Peripheral T-Cell Lymphomas Not Otherwise Specified: Potential Novel Molecular Entities Based on Both Tumor and Microenvironment Cellular Components

INTRODUCTION: The diagnosis of Peripheral T-cell lymphomas not otherwise specified (PTCL-NOS) is currentlybased on an "exclusion criteria" model, since PTCL-NOS lack pathognomonic features. Nevertheless, based on gene expression data, Iqbal et al(Blood 2014) have recently identified two different PTCL-NOS subgroups with different biological and prognostic features that accounts for approximately 80% of the cases and have different biological and prognostic features: one characterized by TBX21 overexpression and T-CD8+ molecular profile; the other by GATA3 overexpression and T-CD4+ profile. Herein, we used a wide comprehensive gene expression profiling (GEP) data set in order to further investigate the molecular features of different PTCL-NOS molecular entities.

MATHERIAL AND METHODS:

A data set were created including samples from 8 main published series (GSE6338, GSE14879, GSE19067, GSE19069, GSE58445 and GSE65823 at http://www.ncbi.nlm.nih.gov/geo/; E-TABM-702 and E-TABM-783 at https://www.ebi.ac.uk/arrayexpress) for a total of 541 patients. R/Bioconductor was used to generate and analyze the gene expression data. We applied the CIBERSORT algorithm (Gentles et al, Nat Gen 2015), which connects specific global expression profiles to the relative prevalence of tissue components (tumour and microenvironment cells).

RESULTS: We first re-classified each sample included in the investigated data set based on previously published signatures (Iqbal et al. Blood 2014, Agnelli et al. Blood 2012). This approach led to a final data set of 144 PTCL-NOS (28%), 127 AITL (23%), 69 ALCL Alk neg (12%), 56 ALCL Alk pos (10%), 59 NK (11%) %), together with 86 healthy T-cell tissues. To our knowledge, this is the largest GEP data set ever described in PTCL so far.

In the 144 PTCL-NOS cases, two main molecular clusters were extracted based on published signature, replicating previous findings: the first was characterized by both GATA3 expression (GATA3+) and T-CD4+ cell origin; the second by TBX21 expression (TBX21+) and T-CD8+ cell origin. Approximately 30% of all PTCL-NOS were characterized by neither GATA3 (GATA-) nor TBX21 expression (TBX21-), and for this reason they were classified as "double negative" PTCL-NOS. Based on data obtained by the CIBERSORT algorithm we found that the contribution of cellular microenvironment components was extremely heterogeneous and variable through the entire PTCL-NOS series. A significant T-CD4+ and T-CD8+ cell enrichment was reported among GATA3+ and TBX21+ groups, respectively. Interestingly, a fraction of GATA3+ PTCL-NOS (n=11, 7.6%) was characterized by a significant γδ T-cell component. Conversely, PTCL-NOS GATA3+ patients without γδ T-cell component signature were characterized by a low non-T-cell microenvironment component. This may reflect the major tumour infiltration due to higher proliferation rate as suggested by the strong GATA3 correlation with MIB1 and MYC expression (p<0.0001 and p=0.01, respectively). PTCL-NOS TBX21+/GATA3- were mainly divided into two different groups: the first characterized by strong plasma cell enrichment and the second by major macrophage contribution. "Double-negative" PTCL-NOS (n=38) were generally characterized by strong T-follicular helper (T-FH) and B-cell signatures contribution.

Based on the clinical data available for 105/144 PTCL-NOS, poor overall survival (OS) was associated with GATA3 expression (p=0.03) [3-y OS 26.4% (range 20-32.8%) vs 47.3% (range 40-54.5%)]. However, such a poor outcome was less evident when we limited the clinical evaluation to patients younger than 60 years (40/105). Among low GATA3 expressors, TBX21 expression was associated with better OS compared to other "double negative" PTCL-NOS (p=0.07) [3-y OS 66.7% (range 51-82.4%) vs 36.4% (range 22-51%)]. Higher CIBERSORT-predicted T-CD8+ contribution was associated with better OS among patients younger than 60 years (p=0.03) [3-y OS 70% (range 55.6-84.5%) vs 30% (range 21.7-38.3%)]. No other variables or clusters were associated with significant impact on OS.

CONCLUSION: Our study based on an innovative computational approach and a large and comprehensive gene expression data set, confirmed the great molecular heterogeneity of PTCL-NOS, suggesting that the current molecular classification of PTCL-NOS may be further improved in the future.