 gene by IgH regulatory elements was found to correlate with the absence of DNA methylation at the CCND1 promoter. Interestingly, this domain of hypomethylated DNA appears to be present not only on the translocated but also on the untranslocated allele. Normal B cells and EBV lymphocytes also displayed extensive DNA hypomethylation and hyperacetylation of histones at the CCND1 loci, although the CCND1 gene is not expressed (4). Hypomethylation up to 90 kb upstream of the CCND1 promoter was eliminated with loss of the translocated chromosome initiated through a gene targeting event.
Somatic cell hybrid fusions indicated that reintroduction of the translocated chromosome but not a nontranslocated chromosome was capable of transhypomethylating a densely methylated CCND1 locus. The Granta cells used in the fusion experiments that reintroduced the t(11;14) chromosome into N1 cells also contained a nontranslocated CCND1 locus that possessed the identical cyclin D1 polymorphism as the translocated CCND1 locus. The recipient N1 cell line possessed a different, distinguishable polymorphism in the cyclin D1 gene (Fig. 5). All of the hybrid clones isolated contained the translocated chromosome. We were unable to isolate clones that contained only the nontranslocated CCND1 locus from Granta cells. Thus, the translocated and nontranslocated CCND1 loci could not be segregated. This result provides further, albeit indirect, evidence that the translocated and nontranslocated cyclin D1 loci are tethered together (at the nucleolus), as suggested by our immunoFISH data (Fig. 9). Alternatively, defects in DNA methylation have been associated with chromosomal instability, both in embryogenesis (39) and in colorectal cancer cell lines (40). Therefore, another possibility is that the hypomethylated nontranslocated CCND1 locus was unstable and was lost during the cell fusion process.
Control fusion experiments indicated that fusion of N1 cells with a B lymphocyte line without a translocated chromosome was unable to hypomethylate the densely methylated cyclin D1 locus of N1 cells in trans. Similar results using other systems demonstrated no trans epigenetic effects with nontranslocated chromosomes (28, 29).
Because the CCND1 locus is extensively hypomethylated in normal B cells, at least three DNA methylation states of the CCND1 loci are implied: (a) normal B cells in where both CCND1 loci are unmethylated and independent; (b) B cells containing CCND1 translocations/insertions, where the translocated locus exerts a transhypomethylating effect on the untranslocated locus; and (c) B cells that have lost the translocated chromosome and where the CCND1 locus is CpG methylated. The second and third DNA methylation states are dependent on localization at the nucleolus, whereas the second DNA methylation state appears to be associated with pairing of the CCND1 loci. Thus, the association of genetic loci with the nucleolus can result in either active (state 2) or inactive (state 3) chromatin, consistent with the results obtained in other systems (21, 41). The mechanisms involved in gene expression or repression by association with the nucleolus remain obscure, but could be related to the ability of the translocated allele to recruit activator molecules that protect the tethered cyclin D1 loci from the general repressive effects of the nucleolus. There may also be subtle differences in nucleolar localization that may be important in epigenetic activation versus repression.
The change in DNA methylation patterns observed in the normal CCND1 locus after loss of the translocated chromosome are consistent with transallele sensing effects (42), which are well studied in Drosophila (transvection; references 42–44) and plants (paramutation) (13–15). Trans effects in both endogenous and transgenic mouse loci have been reported (45–49). Such effects often involve DNA methylation and occur over long distances (45, 46, 48). Transactivation of the endogenous Igf2 gene by trangenes has been reported (49), but given the results reported in this paper, transallelic effects should be considered.
Recent observations in transgenic mice have described paramutation-like effects at the Igf2 (45), U2af1rs-1 (46), c-kit (47), and Rosa26 (48) loci. Paramutation/transvection in human genetic disease associated with the susceptibility to type I diabetes has been suggested at the insulin minisatellite variable number tandem repeat element (50). However, our observed transallelic effects are similar to transvection but differ from paramutation because the DNA transhypomethylation effects are not heritable in the absence of the inducer (translocated) allele (14).
Physical evidence of transcommunication has recently been observed between loci on nonhomologous chromosomes regulated by the T cell enhancers/LCRs (51), and the Igf2/H19 and Wsb1/Nf1 loci (52). CTCF has been demonstrated to be involved in long-distance cis and trans effects at the Igf2/H19 locus (52, 53). Thus, transcommunication may be an important means of regulation in mammalian cells.
At least two nonexclusive mechanisms are implicated in the homology-based transfer of DNA methylation patterns (14). One mechanism involves direct communication between homologous chromosomes, as has been demonstrated in imprinted loci (54) or in mouse embryonic stem cells at the onset of X inactivation (55). Recent work has shown that CTCF is required for this transient X chromosome pairing (56). The nucleolus has also been implicated in silencing of the inactive X chromosome using immunoFISH assays similar to those performed in this study (57). Our FISH and allelic ChIP data provide evidence for direct communication of the CCND1 alleles in MCL cells that we suggest is mediated via CTCF, NPM, and the nucleolus.
Alternatively, small RNA molecules have been shown to direct chromatin modification in trans in plants and Drosophila (13, 14, 58–61). This epigenetic silencing mechanism involves CpG and non-CpG DNA methylation in plants. Tandem and inverted repeats have been shown to be involved in this type of RNA-mediated repression (58–61). More recently, a paramutation-like effect mediated by small RNAs has been described at the mouse c-kit locus (47).
