| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Cell and Tumor Biology |
5ß1 Integrin Overexpressed on the Luminal Surface of Tumor Blood Vessels
1 Cardiovascular Research Institute, Comprehensive Cancer Center, and Department of Anatomy, University of California, San Francisco, California and 2 Protein Design Labs, Inc., Fremont, California
Requests for reprints: Donald M. McDonald, Cardiovascular Research Institute, S1363, University of California, 513 Parnassus Avenue, San Francisco, CA 94143-0130. Phone: 415-476-2118; Fax: 415-476-4845; E-mail: dmcd{at}itsa.ucsf.edu.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
5ß1 is overexpressed on endothelial cells of tumor vessels and is uniformly and rapidly accessible to antibodies in the bloodstream. Here, we determined whether antibodies rapidly gain access to integrin overexpressed on the abluminal (basolateral) surface of endothelial cells through vascular leakiness or whether the rapid accessibility results instead because the integrin is overexpressed on the luminal (apical) surface of endothelial cells due to loss of cell polarity. Using tumors in RIP-Tag2 transgenic mice as a model, we first compared the binding pattern of intravascular anti-5ß1 integrin antibody with the leakage pattern of nonspecific IgG. The distributions did not match: anti-5ß1 integrin antibody uniformly labeled the tumor vasculature, but IgG was located in patchy sites of leakage. We next injected an antibody to fibrinogen/fibrin, which resulted in patchy labeling of tumors that matched the leakage of IgG and the overall distribution of fibrin in tumors. Similarly, injected antibodies to the basement membrane protein fibronectin, a ligand of 5ß1 integrin, or type IV collagen produced patchy sites of leakage instead of uniform labeling of vascular basement membrane. Differences in the kinetics of labeling, which for 5ß1 integrin antibody was near maximal by 10 minutes but for the other antibodies gradually increased over 6 hours, indicated differences in accessibility of their respective targets. Isosurface rendering of confocal microscopic images was consistent with antibody binding to 5ß1 integrin on the luminal surface of endothelial cells. Together, these findings indicate that the rapid accessibility of 5ß1 integrin in RIP-Tag2 tumors results from overexpression of the integrin on the luminal surface of tumor vessels.
Key Words: angiogenesis • immunohistochemistry • pancreas • RIP-Tag2 mice • vascular targeting
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Considering these issues, tumor vasculature is an attractive target for therapeutic agents because of its immediate accessibility and essential role in supporting tumor growth (5). Selective targeting of tumor vasculature is feasible because of the abnormalities of tumor vessels and the presence of molecules that are overexpressed on endothelial cells. Several integrins, including vß3, vß5, and 5ß1, are among the potential targets (6, 7).
Integrin 5ß1 (fibronectin receptor, VLA-5) plays a key role in anchoring the abluminal surface of endothelial cells to their basement membrane, is overexpressed on tumor vessels (7, 8), and is rapidly accessible from the bloodstream.3 After i.v. injection, antibodies rapidly label 5ß1 integrin on tumor vessels.3
These features of 5ß1 integrin raise several questions. If 5ß1 integrin is overexpressed on the abluminal surface of endothelial cells, how can it be rapidly accessible to antibodies in the bloodstream? Because the endothelial barrier limits access to the abluminal surface of endothelial cells, can the rapid accessibility of integrins be explained by the defective barrier function of tumor vessels? Or instead, is 5ß1 integrin on tumor vessels immediately accessible to circulating antibodies because of overexpression on the luminal surface of endothelial cells?
Indeed, there may be two populations of integrins on endothelial cells of tumor vessels, one on the abluminal (basolateral) surface that is accessible from the bloodstream only by leakage across the defective endothelial barrier and the other rapidly accessible because of its location on the luminal (apical) surface. Consistent with the latter possibility, some integrins are expressed on the luminal surface of endothelial cells in vitro (9), but this has not been shown for 5ß1 integrin in tumor vessels.
The goal of the present study was to determine which of two alternatives best explains the rapid accessibility of 5ß1 integrin on tumor vessels: (a) integrins are overexpressed on the luminal surface of endothelial cells, as well as on the abluminal surface, as a reflection of loss of cell polarity, or (b) integrins are overexpressed only on the abluminal surface of tumor vascular endothelial cells but are accessible to antibodies in the bloodstream because they can leak across the defective endothelial barrier of tumor vessels.
