The predictive value of CD8, CD4, CD68, and human leukocyte antigen-D-related cells in the prognosis of cutaneous malignant melanoma with vertical growth phase




To establish the prognostic value of immune system cells that infiltrate melanoma, the authors evaluated the distribution and density of T lymphocyte subsets, macrophages, and dendritic cells in samples of primary cutaneous melanoma from 47 patients with Stage I and II melanoma according to the American Joint Committee on Cancer staging system.


Immunohistochemical demonstrations of CD8 and CD4 lymphocytes, CD68 macrophages, human leukocyte antigen-D-related (HLA-DR) cells, S-100 protein, and melanoma-associated antigens Melan A and HMB-45 were performed. The results were derived from independent histopathologic reviews by two pathologists. The low-density, moderate-density, and high-density groups of cells that infiltrated the base of the tumor during the vertical growth phase were compared with the overall survival rate using the Kaplan–Meier method and the log-rank test. Clinical variables (gender, age, tumor location, Clark level, vascular/lymphatic invasion, and thickness) also were analyzed.


The CD8 lymphocytes exhibited independent statistically positive significance in survival (log-rank test, 8.49; P = 0.01) between patients in different lymphocyte density groups. There was a difference in 5-year survival among patients in the high-density group (78.8%), the moderate-density group (44.4%), and the low-density group (25.0%). The CD4 lymphocytes, which were less numerous than CD8 cells, had similar distribution. There also was a correlation of HLA-DR cells with overall survival (log-rank test, 5.29; P = 0.02). CD68 cell density was not found to be correlated with survival.


The presence and number of infiltrating CD8 lymphocytes as well as the overall occurrence of HLA-DR cells may be considered independent, favorable prognostic factors in melanoma. The current results may be important for identifying other prognostic factors with which to evaluate disease progression and develop immune therapies for patients with melanoma. Cancer 2005. © 2005 American Cancer Society.

In an effort to predict survival better in patients with Stage I and II melanoma, numerous individual clinicopathologic factors and multifactor models have been identified that have independent prognostic value in small and large melanoma cohorts.1–7 Among all of these potentially useful prognostic biomarkers, the new American Joint Committee on Cancer (AJCC) staging system8, 9 incorporates only tumor thickness, ulceration, satellites, and lymph node status because, for most of the other clinicopathologic prognostic factors, there are no reproducible definitions that are easy to apply, and there is a lack of agreement among the world's experts.10 In particular, the presence of inflammation and regression in primary melanomas is a relevant prognostic factor and has been considered a strong indicator for the existence of an antitumor immune response in patients with melanoma.11

Primary AJCC Stage I and II cutaneous melanoma nearly always is associated with a chronic inflammatory infiltrate, which is comprised predominantly of lymphocytes. However, the assessment of tumor-infiltrating lymphocytes (TILs) appears to be subject to great interobserver variability.12 Several authors have suggested the density and pattern of distribution involving the vertical growth phase (VGP) of melanoma or metastatic melanoma deposits in lymph nodes as histologic parameters of prognosis.1, 11, 13–15 The central role of T cells in antitumor immunity is well established, so that several melanoma-associated antigens recognized by T cells have been characterized. T response to these antigens is heterogeneous, because the antigens can be recognized by helper-T lymphocytes, in the context of human leukemic antigen (HLA) II molecules, or by cytotoxic T lymphocytes, in the context of class I molecules. In the light of these findings, numerous studies of immunotherapies and vaccines have evaluated the presence of CD8 (cytotoxic) and CD4 (helper) T cells as an antitumoral activity index of immune system-associated tumoral regression.16–22

Other studies have shown that the location and extent of dendritic cell infiltration and the expression of HLA-D related (HLA-DR) are associated with the clinical response and improved survival.23, 24 In addition, infiltration of macrophages is correlated closely with the depth of tumor and angiogenesis in melanomas,25 and there is controversy regarding the prognostic significance of TILs in solid human tumors.26

To address these points, we quantitatively evaluated CD8, CD4, CD68, and HLA-DR cells in selected areas of tumor base in the VGP. When possible, these cells were counted, and the mean of these counts was correlated with survival in a statistical analysis.



