Cellular Senescence Markers P16 and P21 Are Predictors of Hodgkin

Published OnlineFirst July 21, 2015; DOI: 10.1158/1078-0432.CCR-15-0508

Biology of Human Tumors Clinical Cancer Research Cellular Senescence Markers p16INK4a and p21CIP1/WAF Are Predictors of Hodgkin Lymphoma Outcome Anna Calio1, Alberto Zamo1, Maurilio Ponzoni2, Maria Elisabetta Zanolin3, Andres J.M. Ferreri4, Serena Pedron1, Licia Montagna1, Claudia Parolini1, Vadim E. Fraifeld5, Marina Wolfson5, Hagai Yanai5, Giovanni Pizzolo6, Claudio Doglioni2, Fabrizio Vinante6, and Marco Chilosi1

Abstract

Purpose: There is evidence that Hodgkin Reed-Sternberg Results: Both molecules were independent prognostic factors. (HRS) cells in classical Hodgkin lymphoma (cHL) could dis- A positive staining of one of the two molecules in more than 30% play some molecular and morphologic markers of cellular HRS cells predicted a better EFS (P < 0.01). p16INK4a/p21CIP1/WAF1 senescence (CS). We hypothesized that CS mechanisms may together as a unique categorical variable (both <30%, either have potential prognostic relevance in cHL and investigated <30%, both 30%) sorted out three prognostic groups with whether the expression of the well-established CS biomarkers better, intermediate, or worse outcome either overall or within p21CIP1/WAF1 and p16INK4a by HRS cells might be predictive of I–II, bulky and advanced stages. The presence or the lack of the the probability of event-free survival (EFS). robust expression of p21CIP1/WAF1 and/or p16INK4a deﬁned the Experimental Design: The study analyzed a retrospective prognosis in our series. cohort of 147 patients and the results were validated on a cohort Conclusions: These ﬁndings point to (i) the relevance of of 91 patients independently diagnosed and treated in a different CS-related mechanisms in cHL, and to (ii) the prognostic value institution. p16INK4a and p21CIP1/WAF1 were categorized as dichot- of a simple, reproducible, and low-cost immunohistochemical omous variables (< or 30% of HRS cells at diagnosis) and evaluation of p16INK4a and p21CIP1/WAF1 expression. Clin Cancer evaluated in univariate and multivariate analysis. Res; 1–9. 2015 AACR.

Introduction chemotherapy refined in some instances by radiotherapy (1). However, nonresponder and early relapsed patients still represent Classical Hodgkin lymphoma (cHL) has an incidence of about a therapeutic challenge, and their prognosis remains poor despite 3 cases per 100,000 people per year and is, therefore, quite incremental treatment intensification (2). Why the majority of frequent among lymphomas. To date, available treatments for cHL patients behave so well, and why a small subclass of cHL is cHL achieve cure rates of 70% and 80%, based on standard refractory to treatment remain unanswered questions. Though some clinical and laboratory prognostic markers have been fig- ured out and tested (for example, we have previously described 1Department of Pathology and Diagnostic, Anatomic Pathology Sec- tion, University of Verona, Verona, Italy. 2Vita-Salute San Raffaele sCD30 as a neoplastic mass index and a reliable prognostic University, Pathology and Lymphoid Malignancies Units, IRCCS San marker; refs. 3, 4), it is quite surprising that so far there is no Raffaele Scientific Institute, Milan, Italy. 3Department of Public Health acknowledged biologic feature linked to the degree of malignant and Community Medicine, Unit of Epidemiology and Medical Statis- behavior of cHL (5–7). Indeed, the pathogenesis of cHL has been tics, University of Verona, Verona, Italy. 4Unit of Lymphoid Malignan- cies, Division of Onco-Hematological Medicine, Department of Onco- intensively investigated, revealing a biologic heterogeneity that Hematology, IRCCS San Raffaele Scientific Institute, Milan, Italy. 5The might provide useful information for clinical and prognostic case Shraga Segal Department of Microbiology, Immunology and Genetics, stratification. cHL is a lymphoma derived from B cells (8), Center for Multidisciplinary Research on Aging, Ben-Gurion University of the Negev, Beer Sheva, Israel. 6Department of Medicine, Hematol- characterized by rare tumor cells, the Hodgkin and Reed-Sternberg ogy Section, University of Verona, Verona, Italy. (HRS) cells, surrounded by an abundant reactive microenviron- Note: Supplementary data for this article are available at Clinical Cancer ment. The majority of cells in affected tissues of cHL are a mixed Research Online (http://clincancerres.aacrjournals.org/). infiltrate of different cell types, including macrophages, T and B A. Calio and A. Zamo equally contributed and share first authorship of this article. lymphocytes, plasma cells, neutrophils, eosinophils, and mast cells. Experimental and in situ observations suggest that HRS cells F. Vinante and M. Chilosi share senior authorship of this article. secrete a variety of cytokines, chemokines, and cellular messengers Corresponding Author: Fabrizio Vinante, University of Verona, Piazzale L.A. (9, 10) and actively attract many of the above cells. Scuro 10, 37134 Verona. Phone: 39-045-812-4420; Fax: 39-045-802-7488; Despite the clinical and morphologic heterogeneity of cHL, E-mail: [email protected] only two signaling pathways, JAK-STAT and NF-kB, essentially doi: 10.1158/1078-0432.CCR-15-0508 drive its pathogenesis (10). In cHL, the constitutive activation 2015 American Association for Cancer Research. of NF-kB and JAK-STAT can be caused by multiple genetic

www.aacrjournals.org OF1

Calio et al.

