FOXP3 and survival in urinary bladder cancer

Authors


Amir Sherif, Department of Urology, Karolinska University Hospital, 171 76 Stockholm, Sweden. e-mail: amir.sherif-abdel-hadi@karolinska.se

Abstract

What's known on the subject? and What does the study add?

This is the first study examining FOXP3 expression in invasive urothelial urinary bladder cancer and in their tumour-infiltrating lymphocytes (TILs). The relation of their respective immunohistological expression to survival adds new knowledge in the fields of tumour immunology and prognostic markers.

OBJECTIVE

• To investigate the possible impact of FOXP3 expression in T-cells, as well as in tumour cells, on long-term survival in patients with urinary bladder cancer (UBC) invading muscle.

PATIENTS AND METHODS

• In a retrospective study, tumour specimens from 37 patients cystectomized for T1–T4 UBC during 1999–2002 at the Karolinska University Hospital were examined by immunohistochemistry for tumour expression and/or infiltration of immune cells expressing FOXP3 as well as CD3.

• The results obtained were correlated with clinicopathological parameters, where the primary and secondary outcomes investigated were overall survival and progression-free survival, respectively.

RESULTS

• Infiltration of CD3+ and FOXP3+ lymphocytes (≥3 cells per high-power field) were both correlated with better survival, and this relationship persisted throughout the whole study period (all P < 0.05).

• Patients with FOXP3+ tumour cells had decreased long-term survival compared to those patients with FOXP3 tumours (P < 0.05).

• Despite a limited amount of patient material, the results of the present study indicate that FOXP3 expression, in both lymphocytes and tumour cells, is an important prognostic factor in UBC.

CONCLUSIONS

• FOXP3 expression in UBC cells is associated with decreased long-term survival and thus may be a novel negative prognostic factor in UBC invading muscle.

• By contrast, the presence of FOXP3+ tumour-infiltrating lymphocytes was correlated with a positive prognosis. Because FOXP3 is up-regulated upon activation in human T-cells, FOXP3 may serve more as an activation marker than as a regulatory T-cell indicator in this case.

• These results support the need for larger prospective studies aiming to confirm the results obtained and to examine the underlying mechanisms in detail.

Abbreviations
HR

hazard ratio

TIL

tumour-infiltrating lymphocyte

Treg

regulatory T-cell

UBC

urinary bladder cancer.

INTRODUCTION

Tumour-infiltrating lymphocytes (TILs) are generally considered to be a positive prognostic factor in malignancies [1,2]. In urinary bladder cancer (UBC), the immune system is of special interest because BCG-instillation for the treatment of UBC not invading muscle is aimed at T-cell activation [3].

Tumour immune escape encompasses multiple mechanisms by which tumours manage to evade the immune system (e.g. by tolerance induction, immunosuppressive microenvironment and the involvement of regulatory cell populations) [4,5]. Regulatory T-cells (Tregs) were first defined as a subpopulation of CD4+ T-cells that express the interleukin-2 receptor α-chain (CD25) constitutively [6]. The transcription factor FOXP3 has subsequently been described as a more specific Treg marker. However, as shown previously [7–10], both CD25 and FOXP3 are up-regulated in activated human T-cells, and the results obtained differ if activated human cells with induced FOXP3 expression have functional suppressive capacity.

Increased numbers of CD4+FOXP3+ T-cells have been observed in the peripheral blood of patients with UBC and other forms of cancer [11,12]. It is unclear whether these cells represent a non-specific increase in the Treg pool or an expansion of tumour-specific Tregs or peripherally induced CD4+FOXP3+ cells, nor is it known what impact these cells have on the disease [13–15]. Increased numbers of T-cells expressing Treg markers have also been reported among TILs and in tumour-draining lymph nodes [16,17]. In ovarian carcinoma, increased intratumoural Treg numbers have been linked with reduced survival [18]. However, evidence for a role of Tregs in cancer is conflicting because other studies have failed to show this negative connection and have even pointed towards improved survival in the presence of FOXP3+ TILs in head and neck cancer, as well as in colon cancer [19,20].