From a total of 152 patients with primary skin melanoma who underwent observation at the Oncologic Hospital “Businco” (Cagliari, Italy) between January 1995 and January 2002, 47 patients were selected for further study because they had melanoma that demonstrated VGP, they had no metastases, and they had complete clinical data available, including follow-up until April 2004. The negative lymph node status and the absence of metastases were verified by both clinical and pathologic examination. Selected patients were recruited into a prospective study to evaluate the correlation between the prognostic value of infiltrates and survival with melanoma. The group included 22 men and 25 women, ranging in age from 11–93 years (mean age, 57 yrs; standard deviation, 19.2 yrs). The anatomic location of the primary tumor included 10 tumors located in the trunk, 18 head and neck tumors, and 19 tumors located in extremities. In this group of melanomas, 1 tumor was Clark Level II, 11 tumors were Clark Level III, 24 tumors were Clark Level IV, and 11 tumors were Clark Level V. At the time of last follow-up, 19 patients were alive and 28 patients had died of melanoma within 112 months. Selected surgical specimens were grouped according to the AJCC cancer staging manual,9 as shown in Table 1.

Table 1. Staging and Classification of 47 Patients According to the American Joint Committee on Cancer Staging System
StageClassificationNo. of patients
  1. T: tumor thickness; N: metastatic lymph nodes; M: metastases.


Each tumor was included completely in multiple paraffin blocks. To evaluate prognostic parameters, we selected only sections taken from the block with the largest tumor thickness.

Tumoral areas were identified on sections that were counterstained with hematoxylin and eosin and on adjacent sections that were stained immunohistochemically for melanoma-associated antigens, such as S-100 protein, melan A, and HMB-45. Moreover, we examined hematoxylin and eosin-stained sections for evidence of apoptosis in relation to the infiltrate of TILs. The results were derived from independent histopathologic reviews by two pathologists (C.F. and R.C.) on separate occasions. The study protocol was approved by the Research Ethics Committee at our institutions, and informed consent was obtained from all participants.


The immunohistochemical staining was performed on formalin-fixed and paraffin-embedded archival tissues. Microtome sections measuring 5 μm thick were treated for the immunohistochemical demonstration of CD4, CD8, CD68, HLA-DR, S-100 protein, and melanoma-associated antigens melan A and HMB-45, using the alkaline phosphatase-streptavidin method. The sections were rehydrated in phosphate-buffered saline (PBS) and antigen retrieval was performed by immersion in 0.1% trypsin solution in PBS at 37 °C for 5–10 minutes or by microwave heating for 5 minutes × 4 (total, 20 min) in 10 mM citrate buffer solution (pH 6.0) for melan A antigen. The sections were treated for 45 minutes with 10% normal goat serum or normal horse serum in PBS. Mouse monoclonal antibody to human CD4 (clone 4B12; Novocastra Laboratories, Newcastle-upon-Tyne, U.K.; 1:50 dilution), mouse monoclonal antibody to human CD8 (clone 1A5; Novocastra Laboratories; 1:50 dilution), mouse monoclonal antibody to human macrophages CD68 (clone PG-M1; Dakopatts, Glostrup, Denmark; 1:100 dilution), mouse monoclonal LN3/HLA-DR to human HLA-DR antigen (clone LN3; Clonab®, Biotest, Milan, Italy; 1:5 dilution), rabbit polyclonal antibody to bovine protein S100 (Dakopatts; 1:1000 dilution), mouse monoclonal antihuman HMB-45 (clone HMB-45; Dakopatts; 1:100 dilution), and mouse monoclonal antihuman melan A (clone A103; Dakopatts; 1:100 dilution) were used as primary antisera. Biotinylated antirabbit and antimouse immunoglobulin G were used as secondary antisera. The sections were incubated further in alkaline phosphatase-streptavidin (Vector Laboratories, Burlingame, CA; 1:1000 dilution) for 30 minutes at room temperature, reacted with Fast Red Substrate System (Dakopatts) or with Dako® Fuchsin + Substrate-Chromogen, and then counterstained with Mayer hematoxylin; the sections were rinsed thoroughly in PBS between each step and finally mounted in glycerol gelatin (Sigma Chemical Company, St. Louis, MO). In negative control sections, the specificity of the antisera was tested by replacing the primary antibodies with normal serum, and human tonsil sections were used as positive controls for CD8, CD4, CD68, and HLA-DR antigens.