series of cHL cases, we investigated whether the presence of Translational Relevance CIP1/WAF1 INK4a p21 and p16 might be predictive of the response Hodgkin Reed-Sternberg (HRS) cells in classical Hodgkin to treatment in patients with cHL. lymphoma (cHL) frequently display cellular senescence (CS)–like features, including the expression of p21CIP1/WAF1 INK4a and p16 . We hypothesized that CS mechanisms may have Materials and Methods potential prognostic relevance in cHL and investigated Patients and samples whether the expression of p21CIP1/WAF1 and p16INK4a by HRS A retrospective study was performed in a primary cohort of cells might be powerful predictors of the disease outcome. In patients with cHL who underwent treatment between 1986 and this work, we described that both molecules were indepen- 2012 (n ¼ 147). These cases were collected from the archives of dent prognostic factors in a retrospective cohorts of 147 the Pathology Department of Verona University. A validation patients and in a validation cohort of 91 patients indepen- cohort of cHL patients (n ¼ 91) diagnosed between 1991 and dently diagnosed and treated in different institutions. These 2011 was collected from the archives of the Pathology Depart- findings point to the relevance of CS-related mechanisms in ment of the Vita Salute University in Milan. Case selection was cHL, and to the prognostic value of a simple, reproducible based on the availability of paraffin-embedded formalin-fixed and low-cost immunohistochemical evaluation of p16INK4a lymph node from the initial diagnosis in untreated patients and and p21CIP1/WAF1 expression. of clinical data. All samples were reviewed independently by two expert hematopathologists (M. Chilosi and A. Zamo). Diagnosis of cHL was established on the basis of the WHO 2008 criteria (23). Written informed consent was signed by the patients; when not possible (dead or unreachable patients), lesions interfering with these pathways, including genomic the study material was used after deidentification according to gains of NF-kB activators such as REL and NIK, inactivating the Italian law and with the approval of the institutional ethic mutations of NF-kBinhibitorssuchasIkBa and IkBe, and other committee. All procedures were done in accordance with the lesions causing JAK–STAT pathway dysregulation (10). In addi- Helsinki declaration. tion, dysregulation of both classical and alternative NF-kB pathways can be ascribed to the expression of viral transform- Clinical staging and treatment ing proteins (e.g., LMP1) in EBV-infected HRS in a large Patients were staged according to Ann Arbor criteria (24). number (nearly 40%) of cHL patients (10). Staging procedures included clinical examination, blood tests, Several studies conducted by us and others have demonstrat- chest X-rays, chest/abdomen CT, abdominal ultrasonography, ed that some genes involved in the cell-cycle control are and bone marrow biopsy. More recent patients were also staged dysregulated in cHL (11–14). Among them are the cyclin- by PET-CT. Bulky disease was defined as mediastinal mass exceed- dependent kinase inhibitors p21CIP1/WAF1 and p16INK4a,which ing one third of the thoracic diameter measured at the D5-D6 level are the important components of two major tumor-suppressor and/or extramediastinal mass 10 cm. Unfavorable stages includ- pathways, p53 and Rb (15). Interestingly, these inhibitors are ed bulky disease and III/IV stages. Patients received stage-directed among the main drivers of the irreversible and permanent cell- treatment based on 4 to 6 courses of ABVD chemotherapy with or cycle arrest observed in cellular senescence (CS), a protective without field radiotherapy (25, 26). Complete remission (CR) mechanism against unlimited cell proliferation that could be was defined as the complete disappearance of clinical, physical, induced by various stressful stimuli, including oxidative dam- radiologic, and biochemical abnormalities due to cHL. Partial age, telomere dysfunction, DNA damage, and aberrant onco- remission (reduction of at least 50% for at least 4 weeks of all gene activation (16). Moreover, it has been demonstrated that measurable masses with no new lesions appearing and no pro- senescent cells may acquire a senescence-associated secretory gression at the original sites of the disease) or progression phenotype (SASP), defined as an exaggerated production of (increase of any measured lesion or the appearance of new proinflammatory signals (17, 18). Of note, SASP is dependent lesions) was considered as treatment failure. Clinical assessment on the activation of the NF-kB pathway (19), and in this after treatment was performed every 3 or 4 months for the first manner,senescentcellsareabletoalterthetissuemicroenvi- year, every 6 months for the second and third years and once a year ronment where they home. Another important point is that thereafter. cytotoxic effects of various chemotherapeutic drugs and ioniz- ing radiation are also accompanied by the induction of CS Immunohistochemical staining (20, 21). Finally, there are multiple molecular links between CS All tissue samples were formalin-fixed and paraffin-embed- and cancer, including common genes, microRNAs, and signal- ded according to standard methods. Sections from tissue blocks ing pathways (22). Considering the above, it is plausible to of cHL were immunohistochemically analyzed with the fol- hypothesize that the mechanisms of CS are involved in cHL, lowing primary antibodies: CD30 (clone BerH2; Dako; dilution with potential prognostic relevance. We further suggest that a 1:30); OCT2 (rabbit polyclonal; Santa Cruz Biotechnology; dilu- good prognosis is associated with the ability of HRS cells to tion 1:200); BOB1 (clone C-20; Santa Cruz Biotechnology; acquire a CS-like phenotype, whereas a subgroup of cHL, which dilution 1:1,500); PAX5 (clone 1EW; Novocastra; dilution contains senescence-escaping cells that are more resistant to 1:50); CD15 (clone Carb-3; Dako; dilution 1:30); conventional therapies, is associated with a bad prognosis. p21CIP1/WAF1 (clone SX 118; Dako; dilution 1:50); p16INK4a To address these hypotheses, we analyzed HRS cells for the (clone JC 8; Santa Cruz Biotechnology; dilution 1:50). All cases expression of well-established CS biomarkers, including were tested for the presence of EBV by LMP1 immunostaining p21CIP1/WAF1 and p16INK4a. Furthermore, in a large retrospective (clone CS 1-4; Dako; dilution 1:200) and using in situ

