Even though melanomas account for only 4% of all skin cancers, they cause the greatest number of skin cancer-related deaths worldwide. Over the last few decades an increase in incidence and mortality has been observed in Caucasian populations across the world.1, 2 Clinical outcome in melanoma patients depends on several variables of which tumor thickness is an important factor (according to Breslow).3 The 5 year survival rate for patients with a Breslow thickness <1.5 mm is more than 90%, whereas survival in patients with a Breslow thickness of >3.5 mm is only 50%.4 Other important prognostic factors are, amongst others, gender and age.5–7 Fatal outcome in melanoma patients often results from occurrence of distant metastases, which mostly coincide or are preceded by lymph node metastases. In line with this concept, previous studies demonstrated that patients with a melanoma sentinel lymph node (SLN) metastasis have a worse prognosis than patients without a SLN metastasis.8, 9 However, despite known prognostic parameters, outcome often remains unpredictable and further research to identify additional relevant prognostic markers is warranted.
It has previously been shown that melanomas can elicit an immune response10, 11 and that melanoma cells can effectively be eradicated in vivo by cytotoxic activity of MHC class I antigen restricted CD8+ Granzyme B (GrB+) T-cells.12 Thus, a possible explanation for differences in clinical outcome might be that a proper immune response, although incapable of preventing the primary tumor from growing, is able to prevent the occurrence of lymph node and/or distant metastases. A large number of studies have shown that the cellular immune response plays an important role in the control of melanoma growth and spread.13, 14
Proper activation of a CD8+ T-cell mediated immune response requires antigen to be presented in the context of the appropriate MHC class I molecules. MHC class I molecules are hetero dimers consisting of a transmembrane α-chain, encoded by 3 polymorphic loci HLA-A, -B and -C, which is non-covalently associated with β2-microglobulin (β2m). When a cell looses expression of MHC class I molecules, it can no longer be recognized by CD8+ lymphocytes but it might become susceptible to natural killer (NK) cell recognition.15 Loss of MHC class I antigen expression has been shown to be involved in immune escape and tumor progression12, 16, 17 and has been associated with poor clinical outcome.18, 19 In addition, target cell killing by cytotoxic T-cells (CTLs) requires the help of CD4+ T-helper (Th) cells.20, 21 CD4+ Th cells require presentation of antigens in the context of MHC class II by professional antigen-presenting cells (APCs).22, 23 MHC class II molecules are also heterodimers, but consist of 2 homologous peptides, an α and β chain, each subunit contains 2 extracellular domains, a membrane spanning domain and a cytoplasmic tail. The HLA class II region map into 3 subregions: DR, DQ and DP and are encoded on the HLA region of chromosome 6.24, 25
We have previously shown, in stage III/IV melanoma patients, that presence of activated (GrB+) CTLs in primary biopsy specimens is correlated with a favorable prognosis following treatment with irradiated autologous tumor cell vaccination.26 In the current study, we investigated if presence of GrB+ Tumor Infiltrating Lymphocytes (TILs) in diagnostic primary melanoma biopsies is related to clinical outcome in clinically stage II melanoma patients. Moreover, we investigated whether presence of TILs correlates with expression of MHC class I and MHC class II antigens on tumor cells and/or tumor infiltrating APCs.
Patients, material and methods
Patients and clinical characteristics
Patients primarily diagnosed with stage II melanoma were treated with surgical resection of the primary melanoma and a sentinel lymph node procedure (SNP) in the VU University Medical Center.27 Two groups of patients were selected on the basis of contrasting clinical outcome. Unfavorable outcome was determined by the development of metastasis overtime and favorable outcome was determined by no sign of recurrence at time of last follow-up. Fifteen patients with an unfavorable outcome (median follow up 35 months) were compared to 20 patients with favorable outcome (median follow up 46 months). Patients were further selected to have comparable age, gender distribution and Breslow thickness to avoid a confounding effect of these known prognostic parameters. Paraffin embedded biopsies were kindly provided by the departments of pathology from the following hospitals: Medical Center Alkmaar, Alkmaar, Bovenij Hospital, Amsterdam and VU Medical Center Amsterdam, The Netherlands.
