The Ratio of FOXP3+ Regulatory T Cells to Granzyme B+ Cytotoxic T

Hematopathology / FOXP3+ T CELL–GRANZYME B+ T/NK CELL RATIO

The Ratio of FOXP3+ Regulatory T Cells to Granzyme B+ Cytotoxic T/NK Cells Predicts Prognosis in Classical Hodgkin Lymphoma and Is Independent of bcl-2 and MAL Expression

Todd W. Kelley, MD,1 Brad Pohlman, MD,2 Paul Elson, ScD,3 and Eric D. Hsi, MD1

Key Words: FOXP3; Regulatory T-cells; Granzyme B; Cytotoxic T/NK cells; Hodgkin lymphoma; bcl-2; MAL Downloaded from https://academic.oup.com/ajcp/article/128/6/958/1765062 by guest on 30 September 2021

DOI: 10.1309/NB3947K383DJ0LQ2

Abstract Classical Hodgkin lymphoma (cHL), a neoplasm of ger- We studied the prognostic importance of tumor- minal center B-cell derivation,1 accounts for approximately infiltrating regulatory T lymphocytes (Tregs) and 10% of all lymphomas. Hodgkin–Reed-Sternberg (HRS) cells, cytotoxic T/NK lymphocytes (CTLs) in 98 diagnostic the neoplastic component of cHL, usually constitute less than biopsy specimens from patients with classical Hodgkin 1% of the cellularity of the tumor. The majority of the cells in lymphoma (cHL). Immunohistochemical analysis was the tumor are reactive and consist predominantly of T cells performed for FOXP3 to identify Tregs and for with variable numbers of other inflammatory cells, such as granzyme B (GrB) to identify activated CTLs. Failure- macrophages, neutrophils, eosinophils, and plasma cells. free survival (FFS) and overall survival (OS) were Clinically, cHL has a favorable prognosis with overall survival clinical end points. Patients with fewer than 25 of approximately 80% among all patients using modern thera- FOXP3+ cells per high-power field (HPF) had a mean py. Because of this success and the young age of most patients, ± SD 5-year FFS of 64% ± 7% vs 85% ± 5% for recent efforts have been aimed at reducing therapy to avoid late patients with 25 or more FOXP3+ cells/HPF (P = .05). complications. However, a significant minority of patients will A FOXP3/GrB ratio of 1 or less was associated with have a progressive or relapsing clinical course despite aggres- poor FFS (46% ± 10% vs 86% ± 4%; P < .001) and sive treatment and eventually die of the lymphoma. OS (67% ± 10% vs 93% ± 3%; P < .001). When prior Prognostic scoring systems such as the international prog- available MAL and bcl-2 expression data were included nostic score (IPS) have been developed. The IPS risk stratifies in a multivariate analysis of all clinical and biologic patients according to 7 high-risk features, including hypoalbu- factors, a FOXP3/GrB ratio of 1 or less and tumor cell minemia, anemia, leukocytosis, lymphopenia, stage IV, male expression of MAL and bcl-2 all independently predicted sex, and age of 45 years or older.2 Biologic markers may give poor FFS. This demonstrates the importance of evaluating additional information that results in improved patient risk strat- tumor cell markers and the tumor immune infiltrate ification, and several have recently been identified. For example, when considering biologic prognostic markers in cHL. features of the HRS cells, such as bcl-2 or MAL expression, have been shown to be independent indicators of a poor prognosis.3-5 Characteristics of the nonneoplastic immune cell infiltrate within the lymphoma tumor bed may also be prognosti- cally important. The nature of the infiltrate may be influenced by various cytokines or chemokines elaborated by the tumor cells6-8 and by latent Epstein-Barr virus (EBV) infection.9 In particular, characteristics of the background immune response have been shown to be important in follicular lymphoma.10 Indeed, gene expression profiling studies of cHL also suggest that genes related to the tumor microenvironment are important

