Human Lymphoid Translocation Fragile Zones Are Hypomethylated and Have Accessible Chromatin

Abstract

Chromosomal translocations are a hallmark of hematopoietic malignancies. CG motifs within translocation fragile zones (typically 20 to 600 bp in size) are prone to chromosomal translocation in lymphomas. Here we demonstrate that the CG motifs in human translocation fragile zones are hypomethylated relative to the adjacent DNA. Using a methyltransferase footprinting assay on isolated nuclei (in vitro), we find that the chromatin at these fragile zones is accessible. We also examined in vivo accessibility using cellular expression of a prokaryotic methylase. Based on this assay, which measures accessibility over a much longer time interval than is possible with in vitro methods, these fragile zones were found to be more accessible than the adjacent DNA. Because DNA within the fragile zones can be methylated by both cellular and exogenous methyltransferases, the fragile zones are predominantly in a duplex DNA conformation. These observations permit more-refined models for why these zones are 100- to 1,000-fold more prone to undergo chromosomal translocation than the adjacent regions.

INTRODUCTION

Chromosomal translocations are seminal events in the development of many hematologic malignancies (1). Many hematopoietic malignancies are associated with recurrent reciprocal translocations and their resulting gene fusions or amplifications, which guide diagnosis, assessment of prognosis, and treatment. One example is the BCL2-IGH translocation, t(14;18)(q32;q21), observed in 85% of follicular lymphomas (FL) and 30% of diffuse large B cell lymphomas (DLBCL). A second example is the CCND1-IGH translocation, t(11;14)(q13;q32), observed in almost all mantle cell lymphomas (MCL). Oncogenes near sites of recurrent translocation often confer a selective growth advantage (2). For unknown reasons, human chromosomal translocation breakpoints near these oncogenes frequently cluster within tight DNA zones (often only 20 to 600 bp in length) known as fragile zones (see the green starbursts in the upper portion of Fig. 1A and 2A). These zones are 100- to 1,000-fold more sensitive to breakage than the adjacent DNA regions located, for example, only 20 bp to either side of the fragile zone (2–7). Hence, the DNA sequences surrounding each oncogene are not equally sensitive to breakage. The molecular basis for such an extreme breakage propensity within such a short zone of DNA is unknown, despite our progress on elucidation of other aspects of the mechanism described below (8).

FIG 1.

CG methylation status of CCND1-MTC in multiple human cell lines. (A) Illustration of cyclin gene (CCND1) and translocation breakpoints in CCND1-MTC. Translocation breakpoints near CCND1 (the black block designates the CCND1 gene itself) are distributed over a 340-kb region, as marked by the line with double arrowheads, upstream of the transcriptional start site (marked by the block arrow pointing to the direction of transcription). About 30% of patient breakpoints occur in the 150-bp CCND1-MTC (major breakpoint cluster; depicted as a starburst symbol), and the remaining 70% of events are distributed widely over the surrounding 340-kb region. The MTC fragile zone is ∼110 kb closer to the chromosome 11 (chr. 11) centromere than the CCND1 transcription start site. A detailed view of breakpoint distribution within CCND1-MTC is presented with a triangle positioned at each known patient breakpoint site. Breakpoints identified from der(14) of the t(11;14) have the inverted triangle positioned above the sequence, and those identified from der(11) are positioned below the sequence. The breakpoints within the fragile zone are clustered around the CG DNA sequence motif (highlighted by red). Numbers above the red blocked dinucleotides correspond to the CG sites (left to right) in the red rectangles in panel B. The first CG in the MTC is not included in the numbering because it is a common SNP site as AG instead of CG. (B) CG methylation status of CCND1-MTC determined by sodium bisulfite sequencing in multiple human cell lines. The name of the human cell line is given above each data set. The red rectangles delimit the boundary of the MTC fragile zone. Each row of circles represents the CGs from a single bisulfite-sequenced DNA molecule (allele), where a filled circle represents a methylated CG and an empty circle represents an unmethylated CG. The six numbered CGs in panel A correspond to the six circles within the red boxes in this panel. Three independent PCR amplifications of CCND1-MTC were done with genomic DNA from each pre-B cell line (data from different PCRs are separated by lines). At least 16 molecules were sequenced from each PCR. Molecules with the identical methylation pattern are presented only once, with the number of occurrences (counts) listed to the right of each methylation pattern. The fragile zone (portion within the red box) appears to be hypomethylated relative to the nearby DNA. The regional methylation percentage of each PCR replicate, the methylation percentage of all PCR replicates combined, the P values of the statistical analyses comparing the CG methylation level inside fragile zone and those of both flanking regions outside the fragile zone, and the P values of the statistical analyses comparing the CG methylation level inside the fragile zone and the telomeric side of the outside region are listed under each data set. Statistically significant P values are shown in bold.

FIG 2.

CG methylation status of BCL2-MBR in multiple human cell lines. (A) Illustration of BCL2 and translocation breakpoints in BCL2-MBR. Regions of translocation breakpoints within and near exon 3 of the BCL2 gene are distributed over a 29-kb region (marked by the line with double arrowheads), and the transcriptional start site of the gene is marked by the block arrow pointing in the direction of transcription. The 175-bp BCL2-MBR (leftmost starburst symbol) resides within exon 3 of the gene and accounts for about 50% of the translocation breakpoints involving the BCL2 gene. Two other less frequently used fragile zones, the 105-bp BCL2-icr and 561-bp BCL2-mcr (the two starburst symbols at the right), account for 13% and 5% of BCL2 translocation breakpoints, respectively. A detailed view of the breakpoint distribution within the BCL2-MBR is presented with a triangle at each of the patient breakpoint sites from der(14) of the t(14;18) plotted above the sequence, and those identified from der(18) are plotted below the sequence. The breakpoints within the fragile zone are clustered around the CG DNA sequence motif (highlighted by red). Numbers above the red blocked dinucleotides correspond to the CG sites (left to right) in the red rectangles in panel B. (B) CG methylation status of BCL2-MBR determined by sodium bisulfite sequencing in multiple human cell lines. The name of the human cell line is given above each data set. Each row of circles represents the CGs from a single DNA molecule sequenced, where a filled circle represents a methylated CG and an empty circle represents an unmethylated CG. The BCL2-MBR fragile zone harboring six CGs (one immediately upstream of and five within the MBR) is demarcated by the red rectangle. Three independent PCR amplifications of BCL2-MBR are done with genomic DNA from each pre-B cell line (data from different PCRs are separated by lines). At least 16 molecules are sequenced from each PCR. Molecules with the identical methylation pattern are presented only once, with the number of occurrences listed to the right of each methylation pattern. The fragile zone (portion within the red box) appears to be hypomethylated relative to the nearby DNA. The regional methylation percentage of each PCR replicate, methylation percentage of all PCR replicates combined, and P values of the statistical analyses comparing the CG methylation levels inside the fragile zone and outside the fragile zone are listed under each data set. Statistically significant P values are shown in bold.

