Margin-Infiltrating CD20 B Cells Display an Atypical Memory

Published OnlineFirst September 20, 2013; DOI: 10.1158/1078-0432.CCR-12-3497

Clinical Cancer Imaging, Diagnosis, Prognosis Research

Margin-Inﬁltrating CD20þ B Cells Display an Atypical Memory Phenotype and Correlate with Favorable Prognosis in Hepatocellular Carcinoma

Jie-Yi Shi1,3, Qiang Gao1,3, Zhi-Chao Wang1,3, Jian Zhou1,3,4, Xiao-Ying Wang1,3, Zhi-Hui Min2, Ying-Hong Shi1,3, Guo-Ming Shi1,3, Zhen-Bin Ding1,3, Ai-Wu Ke1,3, Zhi Dai1,3, Shuang-Jian Qiu1,3, Kang Song1,3, and Jia Fan1,3,4

Abstract Purpose: The role of infiltrating B cells in hepatocellular carcinoma has been overlooked for many years. This study is aimed to delineate the distribution, prognostic value, and functional status of B cells in human hepatocellular carcinoma. Experimental design: Immunohistochemistry was used to investigate the distribution and clinical þ significance of infiltrating CD20 B cells in a series of 120 patients with hepatocellular carcinoma. The results were further tested in an independent series of 200 patients with hepatocellular carcinoma. The þ functional status of CD20 B cells was determined by flow cytometry, immunofluorescence, and in vitro coculture assay. þ Results: Infiltrating CD20 B cells were predominantly concentrated in the tumor invasive margin, compared with the peri- and intratumor areas. High density of margin-infiltrating B lymphocytes (MIL-B) þ positively correlated with small tumor size, absence of vascular invasion, and increased density of CD8 T cells (P < 0.05). Survival analyses revealed that increased number of MIL-Bs and their penetration through the tumor capsule were significantly associated with improved overall and recurrence-free survival, and were identified as independent prognosticators for patients with hepatocellular carcinoma (P < 0.05). Importantly, the results were further validated in another independent hepatocellular carcinoma cohort. þ Moreover, we found that MIL-Bs featured an atypical memory phenotype (IgD IgG CD27 CD38 ), expressed surface markers characteristic of antigen-presenting cells, possessed tumor-killing potential by producing IFN-g, interleukin 12p40 (IL-12p40), granzyme B, and TRAIL, and acted in cooperation with þ CD8 T cells. þ Conclusions: The profile of CD20 B cells in situ is a new predictor of prognosis for patients with hepatocellular carcinoma and provides a novel target for an optimal immunotherapy against this fatal malignancy. Clin Cancer Res; 19(21); 5994–6005. 2013 AACR.

Introduction postsurgical recurrence (50–70% at 5 years) renders this Hepatocellular carcinoma is the second leading cause of disease a major challenge. Etiologically, chronic inflamma- cancer-related deaths worldwide, with an estimated tion, mainly resulting from hepatitis virus infection, has 750,000 newly diagnosed cases annually. Liver transplan- long been accepted as a key contributor to hepatocarcino- tation and tumor resection have been proved to be the most genesis (2). effective standard therapies (1). However, the high rate of Recently, evidence has been accumulating that the inflammatory milieu, in particular the composition, localization, and nature of the infiltrating cells, may alter tumor Authors' Affiliations: 1Liver Cancer Institute and 2Biomedical Research biologic behavior and affect disease progression in hepa- Center, Zhongshan Hospital; 3Key Laboratory of Carcinogenesis and Cancer Invasion, Ministry of Education, and 4Institute of Biomedical tocellular carcinoma. In this regard, we and others have Sciences, Fudan University, Shanghai, PR China previously shown that dendritic cells, memory T cells, J.-Y. Shi, Q. Gao, and Z.-C. Wang contributed equally to this work. cytotoxic T cells, and gdT cells dwelling in the tumor bed tend to reduce hepatocellular carcinoma (HCC) growth Corresponding Author: Jia Fan, Liver Cancer Institute, Zhongshan Hos- (3–7), whereas M2 macrophages, N2 neutrophils, Th17 pital, Fudan University, 180 Feng Lin Road, Shanghai 200032, PR China. þ Phone: 86-021-64041990; Fax: 86-021-64037181; E-mail: cells, and Foxp3 regulatory T cells may stimulate HCC [email protected] progression (8–11). These findings are in accordance with doi: 10.1158/1078-0432.CCR-12-3497 the general view that hepatocellular carcinoma is an immu- 2013 American Association for Cancer Research. nogenic tumor, and provide a rationale for designing

5994 Clin Cancer Res; 19(21) November 1, 2013

Margin-Inﬁltrating B Cells in Hepatocellular Carcinoma

þ functional status, and interactions with CD8 T cells in Translational Relevance vitro. The presence and role of B cells in tumor immunity have been ignored for many years. Herein, we detected Patients and Methods þ that the majority of infiltrating CD20 B cells in hepa- Patients and specimens tocellular carcinoma were enriched in tumor margin For the initial cohort, 120 patients with hepatocellular (MIL-Bs) and had an atypical memory phenotype þ carcinoma were randomly selected from patients who had (IgD IgG CD27 CD38 ). Moreover, we found and received curative hepatectomy between 2005 and 2007 in validated that the density of MIL-Bs and their penetra- our institute. The median duration of follow-up was 61.5 tion through tumor capsule were significantly and inde- months (range, 2.0–84.8 months). For validation, a total of pendently associated with prolonged survival and 200 consecutive patients with hepatocellular carcinoma reduced recurrence in patients with hepatocellular car- who had undergone curative hepatectomy between April cinoma. Furthermore, our phenotypic and functional and October 2006 were enrolled, and the median duration assays suggested that B cells were capable of facilitating of follow-up was 59.5 months (range, 2.0–73.2 months). the antitumor responses by serving as antigen-presenting The inclusion and exclusion criteria of patients, postoper- cells, releasing immune-stimulating cytokines, and par- ative surveillance, and treatment modalities have been ticipating in the direct-killing effect through granzyme B described previously (19). Overall survival (OS) was and TRAIL. Thus, our results bring strong evidence that defined as the time between the surgery and death and MIL-B is a powerful predictor for patient prognosis, and recurrence-free survival (RFS) was defined as the time infiltrating B cells may be served as a potential target for between the surgery and recurrence. Patients without recur- immunotherapy in hepatocellular carcinoma. rence or death were censored at the last follow-up. There were no significant differences in the clinicopathologic features between the initial and the validation cohorts (Supplementary Table S1). Informed consent form was immunotherapies for HCC treatment. However, these stud- signed by each patient, and ethical approval was obtained ies mainly focused on the antitumor activities of T cells in from the Zhongshan Hospital Research Ethics Committee cellular immune response. The analysis of the cellular (Shanghai, PR China). component of humoral immunity to tumors might lead to the exploration of new targets for immunotherapy. Immunohistochemistry and evaluation of As the central component of humoral immunity, B lym- immunohistochemical variables phocytes function in antibody production, antigen presen- Immunohistochemistry was conducted as described pre- tation, and proinflammatory cytokines secretion. Initial viously (20), using mouse anti-human CD20 (1:200 dilu- studies in mouse tumor models suggested that B cells tion, clone L26, Dako) and CD8 (1:50 dilution, clone 144B, generally inhibited T-cell responses (12–14). More recently, Abcam) monoclonal antibodies (mAb) as the primary þ the role of CD20 (a B-lymphocyte marker expressed on antibodies. Tumor sections were microanatomically divid- mature B cells but not on plasma cells) lymphocytes in ed into peritumor, invasive margin, and intratumor areas. promoting favorable outcomes in several human cancers The invasive margin was defined as the region within 500 has been reported (15, 16), and the underlying mechanisms mm on each side of the border between the tumor and might include B cells producing tumor-specific antibodies, normal liver tissue (21). The density of positive cells was presenting tumor-specific antigen to T cells, and secreting evaluated as described previously (11). Briefly, five random cytokines that enhance antitumor immunity in the local microscopic fields (magnification, 40) in each area were tumor environment (17). In addition, the presence of selected and captured. Positively stained cells were counted infiltrating B lymphocytes in hepatocellular carcinoma manually by two independent investigators blinded to the has also been documented (18). However, the clinical clinicopathologic data. The density of positive cells was relevance and prognostic significance of infiltrating calculated by averaging. Variations over a range of 5% were B cells in hepatocellular carcinoma, as well as their micro- reevaluated for a consensus result. anatomical distribution and functional status, remain large- ly unknown. Isolation of infiltrating lymphocytes In this study, we first delineated the microanatomical Fresh tumor specimens from 10 patients with hepato- þ distribution of CD20 B cells, and found most B cells cellular carcinoma were obtained immediately after sur- were enriched in the tumor invasive margin, which we gery. The specimens were also divided into peritumor, defined as margin-infiltrating B lymphocytes (MIL-B). invasive margin, and intratumor areas as described above, The relationship between density of MIL-Bs and patient by extending the margin area to 5 mm on each side of the clinical outcome was investigated via immunohistochem- border to obtain enough tissue. The tissues were minced istry of samples from 120 patients with hepatocellular and digested in 1 mg/mL collagenase (Invitrogen) and 50 carcinoma, and then tested in an independent series of UI/mL hyaluronidase (Sigma-Aldrich) for 1 hour. The cell 200 hepatocellular carcinoma cases. In addition, infiltrat- suspension was filtered with a 400-mesh sieve and sep- ing B cells were isolated to investigate their phenotypes, arated by centrifugation through a discontinuous Percoll

