Peritumoral CD1a-Positive Dendritic Cells Are Associated With Improved Survival in Patients With Tongue Carcinoma

Objectives  To determine if survival and recurrence rates for patients with squamous cell carcinoma of the tongue correlate with the degree of dendritic cell (DC) infiltration of the primary tumor or adjacent tongue tissue and if there is an association between tumor or nodal stage and DC infiltration.

Design  Hospital and office medical records were reviewed to obtain 5-year follow-up data. Original pathology specimens were recut and stained for the cell surface markers S100 and CD1a. The number of DCs present in the specimens was quantified microscopically and compared statistically with patient outcome and staging.

Setting  A university hospital.

Patients  All patients who underwent resection of primary squamous cell carcinoma of the tongue from January 1, 1987, through December 31, 1990, for whom 5-year follow-up data and original pathology specimens were available (N=43).

Main Outcome Measures  Time to recurrence, death, or both.

Results  Patients who had greater numbers of CD1a-positive DCs adjacent to tumor had improved survival (P=.02) and decreased recurrence rates (P=.06). The other subpopulations of DCs examined were not associated with survival or recurrence. In addition, the number of CD1a-positive DCs in peritumoral epithelium decreased as the tumor stage increased (P=.01) and if nodal metastases were present (P=.05).

Conclusions  Dendritic cells are antigen-presenting cells that are thought to play a major role in the antitumor immune response. The CD1a surface antigen has been shown to mediate T-cell interactions. The association between CD1a-positive peritumoral DCs and patient outcome suggests an important function for this cell population.

DENDRITIC CELLS (DCs) are arguably the most potent antigen-presenting cells and may be the only cells capable of initiating the adaptive immune response. They elicit a powerful, antigen-specific immune response. They are highly motile and possess multiple veil-like dendritic processes that protrude from the cell surface to help entrap antigens. They originate in the bone marrow and migrate through the bloodstream to the peripheral tissues, where they become activated on injury or in response to cytokines. They entrap particulate antigens directly and engulf soluble antigens by pinocytosis. Then they travel to regional lymph nodes, where the antigens are processed and presented on the cell surface, in antigen-presenting molecules, to T cells and B cells. Dendritic cells express high levels of class I and II cells of the major histocompatibility complex and the alternative antigen-presenting molecule, CD1. A rapidly growing body of literature suggests a pivotal role for DCs in various immune and inflammatory responses, including inducing antitumor immunity.^1-3

The evidence that DCs play an important role in antitumor immunity is convincing. We and others^1,4 have demonstrated in animals that DCs can be combined with tumor antigen to produce vaccines that effectively immunize against tumor and can be used to treat established tumors. Several clinical studies have shown that the presence of a dense DC infiltrate in primary tumor specimens is associated with improved survival for various types of tumors, including lung,^5,6 colon,⁷ gastric,⁸ papillary thyroid,⁹ and nasopharyngeal.^10,11 Few studies, however, have examined the significance of DCs in head and neck squamous cell carcinoma (SCC). We, therefore, studied the relationship of patient outcome to DC infiltration in SCC primary tumors of the tongue. Although most previous studies compared DC infiltration only with survival, we used both survival and recurrence as outcome measures because many host factors can affect survival.

The histological identification of DCs is not straightforward. They are identified using immunohistochemical techniques. Several labeling antibodies have been used, but it is not clear how the staining characteristics of the DCs change as the cells migrate or become exposed to stimuli such as injury, cytokines, particulate or soluble antigens, or T-cell interactions. The cell surface expression of various molecules by DCs is not constant, and changes in expression may reflect functional changes in DCs.¹² We chose to use the antibodies S100 and CD1a in this study because S100 has been used in most of the clinical studies correlating DC infiltration with patient outcome previously,^5-11 and CD1a has been used in some clinical studies³ and basic histological studies to label DCs.^12-14 Neither stain is highly specific for DCs. S100 stains melanocytes and neurons in addition to DCs; CD1a stains macrophages and some B cells. Therefore, we used cell structure to distinguish DCs from other positively staining cell types, as has been done in previous studies. By studying DCs with 2 different staining antibodies—CD1a and S100—we sought to determine if there was any evidence that a particular DC subpopulation might be more closely related to patient outcome and nominal antitumor immunity.

