Detection of submicroscopic lymph node metastases with polymerase ...

ANNALS OF SURGERY Vol. 220, No. 6, 768-774 © 1994 J. B. Lippincott Company

Detection of Submicroscopic Lymph Node Metastases with Polymerase Chain Reaction in Patients with Malignant Melanoma Xiangning Wang, M.D.,* Richard Heller, Ph.D.,* Nancy VanVoorhis, B.A.,* C. Wayne Cruse, M.D.,*.t Frank Glass, M.D.,* t Neil Fenske, M.D.,*`t Claudia Berman, M.D.,* § Jane Leo-Messina, M.D.,*-# David Rappaport, M.D.,*.t Karen Wells, M.D.,* t Ronald DeConti, M.D.,*.§ Lynn Moscinski, M.D.,# Charles Stankard, M.D.,* Chris Puleo, PA-C.,* and Douglas Reintgen, M.D.* II From the Cutaneous Oncology Program, Moffitt Cancer Center, * and the Divisions of Plastic Surgery,t Dermatology, t Medical Oncology,§ and Surgical Oncology, 1 and the Department of Pathology,# University of South Florida, Tampa, Florida

Background The presence or absence of lymph node metastases in patients with malignant melanoma is the most powerful prognostic factor for predicting survival. If regional nodal metastases are found, the 5-year survival for the patient decreases approximately 50%. If the presence or absence of regional nodal metastases will determine which patients receive formal dissections or which patients enter adjuvant trials, then a technique is needed to accurately screen lymph node samples for occult disease. Routine histopathologic examination routinely underestimates the number of patients with metastases. This study was initiated to develop a highly sensitive clinically applicable method to detect micrometastases by examining lymph nodes for the presence of tyrosinase messenger RNA (mRNA). The hypothesis was that if mRNA for tyrosinase is found in the lymph node preparation, that finding is good evidence that metastatic melanoma cells are present.

Methods The assay is accomplished using the combination of reverse transcription and double-round polymerase chain reaction (RT-PCR). The amplified samples are examined on a 2% agarose gel and tyrosinase cDNA is seen as a 207 base pair fragment. Lymph node preparations from 29 patients who were clinically stage I and 11 and undergoing elective node dissections were analyzed both by standard pathologic staining and RT-PCR.

Results Eleven of 29 lymph node (38%) samples from 29 patients with intermediate thickness melanoma were pathologically positive. Nineteen of the 29 lymph node preparations (66%) were RT-PCRpositive, and these included all of the pathologically positive samples, so that the false-negative rate was 0. In a spiking experiment, one SK-Mel-28 melanoma cell in a background of one million normal lymphocytes could be detected, thus indicating the sensitivity of this method. In addition, analysis by restriction enzyme mapping showed that the amplified 207-bp PCR product produced is part of the tyrosinase gene sequence.

768

Occult Melanoma Nodal Metastases

Vol. 220 - No. 6

769

Conclusion The RT-PCR method is an extremely sensitive, reproducible, and efficient technique for the identification of micrometastases in patients with melanoma and could be widely applicable. If clinical correlation is obtained, staging of the melanoma patient becomes more accurate, and treatment becomes more standardized and rational, because all those patients who have evidence of nodal disease can be identified so that they may benefit from more extensive surgery (formal node dissections) or adjuvant therapies. Based on these results, RT-PCR could be a powerful tool to detect micrometastatic melanoma.

