Spatial signatures identify immune escape via PD-1 as a defining feature of T-cell/histiocyte-rich large B-cell lymphoma

Abstract

T-cell/histiocyte-rich large B-cell lymphoma (TCRLBCL) is an aggressive variant of diffuse large B-cell lymphoma (DLBCL) characterized by rare malignant B cells within a robust but ineffective immune cell infiltrate. The mechanistic basis of immune escape in TCRLBCL is poorly defined and not targeted therapeutically. We performed a genetic and quantitative spatial analysis of the PD-1/PD-L1 pathway in a multi-institutional cohort of TCRLBCLs and found that malignant B cells harbored PD-L1/PD-L2 copy gain or amplification in 64% of cases, which was associated with increased PD-L1 expression (P = .0111). By directed and unsupervised spatial analyses of multiparametric cell phenotypic data within the tumor microenvironment, we found that TCRLBCL is characterized by tumor-immune "neighborhoods" in which malignant B cells are surrounded by exceptionally high numbers of PD-L1-expressing TAMs and PD-1+ T cells. Furthermore, unbiased clustering of spatially resolved immune signatures distinguished TCRLBCL from related subtypes of B-cell lymphoma, including classic Hodgkin lymphoma (cHL) and DLBCL-NOS. Finally, we observed clinical responses to PD-1 blockade in 3 of 5 patients with relapsed/refractory TCRLBCL who were enrolled in clinical trials for refractory hematologic malignancies (NCT03316573; NCT01953692), including 2 complete responses and 1 partial response. Taken together, these data implicate PD-1 signaling as an immune escape pathway in TCRLBCL and also support the potential utility of spatially resolved immune signatures to aid the diagnostic classification and immunotherapeutic prioritization of diverse tumor types.

Graphical abstract

Figure 1.

Alterations of PD-L1/PD-L2 correlate with expression of PD-L1 in malignant B cells. (A) Representative FISH images of probes targeting PD-L1 and PD-L2, which lie adjacent on chromosome 9p24.1, and a centromeric region of chromosome 9 (CEP9, aqua) hybridized against individual malignant cell nuclei highlighted by 4′,6-diamidino-2-phenylindole, showing PD-L1/PD-L2 polysomy, copy gain, amplification, and rearrangement (original magnification, ×100). (B) Waterfall plot showing individual TCRLBCL cases (x-axis) and the proportion of malignant cells with the indicated PD-L1/PD-L2 alteration (y-axis); r, relapsed/refractory cases; arrow, 4 cases with high-level amplifications; #, a case with a productive PD-L2 rearrangement (verified by PD-L2 IHC, panel D). (C) Comparison of the proportion of cases with PD-L1/PD-L2 relative gain (ie, copy gain or amplification) across cases of TCRLBCL (n = 33, red), cHL (n = 108, blue), and DLBCL-NOS (n = 21, gray). (D) Representative IHC demonstrating nuclear PAX5 (red) and membranous PD-L1 (brown) coexpression by a malignant B cell in a case of TCRLBCL with PD-L1/PD-L2 amplification (left), and representative IHC demonstrating membranous PD-L2 expression (brown, PAX5-staining was not performed) by a malignant B cell in a case of TCRLBCL, demonstrating rearrangement of PD-L2 (right) (original magification, ×100). (E) Waterfall plot showing individual TRLBCL cases (x-axis) and malignant cell-specific PD-L1 H-score (y-axis). The corresponding genetic classification for each case is indicated. (F) Comparison of PD-L1 H-scores for individual cases (● and ▲) displayed according to those with PD-L1/PD-L2 relative gain (ie, amplification or copy gain; n = 21) vs those without (n = 12). Four cases where >75% of tumor nuclei showed PD-L1/PD-L2 amplification are designated with triangles. (G) Comparison of malignant tumor cell PD-L1 H-scores for cases of TCRLBCL (n = 33, red), cHL (n = 106, blue), and DLBCL-NOS (n = 20, gray). For box-and-whisker plots (F, G) the median (line), 25th and 75th percentiles (boxes), and extreme values (whiskers) are indicated. Statistical comparisons by the 2-sided Mann-Whitney U test (F, G) and Fisher’s exact test with Bonferroni correction for multiple comparisons (C).

Figure 2.

