Presence of a Fluid-Conducting Meshwork in Xenografted Cutaneous and Primary Human Uveal Melanoma

Abstract

purpose. Recently, it was reported that tumor cells themselves generate channels and networks in three-dimensional culture and can be found lining channels (some containing red blood cells [RBCs]) in vivo, and they express endothelial or vascular genes in aggressive uveal melanoma. The implications of these data for current insights in the involvement of angiogenesis in tumor growth, metastasis and therapeutic intervention are considerable. Therefore, this possibility was investigated in the current study.

methods. Thirty human uveal melanomas and 20 xenografts of human cutaneous melanoma were analyzed by Azan histochemistry and immunostaining of endothelial markers. Additionally, in xenografted tumors a tracer study was performed with confocal microscopy and immunoelectron microscopy.

results. Lumina or spaces without endothelial lining containing RBCs were not detected in any lesion. Functional evaluation of the vasculature in xenografts demonstrated rapid tracer appearance both inside and outside blood vessels. Outside blood vessels it spread along matrix networks of arcs and back-to-back loops. Confocal microscopy showed that this extracellular matrix was deposited as stromal sheets around nests of tumor cells. Laminin immunostaining revealed that between sheets surrounding adjacent nests, spaces were present. These spaces were filled, however, with collagen and different types of cells, including cells stained for macrophage markers.

conclusions. Although no evident endothelium-free and RBC-containing channels were present in the tissues examined, there are fluid-conducting spaces in the form of stromal sheets between nests of tumor cells. In this stromal network, blood vessels are embedded. The authors postulate that this extracellular matrix tissue represents a fluid-conducting meshwork.

Malignant tumor growth, survival, and metastasis are facilitated by several different factors, including angiogenesis. ¹ The role of the vasculature in mediating metastatic spread has been supported by studies demonstrating that microvessel density is a negative prognostic factor in several types of tumors, ^{2 3} although in other types of tumors this could not be confirmed. ^{4 5 6 7 8} Concerning both uveal and cutaneous melanoma, the relation between microvascular density and clinical outcome is controversial. ^{9 10 11 12 13 14 15 16 17} In uveal melanoma, certain patterns of extracellular matrix deposition (i.e., arcs and back-to-back loop networks) have been related to rate of metastasis and, hence, prognosis. ^{9 10} Initially, these patterns, identified by conventional periodic acid-Schiff (PAS) staining, were believed to represent blood vessels. Foss et al. ¹⁸ re-evaluated this question and concluded that, instead, they consisted of connective tissue in which blood vessels were present at certain locations only—the so-called fibrovascular tissue. However, a question remained whether the patterns described by Foss et al. were identical with the patterns studied by Folberg. ¹⁹  

Recently, Maniotis et al. ²⁰ suggested that melanoma cells themselves could form a new type of vessel: blood-conducting channels lined by tumor cells that were present in the PAS-positive back-to-back loop networks in both aggressive uveal and cutaneous melanoma. They were also detected by angiography, indicating that channels and the normal blood vasculature were interconnected. This newly described phenomenon was termed vasculogenic mimicry. A recent report ²¹ demonstrated the presence of “mosaic” blood vessels in which endothelial cells and tumor cells alternately form the luminal surface. 

These data have important implications: The existence of blood-conducting channels lined by tumor cells or blood vessel walls consisting of mosaics of tumor and endothelial cells challenges the current concept that a tumor is fully dependent on angiogenesis for growth and metastasis. ² In addition, tumors with vascular channels or mosaic blood vessels would be less sensitive to antiangiogenic or antiendothelial drugs. 

Thus, elucidation of the nature of the extracellular matrix patterns is appropriate. Therefore, we set out to evaluate this issue in primary uveal melanoma. Because previous studies ^{22 23} showed that arclike patterns of matrix deposition are also present in xenografts derived from human cutaneous melanoma cell lines, we re-examined this tissue as well. 

Materials and Methods

Primary Uveal Melanoma

Formaldehyde-fixed, paraffin-embedded tissues from 30 cases of human primary uveal melanoma were obtained from our archive. The uveal melanomas varied from 0.5 to 2.5 cm in diameter (median, 1.7 cm) and included 28 choroidal and 2 ciliary melanoma lesions. Using hematoxylin and eosin (H&E) staining on paraffin sections, we classified the uveal melanomas as 10 spindle cell type, 5 epithelioid, and 15 mixed type. No information on clinical outcome was available. 