OF2 Clin Cancer Res; 2015 Clinical Cancer Research

Cellular Senescence Markers in Hodgkin Lymphoma

hybridization for Epstein-Barr virus early RNAs (EBER; NCL- expression vs. high expression for each molecule), favorable EBV; Novocastra). All staining procedures were performed in an versus unfavorable stage, histotype, sex and, as continuous vari- immunostainer (Bond-Max, Leica). ables, age, Hb, LDH, albumin, fibrinogen, and ESR. To analyze HRS cells were considered positive for p21CIP1/WAF1 when a EFS in multivariate analysis by the Cox's proportional hazards distinct nuclear staining could be shown, whereas the cells were model, the independent variables were chosen on the basis of considered as positive for p16INK4a when either the cytoplasm or their significance in univariate analysis and/or of their recognized the nucleus or both were stained. Ten reactive lymphoid tissues relevance to outcome in cHL. We examined the predictive ability with an abundant reactive follicular component were selected as of the model by applying the Harrell c-statistic (33); a c-statistic controls, including 7 lymph nodes and 3 tonsils. In control near 1 or near 0 indicates greater predictive ability. p21CIP1/WAF1/ samples, low levels of p16INK4a cytoplasmatic immunoreactivity p16INK4a were also considered together as a categorical vari- were observed in dendritic cells and also in epithelial cells in able (both < 30%, either <30%, both 30%) and compared by reactive tonsils; p21CIP1/WAF1 expression was found in scattered the log-rank test. In Supplementary Data, the primary cohort nuclei in reactive follicular component and parabasal epithelial was further studied, including sCD30 (< or 100 U/mL; avail- cells in reactive tonsils. able for only 132 patients in the primary cohort), into the The quantitative evaluation of p16INK4a and p21CIP1/WAF1 in multivariate analysis and comparing p21CIP1/WAF1/p16INK4a HRS was performed on images of selected fields captured with a as a categorical variable within <100 and 100 U/mL sCD30 scanning microscopy device (D-Sight). The score of p16INK4a- subgroups. Results were regarded as statistically significant for and p21CIP1/WAF1-reactive HRS cells was obtained by evaluating P < 0.05. the percentage of positive atypical cells, as well as by comparison with CD30 immunostaining on serial sections. According Results to the percentage of positive HRS cells, four scores were defined: Patient characteristics negative (0), low (1), moderate (2), and high (3). The assigned The main clinicopathologic characteristics of the patients scores were as follows: score 0, devoid of expression or in less (Table 1) were similar between the primary and the validation than 10% positive HRS cells; score 1, 10% to 29% positive HRS cells; score 2, 30% to 59% of positive HRS cells; score 3, 60% of positive HRS cells (27). The concordance rate between the Table 1. p16INK4a and p21CIP1/WAF1 expression level according to presentation two pathologists was 88% for p16 and 92% for p21. Discordant features at diagnosis in primary and validation cohorts cases were reviewed at a multihead microscope and a consensus p16INK4aa p21CIP1/WAF1 was reached for each case. Patients þþ The consistency of score assignment was confirmed on selected N (%) N (%) N (%) N (%) N (%) cases using double immunostaining with CD30 and either Primary cohort p16INK4a or p21CIP1/WAF1. This four-tiered scoring system was All cases 147 77 (52) 70 (48) 76 (52) 71 (48) Male 73 (50) 38 (52) 35 (48) 39 (53) 34 (47) easily reproducible (27). For statistical purposes, it was trans- Female 74 (50) 39 (53) 35 (47) 37 (50) 37 (50) formed in a two-tiered system by considering a 30% cutoff level. Age, y 15–59 129 (88) 68 (53) 61 (47) 65 (50) 64 (50) 60–84 18 (12) 9 (50) 9 (50) 11 (61) 7 (39) Bioinformatic analysis Stage We used the MARQ online tool (http://MARQ.dacya.ucm.es; I–II 46 (31) 26 (57) 20 (43) 28 (61) 18 (39) ref. 28) for searching gene-expression signatures similar to those Bulky 42 (29) 21 (50) 21 (50) 23 (55) 19 (45) reported for cHL. For this purpose, we undertook comprehensive III–IV 59 (40) 30 (51) 29 (49) 25 (42) 34 (58) data mining of DNA microarrays available for cHL, with a special Histology NS 116 (79) 58 (50) 58 (50) 56 (48) 60 (52) focus on studies considering the disease outcome. On the basis of MC 24 (16) 17 (71) 7 (29) 14 (58) 10 (42) rank statistics, MARQ allows retrieving signatures with similar LR 7 (5) 2 (29) 5 (71) 6 (86) 1 (14) patterns of expression by comparing a given list of up-/down- sCD30b 132 67 (51) 65 (49) 63 (48) 69 (52) regulated genes with all arrays allocated in the GEO database <100 83 (63) 41 (49) 42 (51) 34 (41) 49 (59) (http://www.ncbi.nlm.nih.gov/geo/; GEO accession numbers: 100 49 (37) 26 (53) 23 (47) 29 (59) 20 (41) GSE39132, GSE39134; ref. 29). Validation cohort All cases 91 37 (41) 54 (59) 36 (40) 55 (60) Male 50 (55) 17 (34) 33 (66) 22 (44) 28 (56) Statistical analysis Female 41 (45) 20 (49) 21 (51) 14 (34) 27 (66) Data were analyzed using the statistical software Stata 12.1 Age, y 15–59 80 (88) 35 (44) 45 (56) 33 (41) 47 (59) (www.stata.com). Continuous variables were summarized as the 60–84 11 (12) 2 (18) 9 (82) 3 (27) 8 (73) median and range or the mean SEM. Data were summarized by Stage percentages for categorical variables and by medians and ranges I–II 42 (46) 15 (36) 27 (64) 14 (33) 28 (67) for survival times. Main analyses assumed as end-point the event- Bulky 16 (18) 7 (44) 9 (56) 9 (56) 7 (44) free survival (EFS), defined as the time interval between the III–IV 33 (36) 15 (45) 18 (55) 13 (39) 20 (61) beginning of treatment to primary treatment failure, relapse or Histology NS 73 (80) 58 (50) 58 (50) 56 (48) 60 (52) death from any cause or last follow-up. EFS curves were estimated MC 14 (15) 17 (71) 7 (29) 14 (58) 10 (42) – by the Kaplan Meier method (30). In univariate analysis, we LR 5 (5) 2 (29) 5 (71) 6 (86) 1 (14) tested the following variables by log-rank test or, for continuous a- stands for <30% expression; þ stands for 30% expression (HRS cells in the variables, Cox's proportional hazard regression model (31, 32): diagnostic lymph node). p21CIP1/WAF1 and p16INK4a, EBER, LMP1, CD30 (absent or low b15 missing data for sCD30 (see Supplementary Data).