We have previously reported that breakpoints of human lymphoid translocation fragile zones are not randomly distributed (Fig. 1A and 2A). Pro-B/pre-B cell-stage translocations show a breakage propensity for the CG (commonly called CpG but designated CG in this paper) DNA sequence motif within the 20 to 600-bp fragile zones (3). We have proposed a model in which activation-induced deaminase (AID) initiates double-strand breaks (DSBs) by acting at the methylated form of these CG sites in single-stranded DNA (ssDNA) to deaminate the 5-methylcytosine to yield a T-G mismatch (3). The repair at T-G mismatches is slower than at the more typical U-G mismatch that would arise at an unmethylated cytosine. Once the relatively long-lived T-G mismatch is generated, these may be converted to DSBs during the normal repair initiated by TDG/MBD4 glycosylases or, even prior to repair, by the RAG complex or Artemis, both of which can nick at mismatch sites or DNA bubble structures (8). Consistent with what we inferred from patient translocations, we found that methylated CG sites within the fragile zones are essential for the enzymatic activities of AID, RAG, and Artemis to create DNA lesions, thus solving the puzzle of the CG breakage propensity within the fragile zone in a human minichromosome system (8). However, these observations do not address the issue of why these fragile zones, ranging from 20 to 600 bp in size, are more sensitive to DSBs than the other DNA regions surrounding the oncogene in the genome. Namely, why are the CG sites in these fragile zones more sensitive to breakage than the CG sites in DNA only 20 to 100 bp away? The presence of altered DNA structure within the zones represents one possibility, though the in vivo longevity of such altered DNA structures is unclear (4, 9).

Like CG, the WGCW DNA sequence motif (where W = A or T) also appears to be fragile in the germinal-center B cell chromosomal translocations but in larger zones which are ∼2 kb in size rather than in the smaller 20- to 600-bp fragile zones (5). The mechanism of breakage at these larger zones also relies on AID (10). The explanation for why some fragile zones break at the CG motif and others at the WGCW motif is not yet clear, nor are the factors that determine the size or boundaries of the fragile zones (5). Solving these knowledge gaps is our long-term goal.

In an effort to further characterize the properties of these fragile zones, here we determined the pattern of CG methylation at several human lymphoid translocation fragile zones in primary human pre-B cells and in human lymphoma cell lines. We found distinctive hypomethylation zones that match well with the boundaries of the translocation fragile zones. In addition, these human lymphoid translocation fragile zones are accessible to enzymatic modifications by exogenous methylases in vitro (isolated nuclei) and in vivo (in living human pre-B cells), and the fragile zones themselves are more accessible to these enzymes than adjacent DNA regions. Our study results indicate that human lymphoid translocation fragile zones reside within accessible chromatin as duplex DNA and raise the possibility that a greater extent of enzymatic exposure than that represented by the adjacent DNA determines the locations and boundaries of these fragile zones. We discuss possible models for how such distinctive regions could arise.

MATERIALS AND METHODS

Oligonucleotides.

The oligonucleotide primer pairs used in PCR after sodium bisulfite treatment of DNA to avoid CG sites are as follows: for CCND1, AT297 (5′-GGGTTGTTTTTAAGTTTTGGTTATT-3′) and AT298 (5′-CCAATACCCCAAATTCCCTTA-3′); for BCL2, AT301 (5′-GGATGTTATTGGTTATTGAGGAGT-3′) and AT302 (5′-TCAAAAATCTAATCATTCTATTCCCTA-3′); for CRLF2, ZL44 (5′-AAAAGTGTGGAGTTATTTTAAAGAAAAC-3′) and ZL45 (5′-ACCTAACCAACATAATAAAACCCCAT-3′); for MALT1, ZL32 (5′-AAAGGAGTTGATGTATGTGAAGTA-3′) and ZL34 (5′-AAAACAATCACTTAAAAAACCAAAAT-3′); and for MLH1 (11), CPO1842 (5′-TAAATATAAACAAATAATTTCTAAAATAAATA-3′) and CPO1843 (5′-GGAGGGAYGAAGAGATTT-3′). For MAPit (methyltransferase accessibility protocol for individual templates) analysis, CCND1 and BCL2 primers were redesigned to avoid GC and CG sites as follows: for CCND1 MAPit, ZL75 (5′-AAAGGAAAGATTTTATTTAGTGGGAG-3′) and ZL76 (5′-ACCCTTCACCTACTAAAAAACTTATAA-3′); and for BCL2 MAPit, ZL77 (5′-TATTTAAGAAAAATTTGGATGTTATTG-3′) and ZL78 (5′-TCAAAAATCTAATCATTCTATTCCCTA-3′).

Plasmid construction.

The AluI DNA methylase (M.AluI)-expressing plasmid pAlu-IMG was generated by inserting a DNA fragment containing M.AluI optimized for human expression with a c-myc tag on the C terminus upstream of an internal ribosomal entry site (IRES) followed by maxGFP on a pcDNA3 backbone. M.AluI expression is driven by the cytomegalovirus (CMV) promoter on pcDNA3, and the construct also carries a neomycin resistance gene for selection in eukaryotic cells.

Cell culture, transfection, and sorting.