www.aacrjournals.org Clin Cancer Res; 19(21) November 1, 2013 5995

Shi et al.

(GE Healthcare) gradient with densities of 1.06 and 1.08 ELISA analysis g/mL. The lymphocytes enriched at the interface of Per- The concentration of IFN-g in the conditioned superna- coll solutions were harvested and cultured in RPMI-1640 tant from the coculture experiment was determined using a (Invitrogen) supplemented with 10% FBS (Invitrogen), sandwich ELISA kit according to the manufacturer’s instruc- 100 UI/mL penicillin, and 100 mg/mL streptomycin at tions (DuoSet, R&D Systems). 37 Cin5%humidifiedCO2 for further experiments. Immunofluorescence Flow cytometric analysis The desired tumor tissues, obtained from 12 patients with Flow cytometry was used to detect the surface marker and hepatocellular carcinoma and underwent hepatectomy in þ cytokine expression of CD20 B cells isolated from the 10 our institute, were embedded with OCT tissue-freezing patients with hepatocellular carcinoma. Peridinin chloro- medium (Tissue-Tek, Sakura Finetek), and stored at 80C phyll (PerCP)-labeled anti-CD20, phycoerythrin (PE)- immediately after surgery. Tumor samples from another five labeled antibodies to immunoglobulin D (IgD), IgM, IgG, patients with hepatocellular carcinoma were used for þ interleukin 4 (IL-4), IL-6, IL-10, IFN-g, and TRAIL; allophy- CD20 cell sorting. MIL and peritumor-infiltrating B lym- cocyanin (APC)-labeled anti-IL-2, IL-12p40, and CD38; phocytes (PIL-B) were collected on slides using Cytospin 4 and fluorescein isothiocyanate (FITC)-labeled antibodies (Thermo Scientific). Immunofluorescence was performed to HLA-ABC, HLA-DR, CD40, CD80, CD86, CD27, CD8, as described previously (25), using anti-CD20 mouse mAb þ and granzyme B were purchased from BioLegend. CD20 B (1:200 dilution, clone L26, Dako) together with either anti- cells were sorted by flow cytometry from isolated lympho- granzyme B rabbit pAb (1:50 dilution, Abcam) or anti- cytes and then stained with PerCP-labeled CD20 antibody. TRAIL (C92B9) rabbit mAb (1:500 dilution, Cell Signaling þ The viability of CD20 B cells, detected by Trypan blue Technology). Then, two secondary antibodies, goat anti- staining, was about 99% just after sorting, and over 85% mouse IgG-Cy3 and goat anti-rabbit IgG-FITC (Invitrogen), þ after 3-day in vitro culture. For cytokine detection, CD20 B were applied. Images were acquired with a LSM510 Con- cells were stimulated with ODN2006 (a synthetic oligonu- focal Laser Scanning Microscope (Carl Zeiss). cleotide strongly activating B cells, 10 mg/mL, InvivoGen) and CD40L (1 mg/mL, R&D Systems) for 24 hours, followed Statistical analysis by a leukocyte activation cocktail (a mixture containing Data were expressed as the mean SEM, and error bars Phorbol 12-Myristate 13-Acetate, ionomycin, and Brefeldin refer to SEM in figures. The analysis of the association A, 2 ml/mL, BD Pharmingen) before another 4-hour culture between variables was conducted using the Spearman r (22). Then, the cells were permeabilized and incubated with coefficient test, Student t test, c2 test, Fisher exact test, or specific antibodies. Flow cytometry was performed on a one-way ANOVA when appropriate. Univariate and multi- FACSAria II (BD Immunocytometry Systems) according to variate analyses were based on the Cox proportional the manufacturer’s instructions, and analyzed with FlowJo hazards regression model. Survival curves were computed software version 7.6.1(Tree Star). using the Kaplan–Meier method (log-rank test). A two- tailed P value of 0.05 indicated a significant result. All Coculture of immune cells and tumor cells statistical analyses were conducted with SPSS version þ CD8 T cells were sorted from isolated lymphocytes 19.0 (IBM). þ stained with FITC-labeled CD8 antibodies. CD20 B and þ CD8 T cells derived from the hepatocellular carcinoma Results þ margin area, together with a human hepatocellular carci- The distribution patterns of CD20 B cells in noma cell line MHCC97H (97H), were used for coculture hepatocellular carcinoma assays. The 97H cell line was established in our institute (23, Immunohistochemistry was performed on 120 HCC þ 24) from a subcutaneous xenograft of a human metastatic samples to elucidate the distribution of CD20 B cells hepatocellular carcinoma model in nude mice, and in the peritumor, tumor margin, and intratumor areas expressed low HLA-ABC and bare HLA-DR (Supplementary (Fig. 1A). B cells were predominately enriched in the inva- Fig. S1). Briefly, 97H cells (1.0 104 cells/well) were seeded sive margin (median, 192.7 cells/mm2; range, 29.6–820.3 þ þ with CD20 B cells (1.0 104 cells/well), CD8 T cells (1.0 cells/mm2), whereas only small fractions of B cells sporad- þ þ 104 cells/well), or a combination of CD20 plus CD8 ically infiltrated into the peritumor (PIL-Bs; median, 49.5 cells (5.0 103 cells/well for each), respectively, in 96-well cells/mm2; range, 0–354 cells/mm2, P < 0.001) and intra- plates with a mixture of Dulbecco’s modified Eagle Medium tumor (median, 12 cells/mm2; range, 0–257 cells/mm2, P < (DMEM) and RPMI-1640 (1:1, Invitrogen) containing 10% 0.001; Fig. 1B) areas. To confirm the result, flow cytometry þ FBS (Invitrogen), using 97H cells (1.0 104 cells/well) was conducted to detect CD20 B cells in the three micro- without any lymphocytes as control. The cocultures were anatomic areas from 10 patients with hepatocellular carci- þ incubated for 5 days, and the supernatant was collected for noma. Accordingly, the percentage of CD20 B cells within þ ELISA analysis. Tumor cell viability on each day was eval- the total CD45 cell population in the invasive margin (9.4 uated using the CCK-8 assay (Dojindo Inc.) according to the 0.70%) was significantly higher than that in the peritu- manufacturer’s instructions. Six independent experiments mor (4.9 0.57%, P ¼ 0.001) and intratumor areas (1.8 with different patient samples were conducted in triplicate. 0.41%, P < 0.001), respectively (Supplementary Fig. S2).

