Prognostic Value of Tumor-Infiltrating Dendritic Cells in Colorectal Cancer: Role of Maturation Status and Intratumoral Localization

Materials and Methods

Tissue samples. Formalin-fixed, paraffin-embedded tissue samples were obtained from a randomly selected group of 104 patients with colorectal cancer, who underwent curative surgery at Leiden University Medical Center (Leiden, the Netherlands) from 1980 to 1992. All tumor samples were coded to conserve patient confidentially. Clinical information and histopathologic information is shown in Table 1. Patient follow-up was completed until January 2003, with a median follow-up of 5.4 years (range, 0.1-18.6 years; SD, 5.2 years). Tumors were evaluated for quantity and localization of CD4-, CD8-, CD56-, and CD57-positive lymphocytes, lymphocytic infiltration according to Jass criteria on H&E staining, differentiation grade and mucinous characteristics.

Immunohistochemistry. Paraffin sections (4 μm thick) were mounted on aminopropylethoxysilane-coated slides and dried overnight at 37°C. Tissue sections were deparaffinized and rehydrated. Endogenous peroxidase was blocked for 20 minutes in 0.3% hydrogen peroxide/methanol. After washing in PBS, the slides were incubated overnight at room temperature with primary antibodies directed against S-100 protein (rabbit anti-cow, DAKO, Glostrup, Denmark), HLA-DR-DP (clone HL40, Sanbio, Uden, the Netherlands), CD1a (clone O10, Immunotech, Marseilles, France), and dendritic cell-lysosome-associated membrane glycoprotein (CD208, clone 104.G4, Immunotech). All dilutions of antibodies and conjugates were done in PBS containing 1% (w/v) bovine serum albumin. After incubation with these antibodies, sections were washed in PBS. The sections with HLA-DR-DP were incubated with Envision peroxidase anti-mouse (DAKO) for 30 minutes. Sections with S-100 protein were incubated with Swine anti-rabbit labeled with biotin (DAKO) for 30 minutes, washed in PBS, and subsequently incubated with streptavidin biotinylated horseradish peroxidase (DAKO) for another 30 minutes. All sections were then washed in PBS, rinsed in 0.05 mol/L Tris-HCl buffer (pH 7.6) for 5 minutes, and developed in 3,3′-diaminobenzidine with 0.0018% (v/v) hydrogen peroxide for 10 minutes. The reaction was stopped by washing with demineralized water. Subsequently, the sections were incubated at 37°C in a 0.4% pepsin in 0.01 N HCl solution for 20 minutes, after which the pepsin reaction was stopped in demineralized water. After washing in PBS, the tissue sections were incubated overnight at room temperature with a rabbit polyclonal antibody against human laminin (Sigma, St. Louis, MO). After this incubation, the sections were washed in PBS, incubated for 30 minutes with Swine anti-rabbit labeled with biotin, and washed with PBS. Subsequently the sections were incubated for 30 minutes with streptavidin-biotin complex (DAKO) labeled with horseradish peroxidase. After washing in PBS, the sections were developed for 12 minutes at 50°C in a buffered Tris-HCl (pH 7.6) solution containing, per 100 mL, (a) 40 mg 4-chloro-1-naphtol (Merck, Darmstadt, Germany) dissolved in 200 μL dimethylformamide (Baker BV, Deventer, the Netherlands) and 300 μL ethanol (Merck) and (b) 100 μL of a 30% (v/v) hydrogen peroxide (Merck). The reaction was stopped by washing with demineralized water and counterstained with methyl green (Klinipath, Duiven, the Netherlands). The slides were dried overnight at room temperature. The sections were mounted in Kaiser's glycerin (Merck).

A negative control was included for each tumor sample by using PBS/bovine serum albumin instead of S-100 and laminin or HLA-DR-DP and laminin antibodies in the overnight incubations.

Microscopic evaluation of tumor sections. All cell counts were done using a Zeiss Axioscope (Sliedrecht, Netherlands) at a ×200 magnification (×20 objective and ×10 eyepiece). Cases were scored blindly with respect to patient history, presentation, and previous scoring by two independent observers. Dendritic cells were identified by their brown membranous staining pattern, ovoid nuclei, and cytoplasmic flame-like extensions. Structures that were morphologically distinctly distinguishable from dendritic cells, like myenteric plexuses, nerve sheaths, and granulocytes, were excluded. For each section, 25 areas of a representative field of tumor were assessed using an ocular grid comprising a high-power field area of 0.38 mm². Assisted by the laminin staining (stroma, basal membrane-like structure), tumor areas were divided into three anatomic compartments (i.e., tumor epithelium, tumor stroma, and advancing tumor margin). For each case, the total number of dendritic cells per square millimeter tumor or stroma was calculated for all three compartments. A serial H&E-stained section was examined for orientation and confirmation of the histologic diagnosis.