Three cell lines with a pre-B cell phenotype, Nalm-6, Reh, and Bay91, were derived from acute lymphoblastic leukemia (ALL) patients (12–14). BCL2 rearrangement was identified in the Bay91 cell line (14). All three cell lines were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum. Only Reh and Bay91 cells were used in the transfection experiments. For each transfection, 3 × 10⁶ cells in the log growth phase were washed, resuspended in 400 μl of RPMI 1640 medium with 10 μg/ml DEAE-dextran, and electroporated (960 μF, 250 V, in a 4-mm-gap-width cuvette at room temperature) in a Bio-Rad Gene Pulser after adding an equal volume of DNA mix (15). The DNA mix for each transfection contained 10 μg of pAlu-IMG and 3 μg of pMAX-GFP (a green fluorescent protein expression plasmid for transfection efficiency evaluation and fluorescence-activated cell sorter [FACS] analysis) in 400 μl RPMI 1640 medium. The transfected cells were sorted based on the GFP signal 3 days after transfection for further analysis.

Isolation of nuclei and MAPit analysis.

For each MAPit probing assay, nuclei from Reh and Bay91 cells were harvested from 1 × 10⁶ cells under conditions that preserve the integrity of nuclei and chromatin structure as previously described (16). After harvesting, cells were washed twice with phosphate-buffered saline (PBS) and resuspended in cell resuspension buffer (CRB; 20 mM HEPES [pH 7.5], 70 mM NaCl, 0.25 mM EDTA, 0.5 mM EGTA, 0.5% glycerol [vol/vol], 10 mM dithiothreitol [DTT], 0.25 mM phenylmethylsulfonyl fluoride). After pelleting by centrifugation at 1,000 × g, cells were resuspended in cell lysis buffer (CLB; CRB plus 0.08% NP-40), incubated for 10 min on ice, and then washed twice with CRB. One million nuclei were resuspended in 300 μl methylation buffer (MB; CRB plus 160 μM S-adenosylmethionine). A 2-μl aliquot of the nuclei was stained with 0.4% (wt/vol) trypan blue and visualized by light microscopy to confirm their integrity after the nuclei were resuspended in the MB. After prewarming the nuclei to 37°C for 5 min, M.CviPI (New England BioLabs) was added to each reaction at various specified concentrations and the mixture was further incubated for 15 min at 37°C. The methylation reaction was terminated by adding an equal volume of stop buffer (100 mM NaCl, 10 mM EDTA [pH 8.0], 1% SDS), and genomic DNA was extracted. An initial experiment was conducted with 0, 40, 80, and 200 U of M.CviPI, and 200 U of M.CviPI was chosen for further studies. For fragile-zone MAPit assays, three independent experiments were performed and results were pooled for display.

Bisulfite genomic sequencing.

Genomic DNA from cells treated with methylase or left untreated was extracted by using proteinase K digestion, phenol-chloroform extraction, and isopropanol precipitation. Sodium bisulfite treatment was carried out according to the manufacturer's instructions (EZ DNA Gold methylation kit; Zymo Research). Each PCR was carried out using 1 μl of the total 20 μl DNA eluted from a Zymo kit column and the appropriate primers. The PCR products were ligated into TOPO TA cloning vector (Invitrogen), and multiple clones from each ligation were sequenced.

Display of CG and GC methylation status with MethylMapper.

A Web-based, unsupervised hierarchical clustering program named MethylMapper (http://genome.ufl.edu/methyl/) was used to construct CG and GC methylation status plots (17). FASTA format text files of the aligned sodium bisulfite sequencing reads are submitted to MethylMapper for generating CG and GC plots using the default settings. The CG and GC plots are displayed in multiple colors: a red horizontal bar is drawn between any two consecutively methylated CG sites, a yellow horizontal bar is drawn between two consecutively methylated GC sites, a black horizontal bar is drawn between two consecutive sites with no CG or CG methylation, and a gray horizontal bar is drawn between two adjacent sites with different methylation status. The very thin vertical lines represent the CG or GC site. GCG sites are excluded, because both endogenous mammalian DNA methyltransferases (DNMTs) and M.CviPI can methylate the cytosine flanked by guanines and make the classification of CG versus GC methylation of the site ambiguous.

Accessible chromatin fraction calculation.

The fraction of accessible GC motifs in chromatin inside and outside translocation fragile zones was calculated by dividing the total length of DNA with consecutively methylated GC sites (length of the yellow bars) on all molecules sequenced by the combined length of the defined regions (inside or outside fragile zone) from all molecules sequenced. For the CCND1 major translocation cluster (CCND1-MTC), 150 bp was used as the fragile-zone length, and the total length, 267 bp, of the regions outside the CCND1-MTC fragile zone was derived by subtracting 150 bp from the length of the assayed region (from the first GC site to the last GC site on the amplicon). Based on the same principle, 175 bp was used as the BCL2 major breakpoint region (BCL2-MBR) fragile-zone length, and the total length of the regions outside BCL2-MTC fragile zone was 250 bp.

Daudi cell methylome.

The Daudi cell methylome was obtained from Kreck et al. (18) and processed by the use of a custom Perl script, resulting in two WIG files containing CG methylation percentage and read depth, respectively. The WIG files were then uploaded to the Integrative Genomics Viewer (IGV) (19) and displayed along with MYC and BCL6 translocation breakpoints, which were collected from human patients (5).

Statistical methods.

Paired two-tailed Student's t tests were performed to compare CG methylation levels inside and outside fragile zones in cell lines and in primary human pre-B cells to test the null hypothesis that CG methylation levels are the same inside and outside the fragile zones. CG methylation levels in different regions of the each molecule were also compared. When the CG methylation level was higher inside the fragile zone than in other regions (either total outside or telomeric side only), it was defined as “inside high.” When the CG methylation level was lower inside the fragile zone than in other regions, it was defined as “inside low.” Binomial probability tests were performed to test the hypothesis that the probability of “inside high” (and of “inside low”) was equal to 0.5. The Welch two-sample t test, which assumes unequal variances between samples, was used to test the null hypothesis that the same fraction of accessible chromatin exists inside and outside fragile zones on all molecules examined in the in vitro MAPit experiments. Accessible chromatin fractions inside and outside fragile zones within each molecule that contained two or more consecutively methylated GC sites were also compared by a binomial probability test in a way similar to that used for the CG methylation level comparison stated above.

RESULTS

Presence of CG methylation in human lymphoid translocation fragile zones.