958 Am J Clin Pathol 2007;128:958-965 © American Society for Clinical Pathology 958 DOI: 10.1309/NB3947K383DJ0LQ2 Hematopathology / ORIGINAL ARTICLE in predicting prognosis.11 Interest has grown in the role of reg- Paraffin-embedded tissue blocks, obtained at the time of the ulatory T cells (Tregs) in tumor immunity. They have been initial diagnostic procedure, were retrieved from the archives shown to be important modulators of the immune response in of the Division of Pathology and Laboratory Medicine, the setting of transplantation,12-15 autoimmune disease,16 Cleveland Clinic, Cleveland, OH. TMAs were constructed infection,17,18 and neoplasia.19 using two to four 1.0-mm cores per case. In this study, we sought to evaluate the T/NK-cell subset Immunohistochemical staining was performed using anti- composition of cHL cases from our institution. Specifically, bodies to FOXP3 (clone 22510, Abcam, Cambridge, England) we correlated the number of intratumoral Tregs (as deter- to identify regulatory T cells and GrB (clone GrB-7, Accurate mined by FOXP3 expression) and the number of activated Chemical, Westbury, NY) to identify activated CTLs. Staining granzyme B (GrB)+ cytotoxic T/NK lymphocytes (CTLs) was performed using a Ventana Discovery automated with clinical factors and outcome. Furthermore, we also eval- immunostaining system (Ventana, Tucson, AZ). Data were Downloaded from https://academic.oup.com/ajcp/article/128/6/958/1765062 by guest on 30 September 2021 uated the ratio of tumor infiltrating FOXP3+ cells to GrB+ obtained by counting 5 separate 1,000× high-power fields cells (FOXP3/GrB ratio) in predicting survival. (HPFs) and calculating the mean number of positively staining cells per HPF. Counts were performed using an Olympus CX31 microscope (Olympus America, Center Valley, PA). Materials and Methods Hypocellular, fibrous bands associated with the nodular sclerosis subtype were avoided. Appropriate positive and negative Case Selection control stains were generated and evaluated. We identified 98 consecutive cases with paraffin-embedded For quality control purposes and to evaluate interobserv- archival tissue from the initial diagnostic procedure and adequate er reproducibility, 11 (11%) of 98 cases were recounted by a clinical follow-up for inclusion in this study. HIV+ cases were second observer. A comparison of the 2 data sets showed excluded. All patients were initially treated with curative intent excellent reproducibility (FOXP3, Spearman correlation coef- according to the standard of care at the time of diagnosis. ficient, 0.93; P < .001; GrB, Spearman correlation coefficient, However, owing to the long study period (between January 1988 0.94; P < .001). The FOXP3/GrB ratio change did not change and December 2002), standards of care changed, and, conse- significantly in any of the recounted cases. TMAs were qual- quently, the study patients did not receive uniform treatment. The itatively evaluated (positive or negative) for MAL and bcl-2 general treatment regimens were as follows: patients with early- immunoreactivity in HRS cells as described previously.3,5 stage disease received radiation therapy (RT) alone or combination chemotherapy that may have also included RT. Patients with Statistical Analysis advanced-stage disease received combination chemotherapy with OS was defined as the period from initial diagnosis to or without RT. Patients in whom treatment failed after primary death of any cause. FFS was defined as time from initial diag- RT received combination chemotherapy. Patients who experi- nosis to progression of cHL or death of any cause. OS and enced treatment failure after initial chemotherapy received sal- FFS were summarized by using the method of Kaplan and vage chemotherapy and may have also received high-dose thera- Meier. The Wilcoxon rank sum test was used to evaluate asso- py with autologous stem-cell transplantation. These cases have ciations between intratumoral FOXP3+ cells or GrB+ cells been reported in earlier unrelated studies.3 and clinical factors (ie, sex, age, stage, IPS, presence of bulky The medical records were reviewed for clinical character- disease, and histologic subtype). The generalized Wilcoxon istics, including sex, age, stage, IPS, presence of bulky disease, rank sum was used to evaluate outcome. The Cox proportion- presence of B symptoms, number of involved sites, disease al hazards model, with stepwise variable selection, was used progression, and survival. Histologic features were reviewed; to simultaneously assess multiple factors. Significance was set all cases were diagnosed as classical Hodgkin lymphoma at a P value of .05 or less for univariate and multivariate analy- according to the World Health Organization classification ses. The association between the number of intratumoral scheme.20 Histologic subtype (nodular sclerosis, mixed cellu- FOXP3+ cells and GrB+ cells was evaluated using the larity, lymphocyte-rich, lymphocyte-depleted) was noted. Spearman rank correlation coefficient. For convenience, the Survival analysis was performed using failure-free survival numbers of intratumoral FOXP3+ cells and GrB+ cells were (FFS) and overall survival (OS) as clinical end points. dichotomized in the OS and FFS analyses using cut points derived from a recursive partitioning algorithm that identified Immunohistochemical Analysis an optimal FOXP3 cut point of 25 positive cells per HPF. The Tissue microarrays (TMAs) have been shown to be a power- optimal GrB cut point was 7.5 positive cells per HPF. In addi- ful tool for the evaluation of large numbers of samples under iden- tion to evaluating the impact of the absolute number of tical immunohistochemical staining conditions with excellent cor- FOXP3+ and GrB+ cells present, the impact of their relative relation with whole tissue sections, including in lymphoma.21 presence was also assessed. Again using recursive partitioning,