Xenografts in Nude Mice

Human melanoma cell lines 1F6 (nonaggressive), Mel57 (aggressive), ²⁴ and Mel57 stably transfected with VEGF (Mel57-VEGF) (constructed by William P. Leenders in our laboratory) were cultured as previously described. ²⁵ For induction of tumor growth, 2.5 × 10⁶ cells were injected subcutaneously into BALB/c nu/nu mice. A group of 5 (1F6, Mel57-VEGF) or 10 (Mel57) animals were included. Subcutaneous xenografts developed in 17 mice. In the remaining three mice (one injected with Mel57 and two with Mel57-VEGF cells) extensive intraperitoneal outgrowth occurred. The volume of the subcutaneous tumors was estimated by multiplying length, width, and height. When the tumors reached sizes between 100 and 700 mm³, mice were injected intravenously with 100 μL of a 3% (wt/vol) solution of fluorescein isothiocyanate-bovine serum albumin (FITC-BSA; 12 mol FITC/1 mol BSA; Sigma, Brunschwig, Amsterdam, The Netherlands). Tumors were excised 60 minutes after injection and divided into three equal parts: one part was formalin-fixed, the second was divided into two fragments that were fixed by either glutaraldehyde or periodate-lysine-2% paraformaldehyde, and the third was snap frozen in liquid nitrogen. 

Immunohistochemistry

Markers are listed in Table 1 : CD31 (JC/70A), CD34 (QBEnd/10), H and Y antigens (BNH9), thrombomodulin (anti-thrombomodulin) (all from Dako, Glostrup, Denmark), CD31 (Mec 7.46; Hycult Biotechnology, Uden, The Netherlands), 9F1 (Alf Hamann, Hamburg, Germany), and ASD-13 (Karel J. M. Assmann, our laboratory). UEA-1 (Dako), polyclonal antibodies to FVIIIrA (von Willebrand factor [vWF]; CLB, Amsterdam, The Netherlands), and laminin (E2 EHS; Jaap van den Born, Department of Nephrology, University Medical Center, Nijmegen, The Netherlands) were used. Macrophages were detected by FA/11 mAb ²⁶ (Michael J. Smith, Neurobiology Division, Medical Research Council (MRC) Laboratory of Molecular Biology, Cambridge, UK). 

Serial paraffin-embedded 4-μm sections were stained by H&E, PAS, and Azan histochemistry and by immunohistochemistry with polyclonal antibodies to laminin and the markers shown in Table 1 . The distribution of endothelial (Mec 7.46), matrix (laminin), and macrophage (FA/11) markers was evaluated with a standard three-step avidin-biotin complex (ABC) method and development in 3-amino-9-ethyl-carbazole solution (Aldrich, Steinheim, Germany). Sections were counterstained for 45 seconds with Harris hematoxylin (Merck, Darmstadt, Germany) at room temperature and mounted in medium (Imsol-mount; Klinipath BV, Duiven, The Netherlands). 

Immunofluorescence and Confocal Microscopy

Serial 4-μm cryosections of xenograft tissue were fixed in acetone for 10 minutes. Subsequently, binding of Mec 7.46 mAb or anti-laminin antibodies was detected by secondary tetramethylrhodamine isothiocyanate (TRITC)-labeled antibodies (Alexa Fluor 568; Molecular Probes, Leiden, The Netherlands) and mounted in fluorescence mounting medium (VectaShield; Vector Laboratories, Inc., Burlingame, CA). FITC-labeled BSA was visualized in these sections as well. For confocal microscopy 20-μm cryosections of Mel57 and Mel57-VEGF xenografts (n = 3) were fixed in acetone for 10 minutes and stained by a secondary TRITC-labeled anti-rat antibody to detect Mec 7.46 binding and a secondary Cy 5-labeled anti-rabbit antibody (Amersham, Pharmacia Biotech UK, Ltd., Buckinghamshire, UK) to detect laminin. 