5996 Clin Cancer Res; 19(21) November 1, 2013 Clinical Cancer Research

Margin-Inﬁltrating B Cells in Hepatocellular Carcinoma

A C Peritumor Peritumor Peritumor Tumor No tumor capsule Margin area Margin area 500 μm 500 μm Tumor capsule

500 μm 500 μm Case #24 Intratumor Intratumor Tumor Peritumor 500 μm500 μm 200 μm 200 μm

Cells/mm2 Tumor B 0–20 21–40 No penetration 1,000 41–60 900 61–80 800 81–100 400 Case #48 2 101–120 Tumor capsule Peritumor 121–140 200 μm 50 μm 300 272.8 Peritumor cells/mm + 200 192.7

CD20 137.5 127.9 100 Penetration 49.5 25 Case #86 15 Tumor capsule Tumor 12 μ 2 200 m 50 μm Peritumor Margin area Intratumor DE

1,000 P < 0.001

2 800 r = 0.391 Case #16

100 μm100 μm 600 cells/mm + 400 CD20 200 Case #112

100 μm 100 μm 200 400 600 800 1,000 + 2 CD20 CD8 CD8 cells/mm

Figure 1. Immunostaining results of infiltrating lymphocytes in hepatocellular carcinoma (HCC). A, left, a representative case of CD20 immunostaining in hepatocellular carcinoma tissue. The solid-line curve indicates the tumor–normal border, and the area between the dashed curves denotes the margin area. Magnification, 4. Right, immunomap of the same hepatocellular carcinoma lesion. The tumor section is artificially divided into tiles (0.16 mm2 each), and þ þ colored according to CD20 density. B, CD20 B cells are more abundant in the margin area than in the peritumor and intratumor areas (n ¼ 120). Lines indicate 25th, 50th, and 75th percentiles. , P < 0.001. C, representative hematoxylin and eosin (HE) staining of hepatocellular carcinomas with (left) or þ without (right) tumor capsule (top). Magnification, 10. Also shown is the immunostaining of CD20 B cells in cases without (middle) or with (bottom) penetration through the tumor capsule. Magnification, left, 10; right, 40. White lines highlight the tumor capsule. D, serial sections show hepatocellular þ þ þ carcinomas with simultaneously high (top) or low (bottom) densities of marginal CD20 and CD8 cells. Magnification, 20. E, the density of CD20 B cells þ significantly correlates with the density of CD8 T cells in the tumor margin (n ¼ 120; r ¼ 0.391, P < 0.001).

Because the intratumoral area contained the least number of sule (26), was detected (Fig. 1C, top left). Interestingly, we B cells, whereas most B cells were enriched in the tumor found an uneven distribution of MIL-Bs on both sides of the margin, we consequently focused on prognostic value and tumor capsule. In some cases (54.4%, 43/79), MIL-Bs just functional status of MIL-Bs, with PIL-Bs as the control. assembled outside the tumor capsule (Fig. 1C, middle), In 79 of 120 cases (65.8%), a ﬁbrous connective tissue whereas in others (45.6%, 36/79) MIL-Bs were found on surrounding the neoplastic mass, deﬁned as a tumor cap- both sides of the tumor capsule (Fig. 1C, bottom). It seemed

www.aacrjournals.org Clin Cancer Res; 19(21) November 1, 2013 5997

Shi et al.

