Genomic DNA-chip hybridization in t(11;14)-positive mantle cell lymphomas shows a high frequency of aberrations and allows a refined characterization of consensus regions

Mantle cell lymphoma (MCL) represents a distinct clinicopathologic entity that comprises approximately 3% to 10% of non-Hodgkin lymphomas.^1-3  The malignant cells are derived from primary lymphoid follicles and mantle zones of secondary follicles.^4,5  The median survival time is 3 to 5 years and cure is not achieved with current treatment options.^2,6-11  The chromosomal translocation t(11;14)(q13;q32) is characteristic of MCL. This aberration leads to the juxtaposition of the cyclin D1 (CCND1) gene and the immunoglobulin heavy-chain promoter resulting in an overexpression of CCND1.^12,13  CCND1 plays an important role in cell cycle regulation, inducing G₁-S phase transition by binding to cyclin-dependent kinases. Therefore, increased CCND1 expression is likely to contribute to lymphomagenesis.¹⁴  However, in transgenic mice, overexpression of CCND1 was not sufficient for the induction of B-cell lymphomas. Additional, so-called secondary genetic aberrations, as, for example, Mycn or Mycl activation, resulted in the rapid development of lymphomas in these mice.^15,16  Apart from CCND1 overexpression, a number of other cell cycle regulators were demonstrated to have an altered expression pattern in MCL. Examples are decreased expression of the cyclin-dependent kinase inhibitors (CKIs) p16^INK4a and p27, or increased expression of the cyclin-dependent kinases CDK2 and CDK4.^17-19  Similarly, genes involved in the regulation of apoptosis and in DNA repair were found to be aberrantly expressed.

For the comprehensive analysis of secondary genomic aberrations in MCL, several studies using chromosomal banding analysis^20-22  and comparative genomic hybridization (CGH) were performed.^23-27  In contrast to other lymphomas, a high complexity of the karyotype was found. In addition, a characteristic pattern of secondary aberrations (eg, gains on 3q and deletions on 1p, 8p, and 11q) was described. Due to the limited spatial resolution of CGH (approximately 10 megabase pairs [Mbp] for low copy number gains and losses),²⁸  smaller aberrations may have been missed. Also, for most genomic aberrations the characterization of critical regions was not sufficiently accurate to allow the identification of candidate genes.

To overcome these limitations we used the novel technique of matrix-CGH.^29,30  For the detection of genomic aberrations in 53 MCLs, an array containing 812 different DNA probes was designed. With this approach, a genome-wide screening with high spatial resolution was possible.

Between 1984 and 2001 peripheral blood samples (n = 22), lymph node specimens (n = 27), lymphomatous tonsil specimens (n = 1), lymphomatous splenic specimens (n = 1), pleural effusions (n = 1), or bone marrow aspirates (n = 1) from 53 patients with MCL were obtained after they gave their informed consent. The diagnosis was confirmed by standardized clinical, morphologic, and immunologic criteria by expert hematopathologists (P.M. and G.O.). All MCLs had the t(11;14)(q13;q32) as shown by fluorescence in situ hybridization (FISH), chromosomal banding, or Southern blot analysis. Mean age at diagnosis was 63 years (range, 44-82 years); 41 patients were men. Clinical data were available for 36 patients. Median follow-up was 36 months from diagnosis (range, 1-173 months). In addition, for most patients the risk factors at diagnosis according to the International Prognostic Index (IPI) were available.³¹  A comprehensive summary of the clinical data is given in Table 1.

The DNA chip used for this study is based on an array used for the analysis of B-cell chronic lymphocytic leukemia (B-CLL).³²  In addition to 645 bacterial artificial chromosome (BAC) or P1 artificial chromosome (PAC) clones of the CLL study, 167 clones were selected for molecular diagnostics in MCL. These additional clones either contained oncogenes or tumor suppressor genes or covered recurrently aberrant regions in MCL (eg, 8p, 9p, 13q32-13q34). For clone selection, the Clone Registry database (http://www.ncbi.nlm.nih.gov/genome/clone) or BLAST-search (http://www.ncbi.nlm.nih.gov/BLAST/) in genome sequence data were used. The localization of the clones is presented according to the National Center for Biotechnology Information (NCBI) clone registry database (November 2003) or according to data (sequence tagged site [STS] markers or genomic sequences) based on the Map Viewer software of Entrez Genomes (www.ncbi.nlm.nih.gov/mapview) using the Component sequence map. The clones were ordered from the libraries RPCIB753 or RPCIP704 of the RZPD German Resource Center for Genome Research (Berlin, Germany). Altogether, the array contained 812 clones. A total of 209 clones were linearly distributed across the genome at a distance of about 15 megabase pairs and 90 clones covered the sex chromosomes serving as an internal control for the hybridizations. A total of 513 clones covered regions that are known to be recurrently affected in MCL or contained oncogenes or tumor suppressor genes (Figure 1). A complete list of all 812 clones is available at: http://www.informatik.uni-lm.de/ni/mitarbeiter/HKestler/blood_kohlhammer/default.html.