To address the issue, we used antibodies to compare the accessibility of 5ß1 integrin on tumor vessels to that of various other targets on one side or the other of the endothelial barrier. In particular, we determined (i) whether the labeling of 5ß1 integrin matches the distribution of nonspecific IgG extravasated from tumor vessels; (ii) whether the labeling resembles the pattern of leakage of antibodies to fibrin or the historical record of leakage reflected by the distribution of fibrin in tumors; (iii) whether the labeling matches the distribution of antibodies to fibronectin, a principal ligand of 5ß1 integrin, or type IV collagen, an important component of vascular basement membrane on the abluminal surface of endothelial cells; (iv) whether 5ß1 integrin labeling resembles labeling by antibodies to the adhesion molecule CD31 (platelet/endothelial cell adhesion molecule 1) accessible on the luminal surface of endothelial cells; (v) whether antibodies to 5ß1 integrin accumulate in tumors over time more rapidly than other antibodies; and (vi) whether isosurface rendering of confocal microscopic images can reveal whether intravascular anti-5ß1 integrin antibody binds to the luminal surface of tumor vessels.
Our approach, using RIP-Tag2 transgenic mice as a model, was to inject antibodies i.v., allow them to circulate for 10 minutes to 24 hours, and identify their distribution in relation to the vasculature and access to extravascular targets in multistage, multifocal pancreatic islet tumors by using quantitative immunofluorescence and confocal microscopy. Together, the results fit best with the luminal distribution of the integrin as the explanation for the rapid, uniform accumulation of anti-5ß1 integrin antibody in tumors.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Intravenous injection of antibodies and vascular perfusion. After induction of anesthesia with ketamine (87 mg/kg) and xylazine (13 mg/kg) injected i.m., 50 µg of antibody or nonspecific IgG, diluted to 125 µL final volume with 0.9% NaCl (Abbott Laboratories, North Chicago, IL), were injected via a tail vein. A dose of 50 µg was chosen after preliminary studies showed that injection of doses of anti-integrin 5ß1 or anti-fibronectin antibody ranging from 25 to 100 µg gave generally similar results. Anti-type IV collagen antibody, which had an undefined concentration in whole plasma, was diluted 1:25, 1:500, or 1:1,000 (1:25 used for most experiments) and injected in a volume of 125 µL.
One or two of six different antibodies were injected. Integrin 5ß1 was localized by injection of monoclonal rat anti-mouse integrin 5 subunit (anti-CD49e, clone 5H10-27, BD PharMingen, San Diego, CA). This well-characterized rat anti-mouse monoclonal antibody was originally known as clone MRF5 (5H10), http://cbr.med.harvard.edu/investigators/springer/lab/, and is now available commercially (http://www.bdbiosciences.com). Because 5 integrin subunits pair solely with ß1 integrin subunits, anti-5 integrin antibody was treated as if it binds exclusively to 5ß1 integrin and is designated as such.
Alternatively, we injected nonspecific IgG (ChromePure rabbit IgG, Jackson ImmunoResearch, West Grove, PA); polyclonal rabbit anti-human fibrinogen/fibrin (DAKO, Carpinteria, CA), hereafter designated anti-fibrin antibody for simplicity; polyclonal rabbit anti-human fibronectin (Sigma, St. Louis, MO); polyclonal rabbit anti-mouse type IV collagen (LB-1403, Cosmo Bio, Tokyo, Japan); or monoclonal hamster anti-mouse CD31 (Chemicon, Temecula, CA). Some mice received a combination of type IV collagen and CD31 antibodies.
Antibodies were allowed to circulate for 10 minutes or 1, 6, or 24 hours, and the tissues were fixed by vascular perfusion. The chest was opened rapidly, and the vasculature perfused for 3 minutes at a pressure of 120 mm Hg with fixative [4% paraformaldehyde in 10 mmol/L PBS (pH 7.4) at room temperature] from an 18-gauge cannula inserted into the aorta via an incision in the left ventricle. The right atrium was incised to provide an exit for the fixative. After the perfusion, tissues were removed and stored in fixative for 1 hour at 4°C.
Immunohistochemistry. Specimens were rinsed with PBS, infiltrated with 30% sucrose in PBS overnight at 4°C, embedded in Tissue-Tek O.C.T. (OCT, Sakura Finetek, Torrance, CA), and frozen at –80°C. Tissue sections, cut with a cryostat (Microm International GmbH, Walldorf, Germany) at a thickness of 80 µm and dried on glass slides (Superfrost Plus, Fisher Scientific, Pittsburgh, PA), were rinsed of OCT and then incubated in 5% normal goat serum in PBS containing 1% Triton X-100 for 1 hour.
