Distribution Patterns of Dendritic Cells and T Cells in Diffuse Large B-Cell Lymphomas Correlate with Prognoses

The most common subtype of non-Hodgkin's lymphoma is diffuse large B-cell lymphoma (DLBCL; ref. 1). The mainstay strategy for treatment of DLBCL is multidrug chemotherapy. However, long-time survival was achieved in only ∼40% of patients, warranting the development of novel therapeutic strategies such as monoclonal antibodies (2) and tumor vaccines by pulsing autologous dendritic cells with tumor antigens (3, 4). This approach had been extended to equipment of tumor cells with costimulatory molecules or cytokines that activate dendritic cells (3, 5, 6). Furthermore, the use of autogeneic T cells to eradicate lymphoma cells seems to be effective in animal experiments (7). However, little is known about the effect of tumor immunity on DLBCL in human.

In animal models, the mice, which were immunized with irradiated tumor cells and challenged with viable cells of the same tumor, seemed to reject a normally lethal dose of that tumor (3). That finding led to the discovery of the tumor rejection antigen, which is usually specific for an individual tumor. In DLBCL, the tumor-specific antigen is the unique idiotype antigen derived from the B-cell receptor expressed by the neoplastic clone (8). Through the presentation to T cells by self MHC molecules, this idiotype antigen could be potentially recognized as a target by host immunity. Given that the idiotype antigen is so unique and different among individual cases of DLBCL, the expression of MHC molecules in DLBCL cells can be a surrogate marker for lymphoma-specific antigens. It has been found that loss of HLA-DR expression in DLBCL correlated with immune evasion and poor patient survival (9, 10).

Dendritic cells are the most potent antigen-presenting cells. By expressing high levels of MHC and costimulatory molecules, dendritic cells can present tumor-associated antigens to elicit T-cell–mediated tumor destruction (11). Dendritic cells are heterogeneous in human. They are divided into myeloid (mDC/DC1) and plasmacytoid (pDC/DC2) dendritic cells by lineage and their ability to promote T helper subset 1 or subset 2 development, respectively. Dendritic cells also undergo progressive maturation and have been divided into immature [CD1a⁺/CD83^low/dendritic cell–specific lysosome-associated membrane protein negative (DC-LAMP⁻)/dendritic cell–specific intercellular adhesion molecule-3 grabbing nonintegrin positive (DC-SIGN⁺)/cytoplasmic MHC II molecule] versus mature [CD1a^+/−/CD83^high/DC-LAMP⁺/DC-SIGN⁻/surface MHC II molecule] phenotypes (12–15).

The basis of antitumor cellular immunotherapy is that tumor cell antigens can be phagocytized by immature dendritic cells, which then present tumor antigens to tumor-specific T cells following maturation. Both CD4⁺ T cells and cytotoxic CD8⁺ T cells need to participate in this antitumor effect (3, 12). Because dendritic cells elicit both tumor-specific CD4⁺ and CD8⁺ T cells, and both types of T cells are indispensable for lymphoma immunity (10, 16, 17), assessment of CD45RO⁺ effector or memory T cells may reflect the final common pathway of cell-mediated antitumor response. The presence of dense infiltrating CD45RO⁺ T cells in colorectal cancers has been found to correlate with increased survival (18).

In this study, by immunohistochemical staining, we showed for the first time that the presence of CD1a⁺ dendritic cells and increased granzyme B⁺ T cells within tumors and the patterns of denser S100⁺ cells and CD45RO⁺ T cells around the tumor edge correlated with a favorable prognosis. The latter T-cell distribution pattern was associated with tumor expression of HLA-DR. These findings indicate that in ∼20% of DLBCL there is existence of antitumor immune response, and enhancing this effect could be a target for DLBCL immunotherapy.

Tissue samples. Forty-eight cases of primary pretreated DLBCL were recruited from the archival files at the National Cheng Kung University Hospital from February 1995 to December 1998. The specimens were fixed in 10% neutral formalin and embedded in paraffin. Part of the cases were studied previously (19). Studies were carried under a laboratory protocol approved by the institutional review board and were in accordance with the Helsinki Declaration of 1975, which was revised in 1983. Primary nodal and extranodal lymphomas were previously defined (19, 20).

