INTRODUCTION

The line of differentiation of Kaposi's sarcoma (KS) has been long debated, with arguments offered in favor of endothelial, smooth muscle, or undifferentiated mesenchymal lineage. There is now overwhelming evidence that KS shows endothelial differentiation (1, 2, 3, 4, 5, 6, 7, 8) even though there is no consensus on whether the cells of KS have a lymphatic (2, 8) or vascular endothelial (1, 7, 9) phenotype. Recently, a novel monoclonal antibody to vascular endothelial growth factor receptor 3 (VEGFR-3), a tyrosine kinase receptor expressed almost exclusively by lymphatic endothelium in the adult, has been shown to react with a small number of cases of KS (4) and cutaneous lymphangiomas (10). We sought to extend these studies to a large number of well-characterized vascular neoplasms to evaluate possible diagnostic uses of this antibody because a highly sensitive and specific antibody for KS would clearly be a useful diagnostic. At present, antibodies to CD31, CD34, and von Willebrand factor (vWF, Factor VIII related antigen) are commonly used in the distinction of KS from nonvascular neoplasms, but they suffer from lack of sensitivity, in the case of vWF and CD31 (5, 6, 11) and lack of specificity, in the case of CD34 (12, 13, 14). Furthermore, they are of little use in distinguishing KS from benign and malignant vascular tumors (7, 8). In addition, we wanted to determine whether VEGFR-3 defines other vascular lesions in a thematic fashion.

METHODS

Formalin-fixed, paraffin-embedded sections from 70 vascular tumors (24 KS, 16 angiosarcomas (AS), 13 hemangiomas, 6 kaposiform hemangioendotheliomas, 3 intra-abdominal lymphangiomas, 4 malignant endovascular papillary angioendotheliomas [Dabska tumor], 1 epithelioid hemangioendothelioma, 1 splenic littoral angioma, 1 vascular malformation, and 1 case of stasis dermatitis) were retrieved from the consultation files of one of the authors (SWW) and the archives of the Departments of Pathology of Emory University and the University of Washington Medical Center. None of the AS were known to be associated with chronic lymphedema or prior irradiation. Deparaffinized sections were immunostained with a monoclonal antibody raised against VEGFR-3 (9D9F9, 1:1000; laboratory of Dr. K. Alitalo). The production of this antibody has been described in detail elsewhere (4). A subset of cases were also immunostained with antibodies to vWF (F8/86, 1:100; Dako Corp., Carpinteria, CA) and CD31 (JC70, 1:80; Dako). Negative controls consisted of omission of the primary antibody. Sections were subjected to heat-induced epitope retrieval, using a vegetable steamer. Antigens were localized using an avidin-biotin method with 3,3′-diaminobenzidine as a chromogen.

RESULTS

As detailed in Table 1, strong immunoreactivity for VEGFR-3 was seen in 23 of 24 cases of KS (96%; Fig. 1). Strong VEGFR-3 expression was also noted in 8 of 16 (50%) cases of AS; of the positive cases, six had either a prominent lymphocytic infiltrate or hobnail endothelial cells (Figs. 2 and 3). VEGFR-3 was also expressed in all cases of kaposiform hemangioendothelioma and Dabska tumor (Fig. 4). Two of 13 hemangiomas (15%) and 1 of 3 cases of intra-abdominal lymphangioma also contained positive cells. The one vascular malformation studied was focally positive, in an area of thin-walled vessels believed to be morphologically very suggestive of lymphatic differentiation.

TABLE 1 Immunohistochemical Findings
Full size table
FIGURE 1
figure 1

A low power view of a typical case of capillary hemangioma, showing the VEGFR-3–negative tumor vessels and the positive normal lymphatics (A and B). In contrast, all but one case of Kaposi's sarcoma were intensely VEGFR-3 positive (C and D). (A, hematoxylin and eosin, 100 × ; B, anti–VEGFR-3, 200 × ; C, hematoxylin and eosin, 200 × ; D, anti–VEGFR-3, 200 ×).

Full size image
FIGURE 2
figure 2

Low (A) and medium (B) power views of an angiosarcoma, not associated with lymphedema, characterized by a very prominent lymphocytic inflammatory infiltrate. The three angiosarcomas with these features were all strongly VEGFR-3 positive (C). (A, hematoxylin and eosin, 40 × ; B, hematoxylin and eosin, 200 × ; C, anti–VEGFR-3, 200 ×).

