291

Angiogenesis 3: 291±293, 1999.

Ó 2000 Kluwer Academic Publishers. Printed in the Netherlands. Brief communication

A novel method of visualizing vessels in human tumor biopsies Claes Lenander1;2 & Lars Holmgren1 1

Department of Surgery, Ersta Hospital, Stockholm, Sweden; 2 Microbiology and Tumor Biology Center, Karolinska Institutet, Stockholm, Sweden Received 6 September 1999; accepted in revised form 17 April 2000

Today, there exists convincing experimental evidence that tumor growth is depending on induction of de novo blood vessel formation [1, 2]. The morphology of the vascular networks induced by the tumor often displays increased tortuosity, branching and heterogenous vessel diameters [3]. This may in part be explained by the continuous release of angiogenic factors and the lower degree of di€erentiation of vessels found in tumors. Studies of normal and pathological vascularization are often hampered by the diculties in visualizing the three-dimensional structure of the vessels. The current standard method for the evaluation of vascularization of normal and tumor tissues are 5 lm formalin-®xed and paran-embedded sections stained with an endothelialspeci®c marker [4, 5] (Figures 1A and B). This may provide valuable information about the density of vessels but adds very little information regarding the three-dimensional structure of the vessels. Here we describe an easy and fast method to visualize vessels in frozen biopsies from clinical specimen using computerized deconvolution. Biopsies from colon tissue were immediately collected from the surgical specimens after resection and stored at ÿ70  C, until analysis. 90 lm thick sections were cut in a cryostat at ÿ20  C and mounted on positively charged glass slides (Super Frost Plus, Mentzel GlaÈser, Germany). Before immunohistochemical staining, the specimens were ®xed in 3.7% formaldehyde in phosphate bu€ered saline (PBS) at room temperature (RT) or at ÿ20  C in acetone for 10 minutes. In our experience, both ®xation methods worked equally well. After washing in PBS, the tissue sections were permeabilized with 0.1% Triton X-100 (Sigma-Aldrich Sweden AB, Sweden) for 5 min, brie¯y rinsed in PBS and blocked with 10% normal horse serum for 15 min in order to decrease unspeci®c binding. The samples were subsequently incubated at RT for 90 min with a ¯uorescein isothiocyanate (FITC)-conjugated CD31 antibody (5 lg/ml) (Sigma-Aldrich) or the FITC-labeled lectin Ulex Europaeus agglutinin I (5 lg/ml) (Vector Laboratories, Burlingame Ca) mixed with Hoechst 33258 (Sigma-Aldrich) for nuclear counter staining. Finally,

Correspondence to: Claes Lenander, MTC, Ersta sjukhus, Box 4622, S-116 91 Stockholm, Sweden. Tel: +46-8-7146100; Fax: +46-87146665; E-mail: [email protected]

the sections were washed in PBS and mounted with coverslips using an anti-fading reagents (DAKO Carpinteria Ca). Parallel incubations with non-speci®c IgG served as negative controls. Positive signals were analyzed with a computerized ¯uorescence microscope (Leica DM RXA; Leica Microsystems AB, Sweden), using a 10 or a 25´ lens (numerical apertures; 0.3 and 0.75, respectively), connected to a video camera (Hamamatsu Digital Video camera C4742-95; Hamamatsu Photonics K.K. Japan using Snapper DIG-16 video capture card). The system was controlled by the Openlab software image analysis system (Improvision, Image Processing & Vision, Ltd, GB). Images of the

Figure 1. Immunostaining of endothelial cells with an anti CD-31 monoclonal antibody on 5 lm paran sections. (A) Normal colon mucosa, (B) colon adenocarcinoma. Arrowheads = microvessels. Bar, 75 lm.

292

C. Lenander & L. Holmgren

Figure 2. Formation of a composite image from a thick sections by iterative deconvolution. Images are collected in the z-axis of the specimen. Signals that were out of focus were removed by deconvolution from the 2 closest focal planes. All deconvoluted layers are then stacked upon each other to generate a composite image.

Figure 3. Immuno¯uorescence imaging of the vascular architecture of normal and tumor tissue from human colon. Frozen sections (90 lm thick) were stained with an endothelial marker and images were collected and analyzed as described in the text. (A) Normal colonic mucosa where the vessels are arranged in a parallel pattern with uniform diameter. (B) Colon adenocarcinoma showing long tortuous capillaries (arrow). (C) Colon adenocarcinoma with angiogenic `hot spots'. (D) Colon adenocarcinoma. The vasculature have completely lost their regular architecture with an increased vascular density, random branching, and great variation of diameter. (E±F) Normal colonic mucosa (same as A). E, shows a high magni®cation deconvoluted image of small capillaries made from one in-focus layer from the thick tissue and F, shows the composite image made from all deconvoluted layers. Note the appearance of the small microvessels not shown in E (arrows). A±D, bar, 100 lm and E±F, bar, 50 lm. A, E and F are stained with CD31-FITC conjugated antibody and B, C and D with the UEAI-FITC lectin.