Isolation and spotting of DNA probes were performed as described previously with slight modifications.³³  Briefly, DNA was isolated according to standard protocols (Qiagen, Hilden, Germany), sonicated to fragments of 500 to 5000 base pair (bp) in size, and dissolved in 3 × standard sodium citrate (SSC) at a concentration of 0.75 μg/μL in a 384-well plate. Using an Omnigrid microarrayer (Gene Machines, San Carlos, CA), each clone was spotted in 8 replicas onto Corning CMT-Gaps II glass slides (Corning, NY). After printing, the slides were baked at 80°C for 10 minutes, exposed to UV light for cross-linking of the DNA (254 nm/2400 μJ), and stored at room temperature.

Hybridization. Genomic DNA was prepared from the cryopreserved tumor samples using standard protocols including alkaline lysis and affinity chromatography or Trizol reagent. Tumor DNA and reference DNA of a healthy volunteer were differentially labeled with cyanine 3 (Cy3)– and Cy5-conjugated deoxycytidine triphosphate (dCTP) by random primed labeling. Microcon membrane column centrifugation was used to remove unincorporated nucleotides. For hybridization, labeled reference DNA and tumor DNA as well as 70 μg human Cot-1 DNA were precipitated and redissolved in 120 μL Ultrahyb buffer (Ambion, Austin, TX) at 37°C. Denaturation was performed at 75°C for 10 minutes followed by preannealing at 37°C for 30 minutes. After hybridization for 36 hours at 37°C using a GeneTac hybridization chamber (Genomic Solutions, Cambridgeshire, United Kingdom), slides were washed 3 times in 2 × standard saline citrate (SSC), 50% formamide, and 0.1% Tween 20, pH 7.0, at 45°C using a “flow” time of 30 seconds and a “hold” time of 3 minutes. Afterward slides were washed for 2 minutes in 1 × phosphate-buffered saline (PBS), 0.05% Tween 20, pH 7.0, at 25°C. Slides were dried in a centrifuge by spinning for 5 minutes at 1200g.

Image analysis. Image analysis was performed using a dual laser scanner and the GenePix Pro 4.0 imaging software (GenePix 4000 A; Axon Instruments, Union City, CA). The resulting raw data were analyzed using a dedicated software, which was described previously.³³  Fluorescence ratios were normalized by using the median of the fluorescence ratios computed as log₂ values from the 209 DNA control fragments linearly distributed across the genome. Thereafter, the ratios of the 2 color–switch hybridizations were averaged and normalized. Upper and lower thresholds for the identification of genomic imbalances were determined for each individual experiment by calculating the mean ± 3 × the SD of a set of balanced clones. This set of clones consisted of 134 clones, localizing to chromosomes that are not involved in recurrent MCL aberrations (ie, mapping to chromosomes 2, 4, 5, 16, 19, 20, 21, 22). If one of these chromosomes was involved in genomic aberrations, the respective clones were excluded from normalization and the set of “balanced” clones was reduced. An imbalance was scored if at least 2 clones within an aberrant region exhibited ratio values beyond the cutoff levels. The 1p36 region was excluded from the analysis due to a high number of repetitive DNA sequences in this region. The X and Y chromosomes were excluded from the analysis because male tumor DNA was hybridized with control DNA of a female and vice versa to serve as an internal control for the detection of genomic imbalances.

CGH was performed as reported elsewhere.³⁴  Briefly, 1 μg biotin-labeled tumor DNA, 1 μg digoxigenin-labeled control DNA, and 70 μg human Cot-1-DNA (BRL Life Sciences, Gaithersburg, MD) were cohybridized to slides with metaphase cells from blood of a healthy donor. Image analysis was performed using the commercially available image analysis system ISIS (MetaSystems, Altlussheim, Germany). Ratio values of 1.25 and 0.75 were used as upper and lower thresholds for the identification of chromosomal imbalances. Certain chromosomal regions are known to be critical in CGH analysis and, therefore, they were not considered for quantitative image analysis. These regions were the distal part of chromosome arm 1p and the whole chromosome 19 as well as regions with a high content of repetitive sequences (heterochromatin blocks of centromeric regions or the long arm of chromosome Y).³⁵ 

Depending on the availability of sufficient material, additional FISH analyses were performed to further analyze discrepancies between CGH and matrix-CGH results. In 25 MCL cases, selected aberrations (n = 45) were confirmed by FISH. Hybridization was performed as described previously.³⁶  Experiments were evaluated using an epifluorescence microscope (Axioplan; Zeiss). In each case, at least 100 cells were enumerated.