In previous studies using a human minichromosome system, we demonstrated that DNA methylation at CG sites within fragile zones can affect the location of DNA breakage (8). To confirm the existence of methylated CG sites in human lymphoid translocation fragile zones, we used bisulfite sequencing to profile the CG methylation status of major fragile zones in multiple human pre-B cell lines. The Nalm-6, Reh, and Bay91 human pre-B cell lines were chosen because all of the CG-type translocations within the 20 to 600-bp fragile zones appear to occur during the pro-B/pre-B cell stage of B cell development (3). For comparison, we also examined the CG methylation profile from a human nonlymphoid cell line, 293/EBNA1 (20).

The t(11;14)(q13;q32)/CCND1-IGH translocations are found in almost all MCL, with about 30% of the breakpoints (of a total of 104 breakpoints) within a 150-bp major translocation cluster (MTC) located approximately 110 kb closer to the centromere than the CCND1 oncogene (Fig. 1A). Motif analysis indicates that CG sequences in the CCND1-MTC are hot spots for translocation breakage (3, 6). The CCND1-MTC region contains seven CG sites (including a single nucleotide polymorphism [SNP]), which colocalize with the majority of breakpoints. In order to evaluate the methylation status both inside and outside the CCND1-MTC, a PCR amplicon was designed to include seven additional CG sites surrounding the CCND1-MTC region. While many molecules of the CCND1-MTC region sequenced contained methylated CG sites (black circles) across all cell lines evaluated, the CCND1-MTC fragile zone in Reh and Bay91 appeared to be considerably less methylated than the same region in nonlymphoid 293/EBNA1 cells (Fig. 1B). The fragile zone was much less methylated than the outside region on the telomeric side at a statistically significant level (all P < 0.05, Student's t test and binomial test) in all three pre-B cell lines, even though the results do not reach statistical significance with the two flanking regions outside the fragile zone considered together (Fig. 1B).

Similarly to the CCND1-IGH translocation, the t(14;18)(q32;q21)/BCL2-IGH translocations are known to occur in about half of the non-Hodgkin's lymphoma patients, including 80% to 90% of FL and 20% to 30% of DLBCL patients (1). Intriguingly, about 50% of BCL2 breakpoints occur in the 175-bp major breakpoint region (MBR) in the 3′ untranslated region (UTR). Among the 487 sequenced MBR breakpoints, 208 (43%) occurred precisely at CG sites (Fig. 2A), and statistical analysis indicates the proximity of these breakpoints to CG to be very highly significant (3). We also examined the DNA methylation pattern of the BCL2-MBR. The BCL2-MBR contains six CG sites, including five CG sites within the breakpoint boundaries and one CG site one nucleotide away from the most telomeric break in patients. Three additional CG sites immediately adjacent to the BCL2-MBR on the telomeric side were included in the PCR amplicon for assessing DNA methylation (CG sites on the opposite side of the MBR were not included due to limitations in the PCR amplicon size after sodium bisulfite treatment). Hypomethylation in the BCL2-MBR fragile zone was visually apparent only in Reh cells. However, statistical analysis indicates that the fragile zone was hypomethylated compared with the region outside the fragile zone not only in Reh but also in Bay91 pre-B cells (all P < 0.03, Student's t test and binomial test) (Fig. 2B).

CG sites in the CCND1-MTC region and the BCL2-MBR in 293/EBNA1 cells are nearly all methylated, and yet CG methylation analyses of these regions in human pre-B cell lines show hypomethylation at CG sites in the fragile zones to various degrees. Given that cell lines often have increased CG methylation and given that 293/EBNA1 cells have no hypomethylation at all, these findings suggest a hypomethylation of human lymphoid translocation fragile zones in primary cells relative to the adjacent DNA.

Hypomethylation of human lymphoid translocation fragile zones in primary human pre-B cells.

In light of the hypomethylation observed within pre-B cell fragile zones, we decided to evaluate the fragile-zone methylation pattern in primary human pre-B cells, in which CG-type translocations arise. In previous work (3), pre-B cells from the bone marrow of five healthy individuals of similar ages were subjected to FACS analysis, and their genomic DNA was extracted and subjected to sodium bisulfite sequencing. Here we expanded the zone of analysis around the fragile zones on the DNA from those samples.

Similarly to the human pre-B cell line results, many molecules sequenced from human primary pre-B cells contained methylated CG sites (Fig. 3A and B). Interestingly, despite considerable intra- and intersample heterogeneity, these human primary pre-B cells exhibited much clearer hypomethylation within the CCND1-MTC region and BCL2-MBR than what was observed in immortal cell lines. The overall CG methylation level within the fragile zone was much lower than outside the fragile zone. The average methylation level of the six CG sites within the CCND1-MTC region in molecules sequenced from five patients was 37.1%, in contrast to the 74.8% seen outside the fragile zone (P ≤ 0.01) (Fig. 3C). The average methylation frequency in the BCL2-MBR in molecules sequenced from five individuals was 62.5% compared to 92.5% outside the MBR (P ≤ 0.02) (Fig. 3D). In addition to the average methylation frequency, the fragile-zone methylation level was significantly lower than outside the fragile zone in all five individuals (P ≤ 0.02, binomial test) (Fig. 3C and D).

FIG 3.

CG methylation status of CCND1-MTC and BCL2-MBR in human primary tissue. Human pre-B cells were collected from bone marrow of five healthy individuals, and methylation profiles of their fragile zones were determined by sodium bisulfite sequencing. (A and B) Methylation patterns of CCND1-MTC (A) and BCL2-MBR (B) are analyzed and illustrated as described for Fig. 1B and 2B. (C) CG methylation levels outside and inside CCND1-MTC in five individuals. (D) CG methylation levels outside and inside BCL2-MBR in five individuals. The percentages of methylation within and outside the fragile zone marked by the red rectangle were calculated by dividing the total number of methylated sites by the total number of CG sites examined in the region. Counting all molecules, the average methylation percentage of the CCND1-MTC from all five individuals is 37.1%, and it is significantly different from the 74.8% determined for those outside the MTC (P < 0.05). The BCL2-MBR average methylation level from all five individuals is 62.5%, which is significantly lower than the 92.5% determined for those outside the MBR (P < 0.05), when all molecules are considered.