All patients registered in the database of the Department of Otolaryngology–Head and Neck Surgery at the University of Pittsburgh School of Medicine, Pittsburgh, Pa, as having undergone resection of primary SCC of the tongue between January 1, 1987, and December 31, 1990, inclusive, were initially identified (n=69). Patients were included in the study only if 5-year follow-up data could be obtained (n=58) and if original histological slides and tissue blocks were still available (n=47). Also, patients were excluded if they had died of unrelated causes during the follow-up period. Two patients were subsequently identified as having died of other causes during the follow-up period, and 1 patient was unavailable for follow-up. These 3 patients were not included in the analysis of variance (ANOVA) but were included in the Cox analysis. This yielded 43 patients: 23 men and 20 women. The mean age of the patients was 61 years (range, 26-85 years). The following table shows a breakdown of the number of patients for the tumor and nodal stages:

All histological slides were reviewed to confirm the diagnosis and stage and to select representative specimens containing tumor and adjacent tongue tissue for study. Formalin-fixed, paraffin-embedded tissue blocks were serially sectioned for staining with hematoxylin and eosin, CD1a, and S100. The avidin-biotin complex technique (Vector Elite Kit, Vector Inc, Burlingame, Calif) was used for the 2 immunohistochemical markers according to the manufacturer's instructions. Prediluted CD1a monoclonal antibody (Immunotech Inc, Westbrook, Me) was used at a dilution of 1:2 with microwave antigen. S100 (Dako Corp, Carpenteria, Calif) was used at a dilution of 1:300 after pretreatment with protease 24 (Sigma Immunochemicals, St Louis, Mo). Dendritic cells were quantitated based on structure and CD1a or S100 positivity. The DCs were counted in the tumor and in adjacent visible benign squamous epithelium on the same slides. Counts were obtained in 5 high-power fields (×400 magnification). Dendritic cells were counted in areas of greatest staining intensity. Cells were counted as DCs if their structure was characteristic and if a cell nucleus could be identified; dendritic processes alone were not counted. Intratumoral DC counts were not obtained for 1 patient, and peritumoral DC counts were not obtained for 3 patients because the pathology specimens did not yield adequate tissue in these areas.

All patients were observed for at least 5 years (60 months) or until death. The number of months to recurrence, death, or both was recorded. The cause of death, initial tumor stage, and nodal stage were also noted for each patient. Staging criteria of the American Joint Committee on Cancer and the Union Internationale Contre le Cancer were used.

Four different values were obtained for the DC counts: S100-positive DCs within tumor and within peritumoral epithelium and CD1a-positive DCs within tumor and within peritumoral epithelium. These values were compared with 2 outcome measures: survival and recurrence. Cell counts were also compared with tumor stage and lymph node involvement. One-way ANOVA, Cox proportional hazards models, and Kaplan-Meier survival curves were used. All analyses were performed using a Macintosh 8100/80 computer and commercially available statistical software (JMP, SAS Institute Inc, Cary, NC).

Four different subpopulations of DCs were counted for each patient based on the stain (CD1a or S100) and location (peritumoral or intratumoral), as noted in the "Materials and Methods" section. Figure 1, Figure 2, and Figure 3 show tissues stained for DCs from a representative patient examined in this study. The 4 different DC count values were correlated with the time to recurrence and survival using ANOVA. This is comparable to using an unpaired t test when the independent variable is dichotomous, as in this analysis. The ANOVA models were created using dichotomous classifications with recurrence or survival as the sole independent variable. Continuous measures of the 4 separate DC counts were used as the dependent variables in separate analyses. Table 1 and Table 2 present the mean DC counts for patients who did or did not have recurrence (Table 1) and for patients who survived 5 years or died of disease during that time (Table 2). This analysis suggests that peritumoral CD1a-positive DCs are associated with enhanced survival (P=.05) and decreased recurrence (P=.09). None of the other DC subpopulations were associated with either survival or recurrence.