Immunoelectron Microscopy

The endothelial marker 9F1 (Table 1) and laminin were visualized in Mel57 and Mel57-VEGF xenografts (n = 2) both by light and immunoelectron microscopy as described previously. ^{27 28} Binding of the 9F1 mAb was detected with 3,3′-diaminobenzidine (DAB) and anti-laminin polyclonal antibody binding, using ultrasmall gold particles. 

Results

Evaluation of Extracellular Matrix Patterns in Uveal Melanoma and Xenografts

Vessel detection was compared in sections of uveal melanoma lesions. The anti-CD34 mAb consistently stained the vasculature (identified by morphologic characteristics) in all areas of the tumor and was superior to endothelial detection by other antibodies (data not shown). For paraffin-embedded xenograft tissue, endothelial staining by ASD-13 was superior to anti-vWF, 9F1, and Mec 7.46 antibodies (not shown). Mec 7.46, 9F1, and ASD-13 mAb immunostaining of frozen sections of various normal mouse tissues and xenografted tumors demonstrated the specificity of these mAbs for endothelial cells (not shown). Further evaluation of the vasculature on paraffin sections was therefore performed with the anti-CD34 mAb for uveal melanoma and ASD-13 mAb for xenografts. None of the antibodies used stained melanoma cells. In addition, Mel57 cells grown in vitro on microscope slides did not express endothelial markers (not shown). 

Azan staining highlighted extracellular matrix in all 30 uveal melanomas and 20 xenografts, corresponding to the earlier-described PAS-positive patterns, ^{9 10 18} which was confirmed by PAS staining of serial sections (Fig. 1) . Erythrocytes stained an intense red (Figs. 2A 2C) . In 14 (47%) of 30 uveal melanomas, arc, loops and network patterns were present, confirming earlier results. ^{9 10} In all 30 melanomas, erythrocytes were localized in circumscribed lumina in the extracellular matrix that were invariably identified as blood vessels by endothelial CD34 staining and morphologic characteristics in parallel sections (Figs. 2A 2B) . Certain parts of the arced patterns were also detected by the anti-CD34 mAb, but these were not associated with closed loops, lumina, or erythrocytes. Occasionally, erythrocytes were observed outside vessels in areas with hemorrhage and/or necrosis. Otherwise, neither evident circumscribed lumina without endothelial CD34 staining nor erythrocytes outside the blood vasculature were localized. 

All xenografts (including liver and peritoneal metastasis) of human melanoma cell lines Mel57 and Mel57-VEGF contained similar arc, loops and network patterns, ^{10 18} as observed in the 14 uveal melanomas (Fig. 1 and prior paragraph). Occasionally, loops surrounded smaller groups of tumor cells and were more abundantly present than in uveal melanoma. 

Xenografts of the less malignant 1F6 cell line contained an organized structure of extracellular matrix arranged in parallel patterns between groups of tumor cells. ^{10 18} As in uveal melanoma, erythrocytes were exclusively present in lumina surrounded by ASD-13-positive blood vessel endothelium (Figs. 2C 2D) . Especially in the Mel57-VEGF xenografts, necrosis, and hemorrhage were observed in the tumor center. Note that necrosis in human uveal melanoma was hardly observed. Other evident morphologic differences with Mel57 wild-type xenografts were not observed. 

Tumor Perfusion in Xenografts

To study tumor perfusion, we injected FITC-BSA as a tracer intravenously into nude mice carrying melanoma xenografts. The distribution of endothelium and laminin was identical with that observed by endothelial immunostaining and by PAS and Azan histochemistry (Figs. 1 2A 2B 2C 2D 2E 2F and 3C 3D ). Dots of laminin were also observed inside tumor cell nests. In addition, laminin was also present in PAS and Azan-positive patterns in the primary uveal melanoma (Figs. 3A 3B) . In more detail, the amount of laminin related positively with the diameter of the depositions. Based on these findings, we were able to compare the distribution of tracer in relation to the extracellular matrix patterns with fluorescence and confocal microscopy. 