To examine the tissue distribution of antibodies after i.v. injection, sections on slides were incubated for 12 to 15 hours in hamster anti-mouse CD31 antibody to label endothelial cells. This step was omitted on tumor sections from mice injected with anti-CD31 antibody. After several rinses with PBS/0.3% Triton X-100, sections were further incubated for 6 hours with a combination of two fluorescently labeled, species-specific secondary antibodies, one directed against the injected antibody (anti-rat or anti-rabbit) and the other against the CD31 antibody (anti-hamster). Secondary antibodies labeled with FITC, Cy3, or Cy5 (Jackson ImmunoResearch, West Grove, PA) were used at a dilution of 1:400 in PBS in 0.3% Triton X-100. All incubations were at room temperature. Sections were rinsed in PBS and mounted in Vectashield (Vector Laboratories, Burlingame, CA) for examination by fluorescence and confocal microscopy.
For conventional immunohistochemistry, tissue sections on slides were incubated for 12 to 15 hours in a mixture of the CD31 antibody and a second primary antibody. Primary antibodies for localizing 5ß1 integrin, fibrinogen/fibrin, fibronectin, type IV collagen, and CD31 were the same as those used for i.v. injection. All primary antibodies were diluted 1:500 with PBS containing 5% normal goat serum and 1% Triton X-100, except for anti-type IV collagen, which was diluted 1:10,000, and anti-CD31, which was diluted 1:1,000. Thereafter, sections were washed, incubated in secondary antibodies, and processed as for tissues from mice with injected antibodies.
Imaging, fluorescence measurements, and analysis. The relative amount of antibody in tumors following i.v. injection was quantified in digital images acquired with a low-light, externally cooled, three-chip CCD camera (480 x 640 pixel RGB color images, CoolCam, SciMeasure Analytical Systems, Atlanta, GA) on a Zeiss Axiophot fluorescence microscope (Thornwood, NY). Four tumors ranging in diameter from 400 to <1,000 µm were imaged in each RIP-Tag2 mouse (10x objective, 1.0x Optovar, imaged region 960 x 1280 µm). Each group consisted of at least four mice. To minimize the effect of variability associated with immunohistochemistry, tissue sections from at least one mouse from each antibody group and time point were processed together. Images were normalized during acquisition by standardizing the intensity of the background fluorescence to a reference image stained with secondary antibody but no primary antibody, giving in faint background fluorescence. Mean fluorescence intensity values in 80-µm tumor sections were calculated using ImageJ v.1.29 (http://rsb.info.nih.gov/ij/) as described.3 In brief, histograms of fluorescence intensities were obtained from 8-bit gray scale images of tumors (fluorescence intensity ranging from black = 0 to white = 255). Regions of non-tumor tissue were excluded. Values at each intensity for all tumors from each mouse were averaged, and the mean fluorescence intensity for that mouse was calculated. Overall group means were calculated from these values for individual mice. For comparisons among groups of different antibodies and time points, the group with maximal mean fluorescence intensity was assigned a value of 100, and the values for other groups were scaled accordingly and designated normalized fluorescence intensities. The distribution of antibodies in tissue sections was examined with a Zeiss LSM 510 laser scanning confocal microscope with krypton-argon and helium-neon lasers (1,024 x 1,024 pixel RGB color images).
Isosurface rendering. High-resolution confocal image stacks were reconstructed by isosurface rendering using Imaris software (version 3.3.2, Bitplane, AG, Zurich, Switzerland). Isosurface rendering is a computer-generated representation of a specified range of fluorescence intensities in a data set, thereby creating an artificial solid object that makes it possible to visualize the range of interest of a real volume object. Stacks of RGB color confocal images of immunohistochemically stained RIP-Tag2 tumors were imported into Imaris, and voxels with fluorescence intensities above a threshold were assigned to each color channel, relative to a maximum-intensity three-dimensional fluorescence projection. An isosurface was rendered from these voxels, and smoothed with a Gaussian filter, providing a three-dimensional reconstruction in which spatial resolution was conserved. The transparency of the red and green fluorescence channels was then adjusted individually to reveal the relationship between the channels. Renderings were examined after rotation in space, cross-sectioning through multiple planes, and movement of virtual light sources placed to enhance three-dimensional structural elements through shadow casting.
