Detection of Lymphatic Invasion in Primary Melanoma With Monoclonal Antibody D2-40: A New Selective Immunohistochemical Marker of Lymphatic Endothelium

Objectives  To identify the presence of lymphatic invasion in primary cutaneous melanoma using monoclonal antibody D2-40, a marker of lymphatic endothelium, and to correlate the presence of lymphatic invasion with other clinicopathologic characteristics of the tumors.

Design  Retrospective melanoma case series study comparing conventional hematoxylin-eosin staining with D2-40 immunostaining for detection of lymphatic invasion.

Setting  Departments of Pathology and Dermatology, Sunnybrook and Women’s College Health Sciences Center, University of Toronto, Toronto, Ontario.

Patients  Forty-four consecutive cases of primary cutaneous melanoma with a tumor thickness greater than 0.75 mm were examined for presence of lymphatic invasion.

Results  Seven (16%) of 44 melanomas showed the presence of lymphatic invasion under immunostaining with D2-40. In 2 cases, subepidermal lymphatic involvement was present; in 5 cases lymphatic invasion was noted within the tumor, including 1 case of additional lymphatic invasion at the invasive edge of the tumor. Lymphatic invasion was not detected on routine hematoxylin-eosin staining. We observed a trend in the association between lymphatic invasion and 2 markers of tumor aggressiveness, namely, a deeper Clark level and increased frequency of ulceration, which suggests that lymphatic invasion detected with D2-40 may indicate a poor prognosis.

Conclusions  Immunostaining with D2-40 increases the frequency of detection of lymphatic invasion relative to conventional hematoxylin-eosin staining in primary melanoma. Future outcome data will determine the prognostic significance of lymphatic invasion detected by D2-40 immunostaining.

The incidence of malignant melanoma is increasing worldwide.¹ In the United States, this increase is among the highest of all cancers.² Several clinicopathologic characteristics of primary melanoma that predict metastasis and survival have been identified.^3-5 Of those, Breslow tumor thickness is considered the most important prognostic indicator, followed by ulceration⁶; these 2 parameters are used in the staging of primary cutaneous melanoma.^3,4 Several studies suggest that lymphovascular invasion may also be of prognostic significance.^6-9 In 1 study, lymphovascular invasion and ulceration were found to be virtually identical as predictors of relapse and death from primary cutaneous melanoma.⁹ Although lymphovascular invasion is not used presently in the staging of primary cutaneous melanoma,^3,4 it has been suggested that it should be included in any written pathology report because of its potential prognostic importance in localized melanomas.⁴

Lymphovascular invasion in primary melanoma is currently identified by conventional hematoxylin-eosin (H&E) staining as the presence of tumor emboli within vascular channels lined by endothelial cells.^7-9 This technique has several limitations. First, it does not distinguish between lymphatic and blood vessel invasion. Second, flattened spindle cells are frequently observed around nests of melanoma cells. These flattened spindle cells may represent either endothelial cells or melanoma cells, and it is difficult to unequivocally diagnose vascular invasion in these cases. Third, tumor emboli that completely obliterate the lumen of vessels and compress the surrounding endothelial cells are difficult to visualize.

There are no immunohistochemical reagents routinely used to assist in the diagnosis of lymphovascular invasion nor to distinguish lymphatic vessels from arterial capillaries and venules in primary melanoma.^10,11 In fact, in 1 study, double immunostaining with monoclonal antibodies CD31 and PAL-E was used to distinguish lymphatic from blood capillaries by exclusion, with CD31 staining both lymphatic and blood vessels and PAL-E staining blood vessels only.¹¹ A limitation of that study was that PAL-E works only on frozen sections.

Recently, a rabbit polyclonal antibody to LYVE-1, a candidate marker of lymphatic endothelium, was used to assess lymphatic vessel density in cutaneous melanoma.¹² However, the utility of this immunohistochemical reagent in detecting lymphatic invasion in primary melanoma remains to be demonstrated. Furthermore, previous reports that LYVE-1 is present on the endothelial cells of blood vessels¹³ and liver sinusoids (which are also believed to be lined with blood vessel endothelium)¹⁴ confounds the selectivity of this marker for lymphatic endothelium.