Vascular imaging of human biopsies vessels were collected by making optical sections in zaxis every 2 lm (10´ lens) or 0:5 lm (25´ lens) through the whole sample (20±40 pictures/specimen). The Openlab software program operates as an `opto-digital microtome'. Its major function is to produce series of digitized, two-dimensional images obtained from thin optical sections and the out-of-focus haze was subsequently removed using the deconvolution program. Fluorescent signals that were out of focus were removed by deconvolution from the two closest focal planes [6, 7]. All deconvoluted focal layers were then stacked to generate one composite image (Figure 2). The whole process was performed in 10±30 min using a 233-MHz computer (Power Macintosh G3). Finally, the original digital images were saved in a tagged image ®le format (TIFF) and imported into Adobe Photoshop, version 4.0 (Adobe Systems, Mountain View, California). The resulting images of the endothelium of normal and tumor biopsies are shown in Figure 3. In the mucosa of normal colon, capillaries were uniform in their diameter and showed a regular pattern with mostly parallel vessels in cross-sections, and regular honeycomb-like plexus around the mucosal glands (Figure 3A). Analysis of the vasculature of adenocarcinomas revealed striking morphological di€erences of the vessels. Among the features that could be distinguished were increased tortousity of vessels (Figure 3B), the presence of angiogenic hot spots (Figure 3C), and irregularity of vessel diameter and branching (Figure 3D). Each vessel could be followed through the focal planes at 250-fold magni®cation making it potentially possible to analyze protein expression in di€erent structures, e.g. branches or sprouts, of capillaries (Figures 3E and F). Several independent studies have shown that increased counts of microvessels are independent predictors of reduced survival [4, 8]. However, the reproducibility of this method relies on the experience of the investigator's ability to identify and quantify the angiogenic `hot spots' of the specimen. The use of computerized deconvolution of thick specimens makes it possible to relatively simple and rapid get images of vessel morphology in frozen tissue. The resulting images are fully comparable to scanning electron micrograph of vascular cast preparations made from normal tissue and adenocarcinomas from human colon

293 [9]. However, this technique is very labor intensive and preparation of cast corrosions destroy the tissue surrounding the vessels (i.e., the tumor). Our method of vascular imaging of thick sections facilitates the identi®cation of three-dimensional vascular structures in tumor biopsies and may therefore be helpful in the identi®cation of angiogenic hot spots in tumors. This method may also be used in experimental studies to study e.g. tumor vessel morphology in response to antiangiogenic therapy.

Acknowledgements This work was supported by the Swedish Cancer Society and the AÊke Wiberg Foundation.

References 1. Bouck N, Stellmach V, Hsu SC. How tumors become angiogenic. Adv Cancer Res 1996; 69: 135±74. 2. Hanahan D, Folkman J. Patterns and emerging mechanisms of the angiogenic switch during tumorigenesis. Cell 1996; 86: 353±64. 3. Hah-Yukich AA, Nelson AC. Characterization of solid tumor microvasculature: A three-dimensional analysis using the polymer casting technique. Lab Invest 1988; 58: 236±44. 4. Weidner N, Semple JP, Welch WR, Folkman J. Tumor angiogenesis and metastasis ± correlation in invasive breast carcinoma. N Engl J Med 1991; 324: 1±8. 5. Vermeulen PB, Gasparini G, Fox SB et al. Quanti®cation of angiogenesis in solid human tumours: An international consensus on the methodology and criteria of evaluation. Eur J Cancer 1996; 14: 2474±84. 6. Spector DL, Goldman RD, Leinwand LA. Confocal microscopy and deconvolution techniques. In Spector DL, Goldman RD, Leinwand LA (eds): Cells: A Laboratory Manual, Volume 2: Light Microscopy and Cell Structure. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press 1998; 96.1±96.23. 7. Atherton TJ, Wilson S, Morrey RI, Waterfall A. Digital confocal imaging. Research Report RR333, Department of Computer Science, University of Warwick, Coventry 1997. 8. Vermulen PB, Verhoeven D, Fierens H et al. Microvessel quanti®cation in primary colorectal carcinoma: An immunohistochemical study. Br J Cancer 1995; 71: 340±43. 9. Skinner SA, Frydman GM, O'Brien PE. Microvascular structure of benign and malignant tumors of the colon in humans. Digestive Dis Sci 1995; 40: 373±84.