© American Society for Clinical Pathology Am J Clin Pathol 2007;128:958-965 959 959 DOI: 10.1309/NB3947K383DJ0LQ2 959 Kelley et al / FOXP3+ T CELL–GRANZYME B+ T/NK CELL RATIO a FOXP3/GrB ratio of 0.85 was identified as an optimal cut to be 76% ± 4% and 86% ± 4%, respectively. Of 98 patients, point; however, for convenience, this was rounded to 1.0. 70 (71%) remained alive and free of progression at a median of 8.6 years (range, 11 months to 19.9 years). A shorter FFS was associated with an age of 45 years or older (P = .05) and an IPS Results of more than 2 (P = .05). A shorter OS was associated with an age of 45 years or older (P = .001). Sex, stage (I-II vs III-IV), Patient Characteristics presence of bulky disease, and histologic subtype were not sig- Of 98 patients, 50 (51%) were male, 75 (77%) were nificant factors in the univariate analysis of FFS or OS. younger than 45 years (median, 29 years; range, 4-84 years), and 78 (80%) had favorable IPS scores (≤2). Of the tumors, 78 FOXP3+ Tregs and Clinical Factors (80%) were the nodular sclerosis subtype. Of 98 patients, 20 FOXP3 immunostaining demonstrated nuclear staining in Downloaded from https://academic.oup.com/ajcp/article/128/6/958/1765062 by guest on 30 September 2021 (20%) have died and 28 (29%) have progressive diseases a subset of nonneoplastic lymphocytes associated with the and/or died. The 5-year mean ± SD FFS and OS are estimated tumor ❚Image 1❚. FOXP3+ cells did not appear to preferentially

A B

C D

❚Image 1❚ Immunohistochemical stains for FOXP3 and granzyme B (GrB). A, Hodgkin lymphoma case with few FOXP3+ lymphocytes (×400). B, Hodgkin lymphoma case with numerous FOXP3+ lymphocytes (×400). The staining pattern for this transcription factor is nuclear. The Hodgkin–Reed-Sternberg cells are negative. Immunohistochemical stains for GrB show a Hodgkin lymphoma case with few GrB+ T cells (C, ×400) and a case with many GrB+ T cells (D, ×400) and demonstrate the appropriate cytoplasmic staining pattern.

960 Am J Clin Pathol 2007;128:958-965 © American Society for Clinical Pathology 960 DOI: 10.1309/NB3947K383DJ0LQ2 Hematopathology / ORIGINAL ARTICLE localize to HRS cells. All but 1 case (1%) had at least some microenvironment were significantly associated with male sex FOXP3+ T cells in the tumor (range, 0-152 FOXP3+ (P = .006), age of 45 years or older (P = .02), histologic sub- cells/HPF) ❚Table 1❚. Increased FOXP3+ Tregs were associat- type other than nodular sclerosis (P = .001), and, marginally, ed with age younger than 45 years (P = .02), stage I or II dis- with absence of bulky disease (P = .06; Table 1). ease (P = .01), and, marginally, with an IPS of 2 or less (P = .06; Table 1). Regulatory and Cytotoxic T-Cell Subsets and Outcome in cHL GrB+ CTLs and Clinical Factors Cases with fewer than 25 FOXP3+ cells per HPF had a GrB immunostaining demonstrated a cytoplasmic pattern significantly shorter mean ± SD 5-year FFS (64% ± 7%) com- of staining in a subset of small lymphocytes associated with pared with cases with 25 or more FOXP3+ cells per HPF the tumor (Image 1). There was no tendency for GrB+ cells to (85% ± 5%, P = .05) ❚Table 2❚. The number of intratumoral Downloaded from https://academic.oup.com/ajcp/article/128/6/958/1765062 by guest on 30 September 2021 ring HRS cells. All tumors had GrB+ cells present (range, 0.8- GrB+ CTLs was evaluated in the same manner. However, 120 cells/HPF; Table 1). Increased GrB+ CTLs in the tumor there were no significant associations between FFS and OS

❚Table 1❚ Correlation Between Intratumoral FOXP3+ Cells or GrB+ Cells and Clinical Factors*

Factor No. (%) of Cases FOXP3+ Cells/HPF Median P† GrB+ Cells/HPF Median P†

Sex .82 .006 Male 50 (51) 41.8 ± 35.9 32.3 23.2 ± 24.7 18.3 Female 48 (49) 41.6 ± 32.1 38.0 12.1 ± 9.8 10.1 Age (y) .02 .02 <45 75 (77) 46.2 ± 35.1 41.4 16.7 ± 20.5 10.4 ≥45 23 (23) 26.9 ± 24.9 16.4 21.4 ± 16.4 18.2 Stage .01 .37 I or II 62 (63) 47.2 ± 32.5 41.2 16.7 ± 17.9 12.1 III or IV 36 (37) 32.1 ± 34.5 18.0 19.7 ± 22.4 15.8 IPS .06 .38 ≤2 78 (80) 45.4 ± 35.6 38.8 16.7 ± 17.6 12.1 >2 20 (20) 27.1 ± 21.0 21.3 22.2 ± 26.3 14.2 Bulky disease‡ .18 .06 No 56 (58) 35.9 ± 29.2 25.2 21.1 ± 23.1 16.1 Yes 40 (42) 47.8 ± 38.2 41.8 13.7 ± 12.8 10.2 Histologic features .68 .001 Nodular sclerosis 78 (80) 42.4 ± 36.0 36.4 14.9 ± 16.4 10.1 Other 20 (20) 39.1 ± 24.4 33.2 29.1 ± 26.7 25.3