Members of our research group recently reported that the monoclonal antibody D2-40 is a highly selective marker of lymphatic endothelium in sections of both frozen and formalin-fixed paraffin-embedded normal and neoplastic tissues.¹⁵ In a direct comparison of D2-40 and CD31 on paraffin sections of a series of tumors derived from lymphatic endothelium (lymphangiomas) and blood vessel endothelium (hemangiomas), D2-40 stained all lymphangiomas (10 of 10) and no hemangiomas (0 of 10). Conversely, CD31 stained a fraction of lymphangiomas (5 of 10) and all hemangiomas (10 of 10).¹⁵ Tumors thought to be derived from both lymphatic and blood vessel endothelium, including Kaposi sarcoma and a fraction of angiosarcomas, stained positively with both D2-40 and CD31. Our team has also found that D2-40 was useful in detecting lymphatic invasion in paraffin sections of all primary tumors examined, including melanoma, squamous cell carcinoma, and cancers of the breast, prostate, colon, cervix, and endometrium.¹⁶

In the present study, we compare conventional H&E staining with D2-40 immunostaining for detection of lymphatic invasion. We also correlate lymphatic invasion detected by D2-40 immunostaining with other clinicopathologic characteristics of the tumors. Our results demonstrate a clear advantage of immunostaining with D2-40 over H&E for detection of lymphatic invasion in primary melanoma (7 [16%] vs 0 of 44 cases examined) and indicate a trend in the association between lymphatic invasion and 2 other markers of tumor aggressiveness, specifically a deeper Clark level and increased frequency of ulceration.

Forty-four consecutive primary cutaneous melanomas diagnosed between November 2000 and May 2003, with a tumor thickness greater than 0.75 mm, were selected from the files of the Department of Dermatology, Sunnybrook and Women’s College Health Sciences Center, University of Toronto, Toronto, Ontario. Clinicopathologic characteristics, including patient age, tumor site, histologic type and Clark level, cell type, tumor thickness, mitotic index, and the presence of pigment, ulceration, regression, satellite or precursor lesion, actinic damage, and lymphovascular invasion, were obtained from clinical records. The mitotic index was defined as the number of mitoses per 10 high-power fields, and tumor thickness was expressed in millimeters.

Formalin-fixed, paraffin-embedded blocks of tumors of the selected cases were obtained from the Department of Pathology, Sunnybrook and Women’s College Health Sciences Center and cut into 5-μm sections. One section from the deepest area of invasion from each case was stained with D2-40 as follows: The sections were mounted onto glass slides and dewaxed through graded concentrations of ethanol. The sections were incubated in methanol containing 3% hydrogen peroxide to inactivate endogenous peroxidase. They were then incubated with D2-40 (0.1 μg/mL), followed sequentially by biotinylated goat antimouse immunoglobulin antibody (Zymed, San Francisco, Calif) at a 1:200 dilution and a horseradish peroxidase-avidin conjugate (DakoCytomation, Carpinteria, Calif) at a 1:500 dilution. For color development, the sections were incubated with 3,3′-diaminobenzidine or Vector Nova RED (Novocastra, Newcastle-upon-Tyne, England). The whole section was screened for the presence of lymphatic invasion following immunostaining with D2-40 by 2 pathologists, with a concordance of 100% for all cases. An adjacent section was stained similarly for S100 protein using rabbit anti-S100 protein antibody (DakoCytomation) at a 1:250 dilution as the primary reagent and screened for positive expression of S100 protein by melanoma cells.

The relationship between lymphatic invasion detected with D2-40 and other clinicopathologic characteristics of the tumors was tested by the 2-tailed, 2-sample difference in proportions test. Significance was defined as P<.05.