Statistical Analysis

Data-acquisition was performed as follows: 1) First, we identified the advancing edge of the tumor by using tumor markers and drew an imaginary line that separated the tumoral area from the dermal tissue below (Fig. 1A). 2) The adjacent microscopic fields above this imaginary line were evaluated at × 400 magnification (Zeiss AxioPhot2 microscope; Carl Zeiss Vision GmbH, Hallbergmoos, Germany). 3) In this way, we visualized a tumoral area with the base represented by the imaginary line and the height represented by the diameter of the microscopic field (approximately 600 microns). 4) In each microscopic field, the infiltrating CD8 lymphocytes were counted and the average value of the counts was obtained over the entire area considered. The mean of the counts was used for statistical analysis. Patients were divided into 3 groups according to cell density (0–20 cells, > 20–100 cells, and > 100 cells) in relation to the mean value of the number of CD8 TILs for each patient. The overall survival rate was compared between the low-density (0–20 cells), moderate-density (21–100 cells), and high-density (> 100 cells) CD8 TIL groups.

Figure 1.

Four photomicrographs are shown of CD8-infiltrating T lymphocytes. (A) The CD8-infiltrating cell counts were performed in microscopic contiguous fields (at × 400 magnification) of the lower tumoral area, above the black line, which defines the advancing edge of the tumor. The fields had the deepest border on the line. Only lymphocytes in this area were counted. (B) This image corresponds to a detail of Panel A. The arrows indicate the positive lymphocytes nearest to tumor cells. (C) CD8 cells infiltrate the perinodular stroma inside the tumor mass. (D) CD8 lymphocytes diffusely infiltrate close to the tumor cells (Fast Red chromogen). Original magnification × 100 (A, C); × 400 (B); × 200 (D).

Analyses were performed using the SPSS statistical package (version 11.5; SPSS Inc., Chicago, IL). Statistical analyses of the time to death due to malignant melanoma were performed using the product-limit procedure (Kaplan–Meier method), with the date of histologic diagnosis as the starting point; differences between categories were tested by the log-rank test.

The CD4 antigen was present on the helper-T subset and on monocytes and macrophages, which shared the same localization in the tumoral areas. These different cells were assembled in clusters, and the cell borders were indistinguishable; therefore, a count of CD4 lymphocytes was not performed. Moreover, it was not possible to count HLA-DR and CD68 cells because of the difficulty in distinguishing them one by one, especially when they are pigmented. Instead, the overall occurrence of HLA-DR or CD68 cells was scored as negative (−) or sparse (+), and moderate (++) or high (+++). Based on this scoring, we identified two density groups that were classified with a low score if the occurrence of HLA-DR and CD68 cells was negative or sparse and a high score if their occurrence was moderate or high at the base of the VGP. Furthermore, we correlated these data with the survival rate by using the Kaplan–Meier method. In addition, clinical variables such as gender, age, primary tumor location, Clark level of invasion, tumor thickness (AJCC staging system9), and vascular/lymphatic invasion were correlated with survival by using the Kaplan–Meier method.



All sections showed positive immunostaining for at least two of the melanoma markers. Most samples were positive for HMB-45 staining.