GrB, granzyme B; HPF, high-power field; IPS, international prognostic score. * The presence of B symptoms and number of involved sites did not correlate with FOXP3 or GrB results. Data are given as mean ± SD unless otherwise indicated. † Wilcoxon rank sum test. Significant values are in bold type. ‡ Data are missing for 2 patients.

❚Table 2❚ Correlation Between Intratumoral FOXP3+ Cells, GrB+ Cells, and FOXP3/GrB Ratio With Failure-Free Survival and Overall Survival*

Failure-Free Survival Overall Survival

No. (%) No. (%) No. (%) Factor of Cases of Failures 5-y P† of Failures 5-y P†

No. of FOXP3+ cells/HPF .05 .10 <25 43 (44) 19 (44) 64% ± 7% 14 (33) 81% ± 6% ≥25 55 (56) 9 (16) 85% ± 5% 6 (11) 91% ± 4% No. of GrB+ cells/HPF .21 .18 <7.5 32 (33) 5 (16) 84% ± 7% 3 (9) 90% ± 5% ≥7.5 66 (67) 23 (35) 72% ± 6% 17 (26) 84% ± 5% FOXP3/GrB ratio <.001 <.001 ≤1.0 24 (24) 16 (67) 46% ± 10% 13 (54) 67% ± 10% >1.0 74 (76) 12 (16) 86% ± 4% 7 (9) 93% ± 3%

GrB, granzyme B; HPF, high-power field. * The 5-year survival rates are given as mean ± SD. † Generalized Wilcoxon rank sum test. Significant values are in bold type.

© American Society for Clinical Pathology Am J Clin Pathol 2007;128:958-965 961 961 DOI: 10.1309/NB3947K383DJ0LQ2 961 Kelley et al / FOXP3+ T CELL–GRANZYME B+ T/NK CELL RATIO and the number of GrB+ cells present (Table 2). There was Multivariate analysis, in which all clinical and pathologic also no association between the number of intratumoral factors in Table 1 were considered, showed that a low FOXP3+ Tregs and the number of intratumoral GrB+ CTLs FOXP3/GrB ratio was the most significant independent pre- (Spearman r = 0.11; P = .27). dictor of shorter FFS (P < .001; hazard ratio [HR], 5.0) and Because FOXP3 and GrB seem to identify functionally OS (P < .001; HR, 4.17). Other factors identified as independ- distinct populations of immunosuppressive and activated cyto- ently associated with shorter FFS were bulky disease (P = .03; toxic effector lymphocytes, the relative proportion of these HR, 2.43) and an IPS of more than 2 (P = .05; HR, 2.37). An subsets and their relationship to patient outcome was also age of 45 years or older was marginally associated with short- examined. The results of this analysis indicated that the ratio er OS (P = .08; HR, 2.28). The FFS and OS Kaplan-Meier of FOXP3+ T cells to GrB+ T/NK cells was highly correlated curves stratified by the number of intratumoral FOXP3+ cells with outcome. Cases with ratios of more than 1 had signifi- present and the FOXP3/GrB ratio are shown in ❚Figure 1❚. Downloaded from https://academic.oup.com/ajcp/article/128/6/958/1765062 by guest on 30 September 2021 cantly longer FFS and OS compared with cases with ratios of Although assessment of intratumoral FOXP3+ Tregs seems 1 or less (P < .001; Table 2). to predict outcome, it is the combined evaluation of Tregs and

A B

1.0 * 1.0 * * 0.9 ≥25 FOXP3+ cells/HPF (n = 55) 0.9 FOXP3/GrB ratio >1 (n = 74) 0.8 0.8 0.7 *** 0.7 ********<25 FOXP3+ cells/HPF (n = 43) 0.6 ** * 0.6 FOXP3/GrB ratio ≤1 (n = 24) 0.5 * 0.5 0.4 Proportion ** 0.4 Proportion 0.3 0.3 0.2 0.2 0.1 0.1 0.0 0.0 012345678910 012345678910 Years Years

C D

1.0 * 1.0 ≥25 FOXP3+ cells/HPF (n = 55) FOXP3/GrB ratio >1 (n = 74) 0.9 0.9 ****** 0.8 ** 0.8 *** <25 FOXP3+ cells/HPF (n = 43) 0.7 *** * 0.7 FOXP3/GrB ratio ≤1 (n = 24) 0.6 * 0.6 0.5 *** 0.5 Proportion Proportion 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0.0 0.0 012345678910 012345678910 Years Years