Statistical analysis. All quantitative data are presented as means ± SE of values from at least four mice per group. The significance of differences among groups were assessed by ANOVA followed by the Tukey-Kramer honestly significant difference post-test for multiple comparisons (JMP v.5.0.1.2, SAS Institute, Inc., Cary, NC). Ps < 0.05 were considered significant.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
5ß1 integrin antibody binding compared with IgG extravasation. At 10 minutes after injection into the bloodstream, anti-5ß1 integrin antibody uniformly labeled the vasculature of RIP-Tag2 tumors but was largely absent in the surrounding acinar pancreas (Fig. 1A and B). Most of the 5ß1 integrin immunofluorescence colocalized with CD31 staining of tumor vessels (Fig. 1C). In addition to blood vessel staining, irregular patches of extravascular fluorescence were scattered in the stroma (Fig. 1D). These patches were most numerous near the tumor center. Leakage of nonspecific IgG, injected as a marker of immunoglobulin extravasation, had a patchy distribution similar to but fainter than patches of 5ß1 integrin (Fig. 1E and F). Unlike anti-5ß1 integrin antibody, anti-IgG immunoreactivity did not outline tumor vessels. Measurements showed that the overall immunofluorescence for anti-5ß1 integrin antibody in tumors was nearly 4-fold that for nonspecific IgG (Fig. 2).
|
|
5ß1 integrin binding compared with labeling of fibrin. To test whether the patches of IgG staining provided a meaningful indication of leakage sites in RIP-Tag2 tumors, we delineated regions of fibrin deposition and the accessibility of these regions to antibodies in the bloodstream. At 10 minutes after injection, anti-fibrin antibody had a patchy distribution similar to but more intense than nonspecific IgG (Fig. 3A and B). Like IgG, fibrin patches were more abundant centrally than in the tumor periphery. Few were at the interface of tumors and normal pancreas. Fluorescence measurements confirmed that the accumulation of anti-fibrin antibody was greater than IgG but significantly less than anti-5ß1 integrin antibody (Fig. 2). The pattern of fibrin staining after injection of antibody matched the distribution of fibrin revealed by conventional immunohistochemistry (Fig. 3C and D). The similarity of the patterns is important because it showed that there was sufficient antibody to label tissue fibrin despite binding to plasma fibrinogen and that anti-fibrin antibody extravasated in the same scattered regions where fibrin accumulated naturally. The distribution of fibrin in tumors differed conspicuously from the uniform distribution of 5ß1 integrin on tumor vessels (Fig. 3E and F).
|
5ß1 integrin binding compared with labeling of fibronectin or type IV collagen. Conventional immunohistochemistry on tissue sections showed that fibronectin was closely associated with blood vessels in RIP-Tag2 tumors (Fig. 3G and H), as expected for a component of basement membrane (11). To determine whether fibronectin had the same accessibility as its receptor, 5ß1 integrin, we injected i.v. anti-fibronectin antibody. The results were surprising. Unlike the uniform binding of anti-5ß1 integrin antibody (Fig. 4A and B), anti-fibronectin antibody had a patchy distribution at 10 minutes after injection, resembling leakage of nonspecific IgG and anti-fibrin antibody (Fig. 4C and D). Overall, the mean fluorescence intensity of anti-fibronectin antibody in tumors was significantly less than that of antibody to 5ß1 integrin (Fig. 2).
|
Distribution of anti-CD31 antibody labeling. Antibodies to CD31 were injected as a positive control to label this adhesion molecule known to be uniformly expressed on blood vessels in RIP-Tag2 tumors (12) and to be accessible on the luminal surface of tumor vessels (13). The uniform labeling of tumor vessels at 10 minutes after i.v. injection of anti-CD31 (Fig. 4E) resembled the labeling after injection of anti-5ß1 integrin antibody (Fig. 4A); however, blood vessels of normal organs were also labeled. Injection of anti-CD31 and anti-type IV collagen antibodies together revealed the strikingly different distributions of labeling by these two antibodies in tumors (Fig. 4E and F).