Statistical analysis. A group with high and low dendritic cell infiltration was distinguished based on the 75th percentile of the average dendritic cell infiltration scores of all patients. All statistical analyses were done using the SPSS software package (SPSS, Inc., Chicago, IL). Disease-free survival data were analyzed using Kaplan-Meier survival estimation, and the log-rank test was used for comparison of the survival curves. Multivariate analysis with Cox regression was done to assess the prognostic significance of dendritic cell infiltration in the various compartments, patient factors, and tumor features for overall and disease-free survival. Statistical analysis between groups was done using the χ² test for comparing proportions. Correlations between continuous variables were evaluated using Spearman rank correlation test. Ps < 0.05 were considered significant. Ps < 0.10 were considered near significant.

Results

Patient and tumor characteristics. A nonselected panel of 104 primary tumors of colorectal origin was investigated. The patients' characteristics and clinicopathologic variables are shown in Table 1. The panel consisted of about equal numbers of stage II (Duke's B; n = 51, 49%) and stage III (Duke's C; n = 53, 51%) tumors. The average age of the patients was 66.9 years (range, 26.0-85.0 years). Tumor stage correlated inversely with the time of disease-free survival (P = 0.002; data not shown). None of the other patient or tumor characteristics as described in Table 1 correlated with DFS.

HLA class II on tumor epithelial cells. In 29 cases (28%), tumor epithelium stained positive for HLA class II (Fig. 1D). These tumors could not be evaluated for intraepithelial infiltrating HLA class II–positive immune cells. HLA class II expression by the tumor cells correlated significantly with the intraepithelial infiltration of CD1a-positive dendritic cells (P = 0.03). As an internal control, no correlation was found between CD1a-positive dendritic cells and HLA class II expression in tumor stroma and advancing border. No significant relationship was found between HLA class II expression on tumor cells and other tumor-infiltrating immune cells and clinicopathologic or survival variables.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Tumor-infiltrating dendritic cells in human colorectal cancer in a double stain with dendritic cell markers (brown) and laminin (blue). A, intraepithelial CD208-positive mature dendritic cells (brown; closed arrow). Magnification, ×400. B, multiple CD1a-positive dendritic cells (brown) at the advancing margin of tumor and muscularis propria (open arrow). Magnification, ×200. C, S-100-positive dendritic cells (brown) in tumor epithelium and stroma. Note the typical dendritic cells morphology. D, tumor epithelial cells positive for HLA class II. Inset, tumor negative for HLA class II. Only few nonmalignant positive cells, mainly in the stromal compartment, can be observed. T, tumor tissue. S, tumor stroma. M, muscularis propria.

Tumor-infiltrating dendritic cells expressing S-100, HLA class II, CD1a, or CD208 in colorectal cancer tissue. Cells stained with S-100 and HLA class II antibody were found in 104 (100%) and 102 (98%) of 104 primary colorectal tumor samples. Positively stained cells often showed a typical dendritic morphology (Fig. 1C). HLA class II could only be quantified in the epithelial compartment due to the diffuse staining pattern in the stromal and advancing border compartments. In addition, we stained with CD208 (mature dendritic cells) and CD1a antibodies. These cells were found in 75 (72%) of 104 and 94 of 100 (94%) primary colorectal tumor samples, respectively (Table 2). Four samples could not be analyzed for CD1a due to technical procedure failures. Table 2 shows the mean numbers of tumor-infiltrating dendritic cells in different compartments. Dendritic cells, characterized by the expression of CD1a, resided in the advancing tumor margins (Fig. 1B) in relatively high numbers and were evenly distributed between tumor stroma and tumor epithelium, whereas mature dendritic cells expressing CD208 were sparsely present in the tumor epithelium (Fig. 1A) but mainly distributed in the tumor stroma and advancing tumor margin.

Correlation between tumor-infiltrating dendritic cells and clinicopathologic variables. Next, we examined the correlation between the presence of tumor-infiltrating dendritic cells in the different tumor compartments and clinicopathologic variables, including patient age, gender, tumor location, tumor stage, and tumor differentiation (Table 3). Data on S-100 and HLA class II dendritic cells are not shown in the tables but are presented in our previous study (11). For each dendritic cell marker, patients were divided into two groups, high versus low dendritic cell infiltration, with 75th percentile as cutoff point. The infiltration of S-100 dendritic cells and CD208 dendritic cells in the tumor epithelium was significantly higher in right-sided tumors (P = 0.03 and 0.02, respectively).