❚Figure 1❚ Survival curves based on infiltrating FOXP3+ T cells and the intratumoral FOXP3/granzyme B (GrB) ratio. Failure-free (A) and overall (C) survival for patients with ≥25 intratumoral FOXP3+ regulatory T cells per high-power field (HPF) (solid line) and patients with <25 FOXP3+ cells/HPF (broken line). * Cases with an unfavorable low FOXP3 count (<25/HPF) but a favorable high FOXP3/GrB ratio (>1). Failure-free (B) and overall (D) survival for patients with an intratumoral FOXP3/GrB ratio of >1 (solid line) and ≤1 (broken line). Cases with low intratumoral FOXP3 counts but high FOXP3/GrB ratios (designated * in A and C) are more appropriately segregated according to FOXP3/GrB ratios. A, P = .05. B, P < .001. C, P = .1. D, P < .001.

962 Am J Clin Pathol 2007;128:958-965 © American Society for Clinical Pathology 962 DOI: 10.1309/NB3947K383DJ0LQ2 Hematopathology / ORIGINAL ARTICLE activated CTLs that is most important. This is borne out by In support of the prognostic usefulness of this system, survival detailed analysis showing that that there is a subset of patients with curves of the resulting 3 patient subgroups show excellent sep- generally favorable outcome who would be misclassified into the aration (P < .001) ❚Figure 2❚. unfavorable prognosis group based on absolute numbers of Tregs alone (<25 FOXP3+ cells/HPF) but who were more appropriately segregated into the favorable prognosis group based on Discussion FOXP3/GrB ratio (>1). These cases are highlighted in Figures 1A and 1C with asterisks. Notably, all cases with a FOXP3 count of In this retrospective study of 98 newly diagnosed cHL 25 or more cells per HPF had a FOXP3/GrB ratio of more than 1. cases, we have shown that there is important prognostic information gained from evaluating intratumoral T/NK-cell subsets Combined Analysis of FOXP3/GrB Ratio With Tumor and, in particular, the ratio of FOXP3+ Tregs to GrB+ CTLs. Downloaded from https://academic.oup.com/ajcp/article/128/6/958/1765062 by guest on 30 September 2021 Cell MAL and bcl-2 Expression Our results are similar to those obtained by Alvaro et al,22 who Data for expression of MAL and bcl-2 in HRS cells have showed that a higher percentage of Tregs in the tumors of been published previously3,5 and were available for 81 of the patients with cHL was a favorable prognostic marker and that 98 patients in this series. The clinical characteristics of this a subgroup of patients with a high percentage of FOXP3+ subset were not significantly different from the overall series of cells and a low percentage of T cells expressing TIA-1, a 98 patients. In univariate analysis, MAL expression was asso- pan–cytotoxic T-cell marker, had a favorable prognosis. In ciated with decreased FFS (P = .002) and OS (P = .002), keeping with our findings, that study did not identify statisti- whereas bcl-2 expression was not significantly associated with cally significant differences in outcome based on intratumoral outcome (P > .10). In multivariate analysis of this subgroup, GrB content alone. However, the study did not evaluate ratios both of these factors were considered along with the of FOXP3 to TIA-1 or GrB. Indeed, an earlier study found that FOXP3/GrB ratio, age, sex, IPS, stage, presence of bulky dis- a high percentage of intratumoral GrB+ lymphocytes was an ease, and histologic subtype. A FOXP3/GrB ratio of 1 or less, unfavorable prognostic marker in patients with cHL.23 Taken expression of MAL, and expression of bcl-2 were the only sig- together, the trends of these 2 previous studies and the current nificant predictors of an unfavorable FFS ❚Table 3❚. A one are similar and suggest that the presence of increased FOXP3/GrB ratio of 1 or less and expression of MAL were the intratumoral CTLs is associated with a relatively poor progno- only significant predictors of poor OS (data not shown). A sis, whereas increased FOXP3+ Tregs are associated with a point system for assigning risk (low, intermediate, or high) was more favorable outcome. developed based on the significant predictors of FFS in multi- In an independent data set, we have confirmed the some- variate analysis (FOXP3/GrB ratio, MAL, and bcl-2) ❚Table 4❚. what counterintuitive observation that increased intratumoral FOXP3+ Tregs are associated with an improved outcome in

❚Table 3❚ Multivariate Analysis of Failure-Free Survival Low-risk score (n = 32) Parameter Hazard 1.0 * Factor Estimate ± SE Ratio P 0.9 FOXP3/granzyme B ratio 1.75 ± 0.44 4.53 <.001 0.8 Intermediate risk score (n = 22) (≤1 vs >1) 0.7 MAL (positive vs negative) 1.09 ± 0.43 2.97 .01 bcl-2 (positive vs negative) 1.07 ± 0.56 2.91 .05 0.6 High-risk score (n = 27)