Recent studies have shown that tumour cells also can produce FOXP3 [21,22]. This expression is correlated with interleukin-10 and transforming growth factor-β levels [22], and the co-culture of naïve T-cells with FOXP3-expressing tumour cells from a pancreatic carcinoma inhibited proliferation completely [21]. Furthermore, a recent study of breast cancer showed a negative correlation with survival in patients with tumoural expression of FOXP3 [23]. Thus, it appears that FOXP3 has multiple roles in tumoural–immune system interactions. The present study examined the link between FOXP3 expression in TILs and cancer cells, respectively, and survival in UBC.

PATIENTS AND METHODS

The study was approved by the Local Ethics Committee at the Karolinska Institute (EPN- Stockholm region, dnr: 2007/71-31).

STUDY POPULATION AND TISSUE SAMPLES

Paraffin-embedded tumour specimens from 37 patients who underwent radical cystectomy as a result of UBC invading muscle (T1–T4) at Karolinska University Hospital 1999–2002 were examined in the present study. Patient characteristics are presented in Table 1. The mean (median, range) age of patients was 67 (69, 46–81) years at the time of diagnosis. Staging was performed after preoperative radiological examination and by clinical parameters in conjunction with pathological examination of transurethral resection specimens according to the TNM classification of 1997 (Union Internationale Contre le Cancer).

Table 1.  Patient characteristics
Patient numberAge (years)SexcT stagepT stagepN stageSurvival time (years)
  1. *Still alive at the end of the present study. †Diagnosis date set based on mean time from diagnosis to operation in the rest of the material. ‡Patients who were cystectomized for urinary bladder cancer invading muscle in the years 1999–2002 at the institution did routinely not undergo lymphadenectomy (i.e. not even limited bilateral obturator fossa dissection). In a few cases, when there were suspicious nodes, these were dissected. This explains the high percentage of pNX.

177FemaleT2pT2pNX7.79
276FemaleT2pT3pNX0.59
349FemaleT2pT2pN01.92
456MaleT2pT0pNX8.82*
567FemaleT1pT3pNX0.96
671MaleT2pT0pN02.00
779MaleT2pT2pNX0.36
873FemaleT3pT3pNX6.84
958MaleT1pT2pN07.54
1076MaleT3pT3pN21.27
1149MaleT3pT2pN09.42*
1272MaleT2pT3pN01.07
1365MaleT2pT0pNX9.67*
1478FemaleT3pT2pNX1.74
1567FemaleT2pT2pNX2.35
1678FemaleT2pT2pN00.93
1773MaleT2pT3pN13.18
1866MaleT2pT3pNX2.50
1955FemaleT2pT0pN05.43
2059MaleT1pT2pN02.64
2180MaleT2pT3pN13.05
2263FemaleT2pT3pN18.39*
2363MaleT2pT2pN27.48
2481MaleT2pT2pN28.20*
2556MaleT3pT4pN00.13
2678MaleT3pT3pN07.30
2780MaleT2pT3pNX0.27
2846MaleT4pT4pN22.76
2973FemaleT3pT3pN07.91*
3072MaleT3pT3pN07.88*
3177MaleT3pT3pN36.92*
3250MaleT2pT2pN06.87*
3354MaleT2pT0pNX7.35*
3478MaleT2pT0pN16.94
3569MaleT2pT2pN07.14*
3659MaleT2pT0pN00.89
3766MaleT1pT3pN01.55

Disease progression was defined as appearance of pathologically enlarged lymph nodes, local recurrence or distant metastases in patients rendered free of disease at cystectomy. Known metastatic disease was defined as positive pN or pM at pathological examination post-cystectomy. Survival statistics from patient files were correlated with data from the national death registry in Sweden. Follow-up ended on 15 October 2009.

For statistical analysis, patients were stratified into three groups according to clinical staging, where T1 and T4 were separate groups, and T2/T3 were grouped together because of the limited patient material. At Karolinska Hospital, patients who were cystectomized for UBC invading muscle during this time did routinely not undergo lymphadenectomy (i.e. not even limited bilateral obturator fossa dissection). Dissection was performed only in cases where suspicious nodes were found, which explains the high number of pNX in the present study population.