It is also apparent that boundaries of the hypomethylation zones match the patient translocation boundaries of these fragile zones. For CCND1-MTC, the hypomethylation boundary matched the proximal border of the MTC even though the hypomethylation boundary on the distal side was not defined (Fig. 3A). In all five healthy-donor bone marrow samples, the boundaries for the hypomethylation zone corresponded to the fragile zone for BCL2-MBR translocation in lymphoma patients (Fig. 3B, boxed area).

In addition to the BCL2-MBR and CCND1-MTC region, we have recently identified fragile zones in two other recurrent translocations (21, 22). The MALT1-IGH translocations are found in approximately 10% of MALT lymphomas, and the CRLF2-IGH translocations occur in 5% to 7% of all adult and pediatric B cell precursor acute lymphoblastic leukemias (B-ALL) and in 60% of B-ALL in children with Down's syndrome. Fragile zones for both MALT1 and CRLF2 were located 1.4 to 25 kb upstream of their respective promoters, and both showed marked enrichment of CG-type breaks. The CG methylation patterns in the pre-B cells of the two human individuals that we sequenced also showed hypomethylation in the CRLF2 and MALT1 fragile zones, similar to what we observed in the BCL2-MBR and CCND1-MTC region (see Fig. S1 and S2 in the supplemental material). Thus, methylation profiling of primary human pre-B cell fragile zones confirms our observation in immortalized pre-B cells that human lymphoid translocation fragile zones are hypomethylated whereas adjacent DNA is almost completely methylated. These findings suggest the possibility of an accessible chromatin structure in these fragile zones.

Other mature B cell-stage translocation fragile zones also associate with low-methylation regions.

In addition to the pre-B cell-stage translocation fragile zones characterized above, translocations that occur during the germinal center-B cell stage show highly focused clustering of translocation breakpoints as well (1). MYC-IGH and BCL6-IGH translocations are found in a variety of human mature B cell lymphomas, such as Burkitt's lymphomas and DLBCL. AID initiates DSBs at the mouse MYC locus for the MYC-IGH translocation during the murine germinal-center B cell stage (23–25). It has recently become clear that AID targets DNA motifs in the human MYC-IGH and BCL6-IGH translocations as well (5). Chromatin accessibility marked by histone H3K4me3 enrichment has been proposed to explain the mechanism of breakage in these focused and well-defined fragile zones (1). Recently, a base-pair resolution of DNA methylome of one Burkitt's lymphoma cell line, Daudi, has been determined by SOLiD bisulfite sequencing. To determine if these mature B cell fragile zones also show hypomethylation similar to that seen in human pre-B cells, we aligned MYC and BCL6 breakpoints with the Daudi methylome (Fig. 4). It is striking that the boundaries of the hypomethylation zones coincide with the human MYC and BCL6 fragile zones precisely, and these hypomethylation zones define MYC and BCL6 fragile zones better than the presence of H3K4me3 from a human lymphoblastoid cell line, GM12878 (http://hgdownload.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeBroadHistone/). In summary, hypomethylation is not a feature restricted to the pro-B/pre-B cell chromosomal translocation fragile zones but also applies to the two most common translocation fragile zones in human mature B cell malignancies.

FIG 4.

Human MYC and BCL6 breakpoints of translocations coincide with regions of hypomethylation in a human Burkitt's lymphoma Daudi cell line. Known human MYC breakpoints and methylation analysis of Daudi cells (A) and human BCL6 breakpoints and Daudi methylation (B) illustrated using the Integrative Genomics Viewer (IGV). The top section for each panel is an ideogram of the chromosome, with the locations of MYC and BCL6 on chromosomes 8 and 3, respectively, marked by a red vertical bar. Below the ideogram, the specific locations of the region included for the sequencing coverage (blue vertical bars below) and methylation level (red vertical bars below) in the Daudi cell line are indicated by the base numbering of hg19 reference genome. The section below the Daudi methylome tracks shows the gene (black) involved in frequent translocation, range of fragile zones (orange), known actual breakpoints (green vertical bar), and H3K4me3 enrichment (chromatin immunoprecipitation sequencing [ChIP]-seq data) track of GM12878 (a lymphoblastoid cell line derived from an European Caucasian female) from the ENCODE project. Note the matching boundaries of the low-methylation region and the MYC and BCL6 fragile zones.

Human lymphoid translocation fragile zones are accessible to M.CviPI methyltransferase in vitro.

We have established that AID is responsible for initiating most DNA lesions that lead to human B cell translocations (3, 5, 6, 8, 21, 22). In order to deaminate C to U or methyl-C to T in these translocation fragile zones, these zones must be accessible to AID (26, 27). The boundary match between fragile zones and hypomethylation zones provides possible support for such a hypothesis. Decreased CG methylation is often associated with chromatin accessibility, which may be a prerequisite for access of AID.

To directly assess the accessibility of these fragile zones, we probed the fragile zone chromatin structure at the level of single molecules by performing MAPit (methyltransferase accessibility protocol for individual templates) methylation footprinting on human pre-B cell lines. Reh and Bay91 cell lines were used because CG methylation patterns in the fragile zones of those cell lines resemble what we observed in human primary tissues. Nuclei from Reh and Bay91 cells were isolated and probed with M.CviPI, which methylates cytosine at accessible GC sites (Fig. 5A) (28). GC sites within nucleosomes or occupied by nonhistone proteins would be less accessible to M.CviPI and would remain largely unmethylated. After incubation of nuclei with M.CviPI, genomic DNA was extracted and subjected to sodium bisulfite sequencing to obtain the methylation status of CG and GC sites within regions of interest.

FIG 5.