www.aacrjournals.org Clin Cancer Res; 2015 OF3

Calio et al.

cohort. The two cohorts were dissimilar for patient number patients (3 progressed during treatment). Sixteen (18%) of 89 (147 vs. 91), median follow-up (11 vs. 5 years), bulky disease patients relapsed. As a whole, 73 (80%) of 91 patients persisted (42 vs. 16), and adverse event ratio (30 vs. 21%). In the primary in CR and 19 (21%) presented with adverse events. At the end cohort, the median follow-up time from the beginning of treat- of the study, 9 (10%) of 91 patients had died: 7 due to cHL and ment was 11 years for surviving patients (range from 3 months to 2 from apparently unrelated causes. 26 years). A CR was obtained in 144 (98%) of 147 patients (3 progressed during treatment). Forty-one (28%) of 144 patients Prognostic signiﬁcance of HRS senescence markers in the relapsed. As a whole, 103 (70%) of 147 patients persisted in CR primary cohort and 44 (30%) experienced adverse events. At the end of the study, The number of HRS cells expressing p16INK4a and p21CIP1/WAF1 23 (16%) of 147 patients had died: 22 as a direct consequence of on parafﬁn-embedded lymph node samples varied within a wide the disease and one from apparently cHL-unrelated cause. In the range, from rare to the majority of cells. In 48% of the cases, over validation cohort, the median follow-up time was 5 years (range 30% of the HRS cells expressed p16INK4a and p21CIP1/WAF1 from 1 month to 20 years). A CR was obtained in 88 (97%) of 91 (Fig. 1; Table 1).

A B

Figure 1. Expression of CS-related markers p16INK4a and p21CIP1/WAF1 in HRS cells in C D cHL (H&E 20, A and B). The quantitative scoring system used in this study is exempliﬁed in a positive case (column C, E, and G) and a negative case (column, D, F, and H). CD30 (20, C and D); p16INK4a (20, E and F); p21CIP1/WAF1 (20, G and H).

E F

OF4 Clin Cancer Res; 2015 Clinical Cancer Research

Cellular Senescence Markers in Hodgkin Lymphoma

A B 1.00 1.00 p16 ≥30% (n = 37) p16 ≥30% (n = 70) 0.75 0.75 0.50 0.50 p16 <30% (n = 54) p16 <30% (n = 77) 0.25 0.25 Primary cohort Validation cohort P < 0.001 P < 0.001 0.00 0.00 0 5 10 15 20 25 0 5 10 15 20 EFS probability 1.00 1.00 ≥ p21 ≥30% (n = 71) p21 30% (n = 55) 0.75 0.75 0.50 0.50 p21 <30% (n = 36) p21 <30% (n = 76) 0.25 Primary cohort 0.25 Validation cohort P < 0.001 P < 0.001 0.00 0.00 0 5 10 15 20 25 0 5 10 15 20 1.00 1.00 p16/21 ≥30% (n = 42) p16/21 ≥30% (n = 47) 0.75 0.75 p16 or 21 <30% (n = 47) p16 or 21 <30% (n = 25) 0.50 0.50 p16/21 <30% (n = 24)

EFS probability p16/21 <30% (n = 53) 0.25 0.25 Primary cohort P < 0.001 Validation cohort P < 0.001 0.00 0.00 0 5 10 15 20 25 0 5 10 15 years 20

Figure 2. Prognostic signiﬁcance of p16INK4a and p21CIP1/WAF1 in patients with cHL. The Kaplan–Meier analysis of EFS for p16INK4a and p21CIP1/WAF1 and their combined expression according to the expression level at diagnosis in the primary (A) and secondary (B) cohorts. Marks indicate last follow-up.

Results from our univariate analysis indicated that p16INK4a On the basis of our multivariate analysis (Table 2), both and p21CIP1/WAF1 were powerful prognostic factors in terms of p16INK4a and p21CIP1/WAF1 were independent prognostic fac- EFS probability. Their levels of expression at diagnosis in tors, with 30% expression being protective against relapse relation to clinical features are listed in Table 1 and their (HR, 0.25; 95% CI, 0.11–0.57; P ¼ 0.001 and HR, 0.16; 95% CI, frequencies of expression were almost uniformly distributed. 0.06–0.37; P < 0.001, respectively). Bulky and III–IV stages were p16INK4a and p21CIP1/WAF1 were categorized as dichotomous also confirmed to be independent prognostic factors in these variables (<30% or 30% of HRS cells at diagnosis) and Fig. 2A patients (HR, 3.79; 95% CI, 1.54–9.29; P ¼ 0.004 and HR, 2.25; depicts the Kaplan–Meier plot of EFS accordingly. The differ- 95% CI, 1.00–5.07; P ¼ 0.051, respectively). The Harrell c- ence between EFS curves was significant (P < 0.001) for both statistic was equal to 0.82, indicating a good predictive dis- p16INK4a and p21CIP1/WAF1 (Fig. 2A). The EFS probability at crimination and the effectiveness of the Cox's model. Further the plateau was 45% [95% confidence interval (CI), 33–57] comparative analyses considering also sCD30 levels are avail- for lower versus 89% (95% CI, 77–95) for higher p16INK4a able as Supplementary Data (Supplementary Fig. S1; Supple- samples, and 40% (95% CI, 28–53) for lower versus 90% mentary Tables S1 and S2). (95% CI, 79–95) for higher p21CIP1/WAF1 samples. p16INK4a /p21CIP1/WAF1 were also considered together as a unique cate- Prognostic significance of HRS senescence markers in the gorical variable (both < 30%, either <30%, both 30%). This validation cohort combined variable defined three different prognostic groups The prognostic value of p16INK4a and p21CIP1/WAF1 expression associated with statistically significant different EFS probabil- in HRS cells was further confirmed in an independent validation ities either overall or within each stage, the best prognosis being cohort of 91 patients. The results were similar to the primary associated with the robust expression of both molecules cohort (Fig. 2A). p16INK4a or p21CIP1/WAF1 expression in or (Figs. 2A and 3A; Table 3). <30% of HRS cells remained associated with different EFS

www.aacrjournals.org Clin Cancer Res; 2015 OF5

Calio et al.