that, in these 36 cases, MIL-Bs had penetrated the tumor encapsulation, tumor size, tumor number, and vascular capsule and interacted directly with tumor cells. invasion. In addition, the penetration of MIL-Bs through þ þ Serial sections were used to stain for CD20 B and CD8 the tumor capsule independently indicated prolonged OS T cells in the tumor margin. In line with a previous study (HR ¼ 0.44; 95% CI ¼ 0.20–0.97; P ¼ 0.042) as well as RFS þ (27), we found that MIL-Bs always colocalized with CD8 T (HR ¼ 0.47; 95% CI ¼ 0.23–0.97; P ¼ 0.040; Table 2). cells (Fig. 1D). A significant positive correlation (P < 0.001, Considering the positive correlation between CD20 and r ¼ 0.391) was revealed between the densities of the two cell CD8 in the tumor margin, we further evaluated their com- types in tumor margin (Fig. 1E), which implied a potential bined influence on patient outcome. Patients were classified cooperation between them in a tumor-killing effect. into four groups, using their median as the cutoff: I, CD20 low and CD8 low (n ¼ 36); II, CD20 low and CD8 high (n ¼ þ The association of CD20 cells with clinicopathologic 24); III, CD20 high and CD8 low (n ¼ 24); and IV, CD20 features high and CD8 high (n ¼ 36). Differences in both OS (P ¼ We further investigated the relationship of MIL-B density 0.002) and RFS (P ¼ 0.001) were significant among the four with clinicopathologic features, using the median as the groups (Fig. 2D). The 5-year OS (RFS, in brackets) rates were cutoff. Patients with high MIL-B density were prone to have 29% (21%), 32% (24%), 47% (50%), and 64% (69%) for small tumor size (P ¼ 0.004) and absence of vascular groups I, II, III, and IV, respectively. Thus, the data further invasion (P ¼ 0.041) (Table 1). Notably, patients with supported the hypothesis that the cooperation between þ þ MIL-Bs that had penetrated through the tumor capsule CD20 and CD8 cells could potently enhance the antitu- tended to have a single tumor (P ¼ 0.014) and absence of mor effect in hepatocellular carcinoma. vascular invasion (P ¼ 0.035; Table 1). All the correlations were further recapitulated in the validation cohort (Sup- Independent validation plementary Table S1). In contrast, PIL-B density showed no We further validated the prognostic value of MIL-Bs in obvious associations with these clinicopathologic features another cohort of 200 hepatocellular carcinoma cases. In except that high density of PIL-Bs was significantly corre- the validation cohort, a high density of MIL-Bs indicated lated with history of hepatitis (P ¼ 0.022; Table 1). How- significantly prolonged OS (5-year rate 41% vs. 19%, P < ever, the correlation did not reach significance in the val- 0.001) and RFS (5-year rate 55% vs. 33%, P ¼ 0.007), idation cohort (P ¼ 0.097; Supplementary Table S1). compared with patients with low MIL-B density (Supple- mentary Fig. S3A). Also, the penetration of MIL-Bs through þ Prognostic significance of infiltrating CD20 B cells the tumor capsule predicted better OS (5-year rate, 62% vs. The 1-, 3-, and 5-year OS and RFS rates of the 120 patients 31% for penetration and no penetration, respectively, P ¼ with hepatocellular carcinoma were 89%, 66%, and 44% 0.004) and RFS (5-year rate 56% vs. 27% for penetration and 77%, 54%, and 42%, respectively. In univariate anal- and no penetration, respectively, P ¼ 0.005; Supplementary yses, conventional clinicopathologic features that correlat- Fig. S3B). In particular, the multivariate Cox regression ed with a dismal OS and RFS were high alanine amino- model adjusted to established clinical factors authenticated transferase (ALT) level, large tumor size, tumor multiplicity, the density and penetration of MIL-Bs as independent absence of encapsulation, presence of vascular invasion, prognosticators (P < 0.05) for patients with hepatocellular and advanced tumor stages (Supplementary Table S2). As carcinoma (Table 2). expected, a significant positive correlation between the density of MIL-Bs and OS (P < 0.001) or RFS (P < 0.001) The phenotypes of MIL-Bs was detected (Fig. 2A). Patients with above-median level of Flow cytometry showed that PIL-Bs had a na€ve phe- þ þ MIL-Bs had a significantly higher 5-year OS (58% vs. 30% notype (IgD IgM IgG ), whereas MIL-Bs displayed an þ for high or low MIL-Bs, respectively) and RFS rates (63% vs. active mature phenotype (IgD IgM /lowIgG ;Fig.3A). 22% for high or low MIL-Bs, respectively). However, the This antigen-dependent switch from IgM and IgD to IgG density of PIL-Bs was associated with neither OS (P ¼ 0.689) production is a well-known feature of B-cell maturation nor RFS (P ¼ 0.572; Fig. 2B). Intriguingly, the penetration of (28).Inaddition,MIL-BswereCD27 and CD38 ,indic- MIL-Bs through the tumor capsule indicated a markedly ative of an atypical memory phenotype (described in the better OS (5-year rate, 69% vs. 36% for penetration and no discussion; Fig. 3A; ref. 27). penetration, respectively; P ¼ 0.003) and RFS (5-year rate, B cells can influence antitumor immunity by serving as 64% vs. 28% for penetration and no penetration, respec- APCs, which may involve MHC and costimulatory mole- tively; P ¼ 0.003; Fig. 2C). cules. We found that MIL-Bs and PIL-Bs expressed similar Then, variables with P < 0.1 in univariate analysis (includ- intensities of MHC class I (HLA-A, B and C), class II (HLA- ed tumor encapsulation with a borderline significance for DR) molecules, and CD40. However, MIL-Bs expressed RFS; P ¼ 0.115) were adopted as covariates in multivariate significantly higher levels of costimulatory molecules B7- Cox proportional hazards analyses. Multivariate analyses 1 (CD80; P < 0.001) and B7-2 (CD86; P < 0.001) compared revealed that the association of high MIL-B density with with PIL-Bs (Fig. 3A). The data suggested that, compared better OS [HR ¼ 0.44; 95% confidence interval (CI) ¼ 0.26– with PIL-Bs, the MIL-Bs were potentially more potent in 0.76; P ¼ 0.003] or RFS (HR ¼ 0.49; 95% CI ¼ 0.28–0.86; presenting antigen and activating T cells in the tumor P ¼ 0.012; Table 2) was independent of ALT level, tumor microenvironment (29).

5998 Clin Cancer Res; 19(21) November 1, 2013 Clinical Cancer Research

Margin-Inﬁltrating B Cells in Hepatocellular Carcinoma

Table 1. Correlation between CD20þ B cells and clinicopathologic characteristics (n ¼ 120)

MIL-Bs PIL-Bs Penetrationb

Characteristics Low High P Low High P No Yes P Gender Male 48 55 0.067 49 54 0.191 39 28 0.111 Female 12 5 11 6 4 8 Age 50 28 28 1.000 32 24 0.143 19 19 0.447 >50 32 32 28 36 24 17 History of hepatitis No 20 11 0.610 21 10 0.022 12 12 0.601 Yes 40 49 39 50 31 24 Liver cirrhosis No 11 14 0.500 13 12 0.822 10 7 0.681 Yes 49 46 47 48 33 29 ALT (U/L) 40 26 36 0.068 30 32 0.715 20 20 0.423 >40 34 24 30 28 23 16 AFP (ng/mL) 20 22 24 0.707 24 22 0.707 16 14 0.878 >20 38 36 36 38 27 22 Tumor size (cm) 5 37 51 0.004 41 47 0.215 30 28 0.422 >5 23 9 19 13 13 8 Tumor number Single 50 55 0.168 53 52 0.783 36 36 0.014a Multiple 10 5 7 8 7 0 Tumor encapsulation No 22 18 0.439 17 24 0.178 NA Yes 38 42 43 36 Vascular invasion No 30 41 0.041 38 33 0.353 21 26 0.035 Yes 30 19 22 27 22 10 Tumor differentiation I/II 42 43 0.841 44 41 0.547 29 28 0.307 III/IV 18 17 16 19 14 8 TNM stage I 32 40 0.136 38 34 0.456 23 26 0.088 II/III 28 20 22 26 20 10 BCLC stage 0/A 35 44 0.083 44 35 0.083 26 25 0.406 B/C 25 16 16 25 17 11

Abbreviations: ALT, alanine aminotransferase; AFP, alpha-fetoprotein; TNM, tumor–node–metastasis; BCLC, Barcelona Clinic Liver Cancer; NA, not applicable. aFisher exact tests; c2 test for all the other analyses. bc2 test for penetration was only applied in patients with tumor capsule (n ¼ 79).

The antitumor effect of MIL-Bs analysis showed that MIL-Bs mainly produced IFN-g (8.52 B cells can be directed to secrete polarized groups of 0.44%) and IL-12p40 (2.07 0.11%) rather than IL-2, IL- cytokines. Regulatory B cells are characterized by secreting 4, IL-6, or IL-10 (0.05%–0.5%; Fig. 3B), indicating a Be-1 IL-10 or TGF-b1, whereas effector B-cell populations are phenotype. However, PIL-Bs scarcely produced these cyto- grouped into two categories: (i) Be-1 cells producing cytokines (data not shown). All the data supported the notion kines such as IFN-g, IL-12, and TNF-a, and (ii) Be-2 cells that MIL-Bs could facilitate the Th1 immune response, releasing IL-2, IL-4, TNF-a, and IL-6 (30). Flow cytometric leading to strong antitumor activity.

www.aacrjournals.org Clin Cancer Res; 19(21) November 1, 2013 5999

Shi et al.