Temporal changes in antibody accumulation in tumors. To determine the kinetics of antibody leakage and changes in accessibility and distribution over time, we compared the accumulation of nonspecific IgG and antibodies with 5ß1 integrin, fibrin, and fibronectin from 10 minutes to 24 hours after i.v. injection (Fig. 5). Fluorescence from anti-5ß1 integrin antibody was intense at 10 minutes, indicative of immediate accessibility of the integrin, and remained high through 24 hours (Fig. 5A and B). The amount and distribution at 24 hours roughly matched that found by conventional immunohistochemistry (Fig. 5C), which included staining of ducts in the exocrine pancreas. Unlike anti-5ß1 integrin antibody, nonspecific IgG immunofluorescence was stronger at 24 hours than at 10 minutes, indicative of gradual accumulation over time, and still had a patchy distribution at 24 hours (Fig. 5D-F). Similarly, anti-fibrin antibody accumulated over time, but the patches were crisper and the staining more intense than nonspecific IgG, presumably reflecting binding to endogenous fibrinogen and fibrin (Fig. 5G-I). Similarly, the fluorescence intensity of anti-fibronectin antibody increased between 10 minutes and 24 hours, as leakage gave the antibody access to extravascular binding sites; however, even then it lacked the uniform vascular pattern seen with anti-5ß1 integrin antibody or on-section staining for fibronectin (Fig. 5J-L).
|
5ß1 integrin antibody staining changed little from 10 minutes to 6 hours and tended to decrease slightly at 24 hours (Fig. 6). The time courses of accumulation of IgG and antibodies to fibrin and fibronectin were significantly different: their fluorescence was initially low, peaked at 6 hours, and then plateaued (Fig. 6).
|
5ß1 integrin in tumor vessels. The possibility of 5ß1 integrin expression on the luminal surface of tumor vessels was further examined by isosurface rendering with Imaris software, which plots fluorescence intensity in x, y, and z coordinate space while conserving spatial resolution. Isosurface rendering of a high resolution confocal image (Fig. 7A), consisting of a stack of 83 optical slices each 0.85 µm in thickness (total thickness, 
70 µm), showed the relationship of binding of anti-5ß1 integrin antibody to tumor vessels marked by CD31 immunoreactivity. When the blood vessels (green) in the rendered image were made 100% opaque, vessel labeling by anti-5ß1 integrin antibody was masked, but sites of antibody leakage (red) at 10 minutes were clearly marked (Fig. 7B). When the blood vessels were made 100% transparent, the overall distribution of vessel-associated and extravasated 5ß1 integrin was visible (Fig. 7C). Intermediate transparency of the CD31 immunoreactivity revealed the relationship of the 5ß1 integrin signal to endothelial cells (Fig. 7D and E). Unlike the tumor vasculature, vessels of the acinar pancreas had little 5ß1 integrin signal (Fig. 7D). A slice through the rendered isosurface of the tumor vasculature revealed the spatial relationship between anti-5ß1 integrin antibody binding and CD31 immunoreactivity of endothelial cells (Fig. 7F). Antibody labeling of 5ß1 integrin was on the luminal surface (Fig. 7F).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
5ß1 integrin is overexpressed on tumor blood vessels (8, 9) and is rapidly accessible to intravascular antibodies,3 the present study sought to determine whether this rapid accessibility results from binding of antibody to integrin overexpressed on the luminal surface of tumor vessels or from binding of extravasated antibody to integrin overexpressed on the abluminal surface of tumor vessels. Our experiments produced several lines of evidence consistent with the former alternative. The distribution of labeling of tumor vessels by anti-5ß1 integrin antibody at 10 minutes after injection did not match the distribution of extravasated nonspecific IgG or labeling by antibodies to fibrin, fibronectin, or type IV collagen. At this early time point, intravascular anti-5ß1 integrin antibody uniformly labeled tumor vessels, but extravasated nonspecific IgG and antibodies to fibrin, fibronectin, and type IV collagen had patchy distributions in tumors. Fibrin itself had a similar spotty pattern in tumor stroma. Furthermore, unlike 5ß1 integrin labeling that was near its peak at 10 minutes, the other antibodies accumulated in tumors over time, peaked at 6 hours, but never labeled tumor vessels. Finally, isosurface rendering of anti-5ß1 integrin antibody distribution at 10 minutes provided images consistent with the luminal expression of the integrin. Together, the evidence favors the expression of 5ß1 integrin on the luminal surface of tumor vessels as the likely explanation for the immediate accessibility of the integrin to intravascular antibodies.
Technical issues. When the project began, it seemed reasonable to assume that the immediate accessibility and generalized labeling of 5ß1 integrin on tumor vessels resulted from antibody leakage, given that integrins are expressed on the basolateral surface of endothelial cells (14–16), and tumor vessels are abnormally leaky (17, 18). Our approach to test this assumption was to inject antibodies i.v., and follow their distribution and quantify their accumulation in tumors over time using fluorescence and confocal fluorescence microscopy. We reasoned that if vessel leakiness were responsible, intravascular antibodies to 5ß1 integrin and its ligand fibronectin should extravasate and bind in roughly the same pattern. Extravasated nonspecific IgG (molecular weight, 150 kDa), extravasated fibrinogen (molecular weight, 340 kDa), and fibrin in the tumor stroma should also reflect the distribution of leakage.