Antibodies and immunohistochemistry
Paraffin embedded 3 μm tissue sections of primary melanoma biopsies were treated as described previously.26, 28 Lymphocytes were characterized for expression of CD4, CD8, CD56 and GrB using the following antibodies: monoclonal antibody (mAb) GrB7 (mouse IgG2a) specific for human GrB (Monosan, Uden, Netherlands),29 mAb CD4 (mouse IgG1, MEM-241, Monosan), mAb CD8 (mouse IgG1, clone 144B, Dako, Heverlee, Belgium), mAb CD56 (murine IgG2a, MEM-188, Monosan). Expression of MHC class I antigen on melanoma cells was investigated using a polyclonal antibody against β2m (rabbit Ig, Dako), the mAb HCA2 and mAb HC10 preferentially recognizing HLA-B/C locus products30, 31(kindly provided by Prof. Dr. J. Neefjes, Netherlands Cancer Institute, Amsterdam, The Netherlands). The mAb HCA2 reacts preferentially with HLA-A locus heavy chains. Its reactivity contrasts with that of HC10, a mAb with preferential specificity for HLA-B and -C heavy chains. Both HCA2 and HC10 were raised against free class I heavy chains of HLA-A and -B antigens respectively, to obtain mAbs that would still react with denatured class I antigens, as they occur in conventional light microscopically analysis of formalin-fixed, paraffin-embedded sections.
Expression of MHC class II antigen was determined using a mAb recognizing HLA-DR (mouse IgG2b, clone LN3, VUmc, Amsterdam, The Netherlands). For staining with antibodies against GrB, MHC class II, CD8, CD56 and HLA-B/C, antigen retrieval was performed with 10 mM Na-citrate, pH 6 and for staining with antibodies against HLA-A and CD4 antigen retrieval was performed with 10 mM TRIS, 1 mM EDTA, pH 9. No antigen retrieval was required for staining with anti-β2m antibody. Following antigen retrieval, primary antibody was applied and visualization was performed with either the Envision™ horseradish peroxidase system (DakoCytomation, Glostrup, Denmark) for MHC class I, MHC class II, CD8 and CD56, or Power Vision plus™ system (Immunologic, Duiven, The Netherlands) for GrB and CD4 according to manufacturer's instructions.
Characterization of GrB+ TILs
Double staining was performed to determine whether GrB+ TILs are activated cytotoxic CD8+ TILs. To detect double positive (GrB+CD8+) cells, slides were first incubated with GrB7 (mouse IgG2a) as described above followed by a goat-anti-mouse IgG2a -HRP (SouthernBiotech, Birmingham, AL) and subsequently incubated with H2O2. Subsequently, slides were incubated with either anti-CD8 or anti-CD4 antibody as described in the section above followed by goat-anti-mouse IgG1-Biotin (SouthernBiotech). Visualization of CD4 or CD8 was performed with streptavidin-Alexxa488 (Molecular Probes, Invitrogen, Breda, The Netherlands). Images of CD8=GrB= TILs were recorded using a confocal laser scanning microscope a 40× oil objective was used. To visualize tissue compartments as well as the difference between TILs and lymphocytes surrounding the tumor, the gain and offset settings of each individual channel were optimized such that some noise is introduced to show the tissue compartments.
Interpretation of the staining
For interpretation of stainings with antibodies against CD4, CD8, CD56, GrB, MHC class II (HLA-DR), and MHC class I antigens (HCA2, HC10 and β2m), lymphocytes surrounding the tumor served as internal positive control for the staining. The presence of CD4+, CD8+, and GrB+ lymphocytes was scored in 2 groups: cases with positive TILs (>1 per high power field (HPF) and cases without TILs (<1 per HPF). The difference between melanoma cells and lymphocytes was based on morphological examination (see also Fig. 1). Percentages of HLA-A, HLA-B/C and β2m expressing tumor cells were scored semi-quantitatively in steps of 10% from 0 to 100%.32 APCs present in tissue surrounding the tumor area functioned as positive control for HLA-DR staining. Expression of MHC class II antigen was based on HLA-DR staining and tumor cells were recognized based on morphological examination (see Fig. 1). Expression of MHC class II antigen on tumor infiltrating APCs was scored as described above for MHC class I.
Progression free survival (PFS) time was determined from date of first diagnosis of primary tumor until date of diagnosis of disease metastasis. Overall survival (OS) time was determined from date of first diagnosis of primary tumor until date of disease related death. Patients without a recurrence were censored at the last time of follow-up. One patient died of a cause unrelated to the disease and was censored. Differences between the Kaplan Meier curves were analyzed using the Log-rank test. Qualitative variables were analyzed by Pearson's χ2 test, or by Fisher exact test when appropriate. The Mann-Whitney U-test was used to compare group means. All p values were based on two-tailed statistical analysis. If p values ≤0.05 they were considered significant. All analyses were performed using the SPSS statistical software (version 12 SPSS, Inc, Chicago, IL).