MAPit methylation footprinting detects accessibility of the CCND1-MTC and BCL2-MBR in pre-B cell lines. (A) MAPit workflow. MAPit begins with preparation of nuclei and incubation with the GC methyltransferase M.CviPI. Upon termination of the chromatin probing reaction, genomic DNA is extracted and processed using the bisulfite reagent. The bisulfite-treated DNA is PCR amplified using locus-specific primers, and the PCR products are cloned and sequenced. The endogenous CG methylation and M.CviPI footprints (GC methylation) are analyzed and displayed by MethylMapper. (B) CG methylation and M.CviPI footprints of CCND1-MTC in pre-B cell lines Reh and Bay91. Each horizontal bar represents one sequenced DNA strand or molecule, with the same top-to-bottom order of molecules in both CG (left) and GC (right) panels. Vertical bars demarcate individual CG or GC sites (GCG excluded). Consecutively methylated CG and GC sites are connected by red and yellow bars, respectively. Consecutively unmethylated CG and GC sites are connected by black bars. Gray bars connect a methylated site and an unmethylated site. Red rectangles define the location of CCND1-MTC. The spacing of all sites is to scale, and the bar at the lower left under each panel is 147 bp, the length of DNA in a nucleosome core particle. The stars above the GC panel mark the three Alu sites examined in the in vivo M.AluI accessibility assay. (C) CG methylation and M.CviPI footprints of BCL2-MBR in pre-B cell lines Reh and Bay91. All markings are the same as in panel B, with the red rectangles defining the locations of BCL2-MBR. (D) Fractions of accessible chromatin inside and outside of the fragile zones from two regions (CCND1-MTC and BCL2-MBR) in the two cell lines (Reh and Bay91) assessed. (E) Fractions of accessible chromatin fraction inside and outside fragile zones in each molecule from two regions (CCND1-MTC and BCL2-MBR) in the two cell lines (Reh and Bay91) assessed. Values for fractions inside and outside accessible chromatin from the same molecule are connected by a line. Inside high, molecules with higher inside accessible chromatin fraction than outside; inside low, molecules with higher fractions outside accessible chromatin than inside. The numbers of molecules assigned “inside high” and “inside low” and their respective binomial probability (P) values are listed under each plot. The total number of molecules in each condition is less than 32 due to exclusion of molecules that contain no accessible chromatin feature both inside and outside fragile zones.

To optimize the concentration of M.CviPI for chromatin probing, we treated nuclei from Reh cells with 0, 40, 80, and 200 U of M.CviPI and analyzed accessibility to the enzyme in the proximal promoter of MLH1 on individual molecules (see Fig. S3 in the supplemental material). To facilitate pattern recognition, aligned sequences were uploaded into a Web-based hierarchical clustering program, MethylMapper (http://genome.ufl.edu:8080/methyl) (17). MethylMapper generates 3-color images of clustered CG methylation (Fig. 5B and C, left panels) or GC accessibility (Fig. 5B and C, right panels).

The MLH1 promoter is active and is within a hypomethylated CG island (often called a CpG island) in Reh cells. Thus, this region serves as a positive control for this method. M.CviPI footprinting results reveal that this region is highly accessible around its two transcription start sites (TSSs) and partially protected within the second TSS, defining two nucleosome-depleted regions (NDRs) and a possible DNA-binding element, respectively (see Fig. S3 in the supplemental material). Based on this positive control, where 200 U of M.CviPI gives the clearest MLH1 promoter footprinting pattern in Reh cells, we chose to use 200 U in the following experiments performed on the fragile regions in this study.

GCG sites, which can be methylated both by endogenous mammalian DNMTs and by M.CviPI, are omitted in the analysis to avoid ambiguity. As a result, the CG methylation pattern in these figures is slightly different from what we determined in primary cells and cell lines, because omitting GCG sites influences how the endogenous CG methylation pattern is presented. Despite this difference in the presentation results, the natural CG methylation patterns of both the CCND1-MTC and BCL2-MBR in Reh and Bay91 cells are similar to what we found in the human primary cells and immortalized cell lines described above (left panels in Fig. 5B and C).

A high to moderate level of GC methylation in both the CCND1-MTC region and the BCL2-MBR was observed in Reh cells, indicating chromatin accessibility to M.CviPI (Fig. 5B and C, right upper panel). It also appears that there were more horizontal yellow patches inside the fragile zone than outside (Fig. 5B and C; compare regions in the red box and outside the red box), indicating that chromatin inside the fragile zone is more accessible to M.CviPI than the chromatin outside the fragile zone. Similar chromatin accessibility in the human pre-B Bay91 cell line is also apparent (Fig. 5B and C, right lower panel). These higher-accessibility zones colocalized well with the fragile zones (boxed in red) and with the zones of CG hypomethylation observed in Fig. 1B, 2B, and 3. Reh cells appear to have higher accessibility (more GC methylation) and less CG methylation in both the CCND1-MTC region and the BCL2-MBR than Bay91 cells, suggesting that accessibility may correlate with the degree of hypomethylation at CG sites within a fragile zone. Interestingly, a nearly complete lack of methylation at the three CG sites immediately outside the CCND1-MTC fragile zone and a lack of M.CviPI activity in the vicinity were observed in both cell lines (Fig. 5B, right side adjacent to the red box), indicating the potential lack of accessibility to M.CviPI and, presumably, endogenous enzymes immediately outside the fragile zone.

Beyond a visual impression, quantitative analysis of the MAPit results both inside and outside fragile zones provides a more objective evaluation of the chromatin status. The analysis strategy was to consider the DNA with two or more consecutively methylated GC sites on the same molecule sequenced to represent accessible chromatin because single-GC-site methylation may be a random event. The fraction of the DNA with two or more consecutively methylated GC sites in the total region of available GC sites harboring DNA would be a fair reflection of the fraction of accessible chromatin. Therefore, the fraction of accessible chromatin was derived by dividing the sum of the lengths of DNA with two or more consecutively methylated GC sites (defined by the first and the last methylated GC sites) by the total length of DNA in the region analyzed. In the four analyses (two regions in two cell lines), three showed a significantly higher level of open chromatin fraction inside the fragile zone than outside (Fig. 5D) (P < 0.05). The one exception was the CCND1-MTC region in Bay91 cells, which showed a similar trend (inside higher than outside) without reaching statistical significance. Since this method of analysis does not preserve the relationship of the GC methylation status inside the fragile zone to that outside the fragile zone within each molecule, an additional analysis to compare the GC methylation levels inside and outside the fragile zone for each molecule was carried out. If there were no difference between the DNA within and the DNA outside the fragile zone in the levels of chromatin accessibility, the chances of “higher access” to the DNA inside and outside the fragile zones should be equal (0.5). Binomial probability tests were performed to determine whether the probabilities of “inside high” methylation and “inside low” methylation are equal among the molecules containing accessible chromatin. Consistent with the analyses of the fractions of accessible chromatin described above, three of four analyses reached a very significant statistical threshold (Fig. 5E) (P < 0.01), with the majority of the lines being red and with a negative slope (indicating inside high GC accessibility) in all four.