Patients with intraepithelial infiltration of CD208 had a higher disease stage (P = 0.03). Patients with a relatively high CD1a dendritic cell infiltration in the advancing tumor margin had a significantly lower leukocyte infiltration according to the Jass criteria (P = 0.009).

Correlation between tumor-infiltrating dendritic cells and other tumor-infiltrating lymphocytes. Tumor-infiltrating T lymphocytes and natural killer cells are generally considered to be the effector immune cell population in an antitumor immune response. To study whether tumor-infiltrating dendritic cells affect the number and behavior of these tumor-infiltrating leukocytes in colorectal cancer, we evaluated whether dendritic cell infiltration and maturation status correlated to the number of these tumor-infiltrating leukocytes. High infiltration of CD1a dendritic cells in the tumor epithelium was significantly correlated to the intraepithelial infiltration of CD4 lymphocytes (P = 0.006) and showed a trend toward intraepithelial infiltration of CD8 lymphocytes (P = 0.08; nonsignificant; Table 4). The number of infiltrating S-100 dendritic cells in the tumor epithelium was also positively correlated to the number of intraepithelial infiltrating CD4 and CD8 lymphocytes (P = 0.02 and 0.01 respectively; data not shown) as presented in our previous study (11). All other correlations between tumor-infiltrating dendritic cells and other immune cells were not significant.

Prognostic significance of tumor-infiltrating dendritic cells. The number of tumor-infiltrating S-100-positive dendritic cells did not affect the disease outcome regarding overall or disease-free survival. In univariate analysis, patients showing a relatively high number of mature (i.e., CD208-positive) infiltrating dendritic cells in the tumor epithelium had a shorter overall survival (P = 0.004) as shown in Fig. 2B. In addition, patients with relatively high numbers of CD1a-positive dendritic cells in the advancing margin of the tumor had a shorter disease-free survival (P = 0.03) as shown in Fig. 2A. Multivariate analysis revealed that both intraepithelial CD208 and advancing border CD1a were independent from other prognostic patient and tumor characteristics.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Kaplan-Meier survival curves of colorectal cancer patients according to the presence of tumor-infiltrating (A) CD1a-positive dendritic cells and (B) CD208-positive dendritic cells. The 75th percentile of the infiltration number was chosen as a cutoff point to distinguish a group with high and low infiltration (dashed and continuous lines, respectively). AB, advancing border; IEL, intraepithelial.

Discussion

Our study indicated that the presence or absence of different populations of tumor-infiltrating dendritic cells correlated with the clinical outcome of colorectal cancer patients that underwent resection of their primary tumor. This correlation was independent from other clinicopathologic prognostic factors. This stresses the importance of the immune system in tumor development and progression. Furthermore, tumor-infiltrating dendritic cells showed a distinguished infiltration pattern based on their activation and maturation status, which also had an impact on the synchronous infiltration by other leukocytes.

Dendritic cells in general play a key role in the development and maintenance of an immune response. It is therefore plausible to hypothesize that tumor-infiltrating dendritic cells play an important role in a cellular antitumor immune response. They do this by infiltrating, capturing, and processing tumor antigens and subsequent presentation, activation, and recruitment of lymphocytes. Several previous studies already supported this hypothesis by showing a correlation between the quantity of tumor-infiltrating dendritic cells and clinical outcome in different tumor types. Ambe et al. used S-100 protein to detect dendritic cells in colorectal cancer and found a positive correlation between the number of infiltrating dendritic cells and a favorable clinical outcome (10). This observation, however, is in contrast to our recent findings concerning S-100-positive dendritic cells in colorectal cancer (11) but also to findings obtained by others in breast, gastric, ovarian, and tongue carcinomas (9, 12–15).

CD208 is a member of the lysosome-associated membrane glycoprotein family and a highly specific marker for mature dendritic cells (16). CD1a is a nonconventional antigen-presenting molecule expressed on dendritic cell subtypes. It is often being regarded as a marker for immature dendritic cells, which is down-regulated on maturation. However, different in vitro and in vivo studies have shown that CD1a can be present on both mature and immature dendritic cells (7, 17). We used CD208 and CD1a to detect tumor-infiltrating dendritic cells. Although CD1a cannot be regarded as a specific marker for immature dendritic cells, differences with the maturation marker CD208 could be accredited to an immature dendritic cell fraction of CD1a-positive dendritic cells.