The distribution of CD8 TILs (Fig. 1) differed significantly between and within the lesions, as expected: The lymphocytes infiltrated across the base of the VGP of the tumor, adjacent to normal dermis (Figs. 1A and 1B), in 49% of samples; the lymphocytes were present throughout all the tumor in 34% of samples; and the lymphocytes did not infiltrate the melanoma in 17%. Inside the tumor mass, some samples showed CD8 cells infiltrating the perinodular stroma (Fig. 1C) but not close to tumor cells; in other samples, the CD8 lymphocytes infiltrated also close to tumor cells (Fig. 1D). Often, there were pigmented cells with macrophagic morphology near CD8 TILs (Fig. 1B). With regard to apoptosis, only samples with TILs were found to demonstrate apoptotic, melanocytic tumor cells. In these samples, apoptotic cells were rare and scattered.

The CD4 antigen is common to helper lymphocytes and monocytes. CD4 cells were localized predominantly within the stroma around tumor nodules (Fig. 2A). In most samples, CD4 cells, which are recognized as lymphocytes, were colocalized with CD8 TILs and were present in a smaller amount than CD8 cells. CD4 cells with macrophagic morphology often were pigmented and colocalized with CD68 cells (Fig. 2B).

Figure 2.

(A) CD4 cells show mostly macrophagic morphology and are localized predominantly within the stroma around tumor nodules. Scattered and isolated lymphocytes between melanoma cells also are observed. (B) CD68 cells have the same distribution as CD4 cells with macrophagic morphology (Fast Red chromogen). Original magnification × 200 (A), × 400 (B).

HLA-DR cells were numerous in the peripheral tumoral area adjacent to the dermis and gradually spread over the tumor mass (Fig. 3B), in the nodules, and in the stroma surrounding them (Fig. 3C). HLA-DR cells and CD68 macrophages showed a similar distribution throughout the primary tumor, although HLA-DR cells were predominant, and HLA-DR reactivity was stronger. Furthermore, stained cells often were pigmented strongly.

Figure 3.

(A) This image shows positive immunostaining for HMB-45 antigen. (B) This step section was immunostained for human leukocyte antigen-D related antigen. Numerous positive cells are present at the base of the tumor and gradually spread over the tumor mass, surrounding and infiltrating tumor nodules. (C) This image corresponds to a detail of Panel B (Fast Red chromogen). Original magnification × 40 (A, B); × 200 (C).

Statistical Analysis

The correlation between CD8, HLA-DR, and CD68 cell density groups and postsurgical survival status was determined at 10 years of follow-up. Table 2 shows the statistical data from the CD8 TILs density groups as of April 2004, and provides the percentages of mortality for the overall observation period and for 5-year survival according to the density groups. Figure 4 illustrates differences in the survival rate between patients with low (36.2% of patients), moderate (42.5%), and high CD8 TIL density (21.3%). According to our survival analysis, the presence of CD8 lymphocytes infiltrating the base of the tumor mass exhibited a statistically positive, significant correlation with survival (log-rank test, 8.49; P = 0.0144).

Table 2. Statistical Data on CD8 Tumor-Infiltrating Lymphocyte Density Groups and Human Leukemic Antigen-D-Related Density Groups
Density groupRangeNo. of patientsNo. of eventsCensored (%)Median survival (mos)% mortality% 5 yr survival
  1. SE: standard error; TILs: tumor-infiltrating lymphocytes; HLA-DR: human leukocyte antigen-D-related; −: negative; +: sparse; ++: moderate; +++: high.

CD8 TILs       
 Low density0–20171323.5188125 (SE, 10.8)
 Moderate density21–10020114530.65544.4 (SE, 11.2)
 High density> 10010280772078.8 (SE, 13.4)
 Overall 472644.736.756.543.9 (SE, 7.6)
HLA-DR cells       
 Low density−/+1082026800.0
 High density++/+++301550725050.9 (SE, 9.4)
 Overall 402342.53057.545.8 (SE, 7.6)
Figure 4.