The accumulation of antibodies in tumors is governed by convective driving forces for extravasation, vascular surface area, endothelial and other barriers in tumors, and amount and cellular distribution of the targets. The barriers determine which targets can be accessed from the bloodstream (18). The use of immunohistochemistry to localize sites of antibody accumulation in tumors preserved the spatial distribution of leakage and increased the sensitivity of detection through amplification by fluorescent secondary antibodies, which is an advantage over fluorescent dextrans and other tracers used as passive markers of vascular leakage and fluid dynamics in tumor stroma. Fluorescent macromolecules may also be recognized as foreign and internalized by macrophages.
The dose of antibody used in our studies was empirically determined with the end point of having sufficient signal for tissue localization. As only amounts of antibody that exceeded the threshold of detection could be measured, absence of immunofluorescence did not necessarily mean absence of antibody. We cannot exclude that biologically significant amounts of antibody were present where the fluorescence signal was not visible by immunohistochemistry. In addition, antibodies would be expected to become more widely distributed during circulation times longer than 24 hours.
By comparing location and time course of accumulation in tumors, we found that the distribution of labeling and accessibility of 5ß1 integrin differed from the other targets. Integrin 5ß1, fibronectin, and type IV collagen had generally similar distributions in tumors when examined by conventional immunohistochemistry on tissue sections. Injection of antibody to 5ß1 integrin was followed by labeling of tumor vessels that was intense at 10 minutes, did not increase appreciably over time, and resembled conventional on-section staining. By comparison, intravascular anti-fibronectin antibody gave patchy stromal immunofluorescence that was faint initially and gradually increased over 6 hours. Even at 24 hours, labeling by anti-fibronectin antibody did not have a vascular pattern, suggesting that it did not reach many sites of fibronectin expression on tumor vessels readily detected by conventional immunohistochemistry. Consistent with these findings, injection of anti-type IV collagen antibody, which targets a major component of vascular basement membrane in RIP-Tag2 tumors (11, 12), gave patchy staining similar to that obtained after injection of anti-fibronectin antibody.
A potentially confounding issue is that antibodies to fibrinogen and fibronectin would be expected to bind these proteins in blood and decrease the amount of antibody available to bind the corresponding proteins in tumor stroma. Several experiments were done to address this possibility. First, the distribution of injected antibodies to fibronectin was compared with the distribution of anti-type IV collagen antibody. The distributions were remarkably similar and neither resembled the corresponding distribution found by conventional immunohistochemistry on tissue sections. Second, the distribution and intensity of fibrin labeling were compared after i.v. injection of anti-fibrin antibody and on-section immunohistochemistry. The distribution and intensities found with the two approaches matched. Third, labeling of fibrin and fibronectin in tumors was examined at multiple times after injection to determine whether free antibody continued to circulate. Labeling of both fibrin and fibronectin increased over 6 hours after injection of the antibody. Together, these results argue against antibody quenching being a limiting factor in our experiments.
Because of the sensitivity of immunohistochemistry, tumor vessel labeling by anti-5ß1 integrin antibody could have resulted from nonspecific coating by the high dose of antibody. This possibility was tested in pilot studies using antibody-injected tissues preserved by immersion in fixative instead of vascular perfusion. Intravascular antibodies fixed in place with plasma were readily detected by immunohistochemistry. However, this is unlikely to have played a significant role in our experiments because unbound intravascular antibodies were eliminated by vascular perfusion of fixative. In addition, anti-5ß1 integrin antibody did not bind the normal vasculature and none of the other antibodies or nonspecific IgG-labeled tumor vessels.