The incidence of malignant melanoma has been rising steadily. More than 6000 patients die each year from metastatic melanoma in the United States, and it is estimated that by the year 2000, 1 in 75 people in the United States will get the disease.' Once patients are found to have metastatic disease, the strategies for the treatment are changed and the prognosis is poorer.23 Surgical strategies for melanoma care are undergoing change with the advent of preoperative4 and intraoperative mapping techniques5 that identify the sentinel node of the lymphatic basin. It is hypothesized that if the sentinel node is negative, then the remainder of the nodes in the basin also should be negative, and most melanoma patients can be spared the morbidity and expense of a complete node dissection.4 The goal of improving the staging of melanoma patients in the hope ofdefining prognosis and for the early detection of metastatic disease allows a more rational approach to more extensive surgery (complete node dissections) or enrollment on adjuvant therapies that theoretically, because of smaller tumor volumes, are more likely to be successful. A number of methods often are used in clinical laboratories to detect metastatic melanoma. The standard histopathology interpretation with hematoxylin-eosin staining used routinely for the detection of metastatic tumor cells in tissue has the sensitivity to find 1 abnormal cell in a background of 104 normal cells. However, the rate-limiting factor is the number of sections made and examined. Cutting one or two sections from the center of the node samples about 1 / 1000 of the tissue submitted for pathologic examination. For the pathologically suspicious tissue, further immunohistochemical staining with antibodies against S-100 protein or HMB-45 melanoma antigen can help to confirm the diagnosis.6,7 With immunohistochemical staining, the sensitivity of identifying abnormal cells in a background of normal cells is approximately 1:105. Both serial sectioning8 and the routine use of special immunohistochemical stains9 have been shown to increase the yield of occult lymph node

Address reprint requests to Douglas Reintgen, M.D., Program Leader, Cutaneous Oncology, Moffitt Cancer Center, 12902 Magnolia Drive, Tampa, FL 33612-9497. Accepted for publication May 20, 1994.

disease. However, although these techniques have been available for a number of years, they are not used routinely because, again, the rate-limiting step is the number of sections cut and examined. The time and expense involved with these techniques do not permit their widespread use. We previously have reported the application of a cell culture method that is more sensitive than routine histology in detecting melanoma cells in tissue.'0l" With this sensitive cell culture technique, our laboratory previously has shown that 22% of histologically confirmed patients with stage I and II melanoma can be upstaged to stage III disease (nodal metastases). Most significantly, patients who were histologically node negative but culture node positive subsequently demonstrated a poorer disease-free survival compared with those patients who were node negative by both methods. This assay was criticized because of the delay in 4 to 6 weeks in receiving the results and the potential limited applicability to the community hospital setting. Recently, Smith and colleagues'2 reported using reverse transcription coupled with polymerase chain reaction (RT-PCR) to detect melanoma cells in peripheral blood. The procedure involves using PCR to detect the messenger RNA (mRNA) of tyrosinase. Tyrosinase is a key enzyme during melanin synthesis in melanocytes and melanoma cells.13 '4 Our laboratory has modified this method to detect micrometastases in lymph nodes, the most frequent site of melanoma metastases and the most powerful predictor of survival for the melanoma patient. Because melanocytes normally are not present in peripheral blood or lymph nodes, the detection of transcription of the tyrosinase gene is an indication that there are metastatic melanoma cells present. Because the complementary DNA (cDNA) prepared by the reverse transcriptase only represents the actively expressed genes and this cDNA sequence is exponentially amplified by the PCR, this method is highly sensitive and also very specific. The RT-PCR method for early detection of micrometastatic disease is potentially an efficient and powerful tool.

METHODS Specimens Regional lymph nodes from 29 patients with melanoma who were clinically node negative (an average of 5

770

Wang and Others

lymph nodes/patient) and metastatic or primary specimens from 9 patients with other types of cancer (negative controls) were harvested and cut in half. One half ofeach lymph node was sent for standard pathologic evaluation. This consisted of making one or two sections of the central cross-section of the node and staining with H&E. If atypical but not diagnostic cells were identified, immunohistochemical staining with HMB-45 or S-100 was used to characterize the cell population. The other half of the nodal specimen was soaked with calcium and magnesium-free PBS in a petri dish. Surrounding connective iissue was trimmed. The remaining tissue was disrupted into single cells by gently scraping with sterile scalpels. These cells were processed for total mRNA extraction and RT-PCR. Two established cell lines from American Type Culture Collection (ATCC; Rockville, MD) were used for controls, human melanoma-derived cell line SK-MeL28 (ATCC HTB 72) as a positive control and human breast cancer-derived cell line T47D (ATCC HTB 133) as a negative control. Both were cultured separately in Dulbecco's Modified Eagle's Medium (DMEM; Cellgro [Fisher Scientific], Atlanta, GA) supplemented with 5% fetal calf serum (Hyclone, Inc; Logan, UT) for 2 to 3 weeks, then divided into 1 million cells per tube and stored at -70 C for future extraction of RNA.