A n n 1.0 I MIL-Bs low ( = 60) 1.0 I MIL-Bs low ( = 60) II MIL-Bs high (n = 60) II MIL-Bs high (n = 60) 0.8 0.8 II 0.6 II 0.6 OS 0.4 RFS 0.4

0.2 I 0.2 P < 0.001 P < 0.001 I 0.0 0.0 0.0 20.0 40.0 60.0 80.0 100.0 0.0 20.0 40.0 60.0 80.0 100.0 Time after surgery (mo) Time after surgery (mo) B 1.0 I PIL-Bs low (n = 60) 1.0 I PIL-Bs low (n = 60) II PIL-Bs high (n = 60) II PIL-Bs high (n = 60) 0.8 0.8

0.6 0.6 I OS

I RFS 0.4 0.4 Figure 2. Kaplan–Meier curves of II II OS and RFS according to the 0.2 0.2 densities of: A, MIL-Bs; B, PIL-Bs; P P = 0.689 = 0.572 C, penetration of MIL-Bs through 0.0 0.0 tumor capsule; and D, the þ 0.0 20.0 40.0 60.0 80.0 100.0 0.0 20.0 40.0 60.0 80.0 100.0 combination of marginal CD20 þ Time after surgery (mo) Time after surgery (mo) and CD8 cell densities in the margin area. High density of MIL- n C I No penetration ( = 43) I No penetration (n = 43) Bs, the presence of penetration of 1.0 II Penetration (n = 36) 1.0 n II Penetration ( = 36) MIL-Bs and simultaneously high þ þ 0.8 0.8 densities of CD20 and CD8 cells II are signiﬁcantly associated with 0.6 0.6 II both prolonged overall and recurrence-free survival. P values OS RFS 0.4 0.4 are based on log-rank test. I I 0.2 0.2 P = 0.003 P = 0.003 0.0 0.0 0.0 20.0 40.0 60.0 80.0 100.0 0.0 20.0 40.0 60.0 80.0 100.0 Time after surgery (mo) Time after surgery (mo)

D I CD20 low+CD8 low (n = 36) I CD20 low+CD8 low (n = 36) II CD20 low+CD8 high (n = 24) II CD20 low+CD8 high (n = 24) n n 1.0 III CD20 high+CD8 low ( = 24) 1.0 III CD20 high+CD8 low ( = 24) IV CD20 high+CD8 high (n = 36) IV CD20 high+CD8 high (n = 36) 0.8 0.8 IV IV 0.6 0.6 III III OS 0.4 I RFS 0.4 II II 0.2 0.2 P = 0.002 P = 0.001 I 0.0 0.0 0.0 20.0 40.0 60.0 80.0 100.0 0.0 20.0 40.0 60.0 80.0 100.0 Time after surgery (mo) Time after surgery (mo)

In vitro assays showed that the viability of 97H cells was for each type used alone (data not shown). Similarly, þ significantly inhibited, when cocultured with either CD8 ELISAs revealed that the IFN-g level in the supernatant þ T cells or MIL-Bs, suggesting that B cells might exert direct of coculture group with both MIL-Bs and CD8 T cells þ killing effects similar to CD8 T cells. Moreover, the (65.26 5.54 pg/mL) was significantly higher than þ cooperation of MIL-Bs and CD8 T cells was substanti- the groups of either type of lymphocyte used alone ated by the findings that the combination of the two cell (P < 0.05; Fig. 4B), supporting a synergistic effect of the types caused a steeper decrease in tumor cell viability two cell types. during the first 3 days (P < 0.05; Fig. 4A). In addition, the Meanwhile, immunofluorescence on tissues and sorted inhibiting effect of the two cell types used in combination cells demonstrated obviously enhanced expression of gran- lasted for as long as 5 days compared with merely 3 days zyme B and TRAIL in MIL-Bs (Fig. 4C and D), both of which

6000 Clin Cancer Res; 19(21) November 1, 2013 Clinical Cancer Research

Margin-Inﬁltrating B Cells in Hepatocellular Carcinoma

Table 2. Multivariate analyses of factors associated with OS and RFS

OS RFS

Variables HR (95% CI) P HR (95% CI) P All patients in the initial cohort (n ¼ 120) ALT (U/L; 40 vs. >40) NS NS Size (cm; 5 vs. >5) NS 1.73 (1.01–2.97) 0.044 Tumor number (single vs. multiple) NS NS Tumor encapsulation (no vs. yes) 0.51 (0.31–0.85) 0.009 NS Vascular invasion (no vs. yes) 1.88 (1.13–3.15) 0.016 2.45 (1.47–4.08) 0.001 TNM stage (I vs. II/III) NA NA BCLC stage (0/A vs. B/C) NA NA MIL-Bs (low vs. high) 0.44 (0.26–0.76) 0.003 0.49 (0.28–0.86) 0.012 Patients with tumor capsule in the initial cohort (n ¼ 79) ALT (U/L; 40 vs. >40) 2.89 (1.36–6.16) 0.006 NS Size (cm; 5 vs. >5) NS NS Tumor number (single vs. multiple) 3.70 (1.38–9.93) 0.010 3.11 (1.10–8.79) 0.032 Tumor encapsulation (no vs. yes) NA NA Vascular invasion (no vs. yes) NS 2.21 (1.07–4.58) 0.032 TNM stage (I vs. II/III) NA NA BCLC stage (0/A vs. B/C) NA NA Penetration (no vs. yes) 0.44 (0.20–0.97) 0.042 0.47 (0.23–0.97) 0.040 All patients in the validation cohort (n ¼ 200) ALT (U/L; 40 vs. >40) NS NS Size (cm; 5 vs. >5) 1.68 (1.16–2.42) 0.006 NS Tumor number (single vs. multiple) 2.03 (1.25–3.30) 0.004 1.81 (1.03–3.18) 0.038 Tumor encapsulation (no vs. yes) NS NS Vascular invasion (no vs. yes) 1.83 (1.27–2.64) 0.001 1.94 (1.29–2.93) 0.002 TNM stage (I vs. II/III) NA NA BCLC stage (0/A vs. B/C) NA NA MIL-Bs (low vs. high) 0.53 (0.37–0.77) 0.001 0.60 (0.40–0.91) 0.015 Patients with tumor capsule in the validation cohort (n ¼ 93) ALT (U/L; 40 vs. >40) NS NS Size (cm; 5 vs. >5) 1.95 (1.14–3.32) 0.014 1.87 (1.01–3.44) 0.046 Tumor number (single vs. multiple) NS NS Tumor encapsulation (no vs. yes) NA NA Vascular invasion (no vs. yes) NS NS TNM stage (I vs. II/III) NA NA BCLC stage (0/A vs. B/C) NA NA Penetration (no vs. yes) 0.57 (0.34–0.97) 0.038 0.44 (0.23–0.83) 0.011

Abbreviations: ALT, alanine aminotransferase; BCLC, Barcelona Clinic Liver Cancer; CI, confidence interval; MIL-Bs, margin-infiltrating B cells; NA, not applicable; NS, not significant; OS, overall survival; RFS, recurrence-free survival.; TNM, tumor–node–metastasis. could have direct cytotoxicity against tumor cells through Discussion antibody-independent mechanisms. In contrast, PIL-Bs Most studies on the phenotype and functional activity of expressed low levels of granzyme B and TRAIL (Fig. 4C and infiltrating lymphocytes in tumors have been focused on T D). We further confirmed the results by flow cytometry. The lineage, while considerably less is known about infiltrating þ þ percentages of granzyme B (42.2 3.1%) and TRAIL cells B cells. Herein, for the first time, we presented a detailed þ (25.7 1.7%) in total CD20 cells were significantly higher investigation of the distribution, phenotype, and func- þ þ in MIL-Bs than in PIL-Bs (about 4-fold in granzyme B cells, tion of infiltrating CD20 B cells in human hepatocellular þ þ P < 0.001 and 3-fold in TRAIL cells, P ¼ 0.001, respectively, carcinoma. We observed that the CD20 B cells were n ¼ 5; Supplementary Fig. S4). The data implied that the concentrated around the tumor deposit and formed a microenvironment around the tumor margin could facili- dense cell layer in the invasive margin area, which simul- þ þ tate the activation and tumor-killing potential of CD20 taneously contained a high level of CD8 T cells. Impor- cells. tantly, we identified and validated the presence of MIL-Bs

www.aacrjournals.org Clin Cancer Res; 19(21) November 1, 2013 6001

Shi et al.