Distribution of 5ß1 integrin expression. Integrin 5ß1 is highly expressed by a variety of normal cell types of mice, including epithelial cells of pancreatic ducts and smooth muscle cells of the intestinal wall and villi.3 However, in adult mice, most normal blood vessels, except for hepatic sinusoids and high-endothelial venules of lymph nodes, have little or no expression of 5ß1 integrin detectable by conventional immunohistochemistry or injection of anti-5ß1 integrin antibody.3 By comparison, 5ß1 integrin is strongly expressed on blood vessels in multiple different tumors in mice and humans (8, 9).3
Luminal distribution of 5ß1 integrin on endothelial cells. Expression of 5ß1 integrin on the luminal surface of tumor vessels is consistent with several lines of evidence. Immunoprecipitation data supported by immunofluorescence and immunogold labeling indicate that vß3 integrin is expressed on both the apical and basolateral surfaces of cultured human umbilical vein and saphenous vein endothelial cells (9, 19) . Integrins 5ß1, 2ß1, and 6ß1 may be as well (9). LM609, a function-blocking antibody to vß3 integrin that is up-regulated on tumor vessels (6, 20, 21), can target magnetic resonance (MR) contrast media-containing nanoparticles in the bloodstream to tumors (22). Similarly, bacteriophage displaying the cyclic RGD (Arg-Gly-Asp)-containing peptide RGD-4C, which targets vß3, 5ß1, and other integrins with RGD-binding sites, rapidly home to tumor blood vessels (23, 24). RGD-4C phage attach uniformly to the luminal surface of blood vessels in RIP-Tag2 tumors within minutes of i.v. injection (25). Scattered sites of leakage of phage particles resemble the patchy distribution of extravasated antibodies observed in the present study. Further evidence of binding to luminal integrins comes from studies of RGD-targeted 100-nm liposomes used to deliver doxorubicin to tumors in preclinical models (26, 27). The targeted liposomes bind to tumor blood vessels within 2 minutes of injection, when there is comparatively little extravasation. Untargeted doxorubicin-containing liposomes extravasate but do not bind to tumor vessels (26, 27).
Heterogeneity of leakage from tumor vessels. The heterogeneous pattern of leakage of antibodies and nonspecific IgG in RIP-Tag2 tumors was manifested by scattered, irregularly shaped patches of immunoreactivity in the stroma. Leakage sites were consistently more abundant in the core of tumors than at the periphery. Patchy leakiness is a well-documented feature of tumor vessels (18), as shown by extravasation of stealth liposomes (28), fluorescent albumin, and dextrans (29). Leakage, which may occur on one side of a tumor vessel and not the other, results from irregularly distributed endothelial defects that create intercellular holes as large as 5 µm in diameter (28, 30, 31).
Although tumor vessel leakiness provides a route for macromolecules to access the abluminal surface of tumor vessels, as well as tumor cells and other extravascular targets, this access is not uniform and does not reach all regions of tumors (2, 32, 33) . Because of heterogeneous leakage and restricted mobility in the interstitium of tumors (18, 32, 34), macromolecules have limited access to some regions.
The patchy distribution of extravasated antibodies that were so conspicuous by microscopic imaging may not be evident by in vivo imaging. Compared with fluorescence, confocal, and multiphoton microscopy, which have a resolution of 100 nm, MR imaging and computed tomography (CT) have resolutions of 100 to 500 µm, and positron emission tomography, ultrasonography, and optical imaging have resolutions of a few millimeters (25, 35, 36). Therefore, although leakage at the tumor periphery may be visibly greater than in the center (37), small regions of antibody extravasation within the tumor core may be sufficiently blurred in MR or CT images to give the impression of reasonably homogeneous leakage.
We conclude that the rapid and uniform accessibility of 5ß1 integrin on tumor vessels to antibodies in the bloodstream results from overexpression of integrin on the luminal surface of endothelial cells. The integrin is presumed also to be overexpressed on the abluminal surface of the cells, but antibodies have more limited access to that surface because leakage across the endothelial barrier is heterogeneous in tumors. Antibodies targeted to fibronectin and type IV collagen also accumulate in a patchy distribution, despite uniform expression of these proteins in the basement membrane of tumor vessels. Anti-fibrin antibodies accumulate with a similar patchy pattern that matches the distribution of fibrin in the tumor stroma. It remains to be determined whether antibodies that target epitopes on tumor cells or angiogenic growth factors such as vascular endothelial growth factor have a similar nonuniform distribution in tumors. If effective homing to tumor vasculature requires that the target on tumor vessels be uniformly distributed and accessible, 5ß1 integrin is a promising candidate.
|
Acknowledgments |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Douglas Hanahan for providing breeding pairs for our colony of RIP-Tag2 mice; Erin Ator and Mike Mancuso for animal care; Barbara Sennino and Tsutomu Nakahara for help with some of the experiments; Gyulnar Baimukanova for genotyping the mice; Patricia Parsons-Wingerter for helping with the development of the method of quantifying immunofluorescence in tumors; Shunichi Morikawa, formerly at UCSF and now at Tokyo Women's Medical University, Tokyo, Japan, for his help with immunohistochemistry; and Bitplane AG of Zurich, Switzerland, for providing Imaris software for use in this study.
|
Footnotes |
|---|
3 P. Parsons-Wingerter et al. Uniform overexpression and rapid accessibility of 5ß1 integrin on blood vessels in tumors. (2005), Am J Pathol, in press. 