RNA Extraction Total RNA was extracted from the cells with RNA STAT-60 kit (TEL-TEST "B," INC., Friendswood, TX). Briefly, 10 X 106 cells were placed in 1.0 mL phenol and guanidinium thiocyanate solution and homogenized with an 18G, 1½/2-inch needle. After adding chloroform, the homogenate separated into two phases. The upper aqueous phase, which contains RNA, was removed and precipitated with isopropanol. The precipitate was washed with 75% ethanol and solubilized in 10 lL TE buffer (I OmM Tris, 1mM EDTA, pH 8.0).

Primers Two sets of primer sequences for human tyrosinase cDNA and one set of primer sequences for human ,Bactin cDNA were obtained from previous publications or sequence available through Genebank (Table 1 ).8 1' A 284-bp fragment and a 207-bp cDNA fragment were obtained by the extension of outer and nested primers for the tyrosinase gene, respectively. A 146-bp cDNA fragment was produced by the extension of #-actin

primers.

Ann. Surg. * Decernber 1994

Table 1. TYROSINASE-PCR OF NONMELANOMA PATIENT SAMPLES Tissue Source

No.

RT-PCR

B-Actin

Breast primary Metastatic breast Primary colon Metastatic colon

3 2 2 2

-

+ + + +

Reverse Transcription For reverse transcription of tyrosinase mRNA, 1.0 ,uL total RNA extracted from the cells was dispensed in 9.0 ,uL reaction mixture containing PCR buffer (10 mM Tris-HCl, 50 mM KCI, pH 8.3),5.0 mM MgCl2, 1.0 mM each of dNTP (dATP, dCTP, dGTP, dTTP), 1 unit RNase inhibitor, 5 units Moloney Murine Leukemia Virus (M-MLV) reverse transcriptase (Perkin-Elmer-Cetus Corp., Norwalk, CT), and 75 pMol antisense outer primer. The sample is overlaid with two drops of mineral oil, to prevent evaporation, and then the reaction tube was placed into a programmable thermal controller (MJ Research, Inc., Watertown, MA) and incubated at 42 C for 45 minutes, denatured at 99 C for 5 minutes, and then immediately moved to an ice bucket. The identical reaction mixture and incubation program were used to reverse transcribe fl-actin mRNA, except that the primer used was fl-actin downstream primer (BAD).

Polymerase Chain Reaction The following PCR 1 mixture (20,uL) for the extension of outer primers to obtain a 284 bp cDNA fragment was prepared immediately before use and directly added into each reverse transcription reaction tube: 1 X PCR buffer (10 mM Tris-HCl, 50 mM KCI, pH 8.3), 0.625 mM MgCl2, 16.5 qL distilled water, 2.5 units AmpliTaq DNA polymerase (Perkin-Elmer-Cetus Corp or Promega [Fisher Scientific]), and 75 pMol sense outer primer. Heating at 95 C for 3 minutes was followed by a two-step PCR for 30 cycles (95 C for 1 minutes, 62 C for 1 minute). Reaction was concluded at 72 C for 5 minutes to complete the extension. Beta-actin cDNA fragment was amplified with the identical method. However, the primers for tyrosinase gene were replaced by /-actin gene primers. For re-amplification of PCR 1 product with nested primers to obtain 207 bp cDNA fragment, 5 jL of 1:50 diluted PCR 1 product was amplified in a final 25 ,tL reaction mixture containing 1 X PCR buffer (10 mM TrisHC1, 50 mM KCI, pH 8.3),1.5 mM MgC12,0.8 mM each of dNTP, 2.5 units AmpliTaq polymerase, and 150 pMol


Vol. 220 - No. 6

each of nested primers. Two drops of mineral oil also was overlaid. The thermal cycle program was identical to PCR 1. Double PCR without RT was carried out to ensure that the amplified sequence was from mRNA and not from genomic DNA. Reaction buffer for double PCR without RT was the combination of RT and PCR 1 reaction mixture, except the RNase inhibitor and reverse transcriptase were substituted by equal volumes of distilled water. The amplification program was the same. To minimize contamination, all the reaction mixtures were prepared in a fume hood. Powder-free gloves and aerosol-resistant tips were used.