The Kaplan-Meier method was used to estimate the distribution of overall survival.³⁷  Differences between Kaplan-Meier curves were assessed using the log-rank test.³⁸  Odds ratios, together with their 95% confidence limits, were used to compare the risk for genomic aberrations between the V_H subgroups. Results of statistical tests were considered significant if P was less than .05.

In 49 of the 53 cases genomic aberrations were detected by matrix-CGH. The mean number of aberrations per case was 6.7 (range, 0-16). Genomic deletions were more frequent than genomic gains (214 deletions versus 141 gains). The 13q deletion was the most frequent aberration, detected in 33 of 53 cases. Based on cases with small deletions, 2 consensus regions were delineated, one mapping to chromosomal band 13q14 (29 cases) and the other mapping to bands 13q33-13q34 (26 cases). Further frequent deletions affected 11q (23 cases), 1p (20 cases), 9p (19 cases), 8p (18 cases), 6q and 9q (13 cases each) as well as 10p and 17p (11 cases each). The minimal deleted regions localized at chromosomal bands 11q22-11q23, 1p21, 8p21, 9p21-9p22, 6q25-6q26, 9q13-9q22, 9q34, 10p14-10p15, and 17p11-17p13. The most frequent genomic gain mapped to 3q26-3q29 and was present in 29 cases. In addition, overrepresented regions were identified on 8q, 11q (15 cases each), and 12q (9 cases). The consensus regions of genomic gains mapped to 8q24, 11q13 (2 regions), 11q23, and 12q13-12q15. Nine high-level amplifications were seen in 8 of the 53 cases. The amplified regions mapped to chromosomal bands 10p13 including the BMI1 gene (2 cases), 18q21-18q22 including the BCL2 gene (2 cases), 3q27-3q28, 6q14, 8q24, 12p11, and 12q13-12q15. The matrix-CGH data are illustrated in Figure 2.

Forty-five of the 53 cases were also analyzed by chromosomal CGH. By CGH, 203 genomic aberrations were detected in these cases. In contrast, by matrix-CGH a total of 307 gains and losses were identified (genomic gains, CGH: 77, matrix-CGH: 123; genomic losses, CGH: 126, matrix-CGH: 184). Thus, by matrix-CGH, a 50% higher number of aberrations was found. Forty-five of the 104 additional genomic aberrations were analyzed by FISH and in each case the matrix-CGH data were confirmed. Twenty-nine of these 45 aberrations were genomic deletions and 16 were genomic gains. Additional genomic deletions mapping to 1p, 9p, 11q, 13q, and 17p were verified in 3 or more samples each. Additional genomic gains mapping to 3q, 7p, 11q, and 15q were verified in 2 or more samples each. The most prominent difference between matrix-CGH and chromosomal CGH was seen with regard to 11q gains (14 cases vs 1 case), which has not been described as a frequent aberration in MCL. Other remarkable differences were found for 11q deletions (21 vs 12 cases), 17p deletions (10 vs 4 cases), and 12p deletions (8 vs 3 cases). Matrix-CGH and CGH data for the most frequent genomic aberrations are illustrated in Figure 3.

In 46 patients, information on the V_H mutation status was available.³⁹  Genomic aberrations, which were present in more than 5 cases, were evaluated. Using a homology cutoff value of 98% to the nearest germline gene, a division into a V_H-mutated subgroup including 14 patients (30.4%) and a V_H-unmutated subgroup containing 32 patients (69.6%) was possible. The incidence of genomic aberrations in the V_H-mutated and V_H-unmutated subgroups was similar. None of the genomic aberrations was clearly associated with one of the 2 subgroups. The 8q gains were more frequently found in V_H-unmutated patients with an odds ratio V_H-mutated to V_H-unmutated of 0.24 (95% CI, 0.05-1.27).

On chromosomes 8, 9, 10, 11, 12, 13, and 18 consensus regions of less than 15 Mbp were delineated. A list of these regions with the approximate genomic extensions and candidate genes is given in Table 2. In some of these regions, a refined characterization was possible:

-8p21. Eighteen cases of this series had an 8p- deletion. This chromosome arm was covered by 16 different DNA clones. A consensus region of approximately 2.4 Mbp was delineated on band 8p21 (Figure 4A). This region contains several candidate genes including members of the TRAIL receptor family (TNFRSF10B, TNFRSF10C, and TNFRSF10D).