The tracer was located exclusively within the vascular lumina in the 1F6 xenografts (not shown). However, in Mel57 and Mel57-VEGF xenografts, tracer was present both in and outside blood vessels in different tumor areas. Considerable amounts of tracer were found in the extracellular matrix present between tumor cell nests (arc, loops and network patterns ^{10 18} ) as shown by laminin staining (Fig. 2E 2F 2G 2H 2I) . In xenografts of Mel57-VEGF, fluorescence intensity was slightly increased compared with xenografts of Mel57, and tracer also penetrated a larger area around the blood vasculature. In the extracellular matrix, laminin deposition enclosed tumor cell nodules (Fig. 2F) . Often, some space was present between layers of laminin bordering tumor cells (Figs. 2H 2I) . Additional evaluation of 20-μm cryosections by confocal microscopy also failed to produce evidence for a network of open channels in these laminin-containing structures. Instead, regular patterns of matrix deposition were observed, similar to those shown in Figures 1 2 and 3 . Thus, these results indicate that laminin was distributed as septa surrounding nodules of tumor cells and that, where two laminin-enveloped tumor nests lie adjacent, extra space may be present between the laminin layers. 

Examination of Extracellular Matrix Patterns in Xenografts by Immunoelectron Microscopy

Despite the presence of tracer in the extracellular matrix patterns outside blood vessels, we did not observe lumina that contained erythrocytes and showed no endothelial staining at such sites. The nature of the spaces between the sheets of laminin was unclear. Therefore, four xenograft lesions (two Mel57 and two Mel57-VEGF) were processed for electron microscopy and evaluated. In agreement with our light and immunofluorescence microscopy findings, connective tissue containing large amounts of collagen, often surrounded by laminin depositions, was located between groups of tumor cells (Figs. 4B 4C 4D 4E) in which, at certain locations, lumina containing erythrocytes were present (Figs. 4F 4G) . Although it is known that tumor cells, pericytes, or stromal cells cannot always be clearly differentiated from endothelial cells lining lumina in tumors by immunoelectron microscopy, ²⁹ additional immunoelectron microscopy using the 9F1 mAb (Fig. 4F) indicated that lumina containing erythrocytes were lined by endothelial cells. These lumen-lining cells were morphologically different from neighboring tumor cells and had endothelial characteristics, such as Weibel-Palade bodies and tight junctions (Figs. 4G 4H) . Lumina were always surrounded by rims of (probably endothelial) cytoplasm, which were separated from tumor cells by a basal membrane and often also by extracellular matrix (Fig. 4G) . On the basis of these observations, these lumina were classified as putative pre-existent or extracellular matrix-associated new blood vessels. Only in cases of hemorrhage or necrosis were erythrocytes seen outside these lumina. No other true lumina besides those described were detected in the connective tissue present between the nests of tumor cells. No evident fenestrae between endothelial cells or transendothelial holes were observed. 

Different cell types besides tumor cells were associated with the extracellular matrix patterns. Morphologic analysis suggested the presence of macrophages (Fig. 4B) , and endothelial cells (Fig. 4F) . The presence of macrophages was confirmed by additional immunostaining. Macrophages were mostly located between the laminin-positive sheets surrounding tumor cell nests (Fig. 5) and in areas of necrosis. In addition, melanin-laden macrophages were associated with laminin-positive sheets in the uveal melanomas (Figs. 3A 3B ) confirming earlier observations. ³⁰  

Discussion

Maniotis et al. ²⁰ observed endothelial characteristics of tumor cells lining channels, and their observations were supported by the recent report from Bittner et al. ³¹ showing that mRNA transcripts of CD34 and other endothelial markers are expressed in melanoma cells. In this respect, the question arises whether the CD34-positive structures are blood vessels, mosaic vessels (i.e., lined by endothelial and tumor cells ²¹ ), channels lined by tumor cells, extracellular matrix, or a combination of these (for review see Ref. ³² ). Although there are limitations to identifying endothelial cells in histologic sections ³² and we did not double stain melanoma cells for melanoma and endothelial markers, no evident positivity of any included endothelial marker (Table 1) was observed on melanoma cells by light microscopy. Therefore, the identity of nonluminal, CD34-positive cells is uncertain, but they may represent vascular channels lined by tumor cells expressing CD34. ^{20 31} Therefore, using CD34 as a marker for endothelial cells in primary uveal melanoma, we were not able to confirm or rule out the existence of vascular mimicry ²⁰ in primary uveal melanoma. 