PCR Product Analysis The mixture of 9 ,uL PCR product and 1 ,uL gel loading solution (Sigma, St. Louis, MO) was loaded on 2% agarose gel containing ethidium bromide. A 100-bp ladder (Gibco BRL, Inc., Grand Island, NY) served as molecular weight marker. A photograph was taken with Polaroid 667 film by ultraviolet transillumination after electrophoresis.

Restriction Enzyme Mapping Three different restriction enzymes (DdeI, Hinfl, and PvuII) were used to digest RT-PCR2 products. Twenty units each of the enzymes were incubated at 37 C for 2 hours to digest 10 IuL RT-PCR2 products in 20 ,uL reaction mixture. After digestion, resulting fragments were analyzed on a 2% agarose gel, as described for PCR products, and compared with the cut points anticipated for the described tyrosinase gene sequence.

Immunoperoxidase Staining For the eight patients who were histologically node negative but PCR positive, the original blocks were obtained from the lymph node examination, serial sectioned, and stained with immunoperoxidase staining. Standard immunoperoxidase staining techniques with antibodies for S-100 protein and HMB-45 (Biogene, CA) were performed using aminoetheylcarbazole (AEC) chromagen as a developer in an attempt to find the abnormal melanoma cells.

RESULTS Lymph nodes from 29 patients with intermediate thickness melanoma (Clinical stage I and II) were examined by both standard histologic methods and RT-PCR. Routine histopathology detected melanoma cells in lymph nodes from 11 of the 29 patients (38%). A 207-bp

1

2

3

4

5

6

7

8

9

771

10

Figure 1. Tyrosinase RT-PCR2 results from some of our patient samples and cell lines. Lane 1, blank. Lane 2, negative control from breast cancer cell line T47D. Lanes 3 and 4, samples from node negative melanoma patient. Lane 5, sample from primary colon cancer. Lane 6, sample from primary breast cancer. Lanes 7 to 9, samples from lymph nodes of patients with melanoma. Lane 10, positive control from melanoma cell line SK-Mel-28.

tyrosinase gene sequence could be detected (Fig. 1) in lymph nodes of 19 of the 29 patients (66%), which included all 11 pathologically positive samples. Of the 18 patients found to have negative lymph nodes by routine pathology, 8 (44%) were positive by RT-PCR and were upstaged to stage III disease. Ten patients (34%) were found to have lymph nodes negative for melanoma cells by both routine pathology and RT-PCR. No specimens were found to be pathology positive and RT-PCR negative and thus, there were no false-negatives with the RTPCR. In the eight patients who were histologically node negative but RT-PCR positive, no tumor could be identified within nodal tissue by immunoperoxidase techniques that was not discovered by routine histology. Numerous nodal dendritic cells were labeled by S- 100 protein antibody in each case, but the cells did not have the cytologic features of malignant cells, and the sections were negative with the HMB-45 antibody. The detection procedure uses a reverse transcription step followed by two rounds of PCR amplification. The first round of PCR yields a 284-bp DNA fragment. In some samples, the 284-bp fragment was visible on 2% agarose gel after the first round PCR, but the density of the band was very low (Fig. 2). Therefore, to be able to detect low numbers of melanoma cells in the samples, double amplification was necessary. The sensitivity of the assay is increased greatly by using nested primers and double amplification. Eight pathologically diagnosed metastatic tumors from patients with melanoma and nine specimens from patients with other cancers (metastatic disease in lymph

772

Wang and Others

Ann. Surg. December 1994 -

m1 .:. .:

:...::. ....... .........

:..

1 2 :.::

!..::

.:

3 .:.

4 5

6 7 8

..

.1

.::

; .. ....

284

146-. Figure 2. The first round of PCR yielded a 284-bp band that was evident on some of the gels from patients' lymph node samples, but with a low density. Lane designation the same as Figure 1.

nodes and primary tumors) also were examined by the same method. Tyrosinase gene expression was not detected in any of the nonmelanoma lymph nodes and primary tumors (Table 1). However, the mRNA for the gene could be detected in all the metastatic melanoma specimens.