Our allelic ChIP and immunoFISH data suggest that CTCF and NPM participate in transallelic interactions by tethering the cyclin D1 loci from both translocated and nontranslocated cyclin D1 loci together at the nucleolar periphery, as depicted in Fig. 7 E. Our data demonstrate that the 3′ Cα IgH LCR region is sufficient to mediate this effect and that CTCF binding to 3′ Cα LCR HS3 and HS4 correlates with cyclin D1 activation in MCL cells.
Our finding in cyclin D1–deregulated B cell malignancies represents the first report of the association of transvection with human cancer. The critical role of NPM in t(11;14) B cell malignancies is demonstrated by growth arrest and apoptosis selectively in t(11;14) cell lines with NPM shRNA. Our data also suggest a novel approach to therapy of t(11;14) malignancies using agents targeting NPM.
Materials And Methods
Cell culture, electroporation, and retroviral infection.
Cell lines were obtained and maintained as previously described (4). The HBL-2 MCL cell line (37) was provided by S. Dave (National Cancer Institute, Bethesda, MD). Fig. 1 shows details of gene targeting constructs. Electroporations were performed, as previously described (22), using 10–20 μg of linearized targeting vector. Retroviral infections were performed as previously described (62).
FISH (63) and immunoFISH (64) were performed as previously described. For immunoFISH, cells were fixed using 4% paraformaldehyde, methanol/acetic acid (3:1), and methanol/acidic acid, followed by paraformaldehyde with similar results. For two-color immunoFISH, a combination of CCND1 probes from chromosome 11 (9) was labeled by nick translation (63). Three-color immunoFISH used a t(11;14) fusion probe (Vysis; Fig. 1). Cy5-conjugated anti–mouse secondary antibody (Invitrogen) was used with anti-NPM antibody after FISH, according to the supplied protocol. Deconvolution microscopy (Deltavision) was performed using the OHSU core facility. Granta cells were stained with Hoechst dye, and G1 cells were isolated by flow cytometry for immunoFISH analysis. At least 100 cells were assayed per cell line.
Peripheral blood or bone marrow was obtained from MCL patients giving informed consent under an OHSU institutional review board–approved Southwest Oncology Group study. Buffy coat was purified using Ficoll, as previously described (4).
Somatic cell fusions.
Somatic cell fusions were performed using polyethylene glycol, as previously described (65). Hybrids were selected using combined resistance to neomycin and hygromycin or hygromycin and puromycin.
Southern blotting and bisulfite sequencing.
ChIP assays were performed and PCR primers were used as previously described (4). Supplemental materials and methods describes primers and antibodies.
The CCND1 G/A polymorphism was detected by the PCR- RFLP method. A 167-bp fragment of the CCND1 gene at the junction of exon 4/intron 4 was amplified by PCR using 0.1 μg of genomic DNA at an annealing temperature of 64°C. For RFLP analyses, each PCR product was purified by gel extraction and digested with BsrI (ACTGGN/) at 50°C before electrophoresis. The DNA fragments were separated using a 3% 2:1 Nusieve/SeaKem agarose gel. The allele types were determined as shown in Fig. 4. Primer sequences are provided in Supplemental materials and methods.
TUNEL assays were done according to the manufacturer's protocol (Roche). 300–500 nuclei were assayed per sample.
Online supplemental material.
Table S1 summarizes the FISH results demonstrating the absence of t(11;14) in the targeted clones. Fig. S1 shows the results of MSP analysis of parental and cyclin D1− clones. Fig. S2 shows the results of bisulfite sequencing mixing control experiments. Fig. S3 demonstrates FISH analysis of a cyclin D1− 11q− MCL patient. Fig. S4 shows Western blot analysis of proteins assayed by ChIP analysis in the cell lines used. Fig. S5 shows the results of staining of blood from an MCL patient with antibodies to cyclin D1 and NPM. Fig. S6 shows the results of TUNEL assays on control and NPM shRNA–treated cells. Supplemental results describes Fig. 1 D in detail. Supplemental materials and methods provides details about antibodies and PCR primers.
© 2008 Liu et al. This article is distributed under the terms of an Attribution–Noncommercial–Share Alike–No Mirror Sites license for the first six months after the publication date (see http://www.jem.org/misc/terms.shtml). After six months it is available under a Creative Commons License (Attribution–Noncommercial–Share Alike 3.0 Unported license, as described at http://creativecommons.org/licenses/by-nc-sa/3.0/).
Abbreviations used: CCND1, cyclin D1; ChIP, chromatin immunoprecipitation; CTCF, CCCTC binding factor; FISH, fluorescent in situ hybridization; immunoFISH, immunofluorescent FISH; LCL, lymphocyte cell line; LCR, locus control region; MCL, mantle cell lymphoma; MM, multiple myeloma; MSP, methylation-specific PCR; MTC, major translocation cluster; NPM, nucleophosmin; Pol II, RNA polymerase II; shRNA, short hairpin RNA; TUNEL, Tdt-mediated dUTP-biotin nick-end labeling.
H. Liu and J. Huang contributed equally to this paper.
H. Liu's present address is Nimblegen Systems, Madison, WI 53711.
We thank Mark Groudine, William Forrester, Matt Lorincz, Richard Maziarz, Janine Lasalle, and Vicki Chandler for discussions and comments. Rich Fisher, Tom Miller, and the Southwest Oncology Lymphoma Committee provided MCL patient samples.
This work was supported by grants from the National Institutes of Health (HL069133 and HL077818 to W.H. Fleming, and DK56798 to E. Epner), the International Myeloma Foundation (to E. Epner), and the Lymphoma Research Foundation (to E. Epner).
The authors declare no financial conflicts of interest.