In our present study, we did find a significant correlation with clinical prognosis with the dendritic cell markers CD1a and CD208. Moreover, not only the number of infiltrating dendritic cells but also the location of infiltration was of importance. This finding supports the observations of Bell et al. in breast cancer (18) and in colorectal cancer (19). CD1a-expressing dendritic cells in the advancing border and mature (CD208-expressing) dendritic cells in the tumor epithelium both showed a correlation with an adverse outcome. The cause of this negative correlation of tumor-infiltrating dendritic cells on clinical prognosis is speculative. A hypothesis is that antigen-capturing and antigen-processing (CD1a) dendritic cells on their way to the tumor cells are hindered in their function due to inhibiting factors (cytokines, chemokines, etc.) in the tumor environment, whereas maturing dendritic cells in the tumor epithelium are hindered in their way to migrate out to secondary lymphoid tissue after antigen capture and activation possibly by similar inhibitory factors.

It is still unclear how tumor-infiltrating dendritic cells interact with tumor cells and other immune cells in the local tumor area. In our study, we found a distinct infiltration pattern according to the maturation and activation status of the dendritic cells. CD1a-positive dendritic cells, of which presumably a large fraction of immature dendritic cells, tended to be scattered trough the whole tumor area and were vigorously infiltrating the tumor epithelium, whereas mature (CD208) dendritic cells clustered in the peritumoral areas and could hardly be found in the tumor epithelium. It is generally presumed that immature dendritic cells are primarily equipped for a role as antigen-processing cells. Therefore, their presence in the close vicinity of tumor epithelium could be a reflection of their active amass of (tumor) antigens. It is tempting to suggest that the peritumoral clustering of mature dendritic cells reflects a state in which these dendritic cells interact directly with clusters of tumor-infiltrating lymphocytes to generate an antitumor immune response as suggested in the studies of Bell et al. (18) and Suzuki et al. (19).

A relationship between the amount of tumor-infiltrating CD1a-positive dendritic cells and expression of HLA class II on tumor cells has been observed in breast cancer (17). In the present study, we show the same correlation in colorectal cancer. Moreover, our study points out that the intraepithelial dendritic cells fraction is of importance and not the stromal or advancing border dendritic cell fraction. It is still obscure what cellular interactions lead to the found correlation. Further studies are necessary to illuminate these interactions.

In our study cohort, certain tumors contained large amounts of tumor-infiltrating dendritic cells, whereas others contained only small amounts. We could not detect any histopathologic differences between these tumors. A possible explanation could be the immunogenicity of these tumors; for example, microsatellite-instable colorectal tumors are characterized by a more vigorous infiltration of immune cells compared with microsatellite-stable colorectal tumors (20). It is hypothesized that these microsatellite-instable tumors are more immunogenic (i.e., more readily recognizable by the immune system). The differences in quantity and type of tumor-infiltrating dendritic cells are a possible reflection of this phenomenon. This hypothesis needs to be further investigated.

Antigen-pulsed and activated tumor-infiltrating dendritic cells are presumed to interact directly and indirectly with lymphocytes, like CD8- and CD4-positive T lymphocytes (CTL and T-helper cells, respectively), through membrane ligands and by secreting various cytokines, resulting in the activation and recruitment of these lymphocytes. Tumor-infiltrating CTLs are generally present in colorectal tumors and are able to spontaneously lyse colon carcinoma cell lines (21). Different studies have shown that high intraepithelial infiltration of CTL shows a beneficial effect on patient survival. In the present study, we showed that a high intraepithelial infiltration of CD1a-positive dendritic cells is correlated to a high intraepithelial infiltration of CTL and/or T-helper cells. In our previous study, we also found a correlation between tumor-infiltrating S-100-positive dendritic cells and other tumor-infiltrating leukocytes (11). It is presumable that the concerning intraepithelial dendritic cells fraction is S-100 and CD1a positive. The correlation with tumor-infiltrating leukocytes could be an independent phenomenon, reflecting the immunogenicity of the tumor, as mentioned above, but it could also suggest that the lymphocytes are actively recruited and/or activated by these dendritic cells. Bell et al. showed that T-helper cells colocalize with dendritic cells in some breast carcinoma tumors, supporting the hypothesis that tumor-infiltrating dendritic cells interact on a local level with lymphocytes to maintain an immune response. Unfortunately, we did not find supporting evidence for this hypothesis in the form of a correlation between the amount of stromal or advancing border mature dendritic cells and other immune cells in the same tumor compartments.

In summary, tumor-infiltrating dendritic cells in colorectal tumors showed a distinct infiltration pattern based on their maturation status. The presence of immature tumor-infiltrating dendritic cells in the tumor epithelium correlated to a higher infiltration of other immune cells. Furthermore, infiltration of CD208-positive (mature) dendritic cells in the tumor epithelium and CD1a-positive dendritic cells in the advancing borders was correlated to an adverse outcome. Our study showed that functional subsets of tumor-infiltrating dendritic cells affected local tumor cell-immune cell interactions and correlated to the clinical prognosis of colorectal cancer patients.