Full size image
FIGURE 3
figure 3

Two VEGFR-3–positive angiosarcomas were characterized by “hobnailed” endothelial cells and endovascular tufting, as is seen in Dabska's tumor, but showed in addition infiltrative and irregular growth. (A, hematoxylin and eosin, 200 × ; B, anti–VEGFR-3, 200 ×).

Full size image
FIGURE 4
figure 4

All cases of both Dabska's tumor (A and B) and kaposiform hemangioendothelioma (C and D) strongly expressed VEGFR-3. (A and C, hematoxylin and eosin, 200 × ; B, anti–VEGFR-3, 100 × ; D, anti–VEGFR-3, 200 ×).

Full size image

No VEGFR-3 expression was noted in any cases of epithelioid hemangioendothelioma, littoral angioma, or stasis dermatitis. Positive internal controls (normal lymphatic channels) were present in all cases.

vWF expression was seen in 10 of 13 cases of KS (77%); 12 of 13 cases of AS (92%); 4 of 5 kaposiform hemangioendothelioma (80%); and all cases of Dabska tumor, hemangioma, lymphangioma, epithelioid hemangioendothelioma, vascular malformation, stasis dermatitis, and splenic littoral angioma studied. CD31 expression was present in 12 of 13 cases of KS (92%), 13 of 14 cases of AS (93%), and all other studied cases.

The one case of KS that was VEGFR-3 negative was also negative for vWF but was strongly CD31 positive. One case of KS was VEGFR-3 positive but negative for both vWF and CD31. Six of the eight VEGFR-3–positive AS were also strongly positive for both vWF and CD31; one case was vWF negative but strongly CD31 positive, and in one case there was focal and weak vWF expression and strong CD31 expression.

DISCUSSION

The platelet derived growth factor (PDGF) family, including vascular endothelial growth factor (VEGF) and the closely related molecules VEGF-B, VEGF-C, and VEGF-D, play a significant role in angiogenesis and vascular permeability (15). VEGF-C plays a critical role in lymphangiogenesis; in transgenic mice, VEGF-C has the ability to induce both lymphatic endothelial proliferation and lymphatic vessel formation (16, 17). In the adult, expression of the mRNA of the VEGF-C receptor, VEGFR-3 (FLT4), is limited almost exclusively to lymphatic endothelium, although it may be detected in other endothelia at earlier stages of development (18).

Recently, the production of a monoclonal antibody to VEGFR-3, 9D9F9, was reported by Jussila and co-workers (4), who found it to react with lymphatic endothelium but not with capillaries or lymph node high endothelial venules. These authors also found VEGFR-3 expression in paraffin-embedded and frozen sections from eight and three cases of KS, respectively (4). Subsequently, Lymboussaki and co-workers (10) found VEGFR-3 expression in six cases of paraffin-embedded dermal lymphangiomatosis and three cryosectioned intramuscular hemangiomas but not in three paraffin-embedded capillary hemangiomas.

We examined the immunohistochemical expression of VEGFR-3 in 70 cases of paraffin-embedded vascular tumors, including 25 cases of KS. Our results strongly affirm the superb sensitivity of antibodies to VEGFR-3 for KS; the spindle cells in 24 of 25 cases of KS, at all stages of development and including metastatic lesions, were strongly positive for VEGFR-3. The sensitivity of anti–VEGFR-3 in paraffin-embedded sections (96%) compares very favorably to that reported for CD31 (89%) (5), CD34 (93%) (19), and vWF (64%) (20). Our findings of slightly greater sensitivity for CD31 (93%) and vWF (79%) may be the result of our uniform use of heat-induced epitope retrieval. Furthermore, the immunoreactivity was generally present in more than 50% of the tumor cells, with a very low level of nonspecific “background” staining. In all cases, small, normal vascular spaces that were believed morphologically to be consistent with either lymphatics or small venules were also positive; endothelia in clearly identifiable arteries and veins were always negative. Significantly, we did not observe any other normal structure or cell to be positive for VEGFR-3. This is in strong contrast to CD34, which is positive in normal dendritic interstitial cells surrounding blood vessels and nerves and in a variety of nonvascular tumors, including dermatofibrosarcoma protuberans, gastrointestinal stromal tumors, and solitary fibrous tumors (19). CD31 may also be expressed by nonendothelial cells, including monocytes/macrophages (21, 22), very rare carcinomas (23), and epithelioid sarcomas (24). We are in the process of studying VEGFR-3 expression in nonvascular tumors but see no reason to believe that it will be present in more than extremely rare cases.