All samples with detectable tyrosinase gene activity were examined by double PCR without RT. None of these were positive, indicating that the samples did not contain genomic DNA. Beta-actin mRNA was reverse transcribed and amplified from all the tyrosinase gene negative samples. This indicated that the RNA in all the samples was not degraded (Fig. 3). Additional experiments were performed to determine the sensitivity of this method. Decreasing numbers of SK-Mel-28 melanoma cells were added to 10 X 106 lymphocytes isolated from a histologically normal node from a patient with no history of melanoma. The RTPCR method was able to detect approximately 3 melanoma cell in I07 cells (Fig. 4). Mapping with restriction enzymes was used to determine whether the PCR product was obtained from tyrosinase mRNA. The final PCR product was treated with either DdeI, Hinfl, or PvuII. The sequence of tyrosinase mRNA is known, and the size of fragments obtained with these enzymes can be predicted. Figure 5 shows the results of this treatment and illustrates that the cut fragments of the three patient samples are identical and the same as the base pair length anticipated for the tyrosinase gene. The restriction enzyme mapping identified that the RT-PCR2 products are from tyrosinase mRNA.

DISCUSSION The 5-year survival for patients with stage III melaregional nodal involvement is decreased 50%

noma with

Figure 3. Beta-actin RT-PCR1 results from tyrosinase RT-PCR2 negative samples. Molecular weight marker, 1 00-bp molecular ladder. Lane 1, blank (no RNA added). Lane 2, breast cancer cell line T47D. Lanes 3 to 7, samples from tyrosinase RT-PCR2 negative patients.

compared with the survival of node-negative patients. 15"16 Adjuvant therapies have been attempted to increase the survival of the patients with stage III melanoma who have no evidence of disease but are likely to have recurrences. Both the ECOG Interferon Trial'7 and various vaccines trials'8 identify this "high risk for recur-

207

Figure 4. Spiking experiment results. Different numbers of SK-Mel-28 melanoma cells were added to 1 0 million normal lymphocytes. Cell ratios: Lane 1, blank. Lane 2, 10 million normal lymphocytes only. Lane 3, 1/10 X 106 . Lane 4, 2/1 0 X 106 . Lane 5, 3/1 0 X 106 . Lane 6, 4/1 0 X 106.- Lane 7, 5/10 x 106. Lane 8, 10/10 X 106. Lane 9, positive control SK-Mel-28 melanoma cells.


Vol. 220 - No. 6

1

2

3

4

5

6 7

8

9 M

Figure 5. Restriction enzyme mapping. Reverse transcription-PCR2 products from 3 different patient samples were digested by 3 different restriction enzymes. Lanes 1, 4, and 7 show the bands of PCR products digested by Ddel. Lanes 2, 5, and 8 show the bands of PCR products digested by Hinfl. Lanes 3, 6, and 9 show the bands of PCR products digested by Pvull. Molecular marker is a 10-bp ladder. Lanes 1 to 3 are PCR products from one patient, cut with the different restriction enzymes. Similarly, Lanes 4 to 6 and Lanes 7 to 9 are from two different patients, cut with the three different enzymes.

rence" group as resected stage III disease, and if preliminary results hold, adjuvant therapy for melanoma may become a reality. Theoretically, the accurate and early

detection of metastatic disease would allow those stage III patients to be enrolled in the adjuvant trials when systemic tumor burden is small and the likelihood of success is better. Routine histologic examination and immunohistologic staining for the evaluation of metastatic melanoma currently are available in many hospitals. The sensitivity of H&E is reported to be approximately 1 abnormal cell identified in a background of 10,000 cells. Serial sectioning and immunohistochemistry have increased this sensitivity; however, these techniques are expensive and time consuming' -2' and have not been incorporated into the routine screening of lymph node sections in the community hospital. Although the techniques have been available for a number of years and have been reported to increase the yield ofoccult metastases,8,9 they have not been used routinely. The rate-limiting factor for both these techniques is the number of sections of the node made and examined. Because of the time and expense involved, the assays will never be widely adopted. An additional problem with these methods is that they are based on finding changes in morphologic features or expression of specific proteins, processes that are not always accurate and can lead to ambiguous results. Falsenegative results may be obtained because of very few tumor cells in the pathologic samples. We previously reported a cell culture method for de-

