PATIENTS AND METHODS

Established JWCI melanoma cell lines were used as positive controls for the RT-PCR assay.^14,17 Surgical specimens were collected and dissected under stringent “nucleic acid–free” conditions to prevent RNA contamination. SLND was performed in clinical early-stage melanoma patients, as previously described.⁵ Clinical early-stage melanoma patients are defined as those patients with no clinical evidence of regional or metastatic disease. After the patients provided informed consent, lymph nodes obtained from patients undergoing noncancer surgery (cholecystectomy and tonsillectomy) were used as negative controls in the RT-PCR assay.¹⁸ Ten milliliters of blood was collected in sodium citrate–containing tubes from healthy donor volunteers as previously described, and the samples were used as negative controls.¹⁴ The blood was centrifuged using a density gradient solution, and nucleated cells in the blood were collected for RNA isolation, as previously described.¹⁷

Histopathologic examination and MM RT-PCR analysis were performed on each SN as shown in Fig 1. Each SN was bisected, and an 8-μm imprint slide was then prepared from the tissue surface and stained with Diff-Quik I & II (Dade International, Miami, FL) for the pathologist's intraoperative diagnosis. If melanoma cells were identified in the frozen section, a complete lymph node dissection was then performed. Six immediately adjacent frozen sections were cut on the cryostat to a thickness of 12 μm each and stored at −80°C until they were assessed for melanoma cells by RT-PCR. For RT-PCR analysis, we also determined the necessary amount of mRNA that could consistently be isolated from a minimum number of frozen sections of SN. An additional 8-μm frozen section of tissue was examined by Diff-Quik staining. The remainder of the bisected node was then placed in 10% formalin and embedded in paraffin, and 4-μm–thick sections were examined with H&E staining. If no melanoma cells were identified by H&E staining, an adjacent 4-μm–thick section was evaluated by use of antibodies to HMB-45 and S-100 proteins. This paraffin-section evaluation was performed at two different levels separated by approximately 40 μm. In a retrospective analysis, all SNs that were initially histopathologically melanoma-free were reassessed for occult metastases by taking step-sections at two additional levels, separated by 40 μm, and examining them with H&E staining and IHC. The frequency of capsular nevi and nerve cells in histopathologically melanoma-free SNs was determined because these cells may be potential sources of tyrosinase mRNA expression. Lymph nodes obtained from non–cancer patients were confirmed to be cancer-free by H&E staining, with the remainder of the node used for RT-PCR analysis.

TRI-REAGENT (Molecular Research Center, Cincinnati, OH) was used to isolate total RNA from cell lines, primary melanoma biopsy specimens, frozen-section SNs, normal donor blood, and lymph nodes from non–cancer patients as previously described.^17,18 Pellet Paint Co-Precipitant (Novagen, Madison, WI) was used as a carrier for RNA precipitation in the SN frozen sections. The resulting RNA pellet can easily be visualized by its pink color, and, in addition, Pellet Paint does not interfere with the RT-PCR assay. The integrity of all RNA samples was verified by performing RT-PCR and ethidium bromide gel electrophoresis for porphobilinogen deaminase mRNA expression (housekeeping gene). Tissue processing, RNA extraction, RT-PCR assay set-up, and post–PCR-product analyses were performed in separate designated rooms and buildings to prevent cross-contamination, as previously reported.¹⁷

Optimal RT-PCR conditions were carried out as previously described.¹⁷ All RT reactions were carried out with oligo-dT priming, using 1 μg of RNA. The PCR conditions and primer sequences for MAGE-3, MART-1, and tyrosinase were as previously described; optimal annealing temperatures were 55°C, 60°C, and 55°C, respectively.^17,19 Primer sequences designed for each marker spanned at least 1 intron region. PCR reactions were set up in an Omni thermocycler (Hybaid, Middlesex, United Kingdom). The RT-PCR cDNA products of MAGE-3, MART-1, and tyrosinase were 423, 252, and 284 base pairs, respectively. Marker-specific cDNA probes were used for Southern blot analyses of RT-PCR products.¹⁷ Each experiment set-up included respective positive and negative controls for each procedure. If known positive or negative control samples did not yield the expected results, the assay was considered invalid and was repeated for verification. A positive result in the RT-PCR assay is considered to be the expression of at least two mRNA markers, as previously described.¹⁴