IMMUNOHISTOCHEMISTRY

Immunohistochemistry was performed on serial, 4-µm sections from a paraffin-embedded tumour. Heat-based antigen retrieval was performed in Tris-EDTA (pH 9.0) (FOXP3) or citrate (pH 6.0) (CD3). Primary antibodies were monoclonal mouse anti-human CD3 (clone PS1; Vector Laboratories, Inc., Burlingame, CA, USA; dilution 1 : 200) and monoclonal mouse anti-human FOXP3 (clone 259D; BioLegend UK Ltd, Cambridge, UK; dilution 1 : 100). Parallel staining without primary antibody was performed as a negative control, and sections of human spleen as a positive control. Secondary antibody was biotinylated goat anti-mouse IgG (dilution 1 : 200, Vector Laboratories, Inc.). The ABC-ELITE (Vector Laboratories, Inc.) immunohistochemical staining kit was used in accordance with the manufacturer’s instructions, after which the slides were counterstained with Mayer’s haematoxylin.

Three high-power fields (×40) were randomly selected in tumour cell areas, photographed in corresponding areas of consecutive slides for both CD3 and FOXP3, and the number of positive lymphocytes in contact with tumour cells quantified. The presence of FOXP3+ tumour cells was classified as (+) or (−). A tumour was defined as positive if at least 25% of the tumour cells were positively stained, as described previously [23]. Staining was further subdivided as cytoplasmic, nuclear or both. Evaluation was performed by two independent researchers, who were blinded to the clinicopathological data.

STATISTICAL ANALYSIS

For each patient, a mean value of the three measurements on FOXP3 and CD3, respectively, was calculated. Given that there are no standard thresholds for the infiltration of TILs defining clinical outcome in this setting, we used a combination of the minimum P-value method and a twofold cross-validation approach [24] to determine the optimal thresholds for FOXP3 and CD3, respectively. Because both variables by this approach achieved thresholds close to 3, a common threshold for both variables was decided upon. When evaluated, thresholds close to this number did not affect the general conclusions of the analyses. Thus the means were dichotomized into <3 and ≥3, respectively, and the categorized variables used in the analyses. FOXP3, CD3 and FOXP3 in tumour cells (+/−) were individually related to overall survival during the whole study period, during the first 3 years, as well as progression-free survival using the log-rank test and Cox’s proportional hazards model. The Cox’s regression models were adjusted for disease spread (yes/no), sex (male/female), T-group (1/ 2 or 3/ 4), chemotherapy (yes/no) and age (<65 years/≥65 years). In the adjusted analyses on progression-free survival, patient number 28 had to be excluded, otherwise the model did not converge when the T-group was included. When tested, the exclusion of this patient from the other adjusted analyses did not affect the conclusions. P < 0.05 was considered statistically significant.

RESULTS

IMMUNOHISTOCHEMICAL ANALYSIS OF FOXP3 AND CD3 IN UBC

Single stainings with antibodies against CD3 (which is present on all T-cells including regulatory subsets) and FOXP3 was performed on consecutive tumour tissue sections. Both CD3+[25] and FOXP3+ T-cells (confirmed by flow cytometry of UBC; data not shown) are present in the bladder mucosa. In another study [26], only TILs within cancer cell nests were associated with survival, whereas the present study focused on CD3+ and FOXP3+ TILs in contact with tumour cells.

Both CD3+ and FOXP3+ TILs were present in all but a few of the investigated samples, although to varying extents (Table 2). In addition, FOXP3 expression in tumour cells was detected in 17 of 37 patients investigated. Figure 1A,B shows representative images of patients with relatively low and high infiltration of CD3+ TILs, whereas Fig. 2A,B shows the corresponding situation with respect to FOXP3+ TILs in FOXP3- tumours. Examples of FOXP3 expression in tumour cells are shown in Fig. 3. As described previously, cytoplasmic (n= 10), nuclear (n= 3) and cytoplasmic/nuclear (n= 4) FOXP3+ stainings of tumour cells were all observed [21,23].

Table 2.  Immunohistochemistry data
Patient numberTILsTumourSurvival time (years)
FOXP3CD3FOXP3
  1. TILs, Tumour-infiltrating lymphocytes. *Still alive at the end of the present study.