+10p13. Six cases exhibited this aberration; 2 of these were high-level amplifications. In all cases a region of approximately 2.7 Mbp was affected, which contains the BMI1 gene.

+11q13. Genomic gains of this region were identified in 15 cases. Two consensus regions were identified. The proximal region extended over 1.4 Mbp and contained the RELA gene. The distal region included the CCND1 gene and was 5.2 Mbp in size (Figure 4B).

-11q22-11q23. Genomic deletions mapping to 11q22-11q23 were present in 23 of the MCL cases. The minimally deleted region was 1.7 Mbp in size. In all but one of the cases (no. 46), clones RP11-241D13 and RP11-415G10, which contain the ATM gene, as well as 2 additional clones, which were selected for the ATM gene by filter screen, were deleted. In case no. 46, a more distal clone (RP11-347E10) was deleted, whereas the 4 clones containing the ATM gene exhibited a normal copy number. This result was confirmed by FISH analysis with a DNA probe containing the ATM gene (Figure 4C).

-13q14. Deletions in this region were identified in 29 cases. The consensus region was delineated by 2 cases (nos. 14 and 52; Figure 5) and had a size of approximately 400 kilobase pairs. Candidate genes in this region are RFP2 and BCMSUN. This region is similar to the consensus region in B-CLL; however, it has not been well characterized in MCL before.

-13q33-13q34. Twenty-six cases exhibited a deletion containing these chromosomal bands. Four of these cases had 13q33-13q34 deletions without involvement of chromosomal band 13q14. A consensus region of approximately 6.9 Mbp was delineated by one case.

In 36 of the 53 MCL cases clinical data were available. In contrast to previously published studies, in our series -8p was not associated with leukemic MCL; 9 of 24 leukemic patients had an 8p deletion, whereas 5 of 14 patients without a leukemic course exhibited this aberration. Elevated lactic dehydrogenase (LDH) values were significantly associated with an inferior outcome (P < .001). Age older than 60 years and involvement of 2 or more extranodal sites showed a trend toward a shortened survival time (P = .093 and P = .057, respectively). Two genomic aberrations were associated with shortened survival: 13q14- deletions (P = .010) and 8p- deletions (P = .033). The prognostic impact of clinical and genomic parameters is summarized in Table 3.

In this study, we present a dedicated DNA array developed for the molecular diagnostics of genomic aberrations in MCLs. In contrast to other methods for tumor genome analysis, as, for example, FISH or CGH, genomic DNA–chip hybridization provides a comprehensive genome-wide analysis at a high spatial resolution. In comparison to previously published studies in MCL,^23-27  a similar pattern of genomic aberrations was identified by matrix-CGH. However, a higher frequency of genomic gains mapping to chromosome arm 11q and 13q33-13q34 deletions was detected. The 11q gains were present in 15 (28%) of 53 of the cases; in 12 (23%) of 53 cases gains affected chromosomal band 11q13. The 13q deletion has been described as a frequent aberration in MCL in previous studies.^23-27  In this series, 2 consensus regions on 13q14 and 13q33-13q34 were delineated.

In comparison to chromosomal CGH, by matrix-CGH a 50% higher number of genomic aberrations was found. 45 of the 104 additional aberrations were validated using FISH. For each of these 45 aberrations the matrix-CGH findings were confirmed. In some genomic regions, such as 11q, 17p, and 12p, a particularly high frequency of additional aberrations was identified. However, for almost every recurrently affected genomic region, additional gains or losses were found (Figure 3). In most cases, the superior sensitivity of matrix-CGH was explained by the lower spatial resolution of CGH, which is approximately 10 Mbp.²⁸  Therefore, many 11q gains, which had a median size of 6.2 Mbp, were not identified by CGH.

For 46 patients V_H mutation data were available. In contrast to CLL, where almost all patients with 17p- and 11q- deletions have an unmutated genotype,⁴⁰  these aberrations were found commonly in both mutated and unmutated MCL. Although no statistical significance was reached, the highest disparity was found for 8q gains, which were more frequent in the V_H-unmutated group. These data are in line with a previous study, in which FISH was used for genomic analysis.³⁹ 

For several genomic aberrations, consensus regions of 400 kbp to 15 Mbp were delineated. In several of these regions, candidate genes possibly involved in the pathogenesis of MCL are located. The different regions are discussed in the following paragraphs.