773

tecting micrometastases. Twenty-two percent of the patients with stage I and II melanoma (by routine pathologic examination) had their stage of disease increased by culturing out micrometastatic disease in the lymph nodes.'O "' Much like the RT-PCR assay, the entire node is sampled by placement in tissue culture, much different than routine histologic examination that samples at most 1% of the submitted tissue. Although this test offered a more sensitive method than routine histologic examination, on average it takes 4 to 6 weeks to obtain definitive results. There also was the possibility of other cells, such as Langerhans cells, being present to interfere with the characterization of the cell growth and the overall difficulty associated with primary tissue culture. Furthermore, the tissue culture technique is fairly laborious and time consuming. In this report, we described a RT-PCR assay that can detect 3 malignant melanoma cells in a background of 10 million normal lymphocytes, an increase of 1 log in sensitivity. Using this assay, we were able to detect tyrosinase mRNA in all the pathology-positive specimens tested. There were eight pathology-negative specimens that were found to be positive by the RT-PCR method. The paraffin blocks from these eight nodal specimens were serial sectioned and stained with S- 100 protein and HMB-45 immunohistochemical reagents, in the hope of finding the malignant cells with standard methodology. Although all lymph node samples tested had cells that were positive with the S- 100 antibody, none of the cells were correspondingly positive with the HMB-45 stain, and they did not have the cytologic characteristics of malignancy. Even with the expanded technique ofserial sectioning and immunohistochemistry staining, the malignant cells could not be identified. In addition, samples from ten patients were found to be negative by both routine pathology and RT-PCR. By this analysis, these samples are considered true-negatives. This also illustrates that the RT-PCR method is not falsely positive because of the lymph node composition. The specificity of the assay was determined by examining samples from nonmelanoma primary and metastatic tumors and lymph nodes from patients who do not have melanoma. No tyrosinase mRNA was detected in those samples, and the specificity ofthis assay was high as well. To further evaluate the findings, each sample was repeated 2 to 3 times, with similar results. The results were consistent and reproducible. The sensitivity of the assay was determined by double PCR, using nested primers. When evaluating the sample after a single round of PCR amplification, there were no visible bands or very low density bands. However, after double amplification with the nested primers, detection of tyrosinase was possible on 2% agarose gel. During the assay, all precautions must be taken to pre-

774

Wang and Others

vent contamination of RNA samples with genomic DNA. The presence of genomic DNA would lead to false-positive results. The procedures employed during this study were performed with this in mind, and every precaution was taken to ensure the integrity and purity of the RNA samples. To test for potential contaminating DNA, all positive samples were retested without the reverse transcription step. Ribonucleic acid cannot be amplified without first converting to cDNA by reverse transcription. Therefore, if the sample contained only RNA, then performing PCR without the RT step would result in negative results. If DNA was present, the results would be positive, even without the RT step. By adding this step, the possibility of false-postitive results was greatly decreased. Reverse transcription-PCR is a highly sensitive and specific method for the detection of occult metastases. It is economical and relatively simple. Many samples can be tested at the same time, and the entire procedure can be completed in a couple of days. It could be of clinical value for screening metastasis of melanoma not only in lymph nodes, but also at other sites, such as bone marrow and peripheral blood. The RT-PCR assay also can be extended to the detection of other tumors, e.g., breast and colon cancer when a cancer cell-specific mRNA is identified. Further studies, will include clinical follow-up of the patients included in this study. With limited follow-up of a mean of 1 year, one of the eight patients who were histologically node negative but PCR positive has had a recurrence. The rate of recurrence in this entire subgroup should determine if there is a clinical correlation of these findings. If clinical correlation can be proven and effective adjuvant therapies are found, a more sensitive method for the identification of occult metastases will do more than just stage shift25; it will identify a subgroup of the melanoma population who have the most to benefit from the effective adjuvant therapies.

References 1. Balch CM. Changing trends in melanoma. Melanoma Letter 1987; 5:1. 2. Ryan L, Kramar A, Borden E. Prognostic factors in metastatic melanoma. Cancer 1993; 71:2995-3005. 3. Drepper H, BieB B, Hofherr B, et al. The prognosis ofpatients with stage 3 melanoma. Cancer 1993; 71:1239-1246. 4. Norman J, Cruse CW, Wells K, et al. A redefinition of skin lymphatic drainage by lymphoscintigraphy for malignant melanoma. Am J Surg 1991; 162:432-437. 5. Morton DL, Wen DR, Wong J, et al. Technical details of intraoperative lymphatic mapping for early stge melanoma. Arch Surg

1992; 127:392-399.

