Tumor-Infiltrating Cytotoxic T Cells but not Regulatory T Cells Predict Outcome in Anal Squamous Cell Carcinoma

Materials and Methods

Patient selection. Between 1985 and 2001, a total of 112 patients with anal carcinoma were admitted for treatment by combined radiochemotherapy to this University Hospital. For this study, a subgroup of 38 patients was selected according to the criteria given in Fig. 1 . This project received approval from the Interdisciplinary Clinical Research Center of the University Hospital of the Friedrich Alexander University.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Patient selection criteria.

All patients included in the study were HIV negative. Following clinical and rectal examination, a biopsy was taken under general anesthesia. All patients had rectoscopy, including endorectal ultrasound, whenever feasible. Staging was completed by computed tomography scans of the abdomen and pelvis, chest X-rays, and liver ultrasound. Histopathologic diagnosis was established according to WHO criteria (25). T categories were T1 (6), T2 (17), T3 (11), and T4 (4). No patients with small perianal lesions (<2 cm) treated by complete excision were included in this protocol. Twenty-four patients had no evidence of nodal disease, whereas 14 initially presented clinically with involved regional nodes. Nodes of ≥1.5 cm in largest diameter were judged as positive.

Treatment protocol. Radiation treatment was directed to the primary tumor region, inguinal, perirectal, and internal iliac/presacral nodes using a four-field box technique with 10 to 15 MV photons. Dose prescription followed the International Commission on Radiation Units report no. 50 guidelines with single fractions of 180 cGy and a total dose of 5,040 cGy. An additional local boost was applied to larger (>4 cm) primaries either by external multiple-field techniques (540-900 cGy) or interstitial radiation using Iridium-192 low-dose-rate or pulsed dose rate (1,200-1,600 cGy) depending on circumferential extension of the disease. All patients were scheduled for simultaneous chemotherapy with two cycles of 5-fluorouracil at a dose of 1,000 mg/m²/24 hours as a 120-hour continuous i.v. infusion on days 1 to 5 and 29 to 33 and mitomycin C at 10 mg/m²/d on days 1 and 29 as a single i.v. bolus injection. Six weeks after radiochemotherapy, tumor response was evaluated clinically and by means of proctoscopy, endorectal ultrasound, and computed tomography scans of the pelvis. All patients had regular follow-up in 3-month intervals during the first 2 years and in 6-month intervals thereafter. The median follow-up was 25.5 months with a range between 2 and 147 months.

Tissue microarray and immunohistochemistry. Paraffin-embedded biopsy samples obtained from 38 patients with anal squamous cell carcinoma before therapy were processed into a tissue microarray (BioCat, Heidelberg, Germany) using a core diameter of 2 mm (Fig. 2A ). Between two and seven cores were taken per patient (mean = 2.8), resulting in a total of 120 cores. All materials had been submitted for diagnostic purposes and were used in accordance with national ethical principles. Immunohistochemistry of paraffin sections was carried out using a standard streptavidin biotinylated alkaline phosphatase (ABC-AP, DakoCytomation, Hamburg, Germany) method or tyramide signal amplification followed by ABC-AP (only for FoxP3). The following antibodies were used: CD3 (DakoCytomation; Fig. 2B), CD4 (Novocastra, Newcastle upon Tyne, United Kingdom), CD8, CD68 (DakoCytomation), FoxP3 (Abcam, Cambridge, United Kingdom; Fig. 2C), granzyme B (DakoCytomation; Fig. 2D), total caspase-3 (DakoCytomation), and activated caspase-3 (Cell Signaling Technology, Danvers, MA). Using a standard light microscope, images were acquired by a CCD camera, transferred to a Personal computer, and counted using the image analysis software COUNT (Biomas, Erlangen, Germany). Numbers of labeled tumor-infiltrating cells were determined per 100 tumor cells (labeling index). The average number tumor cells counted per patient for each immunohistologic marker was 762 (range, 116-2,862). Between 0 and 17.3 CD3⁺ T cells were detected per 100 tumor cells (median = 3.8). For CD4⁺ T cells, the median was 1.5 per 100 tumor cells (range, 0-11.4); for CD8⁺ T cells, the median was 2.1 per 100 tumor cells (range, 0-9.7); for FoxP3⁺ Treg, the median was 2 per 100 tumor cells (range, 0.3-7.3); for granzyme B⁺ cytotoxic cells, the median was 0.6 per 100 tumor cells (range, 0-9.1); and for CD68⁺ macrophages, the median was 4.4 per 100 tumor cells (range, 0.3-29.2). Average variation of labeling index within the two to seven samples for all patients was quantitatively assessed. Mean labeling index with SD was 0.044 ± 0.018 for CD3, 0.02 ± 0.008 for CD4, 0.025 ± 0.011 for CD8, 0.009 ± 0.003 for granzyme B, 0.064 ± 0.025 for CD68, 0.022 ± 0.007 for FoxP3, and 0.021 ± 0.008 activated caspase-3.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Biopsy specimens were processed into tissue microarrays using a 2-mm needle core (A). Immunohistochemistry was used to identify CD3⁺ T cells (B), FoxP3⁺ regulatory T cells (C, red nuclear labeling), and granzyme B⁺ cytotoxic cells (D, red granular cytoplasmic labeling).