* Wald test. 0.5 0.4 Proportion 0.3 ❚Table 4❚ 0.2 Risk Groups in 81 Cases Based on FOXP3/Granzyme B Ratio, MAL Expression, and bcl-2 Expression 0.1 0.0 5-y Failure-Free 012345678910 Risk Group* No. (%) of Cases Survival ± SE Years Low risk, 0 points 32 (40) 97% ± 5% ❚ ❚ Intermediate risk, 1 point 22 (27) 72% ± 10% Figure 2 Survival curves based on a prognostic scoring High risk, ≥2 points 27 (33) 48% ± 10% system (outlined in Table 4) incorporating characteristics of the tumor microenvironment (FOXP3/granzyme B ratio) and the * Points were determined as follows: FOXP3, >1.0, 0 points; ≤1.0, 2 points; MAL, negative, 0 points; positive, 1 point; bcl-2, negative, 0 points; positive, 1 point. Hodgkin–Reed-Sternberg cells (MAL/bcl-2 expression). P < .001.

© American Society for Clinical Pathology Am J Clin Pathol 2007;128:958-965 963 963 DOI: 10.1309/NB3947K383DJ0LQ2 963 Kelley et al / FOXP3+ T CELL–GRANZYME B+ T/NK CELL RATIO cHL. Furthermore, we have extended these observations to not Tregs because there was no correlation between FOXP3 show that a predominance of intratumoral FOXP3+ T cells to expression and GrB. However, expression of GrB by a subset GrB+ cytotoxic T/NK cells, as manifested by a FOXP3/GrB of intratumoral Tregs cannot be definitively ruled out. Further ratio of more than 1, seems to be a more powerful predictor of investigation will be necessary to make any firm conclusions. survival. We have also shown that there are associations Our data indicate that while increased intratumoral Tregs between these 2 T-cell subsets and clinical factors at diagno- confer a favorable prognosis, the additional information pro- sis: FOXP3 with favorable clinical characteristics (younger vided by GrB assessment allows for the identification of a age and lower stage) and GrB with unfavorable clinical char- subset of patients with low FOXP3 counts but a FOXP3/GrB acteristics (male sex and older age). ratio of more than 1 and a prognosis similar to that of the high The observations in cHL are contrary to expectations FOXP3 subset. These patients would have been misclassified from the paradigm of Tregs in tumor immunology and high- based on FOXP3 analysis alone (cases identified by asterisks Downloaded from https://academic.oup.com/ajcp/article/128/6/958/1765062 by guest on 30 September 2021 light potential differences in their role in cHL vs in solid in Figures 1A and 1C). It is unclear why reduced GrB+ CTLs tumors. Early experiments in mice, which helped to define the are associated with a more favorable prognosis in the context existence of a CD4+/CD25+ suppressive T-cell subset, of the FOXP3/GrB ratio. The cases with a FOXP3/GrB ratio showed that depleting Tregs by injection of anti-CD25 mono- of more than 1, despite low FOXP3 numbers, may represent clonal antibodies in animals inoculated with syngeneic tumors cases in which Tregs are especially inhibitory to bystander led to tumor rejection.24,25 Consequently, intratumoral Tregs CTLs. Effective Treg function would be manifested by low were thought to suppress cytotoxic T-cell responses directed numbers of activated CTLs, reflected by low levels of GrB against the tumor cells, thereby allowing the tumor to escape expression. If this were the case, GrB expression levels might the host immune system, whereas the removal of Tregs simply represent a surrogate marker for the suppressive abili- seemed to result in an effective antitumor response. Clinical ty of local Tregs. By inference, intratumoral Tregs that were findings in patients with ovarian cancer seem to lend support particularly effective against bystander CTLs might be simi- for such a role. Increased Tregs as measured by FOXP3 larly effective in inhibiting the tumor cells themselves. This immunostaining or by reverse transcriptase–polymerase chain hypothesis remains to be tested. reaction for FOXP3 messenger RNA have been associated Characteristics of the tumor immune infiltrate with poor outcome.26,27 These results are the opposite of our (FOXP3/GrB ratio) and the HRS cells (MAL and bcl-2 observations in cHL. Thus, generalizations about the function- expression) are important in predicting prognosis in cHL. al role of Tregs in neoplasia may not be possible. Indeed, the Using these 3 biologic markers, we suggest a strategy for seg- role of Tregs may be tumor-specific, requiring empiric studies regating patients into low-, intermediate-, and high-risk sub- to determine any association with outcome. groups. These findings are preliminary and, because this is a Relatively little is known about Treg function in tumors in relatively small series of patients, they should be interpreted general and especially in cHL. There are several possible accordingly. However, if confirmed in an independent data set, mechanisms that may account for our observations. First, it is this strategy may be useful in assigning risk-adapted treatment conceivable that Treg numbers, or the FOXP3/GrB ratio, are to decrease the incidence of therapy-related adverse events. surrogate markers for an active or intact immune response, thus imparting a relatively favorable prognosis. Second, Treg- From the Departments of 1Clinical Pathology and 3Quantitative 2 mediated suppression of the background inflammation, which Health Sciences, Cleveland Clinic; and Hematologic Oncology and Blood Disorders, the Cleveland Clinic Taussig Cancer Center, is an integral part of cHL, may suppress production of Cleveland, OH. cytokines or other factors important for the proliferation and survival of HRS cells. Because HRS cells may express numer- Address reprint requests to Dr Hsi: Dept of Clinical Pathology, L11, Cleveland Clinic, 9500 Euclid Ave, Cleveland, OH ous surface receptors, including CD30, CD40, NOTCH1, 44195. interleukin (IL)-2R, IL-3R, IL-4R, IL-6R, IL-13R, and tumor necrosis factor receptor family members, the potential path- 28,29 ways are numerous. Another intriguing possibility is that References Tregs in the tumor directly kill or suppress HRS cells. Tregs have been shown to suppress bystander lymphocytes through 1. Marafioti T, Hummel M, Foss HD, et al. Hodgkin and Reed- Sternberg cells represent an expansion of a single clone multiple mechanisms that may involve transforming growth originating from a germinal center B-cell with functional factor β,30,31 a recently described B7 family member (B7- immunoglobulin gene rearrangements but defective H1),32 or GrB itself.33,34 Thus, they may similarly interact with immunoglobulin transcription. Blood. 2000;95:1443-1450. tumor cells in cHL. 2. Hasenclever D, Diehl V. A prognostic score for advanced Hodgkin’s disease: International Prognostic Factors Project on The role of GrB in Tregs is poorly understood. The intra- Advanced Hodgkin’s Disease. N Engl J Med. 1998;339:1506- tumoral GrB+ cells that we quantified in this study were likely 1514.