Ann. Surg. * December 1994

6. Cho KH, Hashimoto K, Taniguchi Y, et al. Immunohistochemical study of melanocytic nevus and malignant melanoma with monoclonal antibodies against S-100 subunits. Cancer 1990; 66:765771. 7. Walts AE, Said JW, Shintaku IP. Cytodiagnosis of malignant melanoma: immunoperoxidase staining with HMB-45 antibody as an aid to diagnosis. Am J Clin Pathol 1988; 90:77-80. 8. International Ludwig Breast Cancer Study Group. Prognostic importance of occult axillary lymph node micrometastases from breast cancer. Lancet 1990; 335:1565-1568. 9. Cochran AJ, Wen DR, Herschman JR. Occult melanoma in lymph nodes detected by antiserum to S-100 proteins. Int J Cancer 1984; 34:159-163. 10. Heller R, Becker J, Wasselle J, et al. Detection of submicroscopic lymph node metastases in patients with melanoma. Arch Surg 1991; 126:1455-1460. 11. Heller R, King B, Baekey P, et al. Identification of submicroscopic lymph node metastases in patients with malignant melanoma. Semin Surg Oncol; 9:285-289. 12. Smith B, Selby P, Southgate J, et al. Detection of melanoma cells in peripheral blood by means of reverse transcriptase and polymerase chain reaction. Lancet 1991; 338:1227-1229. 13. Prota G. Recent advances in the chemistry of melanogenesis in mammals. J Invest Dermatol 1980; 75:122-127. 14. Slominski A, Costantino R. Molecular mechanism of tyrosinase regulation by L-Dopa in hamster melanoma cells. Life Sci 1991; 48:2075-2079. 15. Kwon BS, Haq AK, Pomerantz SH, Halaban R. Isolation and sequence of a cDNA clone for human tyrosinase that maps at the mouse c-albino locus. Proc Natl Acad Sci U S A 1987; 84:74737477. 16. Reintgen DS, Cox EB, McCarty KS, Seigler HF. Efficacy ofelective node dissections in patients with intermediate thickness melanoma. Ann Surg 1983; 198:379-385. 17. Johnson C, Kirkwood J. How does interferon work. ECOG Horizons 1993; 3:1-15. 18. Wallach MK, Bash J, Leftheriortis E, et al. Positive relationship of clionical and serologic responses to vaccinia melanoma oncolysate. Arch Surg 1987; 122:1460-1463. 19. Fisher ER, Swamidoss S, Lee CH, et al. Detection and significance of occult axillary node metastases in patients with invasive breast cancer. Cancer 1978; 42:2025-2031. 20. De Mascarel I, Trojani M, Abadjian G, et al. Ganglions axillaires dans les cancers du sein: comparaison des techniques d'analyse histologrique standard et en coupes macroscopiquement seriees. Bull Cancer (Paris) 1982; 69:451-455. 21. Cochran AJ, Wen D-R, Morton DL. Occult tumor cells in the lymph nodes of patients with pathological stage I malignant melanoma. Am J Surg Pathol 1988; 12:612-618. 22. Trojani M, de Mascarel I, Bonichon F, et al. Micrometastases to axillary lymph nodes from carcinoma of breast: detection by immunohistochemistry and prognostic significance. Br J Cancer 1987; 55:303-306. 23. Wells CA, Heryet A, Brochier J, et al. The immunocytochemical detection ofaxillary micrometastases in breast cancer. Br J Cancer 1984; 50:193-197. 24. Redding WH, Coombs RC, Monaghan P, et al. Detection of micrometastases in patients with primary breast cancer. Lancet 1983; 2:1271-1273. 25. Feinstein AR, Sosin DM, Wells CK. The Will Rogers phenomenon: stage migration and new diagnostic techniques as a source of misleading statistics for survival in cancer. N Engi J Med 1985; 312: 1604-1608.