Classical Hodgkin Lymphoma with Reduced β₂M/MHC Class I Expression Is Associated with Inferior Outcome Independent of 9p24.1 Status

Primary classical Hodgkin lymphomas (cHL) are comprised of a mixed infiltrate of inflammatory/immune cells and small numbers of malignant Hodgkin Reed–Sternberg (HRS) cells (1). HRS cells evade antitumor immunity by multiple mechanisms, including perturbed antigen presentation and augmented PD-1 signaling, that are attributable, in large part, to defined genetic lesions. Recent studies identified HRS cell lines and primary HRS cells with B2M mutations that disrupt expression of the β₂-microglobulin (β₂M)/MHC class I dual protein complex at the cell surface (2). Separate studies defined inactivating alterations of the MHC class II transactivator, CIITA, in cHL (3). Finally, in virtually all cases of cHL, HRS cells have copy number alterations of 9p24.1, a region that includes CD274 (PD-L1) and PDCD1LG2 (PD-L2), and contributes to robust expression of the PD-1 ligands (4).

The biological and clinical significance of the varied immune evasion strategies in cHL are still being elucidated. An initial study on clinical samples from patients with cHL suggested that loss of β₂M protein was, paradoxically, associated with improved clinical outcome (2). In contrast, our recent analysis of diagnostic biopsy samples from patients treated with Stanford V, a standard induction regimen comparable with ABVD (Adriamycin, bleomycin, vinblastine, and dacarbazine), revealed highly significant associations between amplification of PD-L1/PD-L2 and advanced stage disease at presentation and inferior progression-free survival (PFS; ref. 4). The genetic basis of PD-1–mediated immune evasion likely explains the efficacy of PD-1 blockade in cHL. In pilot studies and confirmatory phase II trials, patients with relapsed/refractory cHL treated with PD-1–blocking antibodies had response rates of 65% to 87% and long-lasting remissions (5–7).

We sought to clarify the prognostic significance of perturbed MHC class I and MHC class II antigen presentation by HRS cells and explore the relationship between antigen presentation and 9p24.1 genetic alterations in cHL. Herein, we characterize cell-surface β₂M, MHC class I, and MHC class II expression in a series of uniformly treated cHL patients with long-term follow-up and defined 9p24.1/PD-L1/PD-L2 alterations (4).

The samples used in this study were from a previously described 108-patient series, with Institutional Review Board approval (4). Formalin-fixed paraffin-embedded (FFPE) tumor samples and select pathologic and clinical data from 108 patients with newly diagnosed cHL, treated with the Stanford V chemotherapy regimen + modified involved field radiation, were obtained from the Stanford University (Stanford, CA; ref. 4). Median patient follow-up was 9 years. In this series, 9p24.1/PD-L1/PD-L2 alterations were characterized by FISH as described previously (4, 8).

Immunohistochemical staining for β₂M (Dako, A0072, 1:6,000), MHC class I (Abcam, EMR8-5, 1:6,000), and MHC class II (Dako, CR3/43 M0775, 1:750) was performed using an automated staining system (Bond III, Leica Biosystems) according to the manufacturer's protocol following antigen retrieval (Bond, ER2 solution). Hematoxylin counterstain was subsequently applied.

Staining on two or more adjacent HRS cells was used to determine membrane expression of the antigen presentation proteins, β₂M, MHC class I, and MHC class II. HRS cells were evaluated for the presence of positive membrane staining of each biomarker. If present, the relative intensity of HRS cell membrane expression, relative to adjacent nonmalignant inflammatory cells, was determined. β₂M, MHC class I, and MHC class II IHC were optimized on a test series of cHLs. In a subset of cases, β₂M, MHC class I, and/or MHC class II expression on the vast majority of HRS membranes was equivalent to or greater than that observed on adjacent, nonmalignant inflammatory cells (Supplementary Fig. S1; case 1 for β₂M and MHC class I, case 2 for MHC class II). In a subset, no β₂M, MHC class I, and/or MHC class II expression was detected on the vast majority of HRS cells or cell membranes, despite appropriate internal controls (Supplementary Fig. S1; case 3 for β₂M and MHC class I, case 1 for MHC class II). An additional subset of cases exhibited heterogeneous HRS cell staining, including HRS cells with unequivocally positive but reduced membrane staining, relative to adjacent nonmalignant cells, and those with a combination of reduced and complete loss of staining in a subset of cells (Supplementary Fig. S1; case 2 for β₂M and MHC class I, case 3 for MHC class II).