Overall survival analysis, performed using the Kaplan–Meier method, is illustrated for patients with melanoma who had different numbers of CD8 lymphocytes infiltrating the base of the tumor mass.

The overall occurrence of HLA-DR cells in the peripheral tumor was scored as negative and sparse (25% of patients) or moderate and high (75%). Table 2 shows the statistical data according to the HLA-DR cell-density groups as of April 2004, and indicates the percentages of mortality for the overall observation period and for 5-year survival in the 2 density groups. According to the statistical analysis (Fig. 5), the occurrence of HLA-DR cells infiltrating the base of tumor mass demonstrated a statistically significant, positive correlation with survival (log-rank test, 5.29; P = 0.0214).

Figure 5.

Overall survival analysis, performed by the Kaplan–Meier method, is illustrated for patients with melanoma who had different densities of human leukocyte antigen D-related cells at the base of the tumor mass.

No significant correlation was found between CD68 cell-density groups and survival overall with the Kaplan–Meier method (P = 0.55). The distribution of patients according to the single factors gender, age, anatomic site, Clark level of invasion, tumor thickness, and vascular/lymphatic invasion did not demonstrate any statistically significant correlation with survival (Table 3).

Table 3. Distribution of Clinicopathologic Variables and Survival for 47 Patients with Cutaneous Melanoma
VariableNo. of patientsNo. of events% censored% 5-yr survivalSE (%)P valuea
  • SE: standard error.

  • a

    Log-rank test.

  • b

    Median value.

Gender     0.89
Age     0.75
 < 61 yrsb231343.4839.511.1 
 ≥ 61 yrs241441.675010.2 
Anatomic site     0.84
 Head and neck181138.8943.211.9 
Clark level     0.79
 II, III12741.6741.714.2 
 IV, V352042.8647.18.7 
Thickness     0.24
 T1 and T2168505012.5 
 T3 and T4311938.7143.59.2 
Vascular/lymphatic invasion     0.18


Among all potentially useful prognostic biomarkers, the new AJCC staging system incorporates only tumor thickness, ulceration, satellites, and lymph node status. The majority of the other clinicopathologic prognostic factors have been identified as significantly independent in < 75% of studies with > 450 patients and ≥ 5 years of mean or median follow-up.10 In the current study, the clinical variables tumor thickness, Clark level, vascular/lymphatic invasion, gender, age, and primary tumor site did not demonstrate any statistically significant correlation with survival, most likely because of the low number of patients analyzed. Examination of the apoptotic tumor cells relative to the infiltrating lymphocytes did not show any significant differences, most likely due to the fact that the assessment was performed on sections stained with hematoxylin and eosin, and no phenotypic markers were used, which may have revealed relevant variations.

Malignant transformation may be associated with the expression of molecules on the tumor cells, which are recognized as foreign by the specific immune system and may induce immune responses that can modulate the metastatic potential of cutaneous melanoma. In fact, clinicopathologic correlations have shown that the presence of infiltrates in some tumors, such as malignant melanoma, is associated with a better prognosis compared with histologically similar tumors without infiltrates.27

VGP apparently signals a qualitative change in the biology of a melanoma: There is evidence that the VGP may signify the onset of the metastatic evolution of the disease. Morphologically, the VGP is characterized by the focal proliferation of melanoma cells as a cohesive aggregate in the dermis. The constituent cells are crowded, frequently monomorphous, markedly atypical, generally with high mitotic activity, and individual cell necrosis. Such nodules usually are associated with TILs, which are less prominent than in radial growth phase; have been categorized as brisk, nonbrisk, and absent; and are considered one of the reliable histologic parameters for prognosis.1, 13, 15 The central role of T cells in antitumor immunity is well established, and the effector cells with antitumor activity are predominantly CD8 cytotoxic lymphocytes. In an attempt to verify the discriminant competence of the presence and the number of tumor-infiltrating CD8 cytotoxic cells as independent histopathologic prognostic factors, we evaluated a series of 47 VGP cutaneous melanomas. The localization of CD8 TILs was the same for infiltrating lymphocytes, as described by Clemente et al.13: The CD8 lymphocytes in some samples infiltrated across the base of the VGP (peripheral) of the tumor, adjacent to normal dermis; in other samples, the CD8 lymphocytes were present throughout all of the tumor (diffuse); and, in a few samples, CD8 lymphocytes were absent.