All 44 cases of primary cutaneous melanoma selected for this study had initially been reported as negative for lymphovascular invasion based on screening of sections conventionally stained with H&E. A second independent review confirmed the absence of detectable lymphovascular invasion. Sections of archival paraffin blocks of these tumors were then immunostained with D2-40 and rescreened for the presence of lymphatic invasion with the following results: Subepidermal lymphatics were lined by a single layer of strongly immunoreactive endothelial cells, while the walls of an adjacent blood capillary were negative (Figure, A). In some areas, focal positive immunostaining of the basal layer of the epidermis was also observed (Figure, B). This is the first report that D2-40 also reacts with this cell type.

Positively stained uninvolved intratumoral lymphatics containing lymphocytes were seen also within the tumor mass (Figure, C). Lymphatic invasion defined as the presence of tumor emboli in the lumen of immunopositive lymphatics was observed within the tumor mass (Figure, D), at the invasive edge of the tumor (Figure, E), and in subepidermal lymphatics (Figure, F). In all 3 instances, the tumor emboli obliterating the lumen of the lymphatic vessel would blend in with the surrounding melanoma cells (Figure, D) or appear as solitary nests of melanoma cells (Figure, E and F) in the absence of the surrounding D2-40 immunopositive endothelium. This would explain why lymphatic invasion could not be identified in these cases by H&E alone. There was no staining of the endothelial cells lining an adjacent blood vessel (Figure, E). The number of malignant cells comprising the tumor emboli ranged from 1 to 7. These malignant cells had no distinguishing histologic features from those comprising the bulk of the primary tumor. All malignant cells comprising the tumor emboli stained positively for S100 protein.

Lymphatic invasion was observed in 7 of 44 cases. Of those, 5 displayed tumor emboli in lymphatics within the tumor mass, including 1 where additional lymphatic invasion occurred at the invasive edge of the tumor, and 2 displayed tumor emboli in subepidermal lymphatics. The number of tumor emboli in the positive cases ranged from 1 to 3, and this variation was not associated with any other histopathologic characteristic of the lesion including tumor thickness or the presence of ulceration. Among the 37 of 44 cases that were negative for lymphatic invasion, we observed the presence of up to 3 immunopositive uninvolved lymphatics within the tumor mass in 21 cases.

The correlation between lymphatic invasion and other clinicopathologic characteristics of the tumors is summarized in Table 1. There was no significant association between lymphatic invasion and any other clinicopathologic characteristic of the tumors, including patient age, tumor site, histological type, Clark level, cell type, mitotic index, tumor thickness, or the presence of pigment, ulceration, regression, satellite lesion, precursor lesion, or actinic damage. However, there was an apparent trend in the association between lymphatic invasion and a deeper Clark level (P = .08) or the presence of ulceration (P = .12). Specifically, a Clark level of at least 4 was assigned approximately twice as frequently to melanomas positive for lymphatic invasion (4/7; 57%) as to those negative for lymphatic invasion (9/37; 24%). Similarly, 2 (29%) of the 7 melanomas positive for lymphatic invasion and only 3 (8%) of the 37 negative for lymphatic invasion showed the presence of ulceration. The clinicopathologic characteristics of the 7 positive cases are summarized in Table 2. Four of these cases were assigned a Clark level of 4, and three were given a Clark level of 3. In addition, 2 cases had a tumor thickness of less than 1 mm.

The main objective of the present pilot study was to compare the relative sensitivities of conventional H&E staining and D2-40 immunostaining for the detection of lymphatic invasion in paraffin sections of primary melanoma. For this purpose, we selected cases of primary cutaneous melanoma with a tumor thickness greater than 0.75 mm, which we anticipated would show a higher frequency of lymphatic invasion than thinner tumors. Unexpectedly, we detected no lymphovascular invasion with H&E staining in 44 consecutive cases of primary cutaneous melanoma. In contrast, with D2-40 immunostaining, lymphatic invasion was detected in 7 (16%) of the same 44 cases. We attribute this increase in the sensitivity of detection of lymphatic invasion to the clear demarcation of the lymphatic endothelium surrounding tumor emboli following immunostaining with D2-40. The failure to detect lymphatic invasion on the initial screen of the 44 cases following conventional H&E staining was most likely owing to the difficulty in identifying the lymphatic endothelium circumscribing the tumor emboli. In particular, in the absence of the surrounding immunopositive endothelium, tumor emboli obliterating lymphatic vessels would blend in with the surrounding melanoma cells (Figure, D) or appear as a solitary nest of melanoma cells (Figure, E and F). In those instances, the identification of lymphatic invasion would be impossible in the absence of positive immunostaining of the lymphatic endothelium with D2-40.