Except in the less malignant 1F6 xenografts, intravenously injected tracer was found outside the blood vasculature in arc and loop network matrix ^{10 18} in patterns similar to those observed during angiography by Maniotis et al. ²⁰ and others. ²² These findings indicate that besides blood vessels, spaces accessible to fluids must be present in the extracellular matrix whenever these specific matrix patterns are formed. Azan staining and electron microscopy, however, failed to detect endothelium-free channels or open spaces, whereas laminin and macrophage staining indicated the existence of a compartment in which tracer and macrophages were both present. Additional evaluation of laminin deposition in sectional scanning by confocal microscopy demonstrated the presence of regular matrix patterns, deposited closely around nests of tumor cells and resembling curved sheets. Laminin is just one of the components of the extracellular matrix. The spaces visible in laminin-immunostaining contain other extracellular matrix components, such as collagen (which is clearly shown in Fig. 4 ). The conduction of the intravenously administered tracer is therefore explained by spaces in the extracellular matrix (through the network of collagen, laminin, and other components), which are not visible in electron microscopy and were not observed in this study to contain blood cells such as erythrocytes. 

Our observations indicate that extracellular matrix was deposited in a form resembling curved sheets around nests or nodules of tumor cells. In these structures, blood vessels were embedded. The extracellular matrix pattern was present between spheres of tumor cells, probably deposited as envelopes, resulting in the arc, loops and network patterns detected in two-dimensional analysis. If two spheres of tumor cells lie apposed, either one layer of matrix may be formed or two layers can be deposited separately. Between these two different septa, matrix material is localized consisting of different components (including laminin and collagen ³³ ) in which limited flow may occur (as demonstrated by the appearance of the tracer), and where different types of cells (including macrophages) are present. On the basis of these considerations, we postulate that, in fact, the extracellular matrix arc, loops and network patterns represent a meshwork. 

The presence of tracer outside the vasculature in the extracellular matrix pattern indicates a close relationship with blood vessels. This relationship is confirmed by the enhanced leakage of tracer in the extracellular matrix in xenografts of the VEGF-transfected cell line, which is due to vascular hyperpermeability. ²² Tracer distribution may be simply explained by two hypotheses. On the one hand, it is possible that spaces in the extracellular matrix generated by tumor cells ³⁴ connect to blood vessels and are organized as channels, which has been suggested. ^{20 35} In that case, nonmigratory blood cells such as erythrocytes would be localized in this tissue outside the vasculature, which we did not observe in the xenografted melanoma. On the other hand, interstitial spaces present in the extracellular matrix tissue (i.e., meshwork) may allow limited flow of fluid by enhanced leakage or permeation through the endothelium of the blood vessels and into the surrounding tissue. Considering our results, the latter hypothesis is the more likely, at least in the animal model, and the extracellular matrix arc, loops and network patterns may thus represent a fluid-conducting meshwork. 

In general, fluid movement across vessel walls is governed by differences between blood pressure and local interstitial fluid pressure (IFP). Although IFP is elevated and close to microvascular pressure in different types of solid tumors, including melanoma, ^{36 37 38 39} our and other studies ^{20 22} indicate that fluid can leave the blood stream and form an exudate (without nonmigratory blood cells) that moves across the microvascular arterial wall into the extracellular matrix meshwork tissue. The fluid’s movement is induced by the local higher arterial pressure compared with the IFP. The exudate may be conducted by the meshwork toward the venous microvasculature, where pressure is likely to be lower than the elevated IFP. This difference in pressure would allow re-entry of the interstitial fluid into veins. Alternatively, interstitial fluid may leak into surrounding pre-existent tissue (which has an IFP of approximately 0 mm Hg ^{37 39} ) or vitreous body (oozing out) in case of uveal melanoma. High IFP is maintained by the absence of lymphatics ⁴⁰ and by the limited drainage capacity of the fibrous vascular network. Comparison of the kinetics and composition of the fluid stream and composition of the extracellular matrix to that in normal tissues could yield insight into the role of this phenomenon in tumor biology. The subcutaneous space normally is essentially avascular, however, and cannot be taken as a standard tissue. Nevertheless, our previously reported data have indicate that tracer distribution into subcutaneous tumor tissue is relatively slow, requiring up to 45 minutes to occur. ²²  