A MIL-Bs PIL-Bs

Lymphocyte gate + SSC-A SSC-A CD20 gate Cell count Cell count

FSC-A CD20 PerCP IgD IgM

Figure 3. Flow cytometric analyses of infiltrating B cells. A, representative expressions of surface markers on cells that þ Cell count Cell count Cell count Cell count were gated on CD20 cells. Cells were stained with indicated antibodies or isotype-matched IgG HLA-ABC HLA-DR CD40 controls (gray filled area). Red line, MIL-Bs; Blue line, PIL-Bs (n ¼ 10). B, Be-1/Be-2 cytokine expression profile in stimulated MIL-Bs using the same gating strategy in (A; n ¼ 10). For cytokine detection, þ

Cell count Cell count Cell count Cell count CD20 B cells were stimulated with ODN2006 (10 mg/mL, InvivoGen) and CD40L (1 mg/mL, R&D Systems) for 24 hours, followed CD80 CD86 CD27 CD38 by the addition of a leukocyte B activation cocktail (2 ml/mL, BD Pharmingen) before another 4-hour 0.08% culture. Quadrant gates were 0.42% established on the basis of isotype- matched controls, and the number Lymphocyte gate CD20+ gate in the quadrant indicates the SSC-A SSC-A IL-4 PE IL-2 APC percentage of cells.

FSC-A CD20 PerCP CD20 PerCP CD20 PerCP

0.1% 0.05% 8.37% 2.13% IL-6 PE IL-10 PE IFN- γ PE IL-12p40 APC

CD20 PerCP CD20 PerCP CD20 PerCP CD20 PerCP

þ and their penetration through the tumor capsule as inde- converting resting CD4 T cells to T-regulatory cells. One pendent favorable prognostic factors for patient survival reason for this discrepancy may lie in the functional statuses and tumor recurrence. of B cells in different contexts, as Be-1 and Be-2 have the These findings corresponded with several previous capacity to regulate Th1 or Th2 differentiation by secreting reports that infiltrating B cells were prevalent in human polarized arrays of cytokines (35, 36). cancers, recognizing a wide variety of tumor antigens, To understand how in situ MIL-Bs accumulation influ- associating closely with T cells and other immune cells, enced tumor immunity, we characterized the phenotype of and correlating with favorable outcomes (31, 32). However, these B cells. First, MIL-Bs displayed an active mature þ there were conflicting data suggesting that B cells had a phenotype (IgD IgM /lowIgG ), whereas PIL-Bs exhibited þ þ negative effect on protective antitumor responses and might ana€ve phenotype (IgD IgM IgG ). This phenomenon even facilitate tumor progression (33, 34). For instance, was regarded as that the MIL-Bs were undergoing Ig class Olkhanud and colleagues (34) identified a new subset of B switching in response to antigen exposure and were grad- cells, tumor-evoked regulatory B cells, which resembled Be- ually activated in the tumor margin. Second, MIL-Bs 2 phenotype and promoted breast cancer metastasis by expressed surface markers that were characteristic of APCs,

6002 Clin Cancer Res; 19(21) November 1, 2013 Clinical Cancer Research

Margin-Inﬁltrating B Cells in Hepatocellular Carcinoma

A B 100 110 P value 100 <0.001 80 Day 1 vs. 0.035 90 0.003 Day 2 vs. 0.045 60 80 Day 3 vs. 0.004 0.041 70 40 97H 97H+MIL-Bs 60 97H+MIL-Bs

+ IFN- γ (pg/mL)

Figure 4. % of viability Functional assays for 97H+marginal CD8 20 97H+marginal CD8+ 50 +MIL-Bs tumor-killing effect of MIL-Bs. A, 97H + +MIL-Bs +marginal CD8 97H + results of 97H cell viability from 40 0 +marginal CD8 cocultures performed in groups as Day 1 Day 2 Day 3 indicated. The percentage of C viability was plotted relative to DAPI CD20 Granzyme B Merge control 97H cells. Difference between groups was determined by one-way ANOVA. , P < 0.01; , P < 0.001. Six independent experiments with different patient samples were conducted in Margin tissue Margin triplicate. B, IFN-g production was measured by ELISA in the supernatant of the coculture groups. Results were similar on day 1, 2, or 3. , P < 0.05; , P < 0.01. Six independent experiments were MIL-Bs conducted in triplicate. C and D, immunoﬂuorescence images of margin tissue (n ¼ 12) and þ sorted CD20 B cells (n ¼ 5), showed the coexpression

of CD20 (red) with granzyme B PIL-Bs (green) or TRAIL (green). The top panels in (C) and (D) are for tissue immunofluorescence D (Magnification, 10; Yellow line, DAPI CD20 TRAIL Merge the tumor–normal border), whereas the middle and bottom are for sorted MIL-Bs and PIL-Bs, respectively (Magnification, 100). The low power micrographs show that the percentages of MIL-Bs positive for granzyme B and TRAIL tissue Margin are 43.2 4.3% and 36.2 3.8%, respectively (n ¼ 5), and both percentages of PIL-Bs positive for granzyme B and TRAIL are below

10% (n ¼ 5). MIL-Bs PIL-Bs

including MHC I, MHC II, CD40, CD80, and CD86, so that manipulatedtogenerateanidealantitumorresponse. MIL-Bs were capable of recognizing and presenting tumor- Third, the majority of the MIL-Bs failed to express CD38, associated antigens. The antigen spectrum of the B-cell further indicative of a memory phenotype (37). In addi- antigen receptor (BCR) is not limited to peptide antigens tion, the lack of CD27 expression of MIL-Bs suggested but also includes carbohydrates, phospholipids, nucleic that they might belong to an atypical memory B cell (38). acids, and larger structures, such as virus particles. Because CD38 is a biomarker of plasma cells, and CD27 is a BCR-mediated endocytosis could concentrate small quan- conventional memory B-cell marker with the capacity to tities of speciﬁc antigens and amplify the responses to tumor bind certain surface molecules on activated T helper cells antigens expressed at low levels, B cells could be more easily that induce Ig secretion in humoral immunity (39). It has

www.aacrjournals.org Clin Cancer Res; 19(21) November 1, 2013 6003

Shi et al.