Clinical data were obtained by chart review, including sex, age, serum level of lactate dehydrogenase (LDH), tumor site, Ann Arbor stage, treatment modality, and overall survival in months. There were 27 male and 21 female patients with a mean age of 60.6 years (range, 26-86 years). All the patients with available information were followed up and the duration ranged from 1 to 87 months. All the lymphoma patients were typically treated with a curative cyclophosphamide, doxorubicin, vincristine, and prednisone regimen, as described in our previous study (19). For selected patients, surgical intervention and/or radiotherapy preceded chemotherapy. None of the patients has received rituximab administration.

Immunohistochemical staining. Immunohistochemical staining was done on deparaffinized tissue sections of formalin-fixed material following microwave-enhanced epitope retrieval. After blocking of endogenous activity, sections were then incubated with primary antibodies at room temperature for 2 h. Detection was done with streptavidin-biotinylated peroxidase-conjugated reagents (LSAB⁺ kit, DAKO) with 3-amino-9-ethylcarbazole as the chromogen and hematoxylin for counterstain. The selected primary antibodies, which can be used in formalin-fixed tissues, are summarized in Table 1. The germinal center immunophenotype was determined by the combined expression of CD10 and bcl-6, as described in our previous study (19). Images were photographed using an Olympus DP12 Digital Microscope Camera (Olympus Co.) and processed with Adobe Photoshop version 7.0 software (Adobe Systems, Inc.).

Measurements. Quantitative evaluation of cells positive for CD1a, CD83, and DC-LAMP was assessed by counting at three high-power fields (HPF; 400× magnifications) on higher density areas, as previously described (15, 21). Because most of the cases showed few positive cells, we also used S100 to highlight dendritic cells and monocytic forms (22–24). Only the cells with dendritic cytoplasmic processes were counted at 10 HPFs. For cases showing difference in cell density between intratumoral and peritumoral areas, both areas were evaluated. Staining results were evaluated and semiquantified for HLA-DR, CD10, and bcl-6. For HLA-DR, unequivocal membrane staining in ≥30% of tumor cells was graded as positive (10). The cutoff value was 10% for CD10 and bcl-6 (19). Intratumoral CD45RO⁺ and granzyme B⁺ T cells were counted at 10 fields under oil immersion (×1,000). For cases also showing difference in cell density between intratumoral and peritumoral areas, both areas were evaluated.

EBV in situ hybridization.In situ hybridization study was done to detect EBV-encoded RNA 1 using a PCR-derived digoxigenin-labeled DNA probe (25, 26). The test for nucleotide integrity was done by the use of the RNA-positive control probe (Ventana Medical System, Inc.).

Statistical analysis. Appropriate statistical tests were used to examine the relationships and correlations between variables, including Student's t tests and Pearson or Kendall's τ correlation. The overall survival was measured from initial diagnosis to death from any cause, with follow-up data of surviving patients assessed at the last contact date. Estimates of overall survival distribution were calculated using the method of Kaplan and Meier (27). Time-to-event distributions were compared using the log-rank test (28). Cox proportional hazard model was used to test the simultaneous influence on overall survival of all covariates found to be significant in the univariate analysis (29). The P value referred to is two sided. The analyses were carried using SPSS 13.0 statistical software (SPSS, Inc.).

The clinicopathologic features of 48 lymphoma cases are summarized in Table 2.

Staining patterns of dendritic cells and T cells in DLBCL. Dendritic cells expressing CD1a, CD83, or DC-LAMP were few in number in most DLBCL tissues. The distribution of these cells was mainly aggregated around lymphatics or blood vessels and scarce in tumor beds. Most of the intratumoral dendritic cells showed mDC-type morphology (Fig. 1). The CD1a⁺, CD83⁺, and DC-LAMP⁺ dendritic cell numbers in the tumor bed ranged from 0 to 11 (mean, 0.4; median, 0), 0 to 25 (mean, 3.4; median, 0), and 0 to 5 cells (mean, 0.2; median, 0) per HPF (×400) and the positive rates were 21% (10 of 48), 46% (22 of 48), and 10% (5 of 48), respectively, based on the criteria of unequivocally positive cell(s) in 2 or more separate HPFs or ≥2 positive cells in one HPF. On the contrary, S100⁺ cell numbers ranged from 0 to 79 cells (mean, 23; median, 16) per HPF. Notably, there were 9 of 48 (19%) cases showing a characteristic pattern of denser S100⁺ cells around the tumor edge than in the tumor bed (Fig. 2A). The difference in this rimming pattern was statistically significant (P < 0.05, paired t test). This increased peritumoral accentuation was also noted for CD45RO⁺ T cells in 10 of 48 (21%) cases (Fig. 2B; P < 0.001, paired t test). There were five cases showing peritumoral accentuation for both S100⁺ and CD45RO⁺ cells. In addition, there was positive correlation between the density of S100⁺ cells and that of CD45RO⁺ T cells within the tumor (Pearson correlation coefficient r = 0.296; P = 0.041). The numbers of CD45RO⁺ and granzyme B⁺ T cells (Fig. 2C) in the tumor bed ranged from 2 to 99 (mean, 31; median, 25) and 0 to 76 (mean, 14; median, 5) per HPF (×1,000) and were positively correlated (Pearson correlation coefficient r = 0.353; P = 0.016). No pattern of peritumoral accentuation was noted for granzyme B stain.