The χ² test (or Student's t test for continuous variables) and logistic regression model, using stepwise procedure, were used to assess the correlation of clinicopathologic risk factors and mRNA expression with the histopathology of the SNs. Multivariate analysis, using the Cox proportional hazards regression model, was also performed to examine the association of clinicopathologic risk factors and number of mRNA markers expressed with disease recurrence. Disease-free survival curves were constructed using the Kaplan-Meier method, and the log-rank test was used to test the equality of the curves. P values less than .05 were referred to as significant.²⁰ All statistical analyses were based on the results from the initial analysis of the SN by H&E staining and/or IHC, and RT-PCR.

RESULTS

Previously, we demonstrated the specificity and frequency of the melanoma RT-PCR markers in cell lines and melanoma biopsy specimens.^17,19 The optimal marker combination for the assay was chosen after extensive analysis of several markers. The marker combination was chosen on the basis of specificity, sensitivity, and potential clinicopathologic utility. In the ten primary melanomas assessed, four (40%) expressed MAGE-3, eight (80%) expressed tyrosinase, and ten (100%) expressed MART-1. None of the three mRNA markers were expressed under the assay's optimal conditions in normal lymph nodes (n = 10) or blood (n = 25) obtained from healthy volunteer donors.

Ninety-three SNs were obtained from 72 clinical early-stage melanoma patients who underwent SLND. All of the samples assessed were positive for porphobilinogen deaminase. No single mRNA marker was expressed in all patients with histopathologically proven melanoma cells in the SN, as previously reported from assessments of melanoma biopsy specimens (primary and metastatic).^14,17 Of the 17 patients with histopathologically proven melanoma cells (14 detected by H&E staining, three by IHC alone) in the SN, MAGE-3, MART-1, and tyrosinase were expressed in the SNs of 12 (71%), 15 (88%), and 12 (71%) patients, respectively. Of the 55 patients with histopathologically proven melanoma-free SNs, 24 (44%) expressed MAGE-3, 20 (36%) expressed MART-1, and 16 (29%) expressed tyrosinase. At least two mRNA markers were expressed in 16 (94%) of 17 patients with histopathologically proven metastases and in 20 (36%) of 55 patients with histopathologically proven melanoma-free SNs.

A retrospective analysis of all histopathologically proven melanoma-free SNs (n = 55) was performed by serially resectioning the paraffin-embedded SNs at two different levels and performing H&E staining and IHC analyses. In three (5%) of 55 patients, additional occult melanoma cells were identified; all three cases had SNs that expressed two or more mRNA markers. Therefore, 10% of patients (six of 58) had metastases detected in the SN by IHC alone (three patients in the initial analysis, three patients in the retrospective analysis). Similarly, the single histopathologically positive but RT-PCR–negative node was retested by RT-PCR and shown to remain negative. Capsular nevi, which stained both HMB-45– and S-100–positive, were identified in eight (15%) of 55 patients. Tyrosinase was expressed in only two of these patients. Nerve cells that stained S-100–positive but HMB-45–negative were identified in all SNs.

By univariate analysis, patients with thicker lesions (P < .04), 60 years of age or younger (P = .04), and SNs that expressed MMs (P = .001) were more likely to have histopathologically proven metastases in the SN (Table 1). The mean Breslow thickness in patients with metastases to the SN was 2.1 ± 1.0 mm, compared with 1.5 ± 1.2 mm in patients without metastases. In patients with metastases to the SN, 82% (14 of 17) were 60 years of age or younger, compared with 18% (three of 17) who were older than 60 years of age. MMs were expressed more frequently when SNs contained metastases (94%) than when SNs were melanoma-free (36%). A logistic regression model using a stepwise procedure correlated risk factors (Table 1) with SN metastases. By multivariate analysis, patients at increased risk of metastases to the SN had thicker lesions (P = .03), were 60 years of age or younger (P < .05), and/or expressed MMs (P < .001).