The 8p deletions were analyzed in a previous study showing a close association with leukemic MCL.²⁶  In this study, a consensus region extending from chromosomal bands 8p21-8p23 was delineated. In our series, the region of minimal deletion could be narrowed to a size of 2.4 Mbp (Figure 4A). This region contains a number of candidate genes, for example, several members of the TRAIL receptor family (TNFRSF10B, TNFRSF10C, TNFRSF10D). Deletion of a TRAIL receptor may result in impaired induction of apoptosis. Another candidate gene within this region is RHOBTB2, which has recently been described as candidate tumor suppressor gene in breast cancer.⁴¹  In contrast to the data of Martinez-Climent and coworkers,²⁶  there was no association of 8p deletions with leukemic MCL in our series.

On 10p, a region of 2.7 Mbp was identified, which contains the BMI1 gene, and this gene was amplified in 2 cases. BMI1 encodes for a polycomb family protein. Overexpression can result in transcriptional silencing of the INK4a/ARF locus, contributing to an increase in cell proliferation.^42,43 

For 11q13 gains, 2 consensus regions were identified (Figure 4B). First, in 6 of the 12 cases, this region was 5.2 Mbp in size and contained the CCND1 gene. In MCL an amplification of the CCDN1/IGH fusion gene resulting in a high expression of CCDN1 was described recently.⁴⁴  In this case, CCND1 expression occurred throughout the cell cycle, whereas CCND1 expression shows a cell cycle-dependent oscillation reaching maximum levels in the G₁ phase in tumors only exhibiting the t(11;14).^44,45  Thus, a genomic gain of CCDN1 can be an additional mechanism potentiating the overexpression of CCDN1 induced by the t(11;14). Second, in 6 of the 12 cases, a consensus region of 1.4 Mbp was found, which is located approximately 4.2 Mbp centromeric to the CCDN1 gene. This consensus region contains RELA, a member of the nuclear factor-κB (NF-κB) pathway as well as 2 mitogen-activated protein kinases (MAP4K2 and MAP3K11). Four additional cases exhibited gains mapping to the MLL region in chromosomal band 11q23. In another work, a genomic gain of this region was demonstrated in a case of CLL.⁴⁶ 

On the basis of both genomic and mutational studies, a pathogenetic role in MCL has been suggested for the ATM gene, which is located on band 11q22.^47,48  In our series, a consensus region of approximately 1.7 Mbp was identified on 11q22-11q23. In all but one case, the ATM gene was within the deleted region. In this single case, however, a genomic region approximately 1 to 2 Mbp distal to ATM was deleted. Also, by FISH, no deletion of the clone containing the ATM gene was identified. An 11q deletion without involvement of the ATM gene was already described in another study.⁴⁹  This finding can be interpreted in 2 ways: (1) the consensus region on 11q22-11q23 is located distally to ATM, or (2) a second consensus region containing additional candidate genes is present in bands 11q22-11q23. A gene mapping into this novel consensus region is DDX10, for which a rearrangement was demonstrated in myeloid leukemias.⁵⁰ 

The smallest consensus region was found for band 13q14 and extended over 400 kbp. The 13q deletions are frequent in both MCL and B-CLL.^51,52  Although this region has been well characterized in CLL,^52-56  only few data have been available for 13q14 deletions in MCL.⁵²  Based on our data, a detailed mapping of the consensus region was possible. Similar to published data for B-CLL, BCMSUN and RFP2 are part of the consensus region (Figure 5). Other genes discussed as candidates in CLL, as, for example, BCMS, are located distally to the consensus region in MCL.

Univariate analysis revealed a possible prognostic relevance for some genomic aberrations. Although this association to survival is based on a series of only 36 cases, the validity of our cohort is indicated by the prognostic relevance of established clinical risk factors. The 8p21 deletions, as well as 13q14 deletions, were associated with an inferior overall survival. The prognostic relevance of these aberrations has not been analyzed in a larger study before. Similar to previously published studies using different genomic techniques, no prognostic relevance was found for p53 as well as ATM deletions.^24,57  Previously, also INK4A/ARF deletions were associated with an inferior overall survival.^24,57  In our series, aberrations resulting in a transcriptional silencing of the INK4A/ARF locus (INK4A/ARF deletions or BMI1 gains)^42,43  were also associated with a trend toward shorter survival (P = .052; data not shown).

Apart from expression data,⁵⁷  genomic findings may also become important for risk assessment and for the development of tailored treatment strategies. Therefore, future clinical trials in MCLs should incorporate both gene expression studies and genomic analyses. In comparison to expression studies, which are dependent on fresh or cryopreserved tumor samples, genomic analyses can also be performed using DNA derived from paraffin-embedded tumor tissue.

We gratefully acknowledge Martina Enz, Annett Habermann, Sandra Ruf, and Tatjana Salvi for excellent technical assistance.