Malignant B cells are in frequent contact with PD-L1⁺ TAMs. (A) Representative high-power mIF image demonstrating PD-L1 expression on the surface of CD68⁺ TAMs in contact with a PAX5⁺ malignant B cells. (B) Cell-cell contact map of the mIF image shown in panel A (same scale). PD-L1⁺ TAMs, PD-L1⁻ TAMs (not present in the image), and other cells in contact with malignant cells and PD-L1⁺ TAMs and PD-L1⁻ TAMs not in contact with malignant cells are indicated. (C) Proportion of malignant B-cell contacts for individual cases (●) displayed according to the proportion of contacts with PD-L1⁺ TAMs vs PD-L1⁻ TAMs (n = 26). (D) Low-power, maximum field-of-view mIF image of TCRLBCL shown in panel A. (E) Cell phenotype map corresponding to the mIF image in panel D: PD-L1⁺ TAMs, PD-L1⁻ TAMs, and malignant cells are indicated. (F) TAM density for individual cases (●) displayed according to the density of PD-L1⁺ TAMs and PD-L1⁻ TAMs. (G) Intermediate-power mIF image of the region highlighted in panel D: PD-L1 expression on CD68⁺ TAMs and PAX5⁺ B cells. (H) Cell proximity map of the region highlighted in panel E, showing the spatial distribution of PD-L1⁺ TAMs and PD-L1⁻ TAMs relative to malignant B cells (<75 μm and ≥75 μm). (I) The ratio of PD-L1⁺ TAMs to total TAMs for individual cases (●) displayed according to whether the TAMs are located <75 or ≥75 μm from malignant B cells. For the box-and-whisker plots (C, F, I) the median (line), 25th and 75th percentiles (boxes), and extreme values (whiskers) are indicated. Statistical comparisons by 2-sided Mann-Whitney U test, Wilcoxon matched-pairs signed rank test, and the Kruskal-Wallis test with Dunn’s test for multiple comparisons.

Figure 3.

Malignant B cells are in frequent contact with PD-1⁺ T cells. (A) Representative high-power mIF image demonstrating PD-1 expression on the surface of CD3⁺ T cells in contact with a PAX5⁺ B cells). (B) Cell-cell contact map of the mIF image shown in panel A (same scale). PD-1⁺ T cells, PD-1⁻ T cells, and other cells (solid gray) in contact with malignant cells and PD-1⁺ T cells and PD-1⁻ T cells (outlined) not in contact with malignant cells. (C) Proportion of malignant cell contacts for individual cases (●) displayed according to the proportion of PD-1⁺ T-cell contacts vs PD-1⁻ T-cell contacts (n = 26). (D) Low-power, maximum field-of-view mIF image of TCRLBCL in panel A: PD-1 on CD3⁺ T cells intermixed with PAX5⁺ B cells. (E) Cell phenotype map corresponding to the mIF image in panel D: PD-1⁺ T cells, PD-1⁻ T cells, and malignant B cells. (F) T-cell density for individual cases (●) displayed according to the density of PD-1⁺ T cells and PD-1⁻ T cells. (G) Intermediate-power mIF image of the boxed region in panel D: PD-1 expression on CD3⁺ T cells intermixed with PAX5⁺ B cells. (H) Cell proximity map of the region highlighted in panel E: spatial distribution of PD-1⁺ T cells and PD-1⁻ T cells relative to malignant cells, showing T cells that are located <75 μm from malignant cells and those that are located ≥75 μm from malignant cells. (I) The ratio of PD-1⁺ T cells to total T cells for individual cases (●) displayed according whether the T cells are located <75 or ≥75 μm from malignant B cells. For the box-and-whisker plots (C,F,I), the median (line), 25th and 75th percentiles (boxes), and extreme values (whiskers) are indicated. Statistical comparisons by the 2-sided Mann-Whitney U test, Wilcoxon matched-pairs signed-rank test, and the Kruskal-Wallis test with Dunn’s test for multiple comparisons.

Figure 4.

Spatially defined immune cell neighborhoods distinguish TCRLBCL from cHL and DLBCL-NOS. (A) Representative cell phenotype map generated from a 5-marker mIF panel (CD3, PAX5, CD68, PD-1, and PD-L1), which highlights PD-L1⁺ TAMs, PD-L1⁻ TAMs, PD-1⁺ T cells, PD-1⁻ T cells, malignant B cells, and other cell types (original magnification, ×20). Circles surround the cellular neighborhoods of 3 representative “anchor” cells (yellow with red outline). Neighborhoods are defined by calculating the fractional composition of cell types <75 μm radial distance surrounding each anchor cell. (B) Neighborhood analysis workflow, which included unsupervised k-means clustering of all cell neighborhoods (n = 1 863 437) defined across cases of TCRLBCL (n = 26), cHL (n = 19), and DLBCL-NOS (n = 18) and assigned to 1 of 7 neighborhood clusters. (C) UMAP plot of pooled neighborhood data from TCRLBCL, cHL, and DLBCL-NOS cases with the 7 clusters color-coded. (D) Fractional composition of cell types within each of 7 defined clusters, color-coded. (E) UMAP plot from panel C colored according to the underlying pathologic diagnosis and demonstrating segregation of cell-neighborhoods from cases of TCRLBCL (red), cHL (blue), and DLBCL-NOS (black) into distinct regions. (F) Projections of color-coded neighborhoods (according to panel C) onto the original mIF images for representative cases of TCRLBCL (left, same case as shown in panel A), cHL (top right panel), and DLBCL-NOS (bottom right panel). Malignant cells are denoted by circles. (G-I) Percentage of global and neighborhood-specific cellularity in TCRLBCL comprised of malignant B cells (G), PD-L1⁺ TAMs (H), and PD-1⁺ T cells (I). (G-I) Individual cases (●) and the median (line), 25th and 75th percentiles (boxes), and extreme values (whiskers) are indicated. Statistical comparisons by the Friedman test with Dunn’s test for multiple comparisons. *P < .05; **P < .01; ****P < .0001.