These analyses strongly support the conclusion that human lymphoid translocation fragile zones, such as the CCND1-MTC region and BCL2-MBR, are more accessible than the surrounding DNA to exogenous methyltransferase in vitro and likely more accessible to other soluble nuclear enzymes (such as AID) as well.

Human lymphoid translocation fragile zones are accessible to M.AluI in vivo.

By diffusing the enzyme into harvested nuclei that have a relatively static chromatin structure in the absence of ATP, MAPit measures the steady state of accessibility of these fragile zones. To further measure the chromatin accessibility in the natural physiological environment, we transfected both Reh and Bay91 cells with a plasmid, pAlu-IMG, that expresses M.AluI. Expressed in human cells, M.AluI can methylate C at accessible AGCT sites (C.-L. Hsieh, unpublished data), while C is not normally methylated at AGCT sites in mammalian cells.

In order to identify cells that express M.AluI, we cotransfected pAlu-IMG with pMAX-GFP into human pre-B cell lines (Fig. 6A). The transfected cell population was subjected to FACS analysis for GFP-positive cells 3 days after transfection. Genomic DNA from GFP-positive Reh and Bay91 cells was sodium bisulfite treated and sequenced.

FIG 6.

In vivo expression of M.AluI shows that accessibility of the CCND1-MTC fragile zone is higher than that measured for the surrounding sequences in human pre-B cells. (A) In vivo M.AluI accessibility assay workflow. M.AluI-expressing plasmid pAlu-IMG and pMAX-GFP were cotransfected into Reh and Bay91 cells. M.AluI is able to methylate C in the AGCT context, when the sites are accessible. Transfected cells were sorted based on GFP signal after 3 days of culture to identify M.AluI-expressing cells, before analysis by sodium bisulfite sequencing. (B and C) Methylation status of Reh CCND1-MTC (B) and Bay91 CCND1-MTC (C) after M.AluI treatment. Each row of symbols represents the CG (circles) and AluI (triangles) sites from a single molecule sequenced after sodium bisulfite treatment of DNA. A solid symbol indicates that a site is methylated, while a blank symbol indicates that a site is unmethylated. Red rectangles mark the fragile zone. The percentages at the bottom of the methylation display are the methylation levels for AluI sites analyzed inside and outside the fragile zone. The distances between potential sites of methylation in the CCND1-MTC fragile-zone amplicon are shown at the bottom of the panel, with circles and triangles corresponding to the symbols in panels B and C.

AGCT sequences were less frequent in the genome than two-base sequences such as CG and GC. The BCL2-MBR was not analyzed in this experiment because no AluI sites were present within the MBR. The CCND1-MTC and CRLF2 fragile zones, which contain 3 and 5 AluI sites, respectively, were sequenced to assess M.AluI accessibility. We found that in Reh and Bay91 cells, the combined methylation frequencies of the two AluI sites in the CCND1-MTC fragile zone were 43.8% and 50%, respectively (Fig. 6B and C). The single AluI site outside the fragile zone was methylated less frequently (31.3%) than the two inside sites (43.8%) in Reh cells (Fig. 6B). Also, no methylation was found at the outside AluI site on any of the molecules sequenced from Bay91 cells (Fig. 6C). The CRLF2 fragile zone in Reh cells showed similar results, with 53.8% methylation within the CRLF2 fragile zone versus 30.8% outside (see Fig. S4 in the supplemental material).

The in vivo finding that the average methylation percentage of AluI sites within the fragile zones is higher than outside the fragile zones is consistent with the in vitro findings obtained using the MAPit method described above. This finding clearly indicates that human lymphoid translocation fragile zones are accessible to exogenous methylase in vivo and shows that the fragile zones are more accessible to enzyme modification than their surrounding DNA within the context of the physiologic nuclear environment. The accessibility of the fragile zones may allow nuclear enzymes such as AID to act on these DNA regions, resulting in chromosomal fragility in these zones.

DISCUSSION

Hypomethylation of translocation fragile zones and the mechanism of breakage.

Chromosomal translocations at, for example, the CCND1 locus can occur anywhere over a large region of up to 340 kb, but a disproportionate number (30% of the total number of translocations) occur in a 150-bp zone (0.0004 of the total 340-kb region) (6). Similarly, strikingly narrow fragile zones are present at the BCL2-MBR (175 bp), CRLF2 (311 bp), and MALT1 (86 bp) translocation sites (3, 21, 22). Here, we have shown that the CCND1-MTC, BCL2-MBR, a key region within CRLF2, and the MALT1 fragile zones have some degree of CG methylation but that that amount is notably less than the amount in the DNA immediately adjacent to the fragile zone. In this sense, the fragile zones are hypomethylated relative to the adjacent DNA.

There are three intermediates in the 5-methylcytosine demethylation process, 5-hydroxymethylcytosine (5hmC), which remains a C after sodium bisulfite treatment, 5-formylcytosine (5fC), which becomes a T after sodium bisulfite treatment, and the much less common 5-carboxylcytosine (5caC). We have not done oxidative bisulfite sequencing (to distinguish 5mC from all other forms [5hmC, 5fC, and 5caC]) or reduced bisulfite sequencing (to identify 5fC) (29). It is clear that the hypomethylated fragile zone represents a region that is reduced in an amount corresponding to the sum of 5mC plus 5hmC relative to the neighboring DNA even without these additional tests. We note that deamination of 5mC to T by AID is at least an order of magnitude more efficient than deamination of 5hmC to 5hmU and that AID deamination of the other forms is extremely inefficient or undetectable (30). Biological interpretation of the significance of these less abundant variants of 5mC will have to await further research.