In a mean follow-up period of 12 months, there have been eight recurrences (11%), which include two deaths (3%). In the 35 patients with SNs that were both histopathologically melanoma-free and one or fewer mRNA markers expressed, there were no recurrences. In the 20 patients with SNs that were histopathologically melanoma-free but which expressed two or more mRNA markers, 15% (three of 20) had recurrence of disease. In the 16 patients with histopathologically proven metastases and two or more mRNA markers expressed in the SN, 31% (five of 16) have relapsed. Four of these patients had metastases detected in the SN by H&E staining, and one patient had metastases detected by IHC alone. The patient who had histopathologically proven metastases to the SN but no mRNA marker expression has not relapsed. In the two patients who died of metastatic melanoma, MMs were expressed in the SNs, one of which was histopathologically melanoma-free.

By univariate analysis, MM expression in the SN was a strong predictor of recurrence of disease (Table 2). Multivariate analysis, using a Cox proportional hazards regression model, was also performed to examine the association of the risk factors listed in Table 2 with disease-free survival. Only the number of mRNA markers expressed in the SN was selected by the stepwise procedure as a significant predictor for recurrence of disease (P = .02). The Kaplan-Meier curve in Fig 2 demonstrates that patients whose SNs expressed MMs had a shorter disease-free survival, compared with those whose SNs expressed one or fewer mRNA markers. Patients with histopathologically melanoma-free SNs that expressed MMs, compared with those whose SNs expressed one or fewer mRNA markers, were at increased risk of disease relapse (Fig 3). The patients who were at greatest risk of early disease relapse were those with histopathologically proven melanoma metastases to the SN and MM expression. When patients with histopathologically positive or negative SNs that expressed multiple markers were compared, there was no difference in disease-free relapse (P = .48).

DISCUSSION

Metastasis to the regional lymph nodes is one of the most important prognostic factors in patients with cutaneous melanoma.¹ The 5-year survival rate for localized melanoma (AJCC stages I and II) is approximately 80% but decreases to 40% to 50% with microscopic lymph node involvement (AJCC stage III).^1,2 Accurate staging of the lymphatic basin(s) draining the primary melanoma has become increasingly important because adjuvant therapy, which has been shown to improve survival, is now available.^3,4,21,22 The potential therapeutic benefit of adjuvant therapy for melanoma patients with lymph node metastases stresses the importance of accurate staging of the lymphatic basin.

A more extensive and sensitive analysis of lymph nodes in the lymphatic basin by serial sectioning, IHC, and now RT-PCR can improve the detection of occult tumor cells, compared with conventional sectioning using H&E staining alone.^5-7,23 However, this intense investigation for metastases is not always pathologically feasible when these techniques are performed on all lymph nodes removed by elective lymph node dissection.²⁴ Analysis of the SN, which accurately predicts the histopathology of the remaining lymphatic basin in 99% of cases, resolves the problem of assessing the entire lymphatic basin.^15,16 When SLND is performed at experienced institutions, the SN can be accurately identified in 98% or more of cases.^15,16,25-27 SLND combined with MM RT-PCR analysis provides a powerful tool for staging the draining lymphatic basin in melanoma patients.

In this study, the high specificity of the MM RT-PCR assay was first demonstrated by the fact that none of the mRNA markers were expressed in noncancerous lymph nodes or blood obtained from normal healthy donors. After resectioning all SNs that were initially histopathologically melanoma-free, we identified capsular nevi, a potential source of tyrosinase mRNA expression, in 15% (eight of 55) of patients. This incidence of capsular nevi falls within the 4% to 22% range reported by other institutions.²⁸ However, tyrosinase mRNA was detected in only two of these patients, one of whom subsequently died. This suggests that not all capsular nevi have detectable tyrosinase mRNA by RT-PCR and that the presence of capsular nevi does not exclude the possibility micrometastatic disease. Capsular nevi have a low proliferative/metabolic rate, and, therefore, tyrosinase mRNA may go undetected by RT-PCR because it is being expressed at such a low level or not at all. Nerve cells have also been reported as a source of tyrosinase expression.^19,28 On resectioning the histopathologically melanoma-free SNs, we identified nerve cells in 100% of the cases. However, only 28% of these SNs expressed tyrosinase; thus it is highly unlikely that nerve cells represent the source of tyrosinase mRNA detected in SNs. These findings demonstrate the high specificity of the MM RT-PCR assay under the optimal established conditions.