964 Am J Clin Pathol 2007;128:958-965 © American Society for Clinical Pathology 964 DOI: 10.1309/NB3947K383DJ0LQ2 Hematopathology / ORIGINAL ARTICLE

3. Sup SJ, Alemany CA, Pohlman B, et al. Expression of bcl-2 20. Jaffe ES, Harris NL, Stein H, et al, eds. Pathology and Genetics in classical Hodgkin’s lymphoma: an independent predictor of Tumours of Haematopoietic and Lymphoid Tissues. Lyon, of poor outcome. J Clin Oncol. 2005;23:3773-3779. France: IARC Press; 2001. World Health Organization 4. Rassidakis GZ, Medeiros LJ, Vassilakopoulos TP, et al. bcl-2 Classification of Tumours. expression in Hodgkin and Reed-Sternberg cells of classical 21. Hedvat CV, Hegde A, Chaganti RS, et al. Application of Hodgkin disease predicts a poorer prognosis in patients treated tissue microarray technology to the study of non-Hodgkin’s with ABVD or equivalent regimens. Blood. 2002;100:3935- and Hodgkin’s lymphoma. Hum Pathol. 2002;33:968-974. 3941. 22. Alvaro T, Lejeune M, Salvado MT, et al. Outcome in 5. Hsi ED, Sup SJ, Alemany C, et al. MAL is expressed in a Hodgkin’s lymphoma can be predicted from the presence of subset of Hodgkin lymphoma and identifies a population of accompanying cytotoxic and regulatory T cells. Clin Cancer patients with poor prognosis. Am J Clin Pathol. 2006;125:776- Res. 2005;11:1467-1473. 782. 23. Oudejans JJ, Jiwa NM, Kummer JA, et al. Activated cytotoxic 6. van den Berg A, Visser L, Poppema S. High expression of the T cells as prognostic marker in Hodgkin’s disease. Blood. CC chemokine TARC in Reed-Sternberg cells: a possible 1997;89:1376-1382. Downloaded from https://academic.oup.com/ajcp/article/128/6/958/1765062 by guest on 30 September 2021 explanation for the characteristic T-cell infiltrate in Hodgkin’s 24. Onizuka S, Tawara I, Shimizu J, et al. Tumor rejection by in lymphoma. Am J Pathol. 1999;154:1685-1691. vivo administration of anti-CD25 (interleukin-2 receptor 7. Skinnider BF, Mak TW. The role of cytokines in classical alpha) monoclonal antibody. Cancer Res. 1999;59:3128-3133. Hodgkin lymphoma. Blood. 2002;99:4283-4297. 25. Shimizu J, Yamazaki S, Sakaguchi S. Induction of tumor 8. Poppema S. Immunobiology and pathophysiology of Hodgkin immunity by removing CD25+CD4+ T cells: a common basis lymphomas. Hematology Am Soc Hematol Educ Program. between tumor immunity and autoimmunity. J Immunol. 2005:231-238. 1999;163:5211-5218. 9. Herbst H, Foss HD, Samol J, et al. Frequent expression of 26. Sato E, Olson SH, Ahn J, et al. Intraepithelial CD8+ tumor- interleukin-10 by Epstein-Barr virus–harboring tumor cells infiltrating lymphocytes and a high CD8+/regulatory T cell of Hodgkin’s disease. Blood. 1996;87:2918-2929. ratio are associated with favorable prognosis in ovarian cancer. 