It is important that the small amount of DNA methylation remaining within the fragile zones is sufficient for potential subsequent breakage at the CG sequence motifs that we have documented elsewhere as the mechanism of DNA breakage (8). We have shown that, in a human minichromosome system, the CG sites must be methylated to be prone to a breakage process initiated by AID, but unmethylated CG sites in the same fragile zone are not vulnerable to breakage within the 20 to 600-bp fragile zones at which CG sites are prone to breakage (8). Since AID requires single-stranded DNA substrates (26), the DNA in the fragile zone must at least transiently become single stranded in a manner as yet unknown. (Single-strandedness might conceivably be triggered by the passage of a DNA polymerase through the region, or by some other local cause, and this is a point of future study.) Once AID deaminates the methyl-C to a T, the resulting T-G mismatch is repaired slowly relative to, for example, a U-G mismatch that would arise when AID deaminated an unmethylated C to a U. Because it is longer-lived, the T-G mismatch is more vulnerable to double-strand breakage caused by Artemis or the RAG complex (8).

Most of the hypomethylated fragile regions (CCND1-MTC; BCL2-MBR; BCL2-icr; CRLF2; and the primary fragile zone in MALT1) are less than 200 bp in length. But some, such as the MYC and BCL6 fragile regions, are 2 kb in length. The hypomethylation in the large zones (MYC and BCL6 regions) is in the early portion of their respective coding regions (5). In contrast, the hypomethylated small fragile zones are almost all in nontranscribed regions. We suspect that the breakage sequence motifs in the smaller zones (20 to 600 bp) and the larger zones (∼2 kb) achieve single-strandedness for different reasons, and yet they both manifest hypomethylation. Despite possible differences in the origins of the fragile nature of these zones, it is interesting and likely of fundamental importance that hypomethylation is a common feature among them (see below), in addition to the common requirement for AID to account for their fragile nature.

DNA structure of the fragile zones.

The DNA structure of the fragile zones must be predominantly duplex DNA in order to be a substrate for M.AluI and M.CviPI. DNA methylation enzymes proceed by a mechanism in which they recognize duplex DNA before carrying out a step in which they flip the cytosine out of the duplex in order to methylate it (31, 32). Hence, a nonduplex conformation containing substantial single-strandedness would not be a substrate for methylases. These results do not rule out transient melting of the fragile zones, as we discuss below in relation to AID. But the methylase probing studies indicate that the fragile zones are not single stranded for a large fraction of the time. Transient non-B conformations (called B/A-intermediate DNA) at the CCND1-MTC region and BCL2-MBR have been documented elsewhere, though these are still within the range of duplex conformations (4, 9).

Chromatin accessibility of the fragile zones.

Given the relatively small size of the pre-B lymphoid translocation fragile zones (see Fig. S5 to S8 in the supplemental material), it is interesting that in vivo methylase probing and in vitro methylase probing are able to distinguish the DNA within the fragile zones versus the neighboring DNA. Nucleosome positioning information for human B cells does not suggest any particular positional relationship between the fragile zones and the positioning of nucleosomes in and around the fragile zones that we have studied (nucleosome positioning determined by DNase I hypersensitivity [http://genome.ucsc.edu/ENCODE/]).

Our study results show that the fragile zones are hypomethylated and are not obstructed by inaccessible chromatin or by any stably bound DNA binding proteins, indicating that these zones are accessible. The association of low DNA methylation and accessible chromatin is not surprising. Hypomethylated regions are often associated with accessible chromatin and active histone tail modifications (33–36). But the distinctive feature here is that the fragile zones can be very small, indicating a tight positional association between the hypomethylation and the accessible chromatin.

We have not found RNA polymerase II pausing in the fragile zones (unpublished data). We have also carried out bioinformatics analyses for transcription factor binding sites, noncoding RNA, and histone modifications in the publically available data set from human mature B cells and other human B-linage cells, and known locations for these did not correlate with the fragile zones (see Fig. S9 to S12 in the supplemental material) (37, 38). However, the BCL2-MBR, CCND1-MTC, and MALT1 regions are nearly fully methylated in the H1 human embryonic stem cell line based on the publicly available genome-wide sodium bisulfite sequencing data set (GEO Accession Viewer no. GSM1002649). The findings in H1 cells suggest that these fragile zones likely transition through demethylation between the blastocyst and the pre-B cell stage. It is possible that the factors involved in the hypomethylation boundaries of these fragile zones are present only around the time that the demethylation process takes place during B cell development.

We note that the edges of the predicted locations of nucleosomes or of the internucleosomal regions do not consistently correlate with the edges of the fragile zones, though the GC methylase studies for Reh cells raise the possibility of a preferred location of nucleosomes around CCND1-MTC (see Fig. S13 in the supplemental material).

Proposed mechanism for the role of hypomethylation in translocations.

The close physical proximity of hypomethylation and chromatin accessibility in the fragile zones suggests mechanistic models that have not been previously considered. The model proposed here would require (i) accessible chromatin, (ii) a single-stranded DNA region, (iii) AID presence, and (iv) CG methylation to give rise to DSBs and thus chromosomal translocation. It is possible that prior to pro-B cell development (perhaps in lymphoid-myeloid progenitors), these fragile zones become hypomethylated, perhaps due to obstruction of DNMTs by some DNA binding protein expressed specifically at that stage. Studies of DNA methylation zones have demonstrated corresponding histone modification changes over short distances of 50 or 100 bp (35), explaining why fragile zones could be well demarcated in their accessibility to nuclear enzymes. These zones may remain hypomethylated at the subsequent pro-B stage of development when the DNA binding protein that protects these zones is no longer present, and that is why we detect high chromatin accessibility with no footprint of protein binding in these zones by GC methyltransferase and M.Alu mapping. AID recognizes only ssDNA as a substrate (26, 27); therefore, DNA breathing gives AID the opportunity to deaminate any methyl-C in the fragile zone to a T. When this occurs, a T-G mismatch would arise when the DNA strands reanneal. The T-G mismatch can give rise to a double-strand break if it is cut by the RAG complex or Artemis (2, 8). Hypomethylation of these fragile zones provides accessibility but fewer poorly reparable AID targets. However, as we see from the data, some randomly methylated CG sites remain within the fragile zones. This is because methylation and demethylation processes are dynamic, and the methylation pattern of a DNA region can be heterogeneous due to binding protein occupancy (39–42). These remaining methylated cytosines can be the sites of DSBs (due to AID action) that lead to translocations. Future studies will be needed to dissect the elements of such a speculative model to support it or lead to other models for the interesting observation of hypomethylation within the fragile zones.