In the absence of lymphatic invasion, we detected up to 3 immunopositive uninvolved lymphatics within the tumor mass in 21 (57%) of 37 cases of primary melanoma following immunostaining with D2-40. The sporadic occurrence of uninvolved lymphatics within the tumor mass is consistent with the general view that there is little, if any, active lymphangiogenesis within primary tumors, including melanomas.^11-13,17 In cases positive for lymphatic invasion, we detected tumor emboli within the tumor mass in 5 (71%) of 7 cases, including 1 case where there was additional lymphatic invasion at the invasive edge of the tumor. It is likely that the observed intratumoral lymphatic invasion resulted from melanoma cells invading preexisting intratumoral lymphatic vessels.

Lymphatic invasion is currently not reported in primary melanoma in surgical pathology because no immunohistochemical reagent has gained general acceptance for its use on paraffin sections for this purpose. Our results indicate that D2-40 is a candidate for this application owing to its strong immunostaining of lymphatic endothelium that outlines tumor emboli in lymphatics otherwise indiscernible by H&E. In fact, the increase in the frequency of detection of lymphatic invasion relative to H&E from 0% to 16% following immunostaining with D2-40 in the present pilot study presages a potential prognostic value for this parameter in primary cutaneous melanoma. This prediction will require confirmation in follow-up studies in which lymphatic invasion detected after immunostaining with D2-40 is correlated with outcome. Our results provide a preliminary indication that lymphatic invasion detected with D2-40 may indicate a poor prognosis in that we observed a trend in the association between lymphatic invasion and 2 markers of tumor aggressiveness, namely a deeper Clark level and increased frequency of ulceration.

Another potential direction for future investigation is the usefulness of lymphatic invasion detected with D2-40 as a criterion for selecting patients with primary cutaneous melanoma for sentinel node biopsy. While Breslow tumor thickness is considered the most important prognostic indicator in primary melanoma, other histopathologic characteristics of the tumor are also used in conjunction with this parameter in arriving at this management decision. For example, current literature recommends sentinel node biopsy for melanomas with a tumor thickness greater than 1 mm and those less than 1 mm and a Clark level of 4 or higher and/or the presence of ulceration.^18,19 However, sentinel node biopsy specimens are sometimes positive in melanomas with a thickness less than 1 mm and a Clark level of 3 without ulceration.¹⁹ Additional criteria for selecting candidates for sentinel node biopsy in this subset of patients would be very useful. In our study, 2 of the 7 lesions that were positive for lymphatic invasion had a tumor thickness less than 1 mm, a Clark level of 3, and were not ulcerated (Table 2). It would be important to determine the frequency of sentinel node involvement in this subset of patients. This might lead to a consideration of recommending sentinel node biopsy for melanomas with a tumor thickness less than 1 mm, a Clark level of 3, and lymphatic invasion detected with D2-40.

Correspondence: Lynn From, MD, Department of Pathology, Sunnybrook and Women’s College Health Sciences Center, University of Toronto, 76 Greenville St, Toronto, Ontario, Canada M5S 1B2 (lynn.from@sw.ca).

Accepted for Publication: November 2, 2004.

Acknowlegment: We thank Kevin Kwok, MLT, for skillful technical assistance and Alex Kiss, PhD, for assistance with the statistical analysis of the data.

Financial Disclosure: Drs Kahn and Marks receive royalties from Signet Laboratories Inc and DakoCytomation California Inc from the sale of monoclonal antibody D2-40.