We counted the CD8 TILs in areas at the bottom of the VGP, near the dermis, and classified all tumors into the high, moderate, and low lymphocyte groups. It is interesting to note that there was a significant difference in 5-year survival between patients in the groups with high (78.8%), moderate (44.4%), and low (25%) CD8 lymphocyte density in the series of melanomas without any relation to other clinical variables.

These findings imply that the presence of CD8 TILs can be considered an important, independent prognostic factor. The presence of > 100 CD8 lymphocytes increases the survival rate significantly and leads to a better prognosis, confirming the antitumor activity of cytotoxic T lymphocytes in melanoma. Our current results confirm the work of Clark et al., who identified the significance of TILs in the prognosis of patients with melanoma for the first time in 1989,1 and are quite similar to those published by Clemente et al.13 in terms of 5-year survival between brisk, nonbrisk, and absent categories of TILs. In addition, Tuthill et al.15 found that a brisk host response of TILs was very protective concerning survival in patients with melanoma.

T lymphocytes are critical as mediators of antitumor immunity through the combined activity of CD8 cytotoxic and CD4 helper-T cells; and, in this context, antitumor cytotoxic lymphocyte activity is likely to depend in part on signals provided by CD4 helper cells. Colocalization of CD4 and CD8 lymphocytes was found within lesions, as described by Bernsen et al.,28 although CD4 lymphocytes were found less often than CD8 lymphocytes. However, we did not count CD4 cells, because their antigen is common to both macrophages and lymphocytes; therefore, the count of CD4 cells cannot constitute a valid approach for evaluating these cells as a prognostic parameter. CD4 lymphocytes recognize tumor antigens in association with HLA II molecules that are expressed by activated B lymphocytes, macrophages, and dendritic cells, which contact tumor antigens for their phagocytosis, processing, and activation of T-helper cells. We evaluated the presence and the density of antigen-presenting cells by HLA-DR antigen, and their overall occurrence was found to be correlated significantly with survival. In the light of this finding, it has to be confirmed29 that the tumor antigens are recognized by lymphocytes if they are primed suitably, so the success of the immune response belongs to the mediation of antigen-presenting cells. HLA-DR is also a marker for a subset of dendritic cell precursors as plasmacytoid dendritic cells (P-DCs).30 Increases in these cells have been demonstrated in the peritumoral areas of primary melanomas.31 P-DCs are specialized in the production of type I interferon,32 which may represent an immunologic mechanism to limit melanoma spreading. Therefore, the high density of HLA-DR cells infiltrating the tumor base also appears to be a favorable prognostic factor in view of the role that these cells play as immune-response initiators, as antigen-presenting cells, and as interferon-producing cells.

Several studies have shown that tumor-infiltrating CD68 macrophages contribute to prognosis and are associated with microvascular density.25, 33, 34 The results of the current study with regard to the presence and density of CD68 cells failed to show any correlation with survival, although the involvement of macrophages in response to melanoma cells requires further investigation. The current study results, which were based on small numbers of patients, tend to confirm previous studies but also will require validation in a larger data set for practical use because they may indicate further significant prognostic factors in the evaluation of disease progression and in development of immune therapies for patients with melanoma.


The authors thank Mrs. Maria Itala Mosso and Mr. Massimo Annis for their expert technical assistance.