PD-L1 and tumor-associated macrophages in de novo DLBCL

Abstract

Programmed death-ligand 1 (PD-L1) and its receptor, programmed cell death-1 (PD-1), are important negative regulators of immune cell activation. Therapeutically targeting PD-1/PD-L1 in diffuse large B-cell lymphoma (DLBCL) patients with a single agent has limited activity, meriting a deeper understanding of this complex biology and of available PD-L1 clinical assays. In this study, we leveraged 2 large de novo DLBCL phase 3 trials (GOYA and MAIN) to better understand the biologic and clinical relevance of PD-L1 in de novo DLBCL. PD-L1 was expressed on myeloid cells in 85% to 95% of DLBCL patients (depending on staining procedure), compared with 10% on tumor cells, and correlated with macrophage gene expression. PD-L1 did not identify high-risk patients in de novo DLBCL; it correlated with STAT3, macrophage gene expression, and improved outcomes among a subset of patients. These results may help identify immunologically distinct DLBCL subsets relevant for checkpoint blockade. GOYA and MAIN trials were registered at www.clinicaltrials.gov as #NCT01287741 and #NCT00486759, respectively.

Graphical abstract

Figure 1.

Similar to normal lymph nodes, PD-L1 is expressed by myeloid ICs in DLBCL, with different prevalence and intensity depending on the staining procedure. (A) Membranous immunohistochemical stain for PD-L1 protein (with hematoxylin counterstain) on cells with myeloid/dendritic morphology in normal lymph nodes (original magnification ×400). (B) Representative images of PD-L1 protein staining (SP263; original magnification ×400) among DLBCL patients treated in MAIN using a simplified IHC scoring system capturing PD-L1⁺ ICs or TCs (IHC 1, 1%-5%; IHC 2, 5%-10%; IHC 3, >10%). Yellow arrows represent PD-L1 staining on myeloid cells, and red arrows represent PD-L1 staining on malignant B cells. (C) PD-L1 prevalence and staining intensity among de novo DLBCL patients treated in 2 phase 3 clinical trials (MAIN, GOYA) using 2 different PD-L1 IHC reagents (SP142, SP263). (D) PD-L1 messenger RNA (mRNA) is higher in the ABC DLBCL subset (P = .004; MAIN). Freq, frequency; nRPKM, normalized reads per kilobase million.

Figure 2.

PD-L1 expression correlates with macrophage and STAT3 gene expression. (A) CD274 (PD-L1) mRNA expression inversely correlates with a B-cell gene signature among DLBCL patients treated in MAIN. (B) CD274 mRNA is highly expressed by purified DCs and macrophages compared with resting B cells. PD-L1 protein expression correlates with a macrophage gene signature among DLBCL patients treated in MAIN (C) and GOYA (D). CD274 mRNA correlates with a macrophage gene signature among DLBCL patients treated in MAIN (E) and GOYA (F). CD274 mRNA correlates with STAT3 gene expression (G) and a STAT3 gene signature (H) among DLBCL patients treated in GOYA.

Figure 3.

PD-L1 is not a negative prognostic biomarker in de novo DLBCL and may be associated with better prognosis in some patients, similar to macrophages and STAT3. (A) High PD-L1 protein expression (IHC 2+, 3+) by SP263 is associated with prolonged PFS in de novo DLBCL (MAIN). (B) Forest plot of HRs and 95% CIs for the association of PD-L1 expression and PFS with regard to different PD-L1 IHC reagents (SP142, SP263), the PD-L1 transcript (CD274 mRNA), and distinct DLBCL COO subgroups in MAIN and GOYA. High expression of a macrophage gene signature correlates with prolonged PFS (C) and adds prognostic information to DLBCL COO (D-E). (F) High expression of STAT3 mRNA correlates with prolonged PFS in de novo DLBCL (GOYA). Hazard ratios adjusted for IPI and treatment (MAIN) or IPI, treatment, region, and number of chemotherapy cycles (GOYA). Uncl, unclassified.