Statistical methods. Overall survival and NED survival rates were calculated according to Kaplan-Meier (26). The log-rank test was used to compare survival rates (27) between subgroups of patients. For all immunohistochemical markers, cutoff for definition of subgroups was the median value. Follow-up time was calculated as the interval between end of chemoradiation and last follow-up or death. The following events were defined as end points: death (overall survival) and the appearance of local, nodal, or distant recurrence (NED survival).

Results

Overall survival and NED survival rates for the study population were 83% and 72% at 3 years and 70% and 65% at 5 years, respectively. Influence of clinical as well as immunohistochemical variables on both survival end points is shown in Table 1 .

NED survival was strongly influenced by the intratumoral accumulation of T cells (Fig. 3 ; Table 1). Intratumoral CD3⁺ and CD4⁺ TIL had a significant effect on NED survival: 3-year rates for patients with lower numbers of TIL (≤3.8 CD3⁺ per 100 tumor cells and ≤1.5% CD4⁺ per 100 tumor cells) were 89% and 95%, respectively. Patients with higher numbers of TIL (>3.8% CD3⁺ per 100 tumor cells and >1.5% CD4⁺ per 100 tumor cells), by comparison, had survival rates of 54% (P = 0.02) and 48% (P = 0.01), respectively. Similarly, higher numbers of CD8⁺ TIL (>2.1 per 100 tumor cells) were associated with lower NED survival of 53% versus 89% for patients with lower numbers (≤2.1 per 100 tumor cells), although this did not reach statistical significance (Table 1; Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

NED survival rates in relation to the intratumoral accumulation of T cells for CD3⁺ T cells (top), CD4⁺ T cells (middle), and CD8⁺ T cells (bottom).

No prognostic influence was observed for the numbers of tumor-infiltrating CD68⁺ macrophages and of FoxP3⁺ regulatory T cells (Table 1; Fig. 4 ). The TIL subpopulation showing highest effect on outcome were granzyme B⁺ cytotoxic cells. Patients with comparatively large numbers (>0.6 cells per 100 tumor cells) showed a 3-year NED survival rate of 47%, whereas patients with lower numbers fared significantly better (3-year NED survival rate of 89%, P = 0.008; Table 1; Fig. 4). Expression of total caspase-3 was detected in the neoplastic cells of all cases. The number of tumor cells expressing the activated caspase-3 had no influence on prognosis (Table 1). None of the variables investigated showed a significant effect on overall survival, most likely due to the relatively small number of cases.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

NED survival rates in relation to the intratumoral accumulation of CD68⁺ macrophages (top), FoxP3⁺ regulatory T cells (middle), and granzyme B⁺ cytotoxic cells (bottom).

