Low Density of CD3+ Tumor-Infiltrating Lymphocytes Is Predictive of a Poor Prognosis in Diffuse Large B-Cell Lymphoma Independently of MYC and BCL2 Double Expression Status: A Potential Utility of Immunoscore Based on Whole-Slide Image Analysis

Background: Anti-tumor immunity plays a key role in tumor development and progression and the landscape of tumor immune microenvironment could affect the clinical outcome of patients. The concept of immunoscore based on immune cell infiltration has been actively applied to solid tumors including colon cancer and malignant melanoma and successfully predicted prognosis of patients. However, a potential utility of immunoscore in predicting prognosis of patients with diffuse large B-cell lymphoma (DLBCL) remains unclear. For easy application of immunoscore to daily practice, we investigated the prognostic value of tumor-infiltrating CD3+ T-cell density which is routinely stained for diagnosing malignant lymphomas, and then explored the feasibility of immunoscore application in DLBCL.

Method: One-hundred-four patients with DLBCL treated with R-CHOP therapy, whose tumors were prospectively immunoprofiled for routine clinical diagnosis, were enrolled in this study. Immunohistochemistry was performed on representative whole formalin-fixed paraffin-embedded tissue sections using CD3, CD20, MYC, BCL2, BCL6, CD10 and MUM1. Cell-of-origin was determined using the Hans' algorithm and MYC/BCL2 double expressor (DE) was defined as the co-expression of MYC (in ≥ 40% of tumor cells) and BCL2 (in ≥ 70% of tumor cells). Whole-slide scanning was performed, whole tumor regions were annotated, and images were analyzed with Aperio ImageScope v.12.4.3 software (Leica Biosystems, Nussloch, Germany). The CD3+ density was quantified in three different methods: the total numbers of CD3+ cells/area (mm ²), CD3+ cells/total cells and CD3+ cells/CD20+ cells. Correlation between the CD3 density and clinicopathologic parameters and prognostic impact of the CD3 density were analyzed. To validate the findings, publicly available mRNA and clinical datasets of two DLBCL cohorts were downloaded and analyzed (Schmitz et al. 2018 and Sha et al. 2019). To dichotomizing low versus high CD3 density cases, the optimal cut-off values for overall survival were determined by ROC curve analysis using MedCalc Software v 19.4 (MedCalc Software, Ostend, Belgium). Remaining all statistical analyses were performed using SPSS version 25.0 (SPSS).

Result: All three methods assessing CD3+ cell density showed high concordance (CD3+ cells/area vs CD3+ cells/total cells, k=0.540 p=0.000; CD3+ cells/total cells vs CD3+ cells/CD20+ cells, k=0.394 p=0.000; CD3+ cells/CD20+ cells vs CD3+ cells/area, k=0.739 p=0.000). Patients with low CD3 density had elevated serum LDH (all, p<0.05) and tended to show poor clinical parameters including high Ann Arbor stage (III-IV), high international prognostic index (IPI, 3-5) and DE status. Patients with low CD3 density by all three methods had worse overall survival (OS) and progression free survival (PFS) (all, p<0.05). In univariate analysis, the following clinical variables were predictive of worse OS: Ann Arbor stage IV, Eastern Cooperative Oncology Group performance status (ECOG PS) ≥2, elevated serum LDH, IPI ≥3, DE and low CD3 density by all three methods. In multivariate analysis, low CD3 density by all methods predicted poor OS independent of the DE status, ECOG PS, LDH or Ann Arbor stage (all, p<0.05). Taken together, these findings suggested that evaluating CD3 density by any methods worked well as prognostic indicators and further might work well as immunoscoring methods in DLBCL. Potential utility of CD3+ cell density as an immunoscore observed in our cohort was validated using publicly available dataset generated by Schmitz et al. and Sha et al. In both validation cohorts, patients with low CD3E mRNA expression had elevated serum LDH (p<0.05) and in Sha et al. cohort, low CD3E mRNA expression was correlated with high ECOG PS, high IPI and DE status (p<0.05). Moreover, in both of Schmitz et al. and Sha et al. cohorts, patients with low CD3E mRNA expression tended to have worse PFS (p=0.067 and 0.002, respectively) and OS (both, p<0.05). In multivariate analysis, low CD3E mRNA predicted poor OS, independent of the IPI and DE status, in both validation cohorts.

Conclusion: Low CD3 density was a poor prognostic factor in DLBCL independent of other poor prognostic parameters such as IPI score and DE status. Therefore, evaluating CD3 density might be suitable for immunoscoring DLBCL cases to predict prognosis.