Copy number gain at 12q12-14 may be important in the transformation from follicular lymphoma to diffuse large B cell lymphoma

The purpose of this study was to identify novel areas of genomic copy number change associated with transformation from follicular lymphoma (FL) to diffuse large B cell lymphoma (DLBL). DNA was extracted from tumour cells micro-dissected from paraffin- embedded tissue sections in 24 patients with FL and subsequent transformation to DLBL and 18 patients with de novo DLBL. Tumour DNA was compared to reference DNA using comparative genomic hybridization. Abnormalities common to all 3 groups were gains on chromosomes 4q, 5q, 7q, 11q and X and losses on 3p, 8p and 10q. Copy number changes seen in both transformed and de novo DLBL and not seen in FL were gains on 2p and losses on 1q, 15q and Xq. Gains on 2q, 6p, 7p and 17q and losses on 5p and 8q were specific to transformed DLBL cases. Gain on 12q12-14 was found in 52% of the transformed DLBL cases and was never seen in its follicular counterpart. Patterns of genomic copy number change associated with specific clinical events in NHL have been demonstrated and suggest that gains on 2q, 6p, 7p, 12q and 17q and losses on 5p and 8q may be important in the transformation from low to high-grade disease. © 2001 Cancer Research Campaign http://www.bjcancer.com


Comparative genomic hybridization
Tumour DNA was compared to reference DNA for copy number change using the technique described by Kallioniemi (Kallioniemi et al, 1992. Equal amounts (1 µg) of tumour DNA and normal placental, same sex DNA were labelled with Spectrum Green (Vysis) and Spectrum Red (Vysis) respectively by a standard nick translation reaction to give probe fragment length of 300-3000 bp. 800 ng of each labelled DNA and 60 µg of human COT-1 DNA (Gibco) were cohybridized to normal human metaphase chromosomes (Vysis) at 37˚C for 4 days. Chromosomes were counterstained with 4,6-diamidino-2-phenylindole (DAPI) following post-hybridization washes.

Digital image analysis
3 colour digital images were acquired using a KAF-1400 cooled CCD camera (Photometrics) attached to an epifluorescence microscope (Zeiss Axioskop). Dedicated software, QUIPS (Vysis), was used to calculate the ratio of green (tumour DNA) to red (normal DNA) fluorescence along the length of each chromosome. Mean ratio profiles were calculated using 5 to 8 target metaphases for each biopsy. Abnormalities at 1p32-ter, 16p, 19, 22 and Y were not included as these areas have been found to be unreliable in CGH analysis (Kallioniemi et al, 1994).

Controls
Ratio values of 1.2 and 0.8 were established as thresholds for gain and loss respectively based on results from 3 controls used with every experiment. These consisted of reference male DNA cohybridized with reference female DNA, normal female DNA micro-dissected and extracted from paraffin-embedded tonsil cohybridized with same sex reference DNA and a cell line (MPE 600, Vysis) with established amplifications and deletions against normal opposite sex reference DNA. Ratio values exceeding 1.4 were defined as amplification events.

Survival
Survival curves were calculated using the method of Kaplan and Meier. The statistical significance of differences observed was determined using the log rank test.

Patients
Paired FL and transformed DLBL biopsies were obtained for 24 (9 female, 15 male) patients. Biopsy material of de novo DLBL was obtained for 18 (9 female, 9 male) patients. No previous chemotherapy or radiotherapy had been administered prior to the initial FL or de novo DLBL biopsies with the exception of two patients; one had received radiotherapy and tamoxifen for carcinoma of the breast and subsequently developed FL of the tonsil, the second had achieved a complete remission with PACEBOM chemotherapy for high-grade lymphoma but presented 8 months later with FL.

Survival
There was a trend towards poorer outcome in FL patients who had gain at 12q12-14 in the transformed DLBL biopsy material; median survival from diagnosis and from transformation were 97 months and 23 months respectively in those who developed    enh(12q12-14) at transformation and 116 months and 35 months respectively in those who did not. This difference was not statistically significant.