The extracellular matrix also harbored different types of cells, including macrophages, that were located between sheets of extracellular matrix surrounding nests of tumor cells. This observation supports the idea that migratory cells leave tumor vessels and invade by adhering to matrix components present in stromal sheets, thereby migrating into the spaces between them. This hypothetical mechanism is significant, because tumor-associated inflammatory cells play an evident role in tumor survival. For example, macrophages are associated both with angiogenesis and a poor prognosis in different tumor types, including melanoma. ^{30 41 42 43} If macrophages are able to invade the tumor site by using the extracellular matrix, it is also possible that tumor cells can enter the blood stream through the same route. Furthermore, by formation of extracellular matrix networks, uveal melanoma could acquire an alternative system to drain excess tissue fluids, as a substitute for a lymphatic system that is absent in this type of tumor. ⁴⁰ Also, the presence of the tracer in the extracellular matrix patterns indicates that nutritional components of the blood can reach tumor cells located a long distance from blood vessels. Indeed, the arc, loops and network patterns are abundantly present in those areas in which blood vessel density is low. ²⁰ Most data in the present study were obtained in an animal xenograft model. In this model, however, matrix patterns are formed that have evident morphologic similarities with the matrix patterns present in primary uveal melanoma. Therefore, these considerations may support and explain why uveal and cutaneous melanoma containing the arcs, loops, and network extracellular matrix patterns are associated with a poor prognosis. ^{9 10}  

Close parallels, such as Azan and PAS positivity and the association of laminin and macrophages with these patterns, strongly suggest the presence of a fluid-conducting meshwork in primary uveal melanoma. In this respect, the existence of vascular mimicry ^{20 35} is of substantial importance. Although we did not observe convincing evidence of the presence of vascular channels in the xenografted melanoma on the one hand and we cannot rule out their presence in uveal melanoma on the other hand, it is possible that these blood-conducting channels and a fluid-conducting meshwork are both present in primary uveal melanoma. Furthermore, because of the doubt raised about the nature of the PAS-positive patterns ^{9 10 18 19} and the existence of vascular mimicry, ^{20 35 44} the present study may have further shown the true identity of the patterns. Additional study, however, is needed to elucidate the exact nature of extracellular matrix patterns in uveal melanoma. 

In conclusion, we propose that both in aggressive uveal and cutaneous melanoma, extravascular spaces are present besides the normal blood vasculature. These spaces are located between bordering curved septa of extracellular matrix, surrounding nests or spheres of tumor cells representing a fluid-conducting meshwork and contain different types of cells, including macrophages. There are strong indications that the arcs and loops forming the extracellular matrix are involved in inflammatory cell invasion, nutrition of tumor cells, regulation of local tissue fluid flow and, possibly, metastasis. Indeed, a fluid-conducting meshwork may represent an alternative to a lymphatic system in uveal melanoma. 

 Supported by Grant 98-1816 from the Dutch Cancer Society.

 Submitted for publication November 21, 2000; revised November 19, 2001; accepted November 28, 2001.

 Commercial relationships policy: N.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Endothelial Markers Used for Immunohistochemistry

Antibody	Antigen	Type of Antibody	Tissue Examined
JC70A 45 46	CD31	mAb	Uveal melanoma
QBEnd/10 46 47	CD34	mAb	Uveal melanoma
anti-vWF 18	vWF (FVIIIrA)	pAbs	Uveal melanoma
			Xenografts
BNH9 46	H and Y antigens	mAb	Uveal melanoma
Anti-thrombomodulin 48	Thrombomodulin	mAb	Uveal melanoma
UEA-1 18 *	—	pAbs	Uveal melanoma
MEC 7.46 49	CD31	mAb	Xenografts
9F1 24	Unknown	mAb	Xenografts
ASD-13 50	Unknown	mAb	Xenografts

*  Lectin stained by using peroxidase-linked rabbit anti-UEA-1 PAbs.

Figure 1.

 

Evaluation of extracellular matrix patterns by PAS (A, C) and Azan histochemistry (B, D) in human primary uveal melanoma (A, B) and xenografts of cutaneous melanoma (C, D) in serial paraffin-embedded sections. Bar, 0.1 mm.

Figure 2.

