Molecular Subgroups of Peripheral T-Cell Lymphoma Evolve By Distinct Genetic Pathways

Peripheral T-cell lymphoma (PTCL) is a group of clinically and pathologically heterogeneous non-Hodgkin lymphomas (NHL). Using gene expression profiling (GEP), we have defined molecular classifiers for PTCL subtypes reflecting their pathobiology and oncogenic pathways (Iqbal et al. 2014). We have also shown associations of specific mutations with the molecular subgroups (Wang et al. 2015). Although genomic information is increasing, the pathogenetic mechanisms of PTCLs remain largely unknown. Therefore, we analyzed copy number variation (CNV) and GEP to identify unique genetic abnormalities in the defined PTCL molecular subgroups.

CNV data were generated on fresh frozen or formalin-fixed paraffin-embedded genomic DNA (n=114) on 3 Affymetrix platforms (SNP 6.0, 250K SNP, and OncoScan). Two published cohorts (PTCL-NOS, Hartmann et al. 2010; ALCL, Boi et al. 2013) were included for validation. The gene expression analysis, morphological review and clinical characteristics of these cases have been included in previous studies (Iqbal et al. 2010, 2014).

Angioimmunoblastic T-cell lymphoma (AITL) represents 20% of all PTCL cases. The most recurrent CNV in AITL was chromosome (chr) 5 gain (39%), followed by chr 21 gain (21%). Interestingly, chr 21 gain co-occurred with chr 5 gain (p=0.003). No recurrent losses (≥20%) were identified among these cases. Molecularly re-classified AITL cases from morphologically classified PTCL-NOS cases showed concordant results with bonafide AITL cases. Of the commonly mutated genes, DNMT3A, IDH2, RHOA and TET2, only IDH2^R172Kshowed a significant association (p=0.012) with chr 5 gain. GEP showed enrichment of gene signatures associated with oxidative phosphorylation (PGC-1α target genes) in cases with chr 5 gain.

PTCL, not otherwise specified (PTCL-NOS) is the most common PTCL subtype and cannot be further sub-classified using conventional approaches; however, we have identified 2 molecular subgroups within PTCL-NOS, the GATA3 and TBX21 subgroups which are related to 2 distinct T-helper subsets (Iqbal et al. 2014), by employing GEP. Consistent with earlier observations (Hartmann et al. 2010), PTCL-NOS showed remarkably varied CNVs with nearly 50% of cases showing high CNV frequencies. When correlated with molecular subgroups, distinctive CNVs were observed in the molecular GATA3 and TBX21 subgroups. The GATA3 subgroup displayed a large assortment of CNVs. Complete or partial gain of chr 7 (57%) was the most recurrent gain in these cases. Losses affecting 17p, 10q and 9p21, encompassing tumor suppressors such as TP53 (57%), PTEN (43%) and CDKN2A (43%), were frequent in the GATA3 subgroup. The TBX21 subgroup had significantly fewer CNVs, as none were recurring (≥20%); but gains of 5p or 11p were observed in 14%. Additionally, PTCL-NOS cases with ≥10% abnormal genome had significantly poorer overall survival (p=0.012) compared to those with fewer abnormalities. This finding validates the GEP molecularly defined subgroups, as the GATA3 subgroup displayed more CNVs and has been associated with a worse prognosis compared to the TBX21 subgroup (Iqbal et al. 2014).

We were able to distinguish CNVs characteristic of the different entities, including the co-occurrence of chr 5 and 21 gains specific in AITL. Gain of 1q (complete or partial) was identified in the GATA3 subgroup of PTCL-NOS and anaplastic lymphoma kinase (ALK) (-) ALCL with equal frequencies (~ 36%), but only 16% in ALK(+) ALCL. Complete or partial gain of chr 7 was also observed in ALCL, but at a considerably lower frequency than in the GATA3 subgroup. Additionally, gain of chr 18 or regions of 17q, and loss of 5q or regions on both arms of chr 9, were more frequent in the GATA3 subgroup compared to other entities. The TBX21 subgroup was primarily differentiated from the GATA3 subgroup by presence of fewer CNVs.

Our analysis provides a framework for future investigations into the molecular pathogenesis of PTCL, and highlights potential candidate oncogenes and tumor suppressors deregulated by copy number aberrations. Comparative analysis revealed that certain chromosomal abnormalities are entity-specific. AITL cases with IDH2^R172K also had trisomy 5 suggesting that these oncogenic events cooperate in malignant transformation. Thus, the complexity of PTCL is finally becoming clearer with the integration of high resolution molecular techniques for global genomic analysis.