We devised a 3-tier scoring system to categorize the predominant patterns of β2M, MHC class I, and MHC class II expression by HRS cells in each case. For cases categorized as positive, at least 90% of evaluable HRS cells showed positive membrane staining for the biomarker at levels equivalent to, or greater than, that of adjacent nonmalignant inflammatory cells. For cases categorized as negative, at least 90% of evaluable HRS cells showed no detectable membrane staining for the biomarker relative to nonmalignant inflammatory cells. For cases categorized as decreased, positive membrane staining of HRS cells was present and unequivocally reduced relative to surrounding cells, and/or positive staining was observed in less than 90% of evaluable HRS cells.

Stained slides from the clinical series were scored separately for each of the markers by two independent hematopathologists (S.J. Rodig and G.S. Pinkus), blinded to the clinical data. Kappa statistics were then generated on the two sets of independent scores (Supplementary Table S1). Cases with scores that were found to be discordant between the two independent reviewers were reconciled in a consensus conference. The consensus score was used for the final analysis.

PFS, the time from diagnosis until first progression or death, censored at the time last known alive and progression free, was determined by the method of Kaplan and Meier and compared using log-rank tests. PFS is a preferred metric of response to first-line (induction) therapy, rather than overall survival, as patients who fail induction therapy routinely undergo salvage high-dose chemotherapy and stem cell rescue, a treatment that is curative in a subset of patients. Proportional hazards regression with Firth penalized likelihood were fit using disease stage, PD-L1/PD-L2 amplification, and β₂M, MHC class I, and MHC class II expression. Cox proportional hazards models were compared using likelihood ratio tests, and Wald P values were reported for covariates. Associations between continuous, nominal, and ordinal variables were assessed using Wilcoxon rank-sum, Fisher exact, and Kruskal–Wallis tests, respectively. All P values were two-sided; values ≤0.05 were considered statistically significant.

We optimized the immunohistochemical staining and scoring methods for β2M, MHC class I, and MHC class II on a small series of FFPE diagnostic tissue biopsies from patients with cHL (Supplementary Fig. S1). We next evaluated HRS cell expression of β₂M, MHC class I, and MHC class II in diagnostic biopsy specimens from 108 patients with cHL who received uniform treatment and had long-term follow-up (median, 9 years) and previously characterized 9p24.1/PD-L1/PD-L2 alterations (4). Each cHL was classified as positive, decreased, negative, or not assessable for HRS cell membrane expression of β₂M, MHC class I, and MHC class II. In each case, the HRS cell membrane expression was compared with that of nonmalignant cells within the same tissue section. Representative examples are shown in Fig. 1A. Cases included those with positive HRS cell membrane expression of β₂M, MHC class I, and MHC class II (case 1, Fig. 1A); decreased HRS cell membrane expression of β₂M, MHC class I, and MHC class II (case 2, Fig. 1A); and no HRS cell membrane expression of the three markers (case 3, Fig. 1A). Additional cases were positive for β₂M and MHC class I and negative for MHC class II (case 4, Fig. 1A), or negative for β₂M and MHC class I but positive for MHC class II (case 5, Fig. 1A).

The patterns of β₂M, MHC class I, and MHC class II expression in the cHL series are summarized in Fig. 1B. For β₂M, HRS cells were positive in 16% (17/108), decreased in 27% (29/108), negative in 53% (57/108), and unevaluable in 5% (5/108) of cases (Fig. 1B). For MHC class I, HRS cells were positive in 18% (19/108), decreased in 31% (34/108), negative in 47% (51/108), and unevaluable in 4% (4/108) of cases (Fig. 1B). The association between categories of cell-surface β₂M and MHC class I expression (positive, decreased, or absent) by HRS cells was highly significant across cases (Fig. 1B; P < 0.001), suggesting that B2M alterations might be a structural basis for MHC class I loss (2). Overall, HRS cells in less than 20% of the cases showed positive membrane β₂M or MHC class I staining.

MHC class II expression on HRS cells was positive in 31% (34/108), decreased in 37% (40/108), negative in 30% (32/108), and unevaluable in 2% (2/108) of cases (Fig. 1B). Among the cHLs that were positive for cell-surface MHC class II, only 12% (4/34 cases) were also positive for β₂M/MHC class I; the remainder had decreased/negative β₂M/MHC class I expression [88% (30/34 cases); Fig. 1B]. Consistent with these findings, MHC class II expression was not associated with either β₂M or MHC class I expression (Fig. 1B; P = 0.47 and P = 0.21, respectively).