We used a Cox proportional hazards model to improve statistical efficiency and power by allowing us to consider the continuous nature of the DC count data and to adequately handle the censored nature of the time-to-event data (ie, time to recurrence or survival time). Table 3 shows the P values generated with the Cox proportional hazards model analyzing 1 DC subpopulation at a time. This analysis provided evidence consistent with the ANOVA that suggested that peritumoral CD1a-positive DCs were associated with improved survival (P<.04) and less frequent recurrence (P=.07). We also used the proportional hazards model to simultaneously compare the relationship of all DC subpopulations with recurrence and survival. Theoretically, such a comparison eliminates confounding patient factors, such as nutritional status or concurrent medical disease states, if these factors affect the various DC subpopulations similarly. The P values generated with this model are shown in Table 4. Peritumoral CD1a-positive DCs were again associated with enhanced survival (P=.02) and decreased recurrence (P=.06). Although the latter P value fails to reach significance, its statistical validity is strongly suggested by the lack of any small P values in the other DC subpopulations studied by the same methods.

We related DC counts not only to patient outcome but also to tumor characteristics at presentation. Patients were grouped by tumor stage (T0, T1, T2, T3-4) and lymph node involvement (N0 or N-positive). The 4 DC counts were then compared with the tumor stage and, in a separate analysis, with nodal involvement using ANOVA. Table 5 and Table 6 show the mean DC counts for patients grouped by tumor stage or nodal involvement, respectively. Interestingly, all 4 DC subpopulations tended to decrease as tumor stage increased, or if there was lymph node metastasis, but only the CD1a-positive peritumoral count did so in a statistically significant manner (P=.005 for tumor stage and .05 for node status). This association raised an interesting question: if peritumoral CD1a-positive DCs are associated with tumor stage, and the tumor stage is itself associated with survival, is the association between these cells and patient outcome merely due to an association with the tumor stage, or is there a causal relationship leading to this connection? Perhaps as tumors enlarge, peritumoral CD1a-positive DCs are overwhelmed or suppressed, leading to a less effective immune response and worsened outcome.

Multivariate analyses were performed to compare survival with tumor stage and with all 4 DC counts simultaneously using the Cox proportional hazards model. The following P values were obtained:

They suggest that tumor stage is highly prognostic of survival (P<.001), but that peritumoral CD1a-positive DCs are not (P=.23). A multivariate analysis of 5 variables from a population of 43 patients, however, has weak statistical power. Furthermore, if the tumor stage and CD1a-positive DCs are causally linked to each other, the multivariate analysis may be confounded. Ultimately, the association of CD1a-positive peritumoral DCs with the tumor stage is an important observation that needs to be considered when discussing the influence of DCs on patient outcome.

We sought to further interpret our data by statistically comparing the differentially staining DC subpopulations with each other. Table 7 displays Pearson product moment correlations of the 4 different DC populations we counted. Relatively high correlations exist between the number of CD1a- and S100-positive DCs both in peritumoral tissue (0.31) and within tumor (0.46). This suggests that the CD1a- and S100-positive DC populations fluctuate similarly from patient to patient. On the other hand, CD1a-positive DC counts from within tumor and from peritumoral epithelium do not correlate well (−0.10). This suggests that CD1a-positive DCs around tumor function independently from those cells within tumor.

Kaplan-Meier curves were plotted for both survival and recurrence in our patient population (Figure 4). No recurrence was noted after 41 months, at which time 33% of the patients remained recurrence-free. Five-year survival for our series was 36% overall.

We sought to determine if there was any relationship between DC infiltration and outcome in patients with SCC. Because the natural history of SCC of the head and neck varies by site, we thought that focusing on 1 site—the tongue—would help eliminate differences in outcome related to the location of the primary tumor. We also sought to determine if results differed depending on the immunohistochemical marker used to identify the DCs or on whether DCs were counted from within or around tumor parenchyma. Thus, we wanted to know if there were functional subpopulations of DCs. We found that only 1 of the 4 subsets of DCs examined—CD1a-positive peritumoral DCs—had a statistically significant association with improved prognosis and decreased recurrence rates. These results suggest that this subset of DCs may be functionally distinct from the other subsets examined and may play a role in antitumor immunity.