Discussion

Here, we show that large numbers of tumor-infiltrating CD3⁺ T cells are a significant negative prognostic marker in patients with anal carcinoma. This is unlikely to be an epiphenomenon (e.g., reflecting an inflammatory response to a more aggressive invasive tumor growth) because numbers of CD68⁺ macrophages showed no correlation with disease-free survival. Thus, the observed effect seems to be T cell specific. Further subtyping of T cells revealed that both CD4⁺ and CD8⁺ T cells were associated with an unfavorable prognosis, although this reached statistical significance only for CD4⁺ cells. We, therefore, investigated the possibility that increased numbers of Treg may account for this observation. Treg are CD4⁺/CD25⁺ T cells with the ability to inhibit effector T cells (22, 28). In ovarian cancer, the presence of Treg has been shown to be associated with poor prognosis (24). Treg express various surface antigens, including CD25, CD103, OX-40, CTLA-4, and the glucocorticoid-induced tumor necrosis factor receptor, none of which seems to be entirely specific for Treg (22). The transcription factor FoxP3 is crucial for the development of Treg and is currently considered to be the best Treg marker (29). To identify Treg in tissue sections, we have, therefore, employed immunohistochemistry for the detection of FoxP3. Using this approach, no relationship between the number of tumor-infiltrating Treg and prognosis was observed, suggesting that the effects of Treg observed in ovarian cancer may not be relevant to other tumor models.

Previous studies have identified high numbers of granzyme B⁺/CD8⁺ TIL as a significant predictor of fatal outcome in EBV-associated nasopharyngeal carcinoma as well as in Hodgkin's lymphoma (18, 30). This observation is confirmed here for anal carcinoma. That large numbers of granzyme B⁺ TIL were a statistically significant adverse prognostic factor, whereas this was not the case for CD8⁺ TIL, probably indicates that the state of activation rather than the mere presence of CTL is important (30). The mechanism underlying this observation remains uncertain. It has been speculated that T cell–induced selection of tumor cell clones resistant to apoptosis may explain this phenomenon (19). However, in our study, tumor cells in all cases expressed caspase-3, and the proportion of tumor cells positive for activated caspase-3 showed no relation to prognosis. Thus, in contrast to a previous study of nasopharyngeal carcinoma (19), the absence of activated caspase-3 is no indicator of poor clinical outcome in anal carcinoma. It remains to be determined if other components of apoptotic pathways are disrupted in anal cancer.

Previous studies addressing the possible prognostic significance of TIL have determined T-cell subsets as a proportion of the total TIL. Because our original hypothesis was that TILs represent an immune response directed against the neoplastic cells, we have chosen to determine the numbers of TIL subsets in relation to the number of tumor cells, thus establishing an “effector/target” ratio. Using this approach, we found that the number of tumor-infiltrating CD3⁺, CD4⁺, CD8⁺, or granzyme B⁺ T cells compared with the number of neoplastic cells was extremely small (medians = 3.8, 1.5, 2.1, and 0.6 cells per 100 neoplastic cells, respectively). It remains unclear if such a small number of T cells would be able to mount a sufficient antitumor immune response, assuming that they were directed against the tumor cells. It is conceivable, however, that an immune response insufficient to mediate effective tumor cell killing may still be able to drive the selection of more aggressive tumor cell subclones (e.g., resistant to apoptosis; ref. 18). Negative prognostic effects of large numbers of activated cytotoxic TIL have been reported for nasopharyngeal carcinomas and Hodgkin's lymphomas, both representing EBV-associated tumors (18, 30). By contrast, a positive prognostic effect has been described for other malignancies [e.g., colorectal (8–10), ovarian (11), pancreatic (12), and esophageal carcinoma (13, 14)], which are not known to be associated with a viral infection. It is interesting to note in this context that anal carcinoma is commonly associated with HPV (4). It is possible, therefore, that viral proteins expressed in the tumor cells may render the neoplastic cells more resistant to apoptosis, or that the expression of viral neoantigens in the tumor cells may alter the nature of the local immune responses.

In summary, we show that tumor-infiltrating T cells are an adverse prognostic marker in anal carcinoma. Contrary to expectations, this is not attributable to the presence of Treg but rather to granzyme B⁺ cytotoxic cells. The detailed study of local immune responses in carcinomas is important for the development of immunotherapeutic strategies. Furthermore, the composition of local T-cell infiltrates may be open to therapeutic modulation (e.g., by eliminating Treg; ref. 28), thus providing an additional tool in the treatment of carcinomas.