In 1992, Weidner et al.
20 described the association between the density of vascularization in breast cancer and outcome. Staining sections of breast lesions for the presence of FVIII-RAg, Weidner et al.
20 advised scanning tissue sections to identify the zone of maximum concentration of marker staining. Each discreet focus of staining was to be counted as a separate vessel, even if the morphology suggested the same vessel snaking in and out of the section plane. The observations recorded by Weidner et al. in the breast were challenged subsequently.
21 22 23 24 Evidence both supporting
1 and challenging
2 the relationship of MVD to outcome in a variety of tumor systems has since been reported. The association between high MVD and metastasis has been postulated to represent either increased opportunities for tumor cells to enter the microcirculation through increased exposure to microvessels or to reflect increased angiogenesis by aggressive cancers (i.e., high MVD is an indirect measure of tumor aggressiveness).
25
One would have suspected that the association between MVD and outcome would have been confirmed easily in uveal melanoma, because unlike breast cancer or cutaneous melanoma, which may spread by either lymphatics or by the blood stream, uveal melanoma spreads exclusively by the hematogenous route, because there are no lymphatics within the eye or within uveal melanomas.
26 Nevertheless, Lane et al.
27 and Schaling et al.
28 both failed to discover an association between MVD and outcome. Foss et al.,
8 using FVIII-RAg to label tissue sections of uveal melanomas and the techniques advocated by Weidner et al.
20 for locating hot spots, discovered an association between MVD and adverse outcome but failed to confirm previous observations
11 that PAS-positive looping patterns contribute independently to death in metastatic melanoma.
Mäkitie et al.
9 confirmed the association of both MVD and PAS-positive patterning in choroidal and ciliary body melanoma and death caused by metastatic disease. These investigators pointed out that neither Lane et al.
27 nor Schaling et al.,
28 who failed to discover associations between MVD and outcome, used the method suggested by Weidner et al.
20 Mäkitie et al.
9 also postulated that Foss et al.
8 defined PAS-positive patterns in a fashion different from that used by Folberg et al.
11 and by subsequent groups
29 30 31 who confirmed the association between these patterns and outcome: The distribution of looping PAS-positive patterns in the study by Foss et al.
8 varied from that described by those who found these patterns to be prognostically significant.
After the publication of the work by Mäkitie et al.
9 indicating an independent effect of both MVD and PAS-positive patterns on mortality in ciliary and choroidal melanoma, attention was drawn to the observation that aggressive uveal melanoma cells appear to be genetically deregulated and may express markers not typical for melanoma cells or cells derived from neural crest.
32 33 34 Moreover, the specificity of endothelial cell markers has long been called into question. Mäkitie et al.
9 preferred CD34 to FVIII-RAG in studies of MVD in uveal melanoma. Although the influence of FVIII-RAg staining on MVD was not examined in the present study, FVIII-RAg is not a specific marker for endothelium. The cell line ECV-304 which was originally reported to be a human immortalized vascular endothelial cell (HUVEC) line by virtue of expression of factor VIII, ultrastructural features such as Weibel-Palade bodies, and the formation of tubules on synthetic basement membrane,
35 36 was recently discovered to be a derivative of the human bladder tumor cell line T14.
37 CD34 is a 115-kDa cell surface protein expressed by myeloid and lymphoid progenitor cells, and it has been observed that CD34 may be expressed in endothelial cells during angiogenesis.
38 However, CD34 is also present in lesions derived from the neural crest, such as neurofibromas,
39 cellular blue nevi,
6 and desmoplastic cutaneous melanoma.
4 In some cellular blue nevi, there is diffuse staining of melanocytes by CD34,
6 consistent with our observation of diffuse expression of CD34 in tumor cells in some uveal melanomas
(Figs. 1C 1D) .
In this study, both MVD and PAS-positive patterns exerted independent effects on death caused by ciliary body and choroidal melanoma, as demonstrated by Mäkitie et al.
9 The Cox model developed from our data
(Table 1) differs from previously published models from this data set,
11 in that previously, our classification pertaining to cell type was dichotomous (epithelioid cells absent or present) and cell type did not enter the final model, whereas in the present study, our dichotomous classification was epithelioid predominant or not. With this modification in classification, cell type (epithelioid cell predominant) was entered into the model
(Table 1) , and the “silent” pattern was also entered into the model but with a hazard ratio less than 1, indicating a protective effect by the presence of this feature. Indeed, the silent pattern represents foci in which neither blood vessels nor PAS-positive patterns are present,
40 consistent with improved survival.
In this study, we also investigated the possibility that uveal melanoma cells may label for markers used in the assessment of MVD. Folberg et al.
2 have already shown that the nuclear matrix of some choroidal and ciliary body melanoma cells labels with CD31, but CD31 has not been used in studies of MVD in uveal melanoma. In histologic sections of choroidal and ciliary body melanoma, we identified epithelioid melanoma cells that stained for CD34
(Fig. 1A) . However, to eliminate confusion introduced by the subjective assignment of cells according to histologic lineage, we investigated the labeling of melanoma cells by CD34 by labeling tissue sections for markers expressed in melanoma cells but not on endothelial cells and double labeling for the endothelial cell marker of interest, CD34. S100 protein is a nonspecific marker associated with cells of neural crest lineage, and therefore not expressed on vascular endothelium. HMB45, Melan-A, and micro-ophthalmia transcription factor (MiTF) are more specific for melanoma than S100 protein in the detection of uveal melanoma cells.