The literature contains little information pertaining to CD1a-positive DCs and prognosis in patients with cancer. There is considerable evidence from previous studies, however, that a marked infiltration of tumors with S100-positive DCs is associated with improved prognosis. Improved survival was associated with prominent DC infiltration in lung, colorectal, and gastric tumors.^5-8 Relatively few studies have examined head and neck primary tumors. Patients with papillary thyroid carcinoma and those with nasopharyngeal carcinoma have been shown to have improved survival when marked DC infiltration was present.^9-11 Gallo et al¹⁵ found, in 88 patients with laryngeal SCC, that low, intermediate, and high densities of DC infiltrates were associated with 5-year survivals of 0%, 62%, and 61%, respectively. This study assessed DC infiltration within and around tumor but did not distinguish between the 2 in the survival analysis. Interestingly, these authors found no correlation between the tumor stage and the degree of DC infiltration, in contrast with our findings. Unlike these studies, we did not observe an association between S100-positive DCs and outcome in our patients. Perhaps the biological behavior of SCC of the tongue is different from the tumors examined in the previous studies.

There is also evidence from previous studies^13-17 that multiple subpopulations of DCs exist and that differences in surface marker expression may reflect functional differences among different subpopulations. Modlin et al¹⁶ found significant differences in the DC staining patterns observed when using S100 and OKT6 to analyze skin from 40 patients with leprosy. S100- and OKT6-positive DC subpopulations were found predominantly in different areas and different types of lepromatous lesions. The authors postulated that the 2 populations were functionally different.¹⁶ Our results suggest that the CD1a-positive peritumoral subpopulation of DCs is functionally distinct and is more important to antitumor immunity than the other subpopulations studied. This is an intriguing notion because the CD1a surface antigen has been shown to be a glycopeptide antigen–presenting molecule, closely related to class I molecules (although mapping outside the major histocompatibility locus). CD1a is a target in mixed lymphocyte reactions, graft-versus-host reactions, and skin graft rejection.^12,13 The ability of DCs to stimulate mixed lymphocyte reactions and to activate T cells in vitro has been blocked by antibody to CD1a.¹² There is also evidence that B-cell and macrophage interactions with T cells may involve the CD1a antigen.¹⁴ Some T cells do indeed interact with DCs presenting CD1a-associated epitopes. Potentially, peritumoral CD1a-positive DCs are presenting tumor-derived antigens in the CD1a surface molecule to T cells, thus activating an immune response. This could explain the association between these cells and patient outcome observed in our study.

The intratumoral CD1a-positive subpopulation of DCs was also not associated with outcome in our study. Perhaps the peritumoral cells are immunologically active, but cells that have migrated within the tumor are somehow blocked or sequestered by the tumor. This would also explain the lack of Pearson product moment correlation between the intratumoral and peritumoral CD1a-positive DC counts: the intratumoral cells cannot affect extratumoral cells, and vice versa. They are independent subpopulations. If tumors are able to suppress DC function, this may explain another observation from our study: that DC counts tended to decrease as tumor stage increased (although this trend was statistically significant only for CD1a-positive peritumoral DCs). Perhaps larger tumors suppress the migration or survival of DC subpopulations, impeding the antitumor immune response and thereby worsening the outcome.

Several observations have been made: peritumoral CD1a-positive DCs are associated with improved patient outcome (less recurrence and improved survival) and inversely associated with tumor stage and lymph node metastasis for SCC of the tongue. In addition, there is no association between the number of DCs within the tumor and the number found around it. This raises the possibility of differences in the activity or survival of DCs at these 2 sites. These observations suggest that functionally different subpopulations of DCs exist and that the CD1a-positive peritumoral DCs are especially important in tumor immunity.

Accepted for publication March 19, 1998.

Presented at the annual meeting of the American Academy of Otolaryngology–Head and Neck Surgery, San Francisco, Calif, September 1997.

Corresponding author: Steven A. Goldman, MD, Department of Otolaryngology–Head and Neck Surgery, University of Pittsburgh School of Medicine, 200 Lothrop St, Pittsburgh, PA 15213.