155.3384.337.79
2 11.3313.00+0.59
34.672.33+1.92
426.33260.338.82*
59.331.00+0.96
610.00 11.332.00
76.335.330.36
818.0037.00+6.84
910.338.00+7.54
100.332.67+1.27
115.335.679.42*
120.000.001.07
13 11.673.00+9.67*
141.001.001.74
1524.6713.002.35
160.000.00+0.93
1710.3335.67+3.18
18 11.3321.33+2.50
197.670.33+5.43
209.0014.332.64
215.3363.33+3.05
2212.339.338.39*
232.3316.007.48
244.330.678.20*
2510.337.670.13
265.338.00+7.30
271.670.00+0.27
2825.0021.33+2.76
2952.6762.677.91*
3022.6759.33+7.88*
3122.337.006.92*
328.674.336.87*
337.005.677.35*
343.0030.336.94
350.000.007.14*
362.330.000.89
3731.3334.33+1.55
Figure 1.

CD3+ tumour-infiltrating lymphocytes in urinary bladder cancers. Paraffin-embedded tumour tissue was stained with antibodies against the pan-T-cell marker CD3. (A) Patient number 3 and (B) patient number 1 represent patients with low (<3 cells per high-power field) and high (≥3 cells per high-power field) infiltration, respectively, in FOXP3- tumours (×40 magnification). (C) Kaplan–Meier plot showing overall survival in the two groups. P-value calculated using the log-rank test.

Figure 2.

FOXP3+ tumour-infiltrating lymphocytes in urinary bladder cancers. Paraffin-embedded tumour tissue was stained with FOXP3 antibody. (A) Patient number 36 and (B) patient number 1 represent patients with low (<3 cells per high-power field) and high (≥3 cells per high-power field) infiltration, respectively, in FOXP3- tumours (×40 magnification). (C) Kaplan–Meier plot showing overall survival in the two groups. P-value calculated using the log-rank test.

Figure 3.

FOXP3 expression by tumour cells in urinary bladder cancers. Paraffin-embedded tumour tissue was stained with FOXP3 antibody. (A) Patient number 22 and (B) patient number 28 represent patients with FOXP3- and FOXP3+ tumour cells, respectively (×40 magnification). (C) Nuclear staining of FOXP3+ tumour cells from patient number 3 at ×100 magnification. (D) Kaplan–Meier plot showing overall survival in the two groups. P-value calculated using the log-rank test.

STATISTICAL ANALYSIS

A mean number of ≥3 CD3+ cells correlated with better 3-year survival with the Cox proportional hazards model (hazard ratio, HR, 0.24; P= 0.010, 95% confidence interval, CI, 0.08–0.71) (Table 3) and the log-rank test (P= 0.009, data not shown). This correlation still persisted after the whole study period (HR, 0.42; P= 0.049, 95% CI, 0.18–1.00; P= 0.04 using the log-rank test) (Fig. 1C) (Table 4). Progression-free survival was also significantly shorter in patients with CD3 < 3 (HR, 0.23; P= 0.017, 95% CI, 0.07–0.77) (Table 5).

Table 3.  FOXP3 and CD3 expression in relation to 3-year survival
ExpressionCrude analysesAdjusted analyses*
HR95% CIPHR95% CIP
  1. HR, Hazard ratio (Cox proportional hazards model); TILs, tumour-infiltrating lymphocytes. P-values are two-sided. *Adjusted for known metastatic disease, sex, T-group, chemotherapy and age.

FOXP3 in TILs ≥ 30.300.11–0.820.0190.170.05–0.600.006
FOXP3 in tumour cells1.440.55–3.730.4550.970.34–2.820.962
CD3 ≥ 30.290.11–0.770.0130.240.08–0.710.010
Table 4.  FOXP3 and CD3 expression in relation to survival during the entire study period
ExpressionCrude analysesAdjusted analyses*
HR95% CIPHR95% CIP
  1. HR, Hazard ratio (Cox proportional hazards model); TILs, tumour-infiltrating lymphocytes. P-values are two-sided. *Adjusted for known metastatic disease, sex, T-group, chemotherapy and age.