been documented that the deficiency of CD38 and CD27 In conclusion, we demonstrated that an increased density expressions on B cells might restrict antibody production of MIL-Bs was associated with better survival and reduced but promote cytolytic antitumor responses (40, 41). recurrence in patients who had hepatocellular carcinoma. þ Coincidentally, the CD20 tumor-infiltrating lymphocyte The potential mechanisms included MIL-Bs serving as APC, þ with similar atypical memory phenotype was observed stimulating CD8 T cells by releasing proinflammatory in ovarian cancer as well (27). Furthermore, we demon- cytokines, and exerting a direct killing effect. Thus, an strated that MIL-Bs produced high levels of IFN-g and improved understanding of the mechanism of MIL-B IL-12p40 instead of IL-2, IL-4, IL-6, and IL-10. Taken recruitment to the tumor, MIL-B target antigens, and inter- together, we proposed that the MIL-Bs in hepatocellular actions of MIL-Bs with other immune cells might facilitate carcinoma might belong to theBe-1subset,whichmainly the design of more effective immunotherapies for hepato- participates in cellular-mediated antitumor immunity. cellular carcinoma. We also observed the intense expression of granzyme B and TRAIL on MIL-Bs in hepatocellular carcinoma, which Disclosure of Potential Conflicts of Interest suggested MIL-Bs might be involved in direct cytotoxicity. No potential conflicts of interest were disclosed. The secretion of granzyme B by human B cells required Authors' Contributions stimulation with IL-21 (42), whereas IFN-g stimulation or a Conception and design: J.-Y. Shi, J. Fan Toll-like receptor agonist was necessary for potent TRAIL Development of methodology: J.-Y. Shi, Z.-H. Min, Z.-B. Ding Acquisition of data (provided animals, acquired and managed patients, signaling (43). Therefore, the cytokine milieu in the hepa- provided facilities, etc.): Q. Gao, Z.C. Wang, J. Zhou, X. Wang, Y.-H. Shi, tocellular carcinoma invasive front might be conducive to G.-M. Shi, A.-W. Ke, K. Song the production of granzyme B and TRAIL, and program Analysis and interpretation of data (e.g., statistical analysis, biosta- tistics, computational analysis): J.-Y. Shi, Q. Gao, Z.C. Wang, Z.-H. Min, Z. MIL-Bs toward a status capable of killing tumor cells Dai, J. Fan directly. Writing, review, and/or revision of the manuscript: J.-Y. Shi, Q. Gao, Z.C. Survival analyses showed that the patient containing high Wang þ Administrative, technical, or material support (i.e., reporting or orga- levels of both MIL-Bs and CD8 T cells had a superior nizing data, constructing databases): J.-Y. Shi, J. Zhou, Z.-B. Ding, S.-J. prognosis to those with either cell type alone, suggesting Qiu, J. Fan þ þ cooperative interactions between CD20 and CD8 cells. Study supervision: S.-J. Qiu, J. Fan The close proximity and the similar cooperative effect þ þ Grant Support between tumor-infiltrating CD20 and CD8 lymphocytes This work was financially supported by the Major Program of the NSFC were seen in ovarian cancer (27) as well. In coculture (grant no. 81030038), National Key Sci-Tech Project (grant no. 2012ZX10002011-002), China National Funds for Distinguished Young experiments, a combination of the two cell types induced Scientists (grant no. 812250125), National Natural Science Foundation of a stronger and longer lasting inhibition of tumor growth China (grant nos. 81071992 and 81272730), and the Fok Ying–Tong than either of the two types used alone. Similarly, the Education Foundation (grant no. 132029). The costs of publication of this article were defrayed in part by the combined use of the two cell types increased IFN-g produc- payment of page charges. This article must therefore be hereby marked tion to the largest extent. As such, a strong cooperation advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate between the two types of lymphocytes made it valuable to this fact. explore the mechanisms underlying the interaction occur- Received November 11, 2012; revised August 12, 2013; accepted August ring in the invasive margin area. 17, 2013; published OnlineFirst September 20, 2013.

References 1. Maluccio M, Covey A. Recent progress in understanding, diagnosing, 7. Fu J, Zhang Z, Zhou L, Qi Z, Xing S, Lv J, et al. Impairment of CD4(þ) and treating hepatocellular carcinoma. CA Cancer J Clin 2012;62: cytotoxic T cells predicts poor survival and high recurrence rates in 394–9. patients with hepatocellular carcinoma. Hepatology 2013;58:139–49. 2. Forner A, Llovet JM, Bruix J. Hepatocellular carcinoma. Lancet 8. Shirabe K, Mano Y, Muto J, Matono R, Motomura T, Toshima T, et al. 2012;379:1245–55. Role of tumor-associated macrophages in the progression of hepa- 3. Gonzalez-Carmona MA, Lukacs-Kornek V, Timmerman A, Sha- tocellular carcinoma. Surg Today 2012;42:1–7. bani S, Kornek M, Vogt A, et al. CD40ligand-expressing dendritic 9. Kuang DM, Zhao Q, Wu Y, Peng C, Wang J, Xu Z, et al. Peritumoral cells induce regression of hepatocellular carcinoma by activating neutrophils link inflammatory response to disease progression by innate and acquired immunity in vivo. Hepatology 2008;48: fostering angiogenesis in hepatocellular carcinoma. J Hepatol 2011; 157–68. 54:948–55. 4. Cai XY, Gao Q, Qiu SJ, Ye SL, Wu ZQ, Fan J, et al. Dendritic cell 10. Kuang DM, Peng C, Zhao Q, Wu Y, Chen MS, Zheng L. Activated infiltration and prognosis of human hepatocellular carcinoma. J Cancer monocytes in peritumoral stroma of hepatocellular carcinoma promote Res Clin Oncol 2006;132:293–301. expansion of memory T helper 17 cells. Hepatology 2010;51:154–64. 5. GaoQ,ZhouJ,WangXY,QiuSJ,SongK,HuangXW,etal. 11. Gao Q, Qiu SJ, Fan J, Zhou J, Wang XY, Xiao YS, et al. Intratumoral Infiltrating memory/senescent T cell ratio predicts extrahepatic balance of regulatory and cytotoxic T cells is associated with prog- metastasis of hepatocellular carcinoma. Ann Surg Oncol 2012; nosis of hepatocellular carcinoma after resection. J Clin Oncol 2007; 19:455–66. 25:2586–93. 6. Yi Y, He HW, Wang JX, Cai XY, Li YW, Zhou J, et al. The functional 12. Shah S, Divekar AA, Hilchey SP, Cho HM, Newman CL, Shin SU, et al. impairment of HCC-infiltrating gammadelta T cells, partially mediated Increased rejection of primary tumors in mice lacking B cells: inhibition by regulatory T cells in a TGFbeta- and IL-10-dependent manner. of anti-tumor CTL and TH1 cytokine responses by B cells. Int J Cancer J Hepatol 2013;58:977–83. 2005;117:574–86.