A B 1.00 1.00 p16/21 ≥30% (n = 12) p16/21 ≥30% (n = 22) p16 or 21 <30% (n = 11) 0.75 0.75 p16 or 21 <30% (n = 14) 0.50 p16/21 <30% (n = 20) 0.50 p16/21 <30% (n = 9) 0.25 0.25 Primary cohort stage I−II P = 0.018 Validation cohort stage I−II P = 0.04 0.00 0.00 0 5 10 15 20 25 0 5 10 15 1.00 1.00 p16/21 ≥30% (n = 13) p16/21 ≥30% (n = 6) 0.75 0.75 p16 or 21 <30% (n = 14) p16/21 <30% (n = 6) 0.50 0.50 p16 or 21 <30% (n = 4)

p16/21 <30% (n = 15) 0.25 0.25 EFS probability Primary cohort stage bulky P < 0.001 Validation cohort stage bulky P = 0.137 0.00 0.00 0 5 10 15 20 25 0 5 10 15

p16/21 ≥30% (n = 22) 1.00 1.00 p16/21 ≥30% (n = 14) 0.75 p16 or 21 <30% (n = 19) 0.75 p16/21 <30% (n = 9) 0.50 0.50 p16/21 <30% (n = 18) p16 or 21 <30% (n = 10) 0.25 0.25 Primary cohort stage III−IV P < 0.001 Validation cohort stage III−IV P = 0.008 0.00 0.00 0 5 10 15 20 25 0 5 10 15 20 years

Figure 3. Prognostic signiﬁcance of the combined expression of p16INK4a/p21CIP1/WAF1 in patients with cHL. The Kaplan–Meier analysis of EFS for p16INK4a/p21CIP1/WAF1 at diagnosis within stages I and II (A), II bulky (B), III–IV (C) in the primary (A) and secondary (B) cohorts. Marks indicate last follow-up. 95% CIs are listed in Table 3.

probabilities (P < 0.001). The combined variable p16INK4a and attributed to the fact that even for cHL with a good prognosis, only p21CIP1/WAF1 entailed different, statistically significant EFS prob- a small number of cells in the tumor expressed the CS markers. abilities, associated with a robust or the lack of expression of the Nevertheless, an interesting observation was that the gene-expres- molecules. Here also, the best outcome was associated with a sion profile of cHL with good prognosis was similar to the gene robust expression of both molecules either overall or within each signature of miR-124 overexpression (similarity score of P < stage (Figs. 2B and 3B; Table 3). 0.0001). This tumor-suppressor microRNA is implicated both in The multivariate analysis confirmed that p16INK4a and many types of cancer and in the CS network (22). p21CIP1/WAF1 were independent prognostic variables with higher levels conferring statistically significant protection against relapse (HR, 0.19; 95% CI, 0.05–0.69; P ¼ 0.011 and HR, 0.25; 95% CI, Table 2. Multivariate analysis for event occurrence 0.07–0.85; P ¼ 0.026, respectively). The good predictive discrim- Primary cohort Validation cohort ination ability and the effectiveness of the Cox's model were HR (95% CI) P HR (95% CI) P fi c de ned by a Harrell -statistic of 0.82. (vs.< 30%) p16INK4a 0.25 (0.11–0.57) 0.001 0.19 (0.05–0.69) 0.011 CIP1/WAF1 Similar gene signatures: MARQ-based analysis p21 0.16 (0.06–0.37) <0.001 0.25 (0.07–0.85) 0.026 – – As could be expected, the majority of gene-expression signa- CD30 0.90 (0.45 1.76) 0.752 1.56 (0.55 4.43) 0.408 Stage (vs. I–II) tures similar to cHL include various types of cancer (Supplemen- Bulky 3.79 (1.54–9.29) 0.004 1.34 (0.39–4.55) 0.638 fi INK4a tary Table S3; ref. 7). With regard to our ndings that p16 and III–IV 2.25 (1.00–5.07) 0.051 1.81 (0.62–5.24) 0.277 CIP1/WAF1 p21 are often expressed in cHL, we anticipated that some Agea 1.20 (0.98–1.48) 0.075 1.13 (0.82–1.53) 0.449 similarity should be observed with gene-expression profiles for NOTE: Statistically significant P values are shown in bold type. CS. However, such a similarity was not found. This could be aHR calculated on the basis of an increase of 10 years.