Patients were selected based on comparable important prognostic parameters and the 2 groups with different clinical outcome therefore demonstrated no significant differences regarding gender, age, Breslow thickness and SLN status (see Table I). Although the difference was not significant, patients with an unfavorable outcome were more frequently male than female, which is consistent with previous studies.6
Table I. Tumour Characteristics in Relation to Clinical Outcome
ns, not significant; nd, not determined, due to lack of available tissue, absent means that lymphocytes were only detected around tumour and no infiltrating lymphocytes detected and present means infiltrating lymphocytes detected.
As determined by Mann-Whitney U-test.
Preserved expression of MHC-I (described in m&m) = >50% β2M + >80% HLA-A/B/C.
Presence of activated cytotoxic TILs and CD4+ TILs cells is associated with a favorable clinical outcome
The 2 predefined groups with different clinical outcome were compared and in patients with favorable outcome CD4+ TILs and GrB+ TILs were significantly (p = 0.02 and 0.01, respectively) more frequently detected than in patients with unfavorable outcome (see Table I). Furthermore, patients with a favorable outcome tended to harbor more CD8+ TILs (See Table I, p = 0.08). Staining for different TIL populations are shown in Figure 1.
Double staining procedures (see Fig. 1e) showed that GrB+ TILs were also CD8+ TILs and therefore were considered to be activated cytotoxic TILs (acTILs). In 2/34 cases GrB+ TILs but no CD8+ TILs were detected. These GrB+ cells were CD56− indicating that these cells were not NK cells. In none of the primary melanoma biopsies CD56+ positive tumor infiltrating NK cells were detected.
The presence of acTILs correlated strongly with the presence of both CD4+ and CD8+ TILs (Table II). Thus, based on the analysis of TILs, 2 groups of melanomas could be identified: 1 group of melanomas showing infiltration of acTILs as well as infiltration of CD4+ T-cells and 1 group of melanomas without infiltration of acTILs or CD4+ TILs. As expected from these results a strong significant correlation was observed between individual scoring results for the TIL subpopulations (see Table II).
Table II. Presence of Actils (GrB+) is Correlated with Presence of Several Mediators of Adaptive Immune System
Subsequently, patients were stratified according to the presence or absence of the respective TIL populations and Kaplan-Meier curves were constructed to estimate differences in PFS and OS. The presence of GrB+ (Fig. 2a), CD8+ (Fig. 2b) and CD4+ (Fig. 2c) TILs was significantly associated with a longer PFS. The presence of GrB+ (p = 0.02) TILs was significantly associated with a longer OS and the presence of CD8+ (p = 0.09) and CD4+ (p = 0.08) showed a trend toward a longer OS.
Expression of MHC class I antigen is associated with the presence of acTILs
The mean percentage of tumor cells expressing HLA-A and HLA-B/C was determined at 80% and the mean percentage of tumor cells expressing β2m was 50%. On the basis of the mean percentages of expression of the 3 individual markers (see Fig. 1f-1h), MHC class I antigen expression was considered preserved if both HLA-A and HLA-B/C expression was detected on at least 80% of the tumor cells and β2m was expressed on ≥50% of the tumor cells. Complete loss of MHC class I antigen expression on tumor cells was observed in 1 case for HLA-B/C and in 2 cases for β2m. All other patients showed expression of the 3 different markers ranging from 10% to 100% positive tumor cells. A significant correlation was found between expression levels of HLA-A and HLA-B/C (p = 0.004, Pearson's χ2 test) and between expression of HLA-A/HLA-BC and β2m (p = 0.002, Pearson's χ2 test). More importantly, a strong correlation was detected between expression of MHC class I antigen on tumor cells and presence of acTILs, both for individual markers and for preserved (>50% β2M + >80% HLA-A/B/C) MHC class I antigen expression (see Table II). When comparing the 2 groups with different clinical outcome, it appeared that clinically favorable cases are usually characterized by preserved MHC class I antigen expression on melanoma cells (see Table I). If patients were stratified according to MHC class I antigen expression on tumor cells, preserved MHC class I antigen expression was indeed significantly associated with a favorable PFS (see Fig. 3a, Log rank test; p = 0.005).
Expression of MHC class II antigen on APCs correlates with the presence of CD4+ TILs
Tumor infiltrating MHC class II antigen expressing APCs were detected in 24/34 cases and their presence correlated significantly (p = 0.03) with presence of CD4+ TILs (See Table III) and acTILs (see Table II, p = 0.01). Moreover, presence of infiltrating MHC class II antigen expressing APCs correlated significantly with favorable prognosis (see Fig. 3b). No expression of MHC class II antigen was detected on normal melanocytes (not shown); however, in 9/35 (26%) melanomas MHC class II antigen expression was detected in 10-80% of tumor cells. Expression of MHC class II antigen on melanoma cells correlated significantly with presence of tumor infiltrating CD4+ cells (p = 0.03, see Table III). No correlation with PFS was observed.