10. Dave SS, Wright G, Tan B, et al. Prediction of survival in Proc Natl Acad Sci U S A. 2005;102:18538-18543. follicular lymphoma based on molecular features of tumor- 27. Wolf D, Wolf AM, Rumpold H, et al. The expression of the infiltrating immune cells. N Engl J Med. 2004;351:2159-2169. regulatory T cell–specific forkhead box transcription factor 11. Sanchez-Aguilera A, Montalban C, de la Cueva P, et al. FoxP3 is associated with poor prognosis in ovarian cancer. Tumor microenvironment and mitotic checkpoint are key Clin Cancer Res. 2005;11:8326-8331. factors in the outcome of classic Hodgkin lymphoma. Blood. 28. Re D, Thomas RK, Behringer K, et al. From Hodgkin disease 2006;108:662-668. to Hodgkin lymphoma: biologic insights and therapeutic 12. Rezvani K, Mielke S, Ahmadzadeh M, et al. High donor potential. Blood. 2005;105:4553-4560. FOXP3-positive regulatory T-cell (Treg) content is associated 29. Khan G. Epstein-Barr virus, cytokines, and inflammation: a with a low risk of GVHD following HLA-matched allogeneic cocktail for the pathogenesis of Hodgkin’s lymphoma? Exp SCT. Blood. 2006;108:1291-1297. Hematol. 2006;34:399-406. 13. Barao I, Hanash AM, Hallett W, et al. Suppression of natural 30. Nakamura K, Kitani A, Strober W. Cell contact-dependent killer cell–mediated bone marrow cell rejection by immunosuppression by CD4(+)CD25(+) regulatory T cells CD4+CD25+ regulatory T cells. Proc Natl Acad Sci U S A. is mediated by cell surface–bound transforming growth factor 2006;103:5460-5465. beta. J Exp Med. 2001;194:629-644. 14. Zorn E, Kim HT, Lee SJ, et al. Reduced frequency of FOXP3+ 31. Nakamura K, Kitani A, Fuss I, et al. TGF-beta 1 plays an CD4+CD25+ regulatory T cells in patients with chronic graft- important role in the mechanism of CD4+CD25+ regulatory versus-host disease. Blood. 2005;106:2903-2911. T cell activity in both humans and mice. J Immunol. 15. Sakaguchi S, Sakaguchi N, Shimizu J, et al. Immunologic 2004;172:834-842. tolerance maintained by CD25+ CD4+ regulatory T cells: 32. Yang ZZ, Novak AJ, Stenson MJ, et al. Intratumoral their common role in controlling autoimmunity, tumor CD4+CD25+ regulatory T-cell–mediated suppression of immunity, and transplantation tolerance. Immunol Rev. infiltrating CD4+ T cells in B-cell non-Hodgkin lymphoma. 2001;182:18-32. Blood. 2006;107:3639-3646. 16. Bacchetta R, Passerini L, Gambineri E, et al. Defective 33. Gondek DC, Lu LF, Quezada SA, et al. Cutting edge: contact- regulatory and effector T cell functions in patients with mediated suppression by CD4+CD25+ regulatory cells FOXP3 mutations. J Clin Invest. 2006;116:1713-1722. involves a granzyme B–dependent, perforin-independent 17. Marshall NA, Vickers MA, Barker RN. Regulatory T cells mechanism. J Immunol. 2005;174:1783-1786. secreting IL-10 dominate the immune response to EBV latent 34. Zhao DM, Thornton AM, DiPaolo RJ, et al. Activated membrane protein 1. J Immunol. 2003;170:6183-6189. CD4+CD25+ T cells selectively kill B lymphocytes. Blood. 18. Robertson SJ, Hasenkrug KJ. The role of virus-induced 2006;107:3925-3932. regulatory T cells in immunopathology. Springer Semin Immunopathol. 2006;28:51-62. 19. Knutson KL, Disis ML, Salazar LG. CD4 regulatory T cells in human cancer pathogenesis. Cancer Immunol Immunother. 2007;56:271-285.