OF6 Clin Cancer Res; 2015 Clinical Cancer Research

Cellular Senescence Markers in Hodgkin Lymphoma

Table 3. Probability of EFS (95% CI) according to stage and p16INK4a Previous studies described some relationship between the CIP1/WAF1 /p21 expression at diagnosis expression of cell-cycle protein and patient outcome. For example, No. of EFS % 95% CI in non-Hodgkin lymphoma, the levels of p53/p21CIP1/WAF1 Stage patients (%) plateau (%) Pa (12, 43) and Rb/p16INK4a have been found to correlate with clin- Primary cohort 147 I–II 46 (31) 74 (55–86) ical progression and predict OS. Regarding cHL, Guenova and INK4a p16/21 20 (43) 52 (23–75) 0.018 colleagues (13) reported high-frequency p16 expression in p16/21 14 (30) 78 (36–94) all 44 cases of cHL, but they did not find any statistical prognostic p16þ/21þ 12 (26) 100 — correlation, probably due to the small number of cases and the Bulky 42 (29) 62 (44–75) shorter follow-up. Similarly, Garcia and colleagues (44, 45) – < p16 /21 15 (36) 14 (2 37) 0.001 reported the loss of p16INK4a expression in 73% and in 82.5% p16/21 14 (33) 76 (42–92) p16þ/21þ 13 (31) 100 — of cHL and a trend for association of a shorter OS with low CIP1/WAF1 III–IV 59 (40) 62 (48–74) p21 expression was observed (44). p16/21 18 (31) 21 (6–42) <0.001 In this study, we provide strong evidence that in cHL, an p16/21 19 (32) 76 (47–90) inverse relationship exists between the clinical outcome and the þ þ p16 /21 22 (37) 88 (59–97) expression of p16INK4a and p21CIP1/WAF1, and this relationship Validation cohort 91 is robust and easily detected on routine histologic prepara- I–II 42 (46) 85 (63–99) p16/21 9 (21) 50 (14–79) 0.004 tions. The multivariate analysis on two different series from p16/21 11 (26) 82 (45–95) independent institutions shows that both of the CS markers p16þ/21þ 22 (52) 100 — retain independent prognostic significance among the para- Bulky 16 (18) 66 (36–85) meters analyzed. The prognostic power may be further refined p16 /21 6 (38) 50 (11–80) 0.137 by considering p16INK4a/p21CIP1/WAF1 together as a unique – p16 /21 4 (25) 50 (6 85) categorical double-positive, double-negative, or either-negative p16þ/21þ 6 (37) 100 — III–IV 33 (36) 67 (41–83) variable. Thus categorized, it entails subpopulations with p16/21 9 (27) 44 (14–72) 0.008 dramatically divergent EFS probabilities. As suggested by the p16/21 10 (30) 37 (1–79) biologic role of CS, it is the lack or the presence of the robust p16þ/21þ 14 (42) 100 — expressionofbothCDKIsthatdefine the worst or the best NOTE: Statistically significant P values are shown in bold type; , < 30% or þ, prognosis in our series. This suggeststhataworstprognosis 30% HRS cells expressing the indicated molecule; , either molecule <30%. subgroup of cHL patients could be sorted out to be treated a INK4a CIP1/WAF1 For comparison between p16 /p21 groups. differently. It also points to the potential use of senescence- inducing drugs in cHL treatment. In addition, we show (although only in the primary cohort) that the two CDKIs may be combined with the mass-dependent parameter sCD30 (3, 4) Discussion into an even more refined prognostic index (Supplementary The main hypothesis investigated in this article is that a Fig. S1; Supplementary Tables S1 and S2). subset of HRS cells display a CS-like phenotype, and that The prognostic power of p16INK4a and p21CIP1/WAF1 could be senescence-escaping mechanisms might make HRS cells more variously explained. One of the possibilities is that the inactiva- resistant to common therapeutic regimens, thus affecting tion of the main tumor-suppressor pathways, such as p53/ patients' prognosis. Both hypotheses seem to be supported p21CIP1/WAF1 and pRb/p16INK4a, leads to uncontrolled prolifera- by our data, because we showed a consistent expression of CS- tion and finally to an increased clinical aggressiveness. Indeed, associated markers by HRS cells in many cases of cHL, whereas p16INK4a and p21CIP1/WAF1 act in the pathway of oncogene- the lack of these proteins is associated with a worse outcome. induced CS, and the activation of this mechanism during tumor- Our work demonstrates that in cHL, the neoplastic HRS cells igenesis limits cancer progression (46). As CS was reported to be express markers related to CS, such as the CDK inhibitors induced by various chemotherapeutic drugs (20, 21), it is impor- p21CIP1/WAF1 and p16INK4a. This expression could be related to tant to note that in our study, the CS markers associated with a CS induced by oncogenes (34). Several reports (35–37) have favorable outcome were present before treatment. These observa- demonstrated that p21CIP1/WAF1 and p16INK4a are CS hall- tions are particularly interesting because they present evidence marks and may be useful for the detection of this response with a prognostic value, wherein tumors containing the cells not only in vitro but also in vivo (15, 37). According to our predisposed to CS could be more susceptible to CS induction by hypothesis, HRS cells exhibit a variety of features of senescent anticancer treatment. As such, these tumors could have a better cells, including abnormal cell enlargement, nuclear disorga- prognosis. nization with aberrant expression of cell-cycle regulatory fac- Another possible explanation could be attributed to the tors, abnormal karyokinesis, and telomere aberrant distribu- recent observation that p16INK4a is a major barrier to prolifer- tion (38, 39). Also, the recruitment of inflammatory cells ation in B cells transformed by EBV, a virus that has been around HRS cells, a typical morphologic feature of cHL, can implicated in the pathogenesis of cHL (47). The pathogenic be ascribed in this scheme to the CS-associated "cytokine role of EBV infection in cHL (48) is well known, and EBV has storm" (SASP), a well-known feature characterizing experi- been found in HRS cells in about 40% of cases (10) and EBV mentally induced CS (35). may have a role in inhibiting p16INK4a expression (14). We also Our study provides evidence that the quantitative evaluation observed only a trend between the expression of p16INK4a and ofCSmarkersinHRScellsofcHLcanhaveaprognostic thepresenceofEBV(datanotshown).Notably,thevirallatent significance in line with, and probably grounded on, their membrane protein 1 (LMP1, a marker of EBV presence) blocks biologic role analogously to other neoplasias (40–42). p16INK4a expression (49).

www.aacrjournals.org Clin Cancer Res; 2015 OF7

Calio et al.

In this study, we demonstrate the potential value of immuno- Analysis and interpretation of data (e.g., statistical analysis, biostatistics, histochemical detection of p16INK4a and p21CIP1/WAF1 in diag- computational analysis): A. Calio, A. Zamo, M.E. Zanolin, V.E. Fraifeld, nostic samples for the prediction of cHL outcome. In our series, M. Wolfson, H. Yanai, C. Doglioni, F. Vinante Writing, review, and/or revision of the manuscript: A. Calio, A. Zamo, the cases of cHL exhibiting absence or low expression of the CS M. Ponzoni, A.J.M. Ferreri, V.E. Fraifeld, M. Wolfson, H. Yanai, G. Pizzolo, CIP1/WAF1 INK4a markers p21 and p16 in HRS cells tended to relapse F. Vinante, M. Chilosi or did not respond to therapy. Given the simple and reproducible Administrative, technical, or material support (i.e., reporting or organizing immunohistochemical approach described in this article, we data, constructing databases): A. Calio, S. Pedron, C. Parolini would like to propose a rapid and low-cost method, which could Study supervision: F. Vinante, M. Chilosi be applied in routine diagnosis. Although further and prospective studies are warranted to conﬁrm our data, the results of our study Acknowledgments support the hypothesis that CS features may be characteristic of The authors thank Federica Pedica. HRS cells in the majority of cHL.