16 41
S100 protein is diffusely distributed in uveal melanomas, but is less specific a marker for melanoma cells than MiTF, HMB45, or Melan-A. MiTF, however, is a related to differentiation, and expression of MiTF in cutaneous melanoma has been shown to be associated with a favorable outcome.
42 In preliminary studies in our laboratory, MiTF was not diffusely expressed by melanoma cells in aggressive tumors and was not studied further as a marker for melanoma cells in this study. The distribution of HMB45 and Melan-A in ocular melanocytic lesions has been studied by Heegaard et al.
16 and was found to be comparable to the distribution of S100 protein in melanoma cells.
In exploratory studies, we discovered that S100 protein was more diffusely distributed through the lesion than the other markers, and S100 protein was used to explore the possibility that melanoma cells that express an endothelial cell marker contribute to the calculation of MVD in a hot spot in our series of 200 melanomas. To mirror the study by Mäkitie et al.
9 as closely as possible, we used CD34 as a marker for endothelium in our study of MVD. Mäkitie et al.
9 had demonstrated no significant difference between the calculation of MVD from CD34-stained sections and sections stained for FVIII-RAg, but found that CD34 rendered a reaction product that was easier to interpret. It is noteworthy that not only did melanoma cells colabel for S100 protein and CD34, but also with Melan-A and CD34.
In this study, we graded the coexpression (absent, weak, moderate, intense) of S100 protein and CD34 based on the Pearson correlation coefficient for colocalization within the cell of interest. This may represent too stringent a definition for colocalization. We and our colleagues
32 33 34 43 have shown that aggressive melanoma cells are genetically deregulated cells and that they express inappropriate markers not expected of cells of melanocytic lineage. It is therefore possible that the deregulated melanoma cells may lose expression of S100 protein or Melan-A as they acquire expression of CD34. For example, the expression of MiTF, a melanoma marker, appears to be a marker of differentiation, and there is a tendency to observe reduced expression of MiTF in more aggressive cutaneous melanomas.
42
Nevertheless, by requiring that a cell coexpress a melanoma marker along with CD34, we achieve some measure of certainty that the cell labeling is not an endothelial cell. Not only does CD34 label melanoma cells in uveal melanoma hot spots
(Figs. 1C 1D) , but there also appears to be an association between MVD and the presence of CD34
+ melanoma cells in the hot spot
(Tables 3 4) . Thus, in uveal melanomas, the characteristic known as MVD is not a pure quantification of blood vessels per unit area, but is rather a hybrid measurement of blood vessels and melanoma cells.
It is intriguing to speculate that the inappropriate expression of CD34 on melanoma cells is a marker of aggressive behavior. The observation that there are significantly fewer CD34
+ melanoma cells in hot spots with low MVD than in those with high MVD
(Tables 3 4) suggests that the number of CD34
+ melanoma cells may be a risk factor for metastasis, but we did not quantify the number of S100
+ tumor cells labeled with CD34. In this study, the mere presence of strongly positive CD34
+ melanoma cells in hot spots was not itself associated with death caused by metastatic melanoma.
If high MVD is not necessarily equivalent to a hot spot of vascularity, then it is reasonable to challenge the assumption that high MVD reflects increased angiogenesis. Indeed, in this study results showed that the blood supply to uveal melanomas was complex and heterogeneous. Aside from normal vessels incorporated into the tumor
(Fig. 1B) and vessels within the tumor originating from a presumed angiogenic response
(Fig. 1D) , some vessels are lined in part by endothelial cells and in part by tumor cells, consistent with mosaic vessels
19 (Fig. 1E) . The incorporation of tumor cells into tumor vessels has been well established.
18 44 45 46 In animal models of colon cancer and in human colon cancer specimens, it has been estimated that 13% of tumor vessels are mosaic (contain endothelial cells and tumor cells), representing 4% of the total vascular volume of microcirculation in these tumors. Folkman
47 has suggested that the incorporation of tumor cells into blood vessels (vascular mosaicism) may be a mechanism that enhances metastasis by the shedding of tumor cells into the microcirculation. Finally, in addition to circulation within blood vessels, it is possible that blood and plasma may circulate within components of highly patterned extracellular matrix that is generated by the tumor cells themselves.
2 32 The contribution of patterned matrix to a functional circulation (vasculogenic mimicry) is unclear and controversial.
48
In this study MVD, as defined by Weidner et al.
20 and modified for use in uveal melanoma by Mäkitie et al.,
9 was independently associated with death caused by metastatic melanoma and that PAS-positive looping patterns enter the Cox model along with MVD. However, the colocalization of the putative endothelial cell marker CD34 to melanoma cells in hot spots measured for the calculation of MVD suggests strongly that MVD cannot be equated with angiogenesis. These findings, if replicated with other markers and in other tumor systems, may have important implications for pathologists who associate MVD with outcome and investigators who rely on the associations between MVD and tumor behavior when designing new therapies to target angiogenesis.