Phenotypic and Transcriptional Characterization of Non-Hodgkin Lymphomas from Malawi Defines Targetable Disease Subsets

Outcomes among persons with lymphoma in sub-Saharan African (SSA) remain poor, largely due to resource limitations in diagnostics, therapeutics and supportive care. We hypothesized that subclassification of lymphoma specimens from SSA could improve diagnostic accuracy, thereby altering treatment recommendations and identifying cases responsive to less toxic, targeted agents. The Queen Elizabeth Central Hospital (QECH) in Malawi serves as the teaching hospital for the University of Malawi College of Medicine (UOMCOM) in Blantyre, Malawi's largest city. Pathologic diagnosis of lymphoma at QECH/UOMCOM is based solely on H&E-stained sections. We created tissue microarrays (TMA) from 498 formalin-fixed, paraffin-embedded pediatric and adult nodal/extranodal tumors diagnosed as non-Hodgkin lymphoma (NHL) by H&E from 2004-2014 at QECH/UOMCOM (487 unique patients). Median age was 31 years (range, 1-87; 25% £ 16); 274 males, 189 females, 35 unknown. HIV status was available for 10% (36 reactive; 12 non-reactive). At Brigham and Women's Hospital, we performed immunohistochemistry (IHC) for 28 markers and classified each case per the 2008 WHO Classification. 170 cases (34%) could not be classified, largely due to poor tissue antigenicity or lack of lesional tissue on TMA. The remaining 328 cases (66%) are classified in the Table. 37% of the 133 DLBCL were germinal center B cell-like (GCB), 59% non-GCB, and 4% could not be characterized (Hans criteria). We also performed targeted expression profiling with a previously-published, NanoString-based diagnostic molecular classifier of aggressive B-cell lymphomas. This classifier distinguishes ~85% of pathological BL cases from DLBCL in Western cohorts (Carey CD, J Mol Diagn 2015). Evaluable RNA was available in 64/133 (48%) DLCBL, 38/73 (52%) BL, and 8/9 (89%) BCL-U cases. Unsupervised clustering of transcriptional profiles identified 2 groups (n = 54, 56). All 38 BL cases were classified into Group I (100%; exact binomial 95% CI: 91-100), along with 14 DLBCL (22%; 13-34%) and 4 BCL-U (50%; 16-84%). Group II combined 50 of 64 DLBCL cases (78%; 66-87%), 4 of 8 BCL-U cases (50%; 16-84%), and no BL (0%; 0-9%). Concordance between the classifier and pathologic diagnosis was 84%. Overall, incorporation of phenotypic and transcriptional data allowed reclassification of 83 cases into categories that may be responsive to ibrutinib or idelalisib (MCL; non-GCB DLBCL; CLL/SLL); 75 into categories that could benefit from CHOP (ALCL; DLBCL; PBL; PTCL-NOS); 8 into T-ALL; and 30 additional cases of BL by IHC alone. 18 cases of DLBCL or BCL-U were also found to have a similar transcriptional profile to BL, suggesting that they may benefit from BL-directed therapy. Thus, our findings demonstrate that ancillary phenotypic and transcriptional characterization can reclassify a large fraction of lymphomas in SSA and guide potentially beneficial therapies. Implementation of these tools and enhanced drug access and supportive care are desperately needed to improve patient outcomes across the developing world.