Disclosure of Potential Conﬂicts of Interest Grant Support A. Zamo reports receiving speakers bureau honoraria from Novartis. No This study was funded by the European Comission FP7 Health Research potential conﬂicts of interest were disclosed by the other authors. grant number HEALTH-F4-2008-202047 (to V.E. Fraifeld and M. Chilosi) and by Italian Association for Cancer Research (AIRC, Milan, Italy)/Cariverona Authors' Contributions Foundation (Verona, Italy). Conception and design: F. Vinante, M. Chilosi The costs of publication of this article were defrayed in part by the payment of advertisement Development of methodology: A. Zamo, A.J.M. Ferreri, S. Pedron, L. Montagna, page charges. This article must therefore be hereby marked in M. Chilosi accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Calio, A. Zamo, M. Ponzoni, A.J.M. Ferreri, Received March 9, 2015; revised June 24, 2015; accepted July 4, 2015; C. Doglioni published OnlineFirst July 21, 2015.

References 1. Derenzini E, Younes A. Predicting treatment outcome in classical Hodgkin disease: comparison with retinoblastoma, p53 and MDM2 protein status, lymphoma: genomic advances. Genome Med 2011;3:26. and the presence of Epstein-Barr virus. Mod Pathol 1999;12:1062–71. 2. Diehl V, Thomas RK, Re D. Part II: Hodgkin's lymphoma—diagnosis and 14. Kim LH, Peh SC, Poppema S. Expression of retinoblastoma protein and treatment. Lancet Oncol 2004;5:19–26. P16 proteins in classic Hodgkin lymphoma: relationship with expression 3. Zanotti R, Trolese A, Ambrosetti A, Nadali G, Visco C, Ricetti MM, et al. of p53 and presence of Epstein-Barr virus in the regulation of cell growth Serum levels of soluble CD30 improve International Prognostic Score in and death. Hum Pathol 2006;37:92–100. predicting the outcome of advanced Hodgkin's lymphoma. Ann Oncol 15. Ju Z, Choudhury AR, Rudolph KL. A dual role of p21 in stem cell aging. Ann 2002;13:1908–14. N Y Acad Sci 2007;1100:333–44. 4. Nadali G, Tavecchia L, Zanolin E, Bonfante V, Viviani S, Camerini E, et al. 16. Campisi J. Senescent cells, tumor suppression, and organismal aging: good Serum level of the soluble form of the CD30 molecule identifies patients citizens, bad neighbors. Cell 2005;120:513–22. with Hodgkin's disease at high risk of unfavorable outcome. Blood 17. Rodier F, Coppe JP, Patil CK, Hoeijmakers WA, Munoz DP, Raza SR, et al. 1998;91:3011–6. Persistent DNA damage signalling triggers senescence-associated inflam- 5. Alvaro T, Lejeune M, Salvado MT, Bosch R, Garcia JF, Jaen J, et al. Outcome matory cytokine secretion. Nat Cell Biol 2009;11:973–9. in Hodgkin's lymphoma can be predicted from the presence of accompa- 18. Fumagalli M, d'Adda di Fagagna F. SASPense and DDRama in cancer and nying cytotoxic and regulatory T cells. Clin Cancer Res 2005;11:1467–73. ageing. Nat Cell Biol 2009;11:921–3. 6. Diepstra A, van Imhoff GW, Karim-Kos HE, van den Berg A, te Meerman GJ, 19. Salminen A, Kauppinen A, Kaarniranta K. Emerging role of NF-kappaB Niens M, et al. HLA class II expression by Hodgkin Reed-Sternberg cells is an signaling in the induction of senescence-associated secretory phenotype independent prognostic factor in classical Hodgkin's lymphoma. J Clin (SASP). Cell Signal 2012;24:835–45. Oncol 2007;25:3101–8. 20. te Poele RH, Okorokov AL, Jardine L, Cummings J, Joel SP. DNA damage is 7. Steidl C, Diepstra A, Lee T, Chan FC, Farinha P, Tan K, et al. Gene expression able to induce senescence in tumor cells in vitro and in vivo. Cancer Res profiling of microdissected Hodgkin Reed-Sternberg cells correlates with 2002;62:1876–83. treatment outcome in classical Hodgkin lymphoma. Blood 2012;120: 21. Roninson IB. Tumor cell senescence in cancer treatment. Cancer Res 3530–40. 2003;63:2705–15. 8. Hummel M, Ziemann K, Lammert H, Pileri S, Sabattini E, Stein H. 22. Tacutu R, Budovsky A, Yanai H, Fraifeld VE. Molecular links between Hodgkin's disease with monoclonal and polyclonal populations of cellular senescence, longevity and age-related diseases—a systems biology Reed-Sternberg cells. N Engl J Med 1995;333:901–6. perspective. Aging 2011;3:1178–91. 9. Chilosi M, Pizzolo G. Biopathologic features of Hodgkin's disease. Leuk 23. Swerdlow SH, Campo E, Harris NL, Vardiman JW, editors. WHO classi- Lymphoma 1995;16:385–95. fication of tumour of haematopoietic and lymphoid tissue. 2008. 10. Kuppers R. The biology of Hodgkin's lymphoma. Nat Rev Cancer 2009; 24. Lister TA, Crowther D, Sutcliffe SB, Glatstein E, Canellos GP, Young RC, 9:15–27. et al. Report of a committee convened to discuss the evaluation and staging 11. Doglioni C, Pelosio P, Mombello A, Scarpa A, Chilosi M. Immunohisto- of patients with Hodgkin's disease: Cotswolds meeting. J Clin Oncol chemical evidence of abnormal expression of the antioncogene-encoded 1989;7:1630–6. þ p53 phosphoprotein in Hodgkin's disease and CD30 anaplastic lympho- 25. Hoppe RT. Radiation therapy in the management of Hodgkin's disease. mas. Hematol Pathol 1991;5:67–73. Semin Oncol 1990;17:704–15. 12. Chilosi M, Doglioni C, Magalini A, Inghirami G, Krampera M, Nadali G, 26. Longo DL. The use of chemotherapy in the treatment of Hodgkin's disease. et al. p21/WAF1 cyclin-kinase inhibitor expression in non-Hodgkin's Semin Oncol 1990;17:716–35. lymphomas: a potential marker of p53 tumor-suppressor gene function. 27. Geissinger E, Sadler P, Roth S, Grieb T, Puppe B, Muller N, et al. Disturbed Blood 1996;88:4012–20. expression of the T-cell receptor/CD3 complex and associated signaling þ 13. Guenova M, Rassidakis GZ, Gorgoulis VG, Angelopoulou MK, Siakantaris molecules in CD30 T-cell lymphoproliferations. Haematologica 2010; MR, Kanavaros P, et al. p16INK4A is regularly expressed in Hodgkin's 95:1697–704.