6004 Clin Cancer Res; 19(21) November 1, 2013 Clinical Cancer Research

Margin-Inﬁltrating B Cells in Hepatocellular Carcinoma

13. Oizumi S, Deyev V, Yamazaki K, Schreiber T, Strbo N, Rosenblatt J, 28. Coffman RL, Cohn M. The class of surface immunoglobulin on virgin et al. Surmounting tumor-induced immune suppression by frequent and memory B lymphocytes. J Immunol 1977;118:1806–15. vaccination or immunization in the absence of B cells. J Immunother 29. Lanier LL, O'Fallon S, Somoza C, Phillips JH, Linsley PS, Okumura K, 2008;31:394–401. et al. CD80 (B7) and CD86 (B70) provide similar costimulatory signals 14. Kim S, Fridlender ZG, Dunn R, Kehry MR, Kapoor V, Blouin A, et al. B- for T cell proliferation, cytokine production, and generation of CTL. J cell depletion using an anti-CD20 antibody augments antitumor Immunol 1995;154:97–105. immune responses and immunotherapy in nonhematopoetic murine 30. Lund FE. Cytokine-producing B lymphocytes-key regulators of immu- tumor models. J Immunother 2008;31:446–57. nity. Curr Opin Immunol 2008;20:332–8. 15. Ladanyi A, Kiss J, Mohos A, Somlai B, Liszkay G, Gilde K, et al. 31. Milne K, Kobel M, Kalloger SE, Barnes RO, Gao D, Gilks CB, et al. Prognostic impact of B-cell density in cutaneous melanoma. Cancer Systematic analysis of immune infiltrates in high-grade serous ovarian Immunol Immunother 2011;60:1729–38. cancer reveals CD20, FoxP3 and TIA-1 as positive prognostic factors. 16. Mahmoud SM, Lee AH, Paish EC, Macmillan RD, Ellis IO, Green AR. PLoS ONE 2009;4:e6412. The prognostic significance of B lymphocytes in invasive carcinoma of 32. Nedergaard BS, Ladekarl M, Nyengaard JR, Nielsen K. A comparative the breast. Breast Cancer Res Treat 2012;132:545–53. study of the cellular immune response in patients with stage IB cervical 17. Nelson BH. CD20þ B cells: the other tumor-infiltrating lymphocytes. squamous cell carcinoma. Low numbers of several immune cell sub- J Immunol 2010;185:4977–82. types are strongly associated with relapse of disease within 5 years. 18. Wada Y, Nakashima O, Kutami R, Yamamoto O, Kojiro M. Clinico- Gynecol Oncol 2008;108:106–11. pathological study on hepatocellular carcinoma with lymphocytic 33. de Visser KE, Korets LV, Coussens LM. De novo carcinogenesis infiltration. Hepatology 1998;27:407–14. promoted by chronic inflammation is B lymphocyte dependent. Can- 19. Zhu XD, Zhang JB, Zhuang PY, Zhu HG, Zhang W, Xiong YQ, et al. High cer Cell 2005;7:411–23. expression of macrophage colony-stimulating factor in peritumoral 34. Olkhanud PB, Damdinsuren B, Bodogai M, Gress RE, Sen R, Wejksza liver tissue is associated with poor survival after curative resection of K, et al. Tumor-evoked regulatory B cells promote breast cancer hepatocellular carcinoma. J Clin Oncol 2008;26:2707–16. metastasis by converting resting CD4(þ) T cells to T-regulatory cells. 20. Zhao YM, Wang L, Dai Z, Wang DD, Hei ZY, Zhang N, et al. Validity of Cancer Res 2011;71:3505–15. plasma macrophage migration inhibitory factor for diagnosis and 35. Duddy ME, Alter A, Bar-Or A. Distinct profiles of human B cell effector prognosis of hepatocellular carcinoma. Int J Cancer 2011;129: cytokines: a role in immune regulation. J Immunol 2004;172:3422–7. 2463–72. 36. Harris DP, Goodrich S, Gerth AJ, Peng SL, Lund FE. Regulation of IFN- 21. Halama N, Michel S, Kloor M, Zoernig I, Benner A, Spille A, et al. gamma production by B effector 1 cells: essential roles for T-bet and Localization and density of immune cells in the invasive margin of the IFN-gamma receptor. J Immunol 2005;174:6781–90. human colorectal cancer liver metastases are prognostic for response 37. Choe J, Choi YS. IL-10 interrupts memory B cell expansion in the to chemotherapy. Cancer Res 2011;71:5670–7. germinal center by inducing differentiation into plasma cells. Eur J 22. Blair PA, Norena LY, Flores-Borja F, Rawlings DJ, Isenberg DA, Immunol 1998;28:508–15. Ehrenstein MR, et al. CD19(þ)CD24(hi)CD38(hi) B cells exhibit 38. Wei C, Anolik J, Cappione A, Zheng B, Pugh-Bernard A, Brooks J, et al. regulatory capacity in healthy individuals but are functionally A new population of cells lacking expression of CD27 represents a impaired in systemic Lupus Erythematosus patients. Immunity notable component of the B cell memory compartment in systemic 2010;32:129–40. lupus erythematosus. J Immunol 2007;178:6624–33. 23. Sun FX, Tang ZY, Lui KD, Ye SL, Xue Q, Gao DM, et al. Establishment of 39. Agematsu K, Nagumo H, Yang FC, Nakazawa T, Fukushima K, Ito S, a metastatic model of human hepatocellular carcinoma in nude mice et al. B cell subpopulations separated by CD27 and crucial collabo- via orthotopic implantation of histologically intact tissues. Int J Cancer ration of CD27þ B cells and helper T cells in immunoglobulin produc- 1996;66:239–43. tion. Eur J Immunol 1997;27:2073–9. 24. Tian J, Tang ZY, Ye SL, Liu YK, Lin ZY, Chen J, et al. New human 40. Hagn M, Schwesinger E, Ebel V, Sontheimer K, Maier J, Beyer T, et al. hepatocellular carcinoma (HCC) cell line with highly metastatic poten- Human B cells secrete granzyme B when recognizing viral antigens in tial (MHCC97) and its expressions of the factors associated with the context of the acute phase cytokine IL-21. J Immunol 2009; metastasis. Br J Cancer 1999;81:814–21. 183:1838–45. 25. Ke AW, Shi GM, Zhou J, Wu FZ, Ding ZB, Hu MY, et al. Role of 41. Sanz I, Wei C, Lee FE, Anolik J. Phenotypic and functional heteroge- overexpression of CD151 and/or c-Met in predicting prognosis of neity of human memory B cells. Semin Immunol 2008;20:67–82. hepatocellular carcinoma. Hepatology 2009;49:491–503. 42. Hagn M, Sontheimer K, Dahlke K, Brueggemann S, Kaltenmeier C, 26. Ng IO, Lai EC, Ng MM, Fan ST. Tumor encapsulation in hepatocellular Beyer T, et al. Human B cells differentiate into granzyme B-secreting carcinoma. A pathologic study of 189 cases. Cancer 1992;70:45–9. cytotoxic B lymphocytes upon incomplete T-cell help. Immunol Cell 27. Nielsen JS, Sahota RA, Milne K, Kost SE, Nesslinger NJ, Watson PH, Biol 2012;90:457–67. et al. CD20 þtumor-infiltrating lymphocytes have an atypical CD27- 43. Kemp TJ, Moore JM, Griffith TS. Human B cells express functional memory phenotype and together with CD8þ T cells promote favorable TRAIL/Apo-2 ligand after CpG-containing oligodeoxynucleotide stim- prognosis in ovarian cancer. Clin Cancer Res 2012;18:3281–92. ulation. J Immunol 2004;173:892–9.

www.aacrjournals.org Clin Cancer Res; 19(21) November 1, 2013 6005

Margin-Infiltrating CD20+ B Cells Display an Atypical Memory Phenotype and Correlate with Favorable Prognosis in Hepatocellular Carcinoma

Jie-Yi Shi, Qiang Gao, Zhi-Chao Wang, et al.

Clin Cancer Res 2013;19:5994-6005. Published OnlineFirst September 20, 2013.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-12-3497

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2013/09/23/1078-0432.CCR-12-3497.DC1

Cited articles This article cites 43 articles, 13 of which you can access for free at: http://clincancerres.aacrjournals.org/content/19/21/5994.full#ref-list-1

Citing articles This article has been cited by 11 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/19/21/5994.full#related-urls

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 Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/19/21/5994. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.