 

Evaluation of extracellular matrix and blood vessels in human primary melanoma (A, B) and xenografts of cutaneous melanoma (C, D) in serial paraffin-embedded sections. Extracellular matrix is stained blue, nuclei of tumor cells red, and erythrocytes an intense red by Azan histochemistry (A, C). Arcs and back-to-back loop network patterns were clearly visualized. Blood vessels were detected by CD34 (B) and ASD-13 (D) immunohistochemistry. Parts of the matrix patterns were also CD34 positive (arrow, B). Within the matrix patterns, no lumina were observed, except where blood vessels were present. Erythrocytes were exclusively observed in association with endothelial immunostaining. (E–I) Distribution of intravenously administered tracer (FITC-BSA) in cryosections of xenografts of Mel57 cells was evaluated. In serial sections, the distribution of endothelium (E) and laminin (F) was visualized by immunostaining and demonstrated a pattern similar to that observed by Azan and endothelial immunohistochemistry in paraffin-embedded tissue (see also Figs. 1 3 ). In Mel57 xenografts, the tracer (H, green) was distributed inside and outside the blood vasculature stained red by the endothelial marker mAb Mec 7.46 (G), and along the extracellular matrix, stained red by laminin antibodies (I). (I, arrow) Space between the laminin sheets surrounding coherent groups of tumor cells. CD34-, ASD-13-, Mec 7.46- (E), and laminin- (F) stained sections were counterstained with Harris hematoxylin. Bar, 0.1 mm.

Figure 3.

 

Evaluation of extracellular matrix patterns by PAS (A), laminin (B, D) and Azan (C) immunohistochemistry in human primary uveal melanoma (A, B) and xenografts of cutaneous melanoma (C, D) in serial sections. Melanin-laden macrophages (melanophages white arrow) are located along the PAS- and laminin-positive extracellular matrix patterns. Bar, 0.1 mm.

Figure 4.

 

Immunoelectron microscopy analysis of Mel57 (B–F) and Mel57-VEGF (G, H) xenograft lesions. (A) Low-power overview (toluidin-blue-stained 1-μm section) of Mel57 xenograft tissue. A network was defined as the presence of at least three loops. ¹⁰ (B) Tumor cells (T) were separated by a layer of matrix (M). The central part of (B) is shown in detail in (C). Matrix consisting mainly of collagen fibers was present, whereas no evident lumina were visible. Immunoelectron microscopy shows the deposition of laminin in basal membranes of blood vessels (D, arrowhead; L, lumen). These layers continued into the extracellular matrix sheets (D, arrow), surrounding bundles of collagen (E, C). The box in (D) is magnified in (E). Different types of stromal cells (B, SC), including macrophages (B, MP) were associated with these layers. The nature of other stromal cells was identified by 9F1 immunoelectron microscopy (F, arrow). 9F1 immunostaining demonstrated that lumina present in the extracellular matrix patterns represented blood vessels lined by endothelium, occasionally containing erythrocytes (E). Neither evident lumina nor erythrocytes were found in the extracellular matrix besides those in blood vessels. Further support for the endothelial nature of the lumen-lining cells is provided in (G) and (H): The lumen was lined by a rim of cytoplasm (G, C) which was separated from the tumor cells (G, T) by a basal membrane (G, B), and, often, matrix depositions (G, M) were present between the tumor and endothelial cells. The right luminal wall (G, boxed area) is shown in more detail in (H). Typical endothelial characteristics (i.e., tight junctions; H, arrows), Weibel-Palade bodies (H, solid arrowhead), and a basal membrane (H, open arrowhead) are present. Bars: (A) 0.1 mm; (B–H) 1 μm.

Figure 5.

 

Immunohistochemical evaluation of the presence of macrophages in cryosections of a Mel57-VEGF xenograft. Macrophages, detected by FA/11 immunostaining, were abundantly present and closely associated with the extracellular matrix sheets. (B) Magnification of the boxed area in (A) shows that macrophages were located between the laminin-positive sheets surrounding tumor cell nests in (C). (C, ★) A blood vessel lumen which was detectable by anti-mouse CD31 immunohistochemistry (not shown). All sections were counterstained with Harris’ hematoxylin. Bars, (A) 0.1 mm; (B, C) 50 μm.

The authors thank Henry B. P. M. Dijkman for assistance with confocal microscopy analysis and Geert Poelen and Josianne van Vliet for assistance in performing the animal study.