We next examined the relationship between specific clinical features and β₂M, MHC class I, and MHC class II HRS cell expression in the patients in whom all three antigen presentation pathway components were evaluable (n = 103, Table 1). In this series, there was no significant association between age, stage, B symptoms, or bulky disease and altered β₂M, MHC class I, or MHC class II expression (Table 1). In contrast, there were significant associations between the mixed cellularity cHL (MCHL) subtype and positive β₂M and MHC class I expression, and between Epstein–Barr virus–positive (EBV⁺) cHL and positive β₂M and MHC class I expression (P < 0.001, Table 1), consistent with previous reports (9–12).

We next evaluated a potential association between β₂M, MHC class I, and MHC class II expression and outcome (PFS) in this series of uniformly treated cHL patients with long-term follow-up (Fig. 2). PFS was comparable in patients whose HRS cells had decreased or negative β₂M expression (Fig. 2A, left). However, patients whose HRS cells had decreased/negative β₂M expression had significantly shorter PFS than those whose HRS cells were positive for β₂M (P = 0.037; Fig. 2B, left). Similarly, PFS was comparable in patients with decreased or negative MHC class I expression in HRS cells (Fig. 2A, middle) but significantly shorter than that of patients with positive MHC class I expression in HRS cells (P = 0.031; Fig. 2B, middle). Consistent with these results, reduced (decreased/negative) expression of β₂M and MHC class I had adverse prognostic significance in univariate models (Table 2, P = 0.02). MHC class II expression by HRS cells was not significantly associated with PFS in these patients (P = 0.60; Fig. 2; Table 2). In addition, neither the cHL subtype [MCHL vs. nodular sclerosis Hodgkin lymphoma (NSHL)] nor EBV status was associated with PFS (not shown).

In this series of cHL patients, we found that 9p24.1/PD-L1/PD-L2 amplification was associated with advanced stage disease and inferior PFS (4). Given these findings, we next evaluated the prognostic significance of PD-L1/PD-L2 amplification in patients with defined HRS cell expression of β₂M/MHC class I (Fig. 2C). In patients with positive HRS cell expression of β₂M and MHC class I, PFS was not affected by 9p24.1 amplification, although only 3 patients had 9p24.1 amplification and positive β2M and MHC class I (Fig. 2C, left and middle). In marked contrast, 9p24.1 amplification adversely impacted PFS in patients whose HRS cells had decreased/negative β₂M/MHC class I expression (Fig. 2C, left and middle). Importantly, these results were highly significant despite the small numbers of 9p24.1, β₂M-positive, and MHC class I–positive patients (P = 0.014 and P = 0.013, respectively).

In multivariable models of (i) decreased/negative MHC class I expression and 9p24.1 amplification or (ii) decreased/negative MHC class I expression and advanced stage disease, both features retained adverse prognostic significance (Table 2). When all three features were included in the multivariable model, decreased/negative expression of MHC class I still retained adverse prognostic significance (Table 2, P = 0.05). Similar results were obtained in Cox models containing β₂M rather than MHC class I (Supplementary Table S2).

In this analysis, we found that (i) decreased or absent β₂M and MHC class I expression and decreased or absent MHC class II expression on HRS cells occur in approximately 80% and 70% of cHL cases, respectively; (ii) decreased/absent β₂M/MHC class I expression, but not decreased/absent MHC class II expression, is associated with shorter PFS; and (iii) the prognostic value of decreased/absent β₂M/MHC class I expression is independent of PD-L1/PD-L2 amplification and advanced stage disease.

The B2M subunit is required for assembly of MHC class I on the cell surface of nucleated cells. As expected, we observed a high concordance between β₂M and MHC class I IHC scores. For cases of cHL categorized as negative for β₂M expression, we found that loss of β₂M from the HRS cells was complete. In contrast, HRS cells in certain cases exhibited cytoplasmic expression of MHC class I in the absence of membranous staining (Fig. 1, cases #3 and #5 vs. case #4). These data suggest that loss of β₂M expression is the predominant mechanism for deficient cell-surface MHC class I protein expression in cHL, as previously suggested by genetic studies (2).