OF8 Clin Cancer Res; 2015 Clinical Cancer Research

Cellular Senescence Markers in Hodgkin Lymphoma

28. Vazquez M, Nogales-Cadenas R, Arroyo J, Botias P, Garcia R, Carazo JM, 40. Michaloglou C, Vredeveld LC, Soengas MS, Denoyelle C, Kuilman T, van et al. MARQ: an online tool to mine GEO for experiments with similar or der Horst CM, et al. BRAFE600-associated senescence-like cell-cycle arrest opposite gene expression signatures. Nucleic Acids Res 2010;38:W228–32. of human naevi. Nature 2005;436:720–4. 29. Barrett T, Wilhite SE, Ledoux P, Evangelista C, Kim IF, Tomashevsky M, et al. 41. Chilosi M, Facchetti F, CalioA,Zam o A, Brunelli M, Martignoni G, et al. NCBI GEO: archive for functional genomics data sets—update. Nucleic Oncogene-induced senescence distinguishes indolent from aggressive Acids Res 2013;41:D991–5. forms of pulmonary and non-pulmonary Langerhans cell histiocytosis. 30. Kaplan EL, Meier P. Nonparametric estimation from incomplete observa- Leuk Lymphoma 2014;55:2620–6. tion. J Am Stat Assoc 1958;53:457–81. 42. Martinelli P, Bonetti P, Sironi C, Pruneri G, Fumagalli C, Raviele PR, 31. Cox DR. Regression models and life-tables. J Royal Stat Soc Series B et al. The lymphoma-associated NPM-ALK oncogene elicits a (Methodological) 1972;34:187–220. p16INK4a/pRb-dependent tumor-suppressive pathway. Blood 2011; 32. Vinante F, Rigo A, Tecchio C, Morosato L, Nadali G, Ricetti MM, et al. Serum 117:6617–26. levels of p55 and p75 soluble TNF receptors in adult acute leukaemia at 43. Visco C, Canal F, Parolini C, Andreoli A, Ambrosetti A, Krampera M, et al. diagnosis: correlation with clinical and biological features and outcome. Br The impact of P53 and P21(waf1) expression on the survival of patients J Haematol 1998;102:1025–34. with the germinal center phenotype of diffuse large B-cell lymphoma. 33. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in Haematologica 2006;91:687–90. developing models, evaluating assumptions and adequacy, and measuring 44. Garcia JF, Camacho FI, Morente M, Fraga M, Montalban C, Alvaro T, et al. and reducing errors. Stat Med 1996;15:361–87. Hodgkin and Reed-Sternberg cells harbor alterations in the major tumor 34. Collado M, Serrano M. The power and the promise of oncogene-induced suppressor pathways and cell-cycle checkpoints: analyses using tissue senescence markers. Nat Rev Cancer 2006;6:472–6. microarrays. Blood 2003;101:681–9. 35. Coppe JP, Rodier F, Patil CK, Freund A, Desprez PY, Campisi J. Tumor 45. Garcia JF, Villuendas R, Algara P, Saez AI, Sanchez-Verde L, Martinez- suppressor and aging biomarker p16(INK4a) induces cellular senescence Montero JC, et al. Loss of p16 protein expression associated with meth- without the associated inflammatory secretory phenotype. J Biol Chem ylation of the p16INK4A gene is a frequent finding in Hodgkin's disease. 2011;286:36396–403. Lab Invest 1999;79:1453–9. 36. McConnell BB, Starborg M, Brookes S, Peters G. Inhibitors of cyclin- 46. Collado M, Blasco MA, Serrano M. Cellular senescence in cancer and aging. dependent kinases induce features of replicative senescence in early passage Cell 2007;130:223–33. human diploid fibroblasts. Curr Biol 1998;8:351–4. 47. Skalska L, White RE, Parker GA, Turro E, Sinclair AJ, Paschos K, et al. 37. Shay JW, Roninson IB. Hallmarks of senescence in carcinogenesis and Induction of p16(INK4a) is the major barrier to proliferation when cancer therapy. Oncogene 2004;23:2919–33. Epstein-Barr virus (EBV) transforms primary B cells into lymphoblastoid 38. Chilosi M, Doglioni C, Menestrina F, Montagna L, Rigo A, Lestani M, et al. cell lines. PLoS Pathog 2013;9:e1003187. Abnormal expression of the p53-binding protein MDM2 in Hodgkin's 48. Young LS, Murray PG. Epstein-Barr virus and oncogenesis: from latent disease. Blood 1994;84:4295–300. genes to tumours. Oncogene 2003;22:5108–21. 39. Knecht H, Sawan B, Lichtensztejn D, Lemieux B, Wellinger RJ, Mai S. The 3D 49. Yang X, He Z, Xin B, Cao L. LMP1 of Epstein-Barr virus suppresses cellular nuclear organization of telomeres marks the transition from Hodgkin to senescence associated with the inhibition of p16INK4a expression. Reed-Sternberg cells. Leukemia 2009;23:565–73. Oncogene 2000;19:2002–13.

www.aacrjournals.org Clin Cancer Res; 2015 OF9

Cellular Senescence Markers p16INK4a and p21CIP1/WAF Are Predictors of Hodgkin Lymphoma Outcome

Anna Caliò, Alberto Zamò, Maurilio Ponzoni, et al.

Clin Cancer Res Published OnlineFirst July 21, 2015.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-15-0508

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2015/07/21/1078-0432.CCR-15-0508.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2015/09/30/1078-0432.CCR-15-0508. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.