FOXP3 in TILs ≥ 30.400.17–0.970.0440.280.11–0.750.011
FOXP3 in tumour cells2.251.02–4.990.0451.830.76–4.400.178
CD3 ≥ 30.430.19–0.990.0460.420.18–1.000.049
Table 5.  FOXP3 and CD3 expression in relation to progression-free survival during the entire study period
ExpressionCrude analysesAdjusted analyses*
HR95% CIPHR95% CIP
  1. HR, Hazard ratio (Cox proportional hazards model); TILs, tumour-infiltrating lymphocytes. P-values are two-sided. *Adjusted for known metastatic disease, sex, T-group, chemotherapy and age.

FOXP3 in TILs ≥ 30.350.10–1.150.0830.080.01–0.410.003
FOXP3 in tumour cells2.320.76–7.110.1401.500.41–5.420.540
CD3 ≥ 30.280.09–0.860.0270.090.02–0.480.004

Unexpectedly, FOXP3 expression in TILs was also associated with improved prognosis for 3-year survival (HR, 0.17; P= 0.006, 95% CI, 0.05–0.60) and for the entire study period (HR, 0.28; P= 0.011, 95% CI, 0.11–0.75) (Tables 3 and 4). This correlation was also significant using the log-rank test (P= 0.037) (Fig. 2C) and, furthermore, the patients in the FOXP3 ≥ 3 group had longer progression-free survival (HR, 0.26; P= 0.039, 95% CI, 0.07–0.94) (Table 5).

FOXP3 expression in tumour cells on the other hand, was related to poorer survival (crude HR, 2.25; P= 0.045, 95% CI, 1.02–4.99), although the adjusted P-value using the Cox proportional hazards model was not significant (Table 4). Notably, the log-rank test, which takes the hazards at every observed event time into consideration, also showed a significant negative correlation between FOXP3 expression and survival (P= 0.04) (Fig. 3C). No significant differences were observed between the cytoplasmatic, nuclear and cytoplasmatic/nuclear staining groups with respect to survival (data not shown).

Four patients, with T1–T2 tumours (patient numbers 1, 7, 17 and 20) received BCG treatment at some point before cystectomy. Because BCG evokes an immune response, it was plausible that it may influence the study outcome to some extent; however, these patients did not stand out from the group as a whole with respect to primary or secondary study outcomes.

Interestingly, the three patients with all three negative parameters (CD3 and FOXP3 < 3 and FOXP3+ tumour) had a mean survival of <10 months, whereas, at the other end of the spectrum, the three patients with the most pronounced lymphocyte infiltration (CD3 and FOXP3 > 20) and FOXP3- tumours had a mean survival of >8 years, and two of them were still alive at the end of the present study.

DISCUSSION

For the first time, the present study has shown a negative correlation between FOXP3 expression in tumour cells and survival in UBC, indicating that tumour expression of FOXP3 could be an important prognostic factor in this disease. In addition, FOXP3 expressing TILs were a positive prognostic factor, which could suggest that these TILs are not true Tregs but rather activated T-cells, with up-regulated FOXP3 expression.

The immune surveillance hypothesis proposes that tumour cells are continuously recognized and eliminated by the immune system [27,28]. Tumour immune escape encompasses several mechanisms by which tumours evade the immune system, including the induction/recruitment of regulatory immune cell populations [4,5]. In tumour immune escape, CD4+CD25hiFOXP3+ T-cells comprise the most extensively studied population. However, where activated T-cells are to be expected, both CD25 and FOXP3 should be regarded as being relatively unspecific Treg markers because they are up-regulated in CD4+ T-cells upon activation [7,9].

