Cell-of-Origin Identification and Prognostic Correlation in HIV-Associated Diffuse Large B-Cell Lymphomas: Results of an Italian Multicentric Study

Introduction: Diffuse Large B-cell lymphoma (DLBCL) is an heterogeneous disease in terms of biology, clinical presentation and prognosis. According to the putative Cell-Of-Origin (COO), DLBCL of the general HIV negative population can be classified into three molecular subtypes by gene expression profiling (GEP): the germinal center B-cell type (GCB), the activated B-cell type (ABC) and the unclassifiable type (UC). Immuno-histochemistry (IHC) models routinely surrogate GEP in the COO assignment, identifying GCB and non-GCB type. Hans algorithm is one of the most widely used IHC models and its concordance with GEP varies from 70 to 90% (Meyer PN et al, JCO 2011). NanoString (NS) 20-genes assay has emerged as a feasible technique for DLBCL molecular subgroups identification and shows a high concordance with GEP (Scott DW et al, Blood 2014). DLBCL classification according to COO categories has shown prognostic impact in the HIV negative population, since clinical outcome of non-GCB type is inferior if compared to GCB type in patients (pts) receiving CHOP or Rituximab-CHOP. DLBCL in HIV positive population have shown different biological features. COO subtypes distribution and its prognostic impact has not been extensively studied in HIV setting and still needs to be defined.

Aim of the study: To evaluate the proportion of COO subtypes in HIV-associated DLBCL according to IHC and NS assay, the concordance of these two different diagnostic methods, and the prognostic impact of COO assignment; to analyze BCL-2 and MYC protein expression and their correlation with outcome; to determine the prognostic value of baseline clinical characteristics, such as stage, LDH value and CD4+ lymphocyte count.

Methods: We retrospectively evaluated 66 cases of HIV positive pts with newly diagnosed DLBCL from 2000 to 2016, in 5 Italian Institutions. Histological samples were centrally reviewed by a panel of expert Pathologists for DLBCL diagnosis confirmation, BCL-2 and MYC protein expression and COO assignment according to Hans algorithm (Hans CP et al, Blood 2004). The cut off considered for BCL-2 and MYC overexpression was 70% and 40% respectively. NS Lymph2Cx assay for COO was performed in all pts. Clinical data were gathered from pts medical records.

Results: Pts were mostly male (73%) and median age at DLBCL onset was 45 years (range: 28-83). 80% of pts had advanced stage lymphoma and 34% showed a high-intermediate or high IPI score. HIV first detection was concomitant to DLBCL diagnosis in 30% of pts. At DLBCL diagnosis median CD4+ cell count was 188/microL (range: 8-1172) and 34/66 pts (52%) had detectable HIV viral load. COO assigned by IHC was GCB in 31/66 pts (47%) and non-GCB in 35/66 (53%); COO allocation by NS assay was 57% GCB (n=34/60 pts) and 43% non-GCB (ABC n=11, 18%; UC n=15, 25%). IHC algorithm and NS assay concordantly assigned COO subtypes in 80% of pts with a Cohen's kappa=0,604 (p<0.0005). Twenty pts (30%) had MYC and 23 pts (33%) had BCL-2 overexpression, respectively. An anthracycline-containing first-line chemotherapy was delivered to 97% of pts, with the addition of Rituximab in 86% of cases. Two pts (3%) could receive only palliative treatment. Two-years progression-free survival (PFS) and overall survival (OS) of the entire series were 61% and 56%, respectively, after a median follow-up of 50 months (range 4-184). No difference was found in OS and PFS according to COO subtypes, either determined by IHC or NS. Notably, bcl-2 overexpression had a negative impact on OS (2-years OS 44% vs 70%, p=0.018, HR: 2,34) and PFS (2-years PFS 45% vs 62%, p= 0,12, HR: 1,71), while MYC expression had no significant correlation with survival. A baseline CD4+ count > 200/microL has a protective role both for progression and death (2-years PFS: 62% vs 42%, p=0.02, HR: 0.41; 2-years OS: 72% vs 42%, p=0.002, HR: 0.30). Stage IV disease and elevated baseline LDH were associated with inferior PFS (p=0.0158 and p=0.02) and OS (p=0.0076 and p=0.034).

Conclusions: This analysis of 66 HIV-associated DLBCL confirmed that GCB is the more frequent subtype identified by NS, as previously reported (Baptista MJ et al, Hemat Oncol 2017). We found an high concordance between Hans algorithm and NS in the COO subtypes allocation. However, COO classification showed no impact on pts outcome. FISH assays for detection of MYC, BCL-2 and BCL-6 rearrangements are ongoing in the entire study population.