HLA-DR staining pattern in DLBCL. Tumor cells showed predominantly membranous and occasional cytoplasmic staining for HLA-DR in 18 of 48 (38%) cases (Fig. 2D). The HLA-DR⁺ tumor cells were diffuse in distribution in most cases.

Correlation of dendritic cells with T cells and clinical features. The rimming pattern of denser CD45RO⁺ T cells around the tumor edge and fewer in the tumor bed was correlated with tumor expression of HLA-DR (Kendall's τ correlation coefficient r = 0.446; P = 0.002), presence of CD83⁺ dendritic cells (r = 0.341; P = 0.016) and CD1a⁺ dendritic cells (r = 0.59; P < 0.001) in tumor beds, and the rimming pattern of S100⁺ cells (r = 0.419; P = 0.003). The latter S100⁺ pattern was also correlated with presence of DC-LAMP⁺ dendritic cells (r = 0.363; P = 0.009) and CD1a⁺ dendritic cells (r = 0.418; P = 0.002) in tumor beds. Notably, there was inverse correlation between extranodal lymphoma and CD45RO⁺ T-cell density (r = −0.407; P = 0.001) and between LDH level and presence of CD1a⁺ dendritic cells (r = −0.301; P = 0.012). That is, in extranodal lymphomas there were fewer CD45RO⁺ T cells and dendritic cells, although the decrease of dendritic cells was not statistically significant. Presence of CD1a⁺ dendritic cells was associated with increased granzyme B⁺ T cells in tumor beds (r = 0.327; P = 0.037). For clinical features, tumor stage was positively correlated with LDH level (r = 0.349; P = 0.002).

Correlation of survival with clinicopathologic factors in DLBCL patients. The clinicopathologic features affecting patient survival are summarized in Table 3. In the univariate analysis, the significant parameters related to a favorable prognosis included low-stage tumors (stage I/II), low LDH level (<200 IU/L), expression of HLA-DR by tumor cells, the presence of CD1a⁺ dendritic cells within tumors, the rimming pattern of S100⁺ cells, the rimming pattern of CD45RO⁺ T cells, and the increased granzyme B⁺ T cells within tumors (Fig. 3). In multivariate analysis, the factors significantly associated with better outcome were the presence of CD1a⁺ dendritic cells within tumors (hazard ratio, 10.9; 95% confidence interval, 1.4-84.1), the rimming pattern of CD45RO⁺ T cells (hazard ratio, 3.0; 95% confidence interval, 1.1-8.1), and the increased granzyme B⁺ T cells within tumors (hazard ratio, 3.5; 95% confidence interval, 1.0-12.2). Neither higher number of S100⁺ cells nor denser CD45RO⁺ T cells in tumor beds was related to a favorable prognosis.

With the advent of tumor immunobiology, vaccination for lymphoma treatment was developed (30–32). However, little is known about the antilymphoma immunity in human. In this study, we showed that the presence of CD1a⁺ dendritic cells and increased granzyme B⁺ T cells in tumor beds and the rimming pattern of denser S100⁺ cells and CD45RO⁺ T cells around the tumor edge were correlated with a favorable prognosis for DLBCL patients. Extranodal lymphoma elicited fewer CD45RO⁺ T-cell and dendritic cell infiltrates.

There are several studies drawing the issue about dendritic cells and T cells in B-cell lymphomas (22, 33, 34). However, we found for the first time that the peritumoral distribution of S100⁺ cells and CD45RO⁺ T cells was correlated with a favorable prognosis in DLBCL. This rimming pattern can be a prognostic indicator and, most importantly, a target for immunotherapy. Our study further clarified that increased recruitment of S100⁺ cells and T cells in the tumor bed was not sufficient for a better outcome. Consistent with this regard, it has been noted that patients with T-cell/histiocyte–rich LBCL and “host response” type DLBCL did not have better prognoses (34, 35).