In UBC, pronounced TIL infiltration has been associated with better clinical outcome [29,30]. Infiltration of FOXP3+ lymphocytes in tumour tissue has been reported, and patients with UBC have elevated levels of CD4+CD25hiFOXP3+ cells in peripheral blood [12]. Studies in other malignancies have differed with respect to the prognostic significance of FOXP3+ TILs, and both negative [18,31] and positive [19,20] correlations with survival have been reported. One explanation for this apparent discrepancy could be that, in the former situation, FOXP3 expression represents the presence of true Tregs, whereas, in the latter case, FOXP3 marks an activated T-cell population. In accordance with this reasoning, the observed positive relationship between FOXP3+ TILs and survival could suggest that, at least in the case of the TILs in UBC, FOXP3 may be regarded as an activation marker. Although the mechanisms behind the impact of this population on survival are unknown, the fact remains that FOXP3+ TILs constitute a positive prognostic factor in the present study. It would be of great interest to study the function of these FOXP3+ TILs in more detail to determine their functional suppressive capacity. Alternatively, an epigenetic approach to determine the ‘Treg-commitment’ of the FOXP3+ TILs could be employed [7]. Unfortunately, this was not possible in the present study as a result of the nature of the retrospective material.

The implications of FOXP3 expression in tumour cells [21–23] are not completely understood, although it has been linked with the production of anti-inflammatory cytokines [22], and co-culture of naïve T-cells with FOXP3+ pancreatic carcinoma cells was reported to inhibit T-cell proliferation completely [21]. To add to the complexity, two very recent studies on breast cancer link tumoural FOXP3 expression with survival [23,32] but, whereas Merlo et al. [23] report a negative correlation with survival in general, Ladoire et al. [32] describe a positive correlation between tumoural FOXP3 and survival in patients with HER2+ tumours who have received neoadjuvant therapy. In the case of UBC, the present study demonstrates that patients with FOXP3+ tumours have significantly reduced survival compared to patients with tumours that are FOXP3-.

The secondary outcome, progression-free survival, showed no significant correlation with FOXP3 expression in tumour cells. This apparent discrepancy might be a result of the relatively low number of patients, which limits the power of the present study. Alternatively, given that the biological mechanisms behind this relationship are unknown, it could be speculated that overall survival is affected through a process unrelated to tumour progression. In tumour immunotherapy trials, decreases in tumour volume have not been a prerequisite for survival benefit and disease progression has been recorded before tumour shrinkage, leading to a questioning of the classical RECIST criteria as the sole clinical end point [33].

In accordance with previous studies, we found no apparent differences in survival between patients with different subcellular tumour FOXP3 stainings [23]. The fact that many tumours display cytoplasmic staining without nuclear components may be a result of defects in the nuclear localization signals of FOXP3, possibly because of acquired mutations. A recent study by Hancock and Özkaynak [34] suggests that FOXP3 achieves nuclear transport by co-transporting with other nuclear factors. Accordingly, it is plausible that the cytoplasmic localization of FOXP3 in tumours could be a result of a lack of expression or defects of these specific proteins in the tumour tissue.

In a previous study [35], we found that TILs from UBC were unresponsive against autologous tumour extract. Because reactivity was detected in draining lymph nodes, this is suggestive of active immune suppression rather than non-responsiveness. It is tempting to speculate that FOXP3 expression in the tumour cells leads to the induction of immunoregulatory genes and, subsequently, to immune suppression, resulting in the observed negative effect on survival in these patients. We have started a prospective study aiming to further evaluate the impact and function of FOXP3 in lymphocytes and tumour cells in UBC patients.

In conclusion, the results obtained in the present study suggest that, in UBC, FOXP3+ TILs could represent a positive prognostic factor, thus not representing an immune escape mechanism. Instead, tumoural FOXP3 expression appears to be a negative prognostic factor and may represent a novel tumour escape mechanism in UBC. These results clearly warrant further studies that aim to consolidate the importance of FOXP3 expression in UBC prognosis. Future research will clarify the mechanisms involved in the induction of FOXP3 in tumour cells, as well as the consequences that expression has on the interaction of the tumour with the immune system. A better understanding of tumour escape mechanisms will provide a basis for future clinical applications and immunotherapy.

ACKNOWLEDGEMENTS

This work was supported by the Swedish Cancer Foundation, Söderbergs Foundation (Stockholm, Sweden), Abbot Scandinavia, the John Hedin Cancer Research Foundation (Stockholm, Sweden) and the Wallenberg Foundation (Stockholm, Sweden).

CONFLICT OF INTEREST

None declared.

Ancillary