Regions of acquired uniparental disomy at diagnosis of follicular lymphoma are associated with both overall survival and risk of transformation

Follicular lymphoma (FL) is the most common indolent lymphoma and has marked heterogeneity in outcome,^1-3  with frequent transformation to aggressive lymphoma affecting dramatically on overall survival (OS).^4,5  The genomic hallmark of FL is t(14;18)(q32;q21) with many secondary abnormalities described,^6,7  although few have confirmed prognostic value.^6,8,9  The target genes remain largely unidentified.

Single nucleotide polymorphism (SNP) arrays allow detection of copy-neutral loss of heterozygosity (LOH) undetectable by previous methods, termed acquired uniparental disomy (aUPD). We and others have described aUPD in several hematologic malignancies,^10,11  including FL.^12,13  It is due to mitotic recombination or nondisjunction and may render a cell homozygous for a preexisting abnormality leading to clonal selection. Both FL series profiled to date have shown frequent, nonrandom regions of aUPD; however, its prognostic implication has not been addressed. We aimed to clarify the incidence of aUPD in 185 diagnostic FL patients and correlate this with clinical outcome.

Tumor-extracted DNA from 185 untreated patients with FL presenting between 1974 and 2001 was obtained through the Lymphoma/Leukemia Molecular Profiling Project (LLMPP). Clinical data were available in 169 (93%) of 182 cases (Table S1, available on the Blood website; see the Supplemental Materials link at the top of the online article), transformation data were available for 141 cases, and 39 patients transformed to aggressive lymphoma. Research ethics committee approval was obtained from the London School of Medicine before initiation of the study.

SNP array genotyping was performed with the use of the 10K 2.0 GeneChip (Affymetrix, Santa Clara, CA) as previously described.¹⁴  Signal intensity data were analyzed by GeneChip DNA analysis software, and GTYPE was used for calling genotypes. GOLF, an in-house software was used for copy number (CN) and UPD estimation (http://bioinformatics.cancerresearchuk.org/∼cazier01/). In the absence of germline controls the definition of aUPD was based on runs of consecutive homozygous markers defined as maximum 2 heterozygous calls in 50 consecutive SNPs. This was described by Gupta et al,¹⁵  and briefly this stringent criterion is based on 10K 2.0 data available from 90 independent germline samples from the ColoRectal Tumour Gene Identification (CORGI) study consortium giving a false-positive rate of 3.3%.¹⁶  CN was determined based on the log₂ ratio of signal intensity from the lymphoma sample versus the pooled signal intensity of 10 unrelated control DNA samples. A ratio below 0.75 and above 1.25 on at least 3 consecutive SNPs was defined as a loss and a gain, respectively. The data were also analyzed with the use of Partek Genomics Suite (Partek, St Louis, MO). Sex chromosomes were excluded from analysis. SNP and gene annotations used National Center for Biotechnology Information (NCBI) genome build 35.

Mutation screening was performed for the coding region of ID4, BRD2, HLA DQB1, SOCS1, CDK4, DYRK2, PTEN, and specific exons of NFKB2¹⁷  (Table S2; protocols available on request).

The Fisher exact test and the Mann-Whitney U test were used to determine the association with clinical characteristics. OS was defined as the time from diagnosis to death or, for patients remaining alive, the time from diagnosis to last contact. Progression-free survival (PFS) was defined as the time from diagnosis to first clinical progression, transformation, or death from any cause or, for patients remaining alive and disease-free, the time from diagnosis to last contact. Transformation to aggressive lymphoma was defined histologically or by clinical criteria. Kaplan-Meier survival estimates were obtained, and the log-rank test was used to compare differences between the groups with UPD versus those without. Multivariate Cox regression was used to determine whether regions of recurrent abnormality remained independently predictive of PFS and OS after adjusting for the International Prognostic Index (IPI). Statistical significance was set at P values less than .05.