There is debate in the literature, as well as a great deal of contradictory evidence, about whether the cells of KS show lymphatic or vascular endothelial differentiation. Some authors have favored lymphatic endothelial differentiation, based on ultrastructural findings of discontinuous basal lamina and the absence of dendritic pericytes (25), and immunohistochemical findings such as absent or only weak vWF and Ulex europaeus lectin expression (2, 6, 26). Other authors have claimed blood vessel endothelial differentiation in KS, on the basis of immunohistochemical expression of vWF in frozen sections (7, 9), the demonstration of uniform laminin investment of the spindle cells (2) or the supposed absence of CD31 on both KS and lymphatic endothelium (27). In attempting to unravel these often contradictory claims, it is important to note that whereas early studies suggested that vWF, CD34, and CD31 were expressed only by blood vessel endothelium, more recent studies have shown weaker but demonstrable levels of all three antigens in lymphatic endothelium (4, 10, 28). This implies that the presence or absence of these antigens is not an absolute indication of differentiation in a vascular tumor. Our finding of VEGFR-3 expression in almost all cases of KS may support the concept that KS shows lymphatic endothelial differentiation. An alternative is that KS may recapitulate an immature endothelial cell phenotype, thereby explaining VEGFR-3 positivity (18).

We had hoped that the finding of VEGFR-3 expression might be useful in separating KS from other vascular tumors with which it may be confused, including spindled variants of AS and kaposiform hemangioendothelioma. Unfortunately, among vascular tumors, VEGFR-3 expression is not specific for KS. We were able to find VEGFR-3 expression in 8 of 16 AS, with a staining intensity comparable to KS. It is interesting that of these eight positive cases, six were characterized by highly unusual appearance. Three displayed a striking lymphocytic infiltrate, although none arose in the setting of lymphedema. Two were characterized by small cuboidal endothelial cell reminiscent of those of the Dabska tumor yet grew in an infiltrative and irregular pattern similar to a conventional AS. Another was a highly spindled AS, with a prominent lymphocytic infiltrate, arising from the peritoneal surface, which in many areas out of context resembled KS. One could therefore mount a reasonable argument that these AS represent one or more distinct subsets perhaps linked by lymphatic endothelial differentiation. That some AS might display lymphatic differentiation was implied by the early term lymphangiosarcoma. This term was later supplanted by the generic angiosarcoma because of the inability to separate precisely lymphatic from capillary vascular endothelial differentiation by light microscopy. However, the diverse clinical settings in which AS develop (cf. postirradiation, lymphedema associated), buttressed by the results of our study, lend support to the idea that AS comprise several phenotypic and possibly biologic subsets. None of the AS in the current study were known to have been associated with chronic lymphedema or prior irradiation.

We have also found uniform expression of VEGFR-3 in two rare pediatric vascular tumors of intermediate malignancy, kaposiform hemangioendothelioma and Dabska tumor. This finding is notable because both of these tumors have been postulated to show lymphatic endothelial differentiation, on the basis of their characteristic infiltration by lymphocytes (29, 30), the association of the former with lymphangiomatosis (29, 31), and their reported usual absence of vWF expression (29, 31). We observed vWF expression in 80% of kaposiform hemangioendothelioma and all Dabska tumors; again, this difference may be the result of our use of heat-induced epitope retrieval.

We also saw expression of VEGFR-3 in only 2 of 13 cases of capillary hemangioma, in general agreement with the previously reported results of Lymboussaki et al. (10). We found VEGFR-3 positivity in only one of the three intra-abdominal lymphangiomas, as compared with the uniform positivity previously noted in dermal lymphangiomatosis (10); possibly this may reflect lower production of this protein in intra-abdominal tumors as compared with dermal ones.

In conclusion, we have shown immunohistochemical expression of VEGFR-3 in 96% of cases of KS, all cases of kaposiform hemangioendothelioma and Dabska tumor, and a significant subset of AS. The sensitivity of anti–VEGFR-3 for KS is higher than that of the routinely used vascular markers CD31, CD34, and vWF and suggests a valuable role for this antibody in the diagnosis of difficult cases of KS. In particular, this antibody could serve to distinguish various fibrovascular proliferations (e.g., stasis changes, chronic ulcers with surrounding reactive fibrosis) from KS. Our findings also support the concept that the KS, kaposiform hemangioendothelioma, and the Dabska tumor have a lymphatic endothelial phenotype. AS expressing this receptor protein may represent tumors displaying lymphatic differentiation. Whether these phenotypic differences will translate into biologic ones requires additional study.