The lymphomas in extranodal sites showed fewer CD45RO⁺ T-cell infiltrates. This finding was interesting but not found to be significant on univariate and multivariate analyses for outcome. The reason may be likened to that higher T-cell infiltrates in tumor beds did not necessarily correlate with a better outcome. In addition, it seems that nodal lymphomas have a different pattern of T-cell infiltration. The inclusion of extranodal cases, which typically have far fewer infiltrating T cells at some tissue sites, may be a confounding variable in the multivariate analysis. In contrast to no role of intratumoral CD45RO⁺ T-cell infiltrates, granzyme B⁺ T-cell infiltrates within tumors were significant for better outcome.

It may be of concern that the numbers of cells stained for dendritic cell markers were relatively few and the S100 was not entirely specific as a dendritic cell surrogate. It has been found that the sensitivity of immunostaining on paraffin-embedded tissue sections was lower than on frozen tissue sections (36). Furthermore, the dendritic cell numbers in this study were approximate to those in previous studies (15, 37). Dendritic cells are heterogeneous in human. Given the shifts in intensity in dendritic cell marker expression during dendritic cell differentiation and maturation during nodal transit, there is no one marker that detects all dendritic cell stages reliably in paraffin sections. Therefore, we have included several different markers and found that S100, which was counted only in dendritic forms, can give an independent and perhaps broader measure of the dendritic cell number at all maturation stages.

Tumor stage and LDH level are two well-recognized factors for prognosis (1). As shown in univariate analysis (Table 3), we further found HLA-DR, CD1a⁺ dendritic cells, granzyme B⁺ T cells, and the rimming patterns of S100⁺ cells and CD45RO⁺ T cells to be relevant for prognosis, although only CD1a⁺ dendritic cells, the rimming pattern of CD45RO⁺ T cells, and intratumoral granzyme B⁺ T cells were significant in multivariate analysis. This was likely related to the general association of HLA-DR and the S100 rimming pattern with the presence of CD1a⁺ dendritic cells in tumors. Furthermore, the majority of cases (70%) containing CD1a⁺ immature dendritic cells also recruited CD83⁺ mature dendritic cells in tumors. Because neoplasia is a chronic process and prolonged activation of mature dendritic cells induces their “exhaustion” or “paralysis” (14, 38, 39), the finding of CD1a⁺ dendritic cells in tumor may thus indicate the importance of persistent recruitment of young dendritic cells into tumors to maintain the antitumor immunity. This finding may also underscore the phagocytic activity of CD1a⁺ dendritic cells for the contribution of antitumor immune reactions (21, 40). The peritumoral recruitment of immature or mature dendritic cells has been found in cutaneous melanomas and breast carcinomas (15, 21). However, in those studies there were no survival data to highlight the clinical significance of the dendritic cells. To the best of our knowledge, this is the first study drawing the prognostic implication of dendritic cells on the DLBCL patients.

Previous studies have shown that tumors prevent dendritic cell maturation and increased numbers of immature dendritic cells in the tumors may be associated with worse outcomes due to the tolerizing effects of these cells (37, 41). In this study, the majority of DLBCL containing CD1a⁺ immature dendritic cells also recruited CD83⁺ mature dendritic cells in tumors and bore a favorable outcome, and the numbers of CD83⁺ dendritic cells were greater than those of CD1a⁺ dendritic cells. Taken together, it suggests that maturation and differentiation of intratumoral dendritic cells can be achieved in a subset of DLBCL and might serve as a therapeutic target. On the other hand, the different findings of the correlation of immature dendritic cells with tumor prognosis may result from the heterogeneity and diversity of dendritic cells as well as the different markers used for detection.

Lost or low expression of HLA-DR on DLBCL cells has been found to be correlated with poor patient survival (10). The decreased tumor immunosurveillance was considered as a result of diminished tumor-infiltrating CD8⁺ T cells. In parallel, we also found a similar result for HLA-DR to correlate with a better outcome and with rimming pattern of CD45RO⁺ T cells.

In conclusion, we showed that in ∼20% of DLBCL cases, the presence of CD1a⁺ dendritic cells within tumors and the rimming effects of S100⁺ cells and CD45RO⁺ T cells were correlated with a favorable prognosis. The latter T-cell distribution pattern was associated with tumor expression of HLA-DR. Taken together, it may imply that adjuvant dendritic cell and T-cell immunotherapy may further boost treatment responses.

We thank Mi-Chia Ma (Associate Professor, Department of Statistics, National Cheng Kung University, Tainan, Taiwan) for advice with statistical interpretation.