In contrast to prior reports in cHL, we observed unequivocally reduced, but not completely absent, β₂M, MHC class I, and class II membrane expression on HRS cells relative to normal, surrounding inflammatory cells in a significant subset (30%–40%) of cases (2, 13). We used a 3-tier scoring system, positive, decreased, and negative, to capture the heterogeneity of β₂M/MHC class I and MHC class II protein expression. Cases with reduced, but not completely absent, expression of the β₂M/MHC class I, and MHC class II proteins may have single-allele loss or a heterozygous-inactivating mutation in genes directly encoding these proteins (such as single-allele inactivation of B2M or single-copy loss of the HLA locus) or alterations in genes encoding critical transcriptional regulators (such as NLRC5 and CIITA; refs. 3, 14, 15). We also observed cases, classified as decreased, with positive expression in only a subset of HRS cells, suggesting that subclones with distinct characteristics can exist within a single case.

In contrast to a recent study (2), we found that cHL patients with decreased or negative β₂M expression have inferior outcomes (Fig. 2; Table 2). Unlike the prior study, patients in this series received uniform treatment at a single institution and had prespecified clinical follow-up for a median of 9 years. In further contrast, this series includes a representative mix of early- and advanced stage disease and NSHL and MCHL. As previously reported for cHL, we found that patients with advanced stage disease had inferior PFS, and patients with MCHL and/or EBV⁺ cHL had significantly higher β₂M/MHC class I expression on HRS cells (4, 9–12). Given these characteristics, we believe that the current series provides a comprehensive framework for analyzing the prognostic significance of β₂M, MHC class I, and MHC class II expression in newly diagnosed cHL patients treated with standard induction therapy. In addition, these data highlight the likely biological importance MHC class I–mediated antigen presentation by HRS cells to cytotoxic T cells for optimal clinical response to nonimmune therapy. That MHC class I expression is associated with a more favorable outcome to standard therapy in cHL is consistent with the findings in multiple other tumor types (16–19).

The paucity of β₂M/MHC class I expression on HRS cells also prompts speculation regarding additional mechanisms of action of PD-1 blockade in cHL. In other tumor types, clinical responses to standard chemotherapy and immunotherapies, including checkpoint blockade, have been associated with the presence of CD8⁺ T cells within the tumor microenvironment, underscoring the importance of β₂M/MHC class I–mediated antigen presentation by malignant cells (16, 20–22). Patients with relapsed/refractory cHL who are treated with PD-1–blocking antibodies have response rates of 65% to 87% and long-lasting remissions. Yet, the complete loss of MHC class I and β₂M expression by HRS cells in approximately one half of cHLs suggests that CD8⁺ cytotoxic T cell–mediated killing of HRS cells may not be the only mechanism of antitumor immunity augmented by PD-1 blockade. Given our findings, the significance of impaired antigen presentation by HRS cells on the quality and durability of clinical responses to PD-1 inhibitors will be of great interest.

M.A. Shipp reports receiving a commercial research grant from and is a consultant/advisory board member for Bristol-Myers Squibb. S.J. Rodig reports receiving a commercial research grant from and has received speakers bureau honoraria from Bristol-Myers Squibb. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.G.M. Roemer, R.H. Advani, A.H. Ligon, B. Chapuy, M.A. Shipp, S.J. Rodig

Development of methodology: M.G.M. Roemer, C.D. Carey, D. de Jong, M.A. Shipp, S.J. Rodig

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.G.M. Roemer, R.H. Advani, Y. Natkunam, C.F. Connelly, C.J. Pak, R.T. Hoppe, S.J. Rodig

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.G.M. Roemer, R.H. Advani, R.A. Redd, G.S. Pinkus, A.H. Ligon, C.D. Carey, B. Chapuy, R.T. Hoppe, D.S. Neuberg, M.A. Shipp, S.J. Rodig

Writing, review, and/or revision of the manuscript: M.G.M. Roemer, R.H. Advani, R.A. Redd, G.S. Pinkus, Y. Natkunam, A.H. Ligon, C.J. Pak, B. Chapuy, R.T. Hoppe, D.S. Neuberg, M.A. Shipp, S.J. Rodig

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.H. Advani, C.J. Pak, S.E. Daadi

Study supervision: R.H. Advani, M.A. Shipp, S.J. Rodig

This work was supported by the NIH (R01 CA161026; to M.A. Shipp), the Miller Family Fund(to M.A. Shipp), Leukemia & Lymphoma Society (to S.J. Rodig), International Immune Oncology Network of Bristol-Myers Squibb(to S.J. Rodig and M.A. Shipp), and the Center for Immuno-Oncology(to S.J. Rodig).