DISCUSSION
CGH has clearly demonstrated patterns of genomic gains and losses, some of which are common to all groups, with others being specific to histological subtype. The study of NHL by CGH to date suggests that amplification and deletion of oncogenes and tumour suppressor genes respectively may be more important in the pathogenesis of NHL than previously thought (Werner et al, 1997).
Gains of chromosomal material (mean 4.9, range 1-14) were observed more frequently than losses (mean 2.2, range 0-11), a finding consistently demonstrated in all published series of CGH and NHL. The most common abnormalities found, irrespective of subtype, were gains on 4q, 5q, 7q, 11q and X and losses on 3p, 8p and 10q. All subtypes of NHL have developed from the same haematopoietic lineage and it would be reasonable to anticipate certain similarities across the spectrum of this disease. Overrepresentation on the X chromosome occurs in 21-50% of FL and DLBL in reported CGH series Dierlamm et al, 1996;Joos et al, 1996;Monni et al, 1996;Avet-Loiseau et al, 1997) and was the most common aberration found in this study. This correlates with well established cytogenetic data showing that additional X chromosomes are the most frequent gains in NHL (Fifth International Workshop on Chromosomes in Leukaemia-Lymphoma, 1987;Offit et al, 1991;Hammond et al, 1992), with possible oncogenic sites at Xp22 and Xq28 (Goyns et al, 1993). Losses on 8p and 10q have also been reported in NHL (Goodacre et al, 1994;Avet-Loiseau et al, 1996;Monni et al, 1996) and may represent novel tumour suppressor gene sites.
Comparison of FL and transformed DLBL paired biopsies from the same patient provides a unique picture of the evolution of this disease. Abnormalities found in the FL series were often present in the transformed tissue, supporting the theory that the DLBL has evolved directly from the initial FL clone. The transformed cases were more complex, with a greater number of gains and losses (Table 1). This may be a consequence of the accumulation of genetic aberrations with repeated cell divisions (Yunis et al, 1987) until a critical threshold is reached, beyond which the transformed DLBL develops. However, certain abnormalities were repeatedly found in the transformed tissue that were never present in the FL counterpart, suggesting that specific, non-random events are necessary for transformation to occur.
Gain on 2p and losses on 1q and 15q were present in both the de novo and transformed DLBL biopsies but were never present in FL. High level gain at 2p13-16 has been identified as a common finding in high-grade NHL (Houldsworth et al, 1996;Joos et al, 1996;Werner et al, 1997) resulting from amplification of the REL proto-oncogene at this site (Houldsworth et al, 1996;Joos et al, 1996). The specific abnormalities present in the DLBL tissue irrespective of its origin may confer its histological phenotype.
Over-representation on 2q, 6p, 7p and 17q and loss on 5p and 8q were only ever found in the transformed DLBL cases. The histological appearance of de novo and transformed DLBL is usually identical, but the clinical behaviour is quite different. It is possible that the specific genetic abnormalities present in the transformed cases reduce their susceptibility to cytotoxic drugs or ionizing radiation.
Over-representation of 12q12-14 was found in 12 of 23 transformed DLBL biopsies and was never present in the FL counterpart. This gain was also found in 2 of the 18 de novo DLBL cases. Both of these cases had been highlighted in the pathological review at the beginning of the study as having histological features suggestive of transformation from an occult FL. Gain at 12q12-14 is one of the most common abnormalities found in the published CGH data in high-grade NHL (Monni et al, 1996;Joos et al, 1996;Rao et al, 1998). It was not stated in these studies, whether the cases were de novo or transformed high-grade NHL. Emerging data suggest that amplification of an oncogene at this site might also be important in other human tumours (Elkahloun et al, 1996), including sarcoma (Khatib et al, 1993), glioma (Reifenberger et al, 1994), testicular germ cell tumour (Riou et al, 1995), prostatic carcinoma (Sattler et al, 1999) and liposarcoma (Knuutila et al, 1998). The 12q amplicon has been found to be highly complex, exhibiting discontinuous regions of amplification (Wolf et al, 1997). There are a number of candidate genes in this region which include MDM2, CDK4, CDK2, GL1, ASA, and GAD153. MDM2 has been shown to be over-expressed in some cases of high-grade NHL (Finnegan et al, 1994). It has a central role in the stabilization of p53 and could theoretically reduce susceptibility to chemotherapy. More recently, a gene encoding the human BAX inhibitor, BI1 has also been identified at this site. When over-expressed in mammalian cells, BI1 has been shown to suppress BAX-induced apoptosis and can interact with BCL2 which is known to be aberrantly expressed in nearly all FCL (Xu and Reed, 1998).
In contrast to others (Monni et al, 1997;Werner et al, 1997), we found only one case in which there was low-level copy number gain at 18q21, the site of the BCL-2 gene. The former group, in particular, have frequently found high-level amplification at this locus in DLBL. Werner and colleagues identified two such cases in a series of 62 follicular and diffuse lymphomas (Werner et al, 1997). Such inconsistency may reflect the relatively smaller number of cases in our study or population differences. At other loci, our data are in complete agreement with these studies, with high-level amplifications at 2p and on the X chromosome being frequently observed abnormalities.
If a specific genetic abnormality conferred drug resistance in transformed DLBL, those patients found to have the abnormality would be expected to have a poorer outcome. In this study, a trend towards worse survival was observed in patients with gain at 12q12-14 present in the transformed DLBL DNA (23 months compared with 35 months in those without the aberration). Patients with transformed DLBL have a very poor outcome because the duration of response to currently available treatment modalities is limited. If amplification of a gene contributing to this clinical behaviour could be identified, it would increase our understanding of the molecular mechanisms driving the transformation event and might also provide a target for novel therapeutic strategies of the future.

ACKNOWLEDGEMENT
This work has been supported by Yorkshire Cancer Research.