Received 7/28/04. Revised 12/18/04. Accepted 1/24/05.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
vß3 antagonists promote tumor regression by inducing apoptosis of angiogenic blood vessels. Cell 1994;64.
5ß1 with the central cell-binding domain of fibronectin. Am J Pathol 2000;156:1345–62.5ß1 and 2ß1 in human umbilical vein endothelial cells. J Cell Biochem 1997;64:505–13.[CrossRef][Medline]
vß3 blocks human breast cancer growth and angiogenesis in human skin. J Clin Invest 1995;96:1815–22.
vß3. Important Adv Oncol 1996;1:69–87.
Vß3-targeted magnetic resonance imaging. Nat Med 1998;4:623–6.[CrossRef][Medline]
v integrins as receptors for tumor targeting by circulating ligands. Nat Biotechnol 1997;15:542–6.[CrossRef][Medline]
This article has been cited by other articles:
|
|
 |
|
  T. Okazaki, A. Ni, O. A. Ayeni, P. Baluk, L.-C. Yao, D. Vossmeyer, G. Zischinsky, G. Zahn, J. Knolle, C. Christner, et al. {alpha}5{beta}1 Integrin Blockade Inhibits Lymphangiogenesis in Airway Inflammation Am. J. Pathol., June 1, 2009; 174(6): 2378 - 2387. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Z. Behm, B. A. Kaufmann, C. Carr, M. Lankford, J. M. Sanders, C. E. Rose, S. Kaul, and J. R. Lindner Molecular Imaging of Endothelial Vascular Cell Adhesion Molecule-1 Expression and Inflammatory Cell Recruitment During Vasculogenesis and Ischemia-Mediated Arteriogenesis Circulation, June 3, 2008; 117(22): 2902 - 2911. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Sawada, A. K. Mitra, A. R. Radjabi, V. Bhaskar, E. O. Kistner, M. Tretiakova, S. Jagadeeswaran, A. Montag, A. Becker, H. A. Kenny, et al. Loss of E-Cadherin Promotes Ovarian Cancer Metastasis via {alpha}5-Integrin, which Is a Therapeutic Target Cancer Res., April 1, 2008; 68(7): 2329 - 2339. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Umeda, S. Kachi, H. Akiyama, G. Zahn, D. Vossmeyer, R. Stragies, and P. A. Campochiaro Suppression and Regression of Choroidal Neovascularization by Systemic Administration of an {alpha}5beta1 Integrin Antagonist Mol. Pharmacol., June 1, 2006; 69(6): 1820 - 1828. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. J. Yao, M. G. Ozawa, A. S. Varner, I. M. Kasman, Y. H. Chanthery, R. Pasqualini, W. Arap, and D. M. McDonald Antiangiogenic therapy decreases integrin expression in normalized tumor blood vessels. Cancer Res., March 1, 2006; 66(5): 2639 - 2649. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Nakahara, S. M. Norberg, D. R. Shalinsky, D. D. Hu-Lowe, and D. M. McDonald Effect of Inhibition of Vascular Endothelial Growth Factor Signaling on Distribution of Extravasated Antibodies in Tumors Cancer Res., February 1, 2006; 66(3): 1434 - 1445. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. M. McDonald and P. Baluk Imaging of Angiogenesis in Inflamed Airways and Tumors: Newly Formed Blood Vessels Are Not Alike and May Be Wildly Abnormal: Parker B. Francis Lecture Chest, December 1, 2005; 128(6_suppl): 602S - 608S. [Full Text] [PDF]
|
|
|
|
|
|
P. Parsons-Wingerter, I. M. Kasman, S. Norberg, A. Magnussen, S. Zanivan, A. Rissone, P. Baluk, C. J. Favre, U. Jeffry, R. Murray, et al. Uniform Overexpression and Rapid Accessibility of {alpha}5{beta}1 Integrin on Blood Vessels in Tumors Am. J. Pathol., July 1, 2005; 167(1): 193 - 211. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Prevention Journals Portal | Cancer Reviews Online |
| Annual Meeting Education Book | Meeting Abstracts Online |