Table III. MHC Class II Expression on Both Tumor Cell and Infiltrating Cells is Correlated with Presence of CD4+ TILS
As determined by Pearson's χ2 test of Fisher exact test, when appropriate.
MHC class II expression on tumour cells
MHC class II expression on APCs
In this report we have shown that diagnostic biopsies of clinically stage II melanoma patients with a favorable outcome are characterized by presence of activated (GrB+ and CD8+) TILs and T helper (CD4+) TILs, whereas these TILs are absent in melanoma biopsies of most patients with unfavorable clinical outcome. Moreover, we have shown that presence of acTILs and CD4+ TILs correlates with preserved MHC class I antigen expression on tumor cells and with the expression of MHC class II antigen on tumor infiltrating APCs and with MHC class II antigen expression on tumor cells.
Our results are in accordance with previous studies demonstrating that presence of lymphocytes is associated with a favorable clinical outcome in melanoma biopsies.33–35 Furthermore, our data demonstrate that melanoma infiltrating T-lymphocytes consist of CD4+, CD8+ and GrB+ T-lymphocytes populations. In 2/34 cases GrB+ TILs but no CD8+ TILs were detected. Most likely these cells are activated TILs which lost their detectable expression of CD836 because no NK cells and no CD4+GrB+ TILs were detected. Activation of cytotoxic T-lymphocytes depends on specific MHC class I restricted antigen recognition and on the presence of co-stimulatory expressing and cytokine producing CD4+ T helper cells.21, 37
We observed a strong correlation between presence of TILs and intact MHC class I antigen expression on melanoma cells. This association, which has previously also been observed in solid tumors38, 39 and in lymphomas,32, 40 supports the idea that loss of MHC class I antigen expression results in failure to mount a cellular, melanoma antigen-specific immune response and as such may be a possible powerful immune escape mechanism.18, 41 Alternatively, interferon-γ (IFNγ) upregulates transcription of MHC class I and II molecules42 and therefore, the association between decreased MHC class I antigen expression and absence of TILs may be caused by lack of TIL derived IFNγ in the tumor environment. Of note, even in cases with apparent intact MHC class I antigen expression specific HLA haplotype losses may interfere with proper antigen recognition,43 because the used antibodies do not allow to identify these MHC class I antigen alterations.
Even though we did not demonstrate that the infiltrating TILs are specifically directed against the melanoma cells, it can reasonably be assumed that the TILs are part of an intact anti-tumor response, assuming the activated status (GrB+) strongly suggests that these TILs are functionally active. This opinion is further strengthened by our previous report demonstrating tumor antigen-specific TILs are present in melanoma patients.14 In addition, in the large majority of cases with acTILs also CD4+ TILs and MHC class II antigen expressing APCs were present. In an earlier report, presence of CD8+ and CD4+ TILs was also closely correlated.12 However, no correlation was detected between expression of MHC class II antigen and the presence of TILs. This discrepancy with our results might be explained by the fact that only the presence of CD8+ TILs was investigated whereas in our study GrB+ TILs were analyzed.
Thus, in patients with favorable outcome, all ingredients for an effective cellular immune response appear to be present. Still, even though this immune response is apparently unable to effectively eradicate the primary tumor, because the tumor would not disappear without adequate surgery, the presence of a cellular immune response appears to be able to prevent the occurrence of distant metastasis. In patients with an unfavorable outcome with eventually the occurrence of metastases, these ingredients for an effective cellular immune response were not observed, further suggesting that the cellular immune response plays a functional role in preventing melanoma dissemination.
Expression of MHC class II antigen in melanoma cells was detected in 25% of cases. Although controversial, most reports described that expression of MHC class II antigen is associated with an unfavorable outcome.44 No such association was observed in our series, but this might be caused by the relatively low number of cases studied. However, its possible effect appears to be limited when compared with the prognostic value of activated TILs.
Because patients were selected based on comparable established prognostic parameters (SLN status, Breslow thickness, gender and age), the differences in clinical outcome between patients with favorable and unfavorable outcome in this study are most likely to be directly related to differences in the cellular immune response.
We conclude that the presence of acTILs and CD4+ TILs and putative intact antigen presentation in the context of MHC class I and MHC class II antigen presentation in primary melanomas predicts a highly favorable outcome in clinically stage II melanoma patients. These data strongly support the notion that the cellular immune response is a major factor in preventing melanoma cell dissemination.