To determine the sensitivity of the MM RT-PCR assay in detecting occult melanoma cells, expression in the SN must be compared with conventional histopathology, which is the current gold standard for diagnosis. No single mRNA marker was expressed in all SNs with histopathologically proven metastases. Single-marker RT-PCR assays with 100% sensitivity may be compromised by a low specificity.^7-10 Because melanoma is heterogeneous in tumor gene expression, it is not surprising that a single-marker assay is unable to detect 100% of occult melanoma cells. In this study, we demonstrated that the use of MMs improves the sensitivity of the assay in detecting melanoma cells. Additionally, the level of mRNA expression of a specific gene by melanoma cells diluted among the mRNA of normal cells may be too low for detection, even when amplified by RT-PCR. A sampling error may also occur when the lymph node is assessed for histopathologic and RT-PCR analysis. The sensitivity in detecting micrometastases even with the combined sensitive techniques of IHC and RT-PCR can still be limited by the number of serial sections assessed, as was demonstrated in this study. To minimize this sampling error, each SN was bisected and serially sectioned for both histopathologic and molecular analysis as shown in Fig 1. Bisecting and serially sectioning the SN is logistically more rational, compared with “arbitrarily” sampling a piece of lymph node for molecular and histopathologic analysis. The likelihood of 100% concordance in comparing histopathology to RT-PCR, using the technique of “arbitrarily” sampling half of the SN for each type of assay, is very low. This error would be further compounded when using a single-marker assay, whose frequency of expression is less than 100% in melanoma biopsies. From our experience at JWCI, the majority of SNs from early-stage melanoma patients that are positive for melanoma will have only small micrometastases or occult melanoma cells. By combining the MM RT-PCR assay with the techniques of bisecting and serially sectioning the SN, the sensitivity of this assay in detecting histopathologically proven metastases in the SN was 94%.

The clinicopathologic importance of melanoma cells detected in the SN by RT-PCR has not been clearly established to date. However, Shivers et al,⁷ recently demonstrated a significantly decreased disease-free survival and overall survival in patients with histopathologically negative SNs that expressed mRNA tyrosinase, compared with those without tyrosinase expression. In our study, we demonstrated that patients whose SNs expressed MMs in comparison to one or fewer mRNA markers were more likely to have metastases to the SN (Table 1). In addition, patients with histopathologically melanoma-free SNs with MM expression were at greater risk of disease relapse, compared with those expressing one or fewer mRNA markers. By multivariate analysis, MM expression in the SN was also the strongest predictor of disease relapse, even when compared with the histopathology of the SN (P = .02). These findings indicate that MM expression in the SN correlates with a poor prognosis. Therefore, the 36% (20 of 55) of patients in this study with histopathologically melanoma-free SNs are at increased risk of disease relapse. It is this subgroup of patients who may benefit from aggressive surveillance for disease recurrence and stratification into more appropriate adjuvant therapy. This should ultimately improve survival of this patient subpopulation. In addition, the 64% (35 of 55) of patients with histopathologically proven melanoma-free SNs that expressed one or fewer mRNA markers represent a low-risk group for disease relapse who may be spared from the side effects of postadjuvant therapy.

In summary, H&E staining and IHC underestimate the true incidence of melanoma metastases. The MM RT-PCR assay combined with SLND and serial sectioning of the bisected SN is a more rigorous and highly accurate approach leading to ultrastaging of the tumor-draining lymphatic basin of melanoma patients. Our early findings indicate that MM expression in the SN not only more accurately reflects melanoma micrometastases, compared with H&E and IHC, but is also a more powerful predictor of disease relapse.