This study of 185 diagnostic FL cases addresses the clinical effect of aUPD, using the well-characterized LLMPP cases.¹⁸  It confirms earlier reports that aUPD is frequent at diagnosis of FL, occurs nonrandomly at recurring chromosomal locations,^12,13  and now shows that aUPD is clinically important. SNP genotype call rates were greater than 90% for 182 of 185 cases analyzed. Abnormalities, including CN and copy-neutral changes, were detected in 118 (65%) of 182 cases (Figure 1). The number of abnormalities ranged from 0 to 9 (median, 1) with median LOH size of 48 Mb (range, 3.4-241 Mb). The 20 cases with more than 3 abnormalities at diagnosis had inferior OS (P = .03) compared with those with no more than 3 abnormalities (n = 149) (Figure 2A). aUPD was seen in 96 (81%) of 118 abnormal cases. Sites of recurrent aUPD were detected on chromosome 6p (n = 25), 16p (n = 22), 1p (n = 16), 12q (n = 17), 10q (n = 8), and 6q (n = 8). aUPD6p was observed in 25 (14%) of 182 cases (Table S3), and, as shown previously,¹³  aUPD6p was associated with aUPD1p (P = .002). aUPD16p occurred in 22 (12%) of 182 cases (Table S4) and in 10 of 22 cases transformed to aggressive lymphoma (P < .03).

aUPD16p correlated with poorer PFS (P = .01) (Figure 2B) but not OS and remained significant after adjusting for IPI (hazard ratio [HR], 1.80; confidence interval [CI], 1.09-2.98; P = .02). The association of aUPD16p with transformation (P = .03) represents a potential new risk factor for this event with a trend seen toward shorter time to transformation with aUPD16p (HR, 1.94; CI, 0.92-4.13; P = .08). Loss of 1p36 has been previously associated with poorer OS in FL.⁶  aUPD1p was present in 16 (9%) of 182 cases, and analysis showed that patients with aUPD1p36 (n = 14) had inferior OS (P = .01; Figure 2C) and PFS (P = .05 log-rank test). On multivariate analysis aUPD1p36 correlated with poorer OS after adjusting for IPI (HR, 2.10; CI, 1.00-4.43; P = .05) but not for PFS (HR, 1.70; CI, 0.90-3.18; P = .10). aUPD12q was present in 17 (9%) of 182 cases, 10q abnormalities (loss [n = 9] and aUPD [n = 8]) were seen in 17 (9%) of 182 cases, and chromosome 6q abnormalities were identified in 21 (11%) of 182 cases (loss [n = 13] and aUPD [n = 8]). Loss of 6q and inferior outcome have been shown previously in FL⁹ ; no significant associations were seen here for OS, PFS, or transformation with 6p, 6q, 10q, or 12q abnormalities. Recurrent regions of gain were identified on chromosome 2p (involving REL), chromosome 7, 8q, 12q (involving MDM2), and 18q, areas of gain previously described in FL.^19-21  No statistically significant associations were seen here with clinical outcome. The regions of aUPD, copy number loss, and gain were verified with the use of Partek Genomic Suite.

The recurring regions of abnormality seen here showed considerable overlap with previous analysis that used cell-sorted FL samples, but the frequency of abnormalities was higher.¹³  This finding suggests that in FL purified tumor populations are preferred for accurate genotyping because DNA from nontumor cells will interfere with the detection of abnormalities. It is probable that some of the “normal” cases in this study may have low malignant cell content, below the threshold of detection of this array. Identification of gene lesions made homozygous has proven difficult in FL; in this series none of the target genes screened were mutated (Table S1), and it is probable that other selective mechanisms are important.²²  The regions identified here were invariably large, pointing toward clonal events, but this may be an underestimate. Comparison of tumor and germline DNA is required to distinguish smaller somatic UPD from inherited events.²³  This study focuses on recurrent regions of LOH making it less likely to detect a rarely reported constitutional abnormality.²⁴  The use of higher resolution platforms may identify cases with more localized aberrations suitable for further characterization by next-generation sequencing.

This work was supported by grants from Cancer Research UK, the Medical Research Council, and the National Institutes of Health (SPEC grant 5UO1CA114778). D.O. is supported by a Medical Research Council Clinical Research Fellow grant. C.O. is a Cancer Research UK Barts-Cambridge Molecular Pathology Clinical Research Fellow.

National Institutes of Health

Contribution: D.O. designed the study, performed research, analyzed data, and wrote the paper; C.O., Y.Y., and D.W. performed research; M.G., R.W., J.G., B.D.Y., and J.-B.C. analyzed data; A.R., G.O., L.M.R., H.H., N.A.J., E.C., W.C.C., and R.D.G. collected data; and L.M.S., T.A.L., and J.F. designed the study and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Derville O'Shea, Centre for Medical Oncology, Barts and The London School of Medicine, Charterhouse Square, London EC1M6BQ, United Kingdom; e-mail: derville.oshea@cancer.org.uk.