High-Resolution Genomic Copy Number Profiling of Primary Intraocular Lymphomas Using SNP Microarrays

Abstract 1354

Primary intraocular lymphoma (PIOL) is a rare form of non-Hodgkin lymphoma (NHL), accounting for less than 1% of all NHL cases. Because 60–80% of PIOLs develop in the central nervous system (CNS), 25% of CNS lymphomas (CNSL) are accompanied by intraocular lesions, and both are B-cell lymphomas, PIOL has been considered as a subtype of primary CNSL (PCNSL). However, there is little genetic evidence to support this classification. The purpose of this study was to evaluate the relevance of the current PIOL classification by genomic profiling of our PIOL series.

Intraocular lymphoma (IOL) was diagnosed using the following criteria: (1) typical eye involvement; (2) identifying lymphoma cells in the vitreous fluid; (3) detecting the clonality of infiltrating cells in the vitreous fluid by PCR analysis of the IgH gene or by flow cytometry for the surface light chain expression. Patients who had either (1) and (2) or (3) were diagnosed with IOL. IOL without any other systemic lesion was defined as PIOL. Other IOLs were classified as IOL with a CNS lesion (IOCNSL) and secondary IOL (SIOL) following another systemic lymphoma.

Genomic DNA extracted from the vitreous fluid was subjected to SNP-array karyotyping using GeneChip 250K Nsp arrays (Affymetrix, Santa Clara, CA, USA). Genetic lesions were sensitively detected by a hidden Markov model-based algorithm and were compared with those of PCNSLs examined by array-based CGH (Blood. 2011; 117, 1291–1300).

We examined the samples of 53 IOL patients including 31 with PIOL, 14 with IOCNSL, and 8 with SIOL. Among these patients, 30 samples (12 PIOL, 12 IOCNSL, and 6 SIOL) could be analyzed for SNP array.

In PIOL, 1q32.1–3 containing MDM4, ELK4, IL10 was the region of the most frequent gain (>75%). Regions of frequent gain, detected in 75% of the patients, were located on 1q25.1–3 (containing ABL2 etc.), 1q23.3 (RXRG, LMX1A etc.), 1q42.12 (PARP1 etc.), 1q43 (RYR2), 1q44 (PGBD2, ZNF692 etc.), 12q13.3,(STAT6, GLI1 etc.), 12q23.2-24.2 (TBX3,PRKAB1 etc.), 18q21.31-22.1 (BCL2 etc.), and 19q13.32-13.43 (BAX etc.). 15p13-q11.2 (POTE15 etc.) was the region of the most frequent loss (7/12; 58%).

In IOCNSL, regions of the most frequent gain included 7q32.3, 7q34, 12 q13.13-q13.2, 12q23.3, 12q24.21, and 12q24.23 (10/12; 83%). Regions that were gained in 75% (9/12) sites included 1q25.1-3, 7q11.22-23, 7q22.1, 12p13.32-p13.31, 12q13.13-q13.2, 12q24.31, and 18q21.1. 6q24.1–2 and 9q21.3 were regions of the most frequent loss (7/12; 58%); the latter contained CDKN2A.

In SIOL, 7q34 and 12q13.13–2 were regions of frequent gain; they were gained in 100% of the patients and also in 83% of IOCNSL. Regions of the most frequent loss included 5p15.33, 6q27, 8p23.3, 8q24.3, 9q34.3, and 16q24.2-24.3 (5/6; 83.3%).

Next, we compared the findings of PIOL with those of PCNSL, IOCNSL, and SIOL to identify the genetic characteristics of PIOL.

In an overview of copy number alterations, a significant difference between PIOL and PCNSL was detected on chromosome 18. In PIOL, this region revealed copy number gain. High frequency was particularly detected in 18q21.32-q22.1 containing BCL2. In contrast, chromosome 18 generally exhibited copy number loss in PCNSL patients, and it is worth mentioning that no gain of BCL2 was detected in them. In total, among the frequent gain (>75%) regions in PIOL, only 1 region (10%), 19q13.32–33 was identical to the frequent gain (>50%) region in PCNSL. Interestingly, chromosome 18 generally revealed gains in IOCNSL and only 1 of the frequent gain (>75%) regions in IOCNSL overlapped those (>50%) in PCNSL. Deletion of CDKN2A, which was most frequently (66.7%) observed in PCNSL, was not detected in any PIOL patient.

In PIOL, among 10 regions of frequent gain (>75%), 2 regions (20%) were identical to the regions of frequent gain (>75%) in IOCNSL. Another 2 regions (20%) were identical to those in SIOL. Among 4 regions of frequent loss in PIOL, 2 regions (50%) were identical to those of IOCNSL.

In conclusion, the genomic profile of PIOL, also of IOCNSL, showed a remarkable contrast to that of PCNSL. On the other hand, we could not find significant differences between PIOL and IOCNSL. Thus, PIOL with or without CNS lesions may be a B-cell lymphoma that is genetically distinct from PCNSL.