| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
5ß1 Integrin on Blood Vessels in Tumors
From the Department of Anatomy,* Cardiovascular Research Institute, Comprehensive Cancer Center, University of California, San Francisco; and Protein Design Labs, Incorporated, Fremont, California
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
5ß1 is among the proteins overexpressed on tumor vessels and is a potential target for diagnostics and therapeutics. Here, we mapped the distribution of 5ß1 integrin in three murine tumor models and identified sites of expression that are rapidly accessible to intravascular antibodies. When examined by conventional immunohistochemistry, 5ß1 integrin expression was strong on most blood vessels in RIP-Tag2 transgenic mouse tumors, adenomatous polyposis coli (apc) mouse adenomas, and implanted MCa-IV mammary carcinomas. Expression increased during malignant progression in RIP-Tag2 mice. However, immunoreactivity was also strong in normal pancreatic ducts, intestinal smooth muscle, and several other sites. To determine which sites of expression were rapidly accessible from the bloodstream, we intravenously injected anti-5ß1 integrin antibody and 10 minutes to 24 hours later examined the amount and distribution of labeling. The injected antibody strongly labeled tumor vessels at all time points but did not label most normal blood vessels or gain access to pancreatic ducts or intestinal smooth muscle. Intense vascular labeling by anti-5ß1 integrin antibody co-localized with the uniform CD31 immunoreactivity of tumor vessels and contrasted sharply with the patchy accumulation of nonspecific IgG at sites of leakage. This strategy of injecting antibodies revealed the uniform overexpression and rapid accessibility of 5ß1 integrin on tumor vessels and may prove useful in assessing other potential therapeutic targets in cancer.
The accessibility of vascular targets is governed by their cellular location and relationship to the endothelial barrier. In normal vessels, the barrier function of the endothelium restricts the extravasation of macromolecules, limiting access to pericytes and the abluminal surface of endothelial cells. In tumors, defects in the endothelial monolayer and other abnormalities make blood vessels leakier than their normal counterparts.17-20 Vascular leakiness makes it possible to reach targets in tumors that would normally be isolated by the endothelial barrier. However, not all tumor vessels are equally leaky, leakiness is heterogeneous along individual tumor vessels, and the amount of leakage is limited by the high interstitial fluid pressure in tumors.19,21,22 The effectiveness of molecular targets in tumors for macromolecular therapeutics in the bloodstream is thus determined by their amount of expression, cellular distribution, and accessibility.
Integrin 5ß1 and its ligand fibronectin have been found to be overexpressed in blood vessels of human and mouse tumors.13
The oncofetal isoform of fibronectin with an extra domain B (ED-B) is up-regulated in tumor vessels and is being explored as a diagnostic and therapeutic target.23
Selective antagonists targeted to 5ß1 integrin reverse tumor growth in preclinical models13,14,24
and have entered clinical trials.25
Selective inhibitors of 5ß1 integrin are thought to lead to endothelial cell apoptosis by anoikis and other mechanisms.13,14
The present study sought to obtain a better understanding of the distribution of 5ß1 integrin expression and to identify which sites are accessible to antibodies in the bloodstream. Specifically, we 1) compared the cellular distribution of 5ß1 integrin in tumors and normal organs, 2) examined the accessibility of 5ß1 integrin on tumor blood vessels to antibodies in the bloodstream, and 3) determined the extent and uniformity of expression of 5ß1 integrin on blood vessels in different mouse tumor models and during tumor progression.
Our approach was first to use conventional immunohistochemistry to determine the distribution of 5ß1 integrin in mice. Next, we tested the accessibility of the integrin by injecting the same anti-5ß1 integrin antibody intravenously and examining sites of binding 10 minutes to 24 hours later. Using these two approaches, we compared the distribution and accessibility of 5ß1 integrin in pancreatic islet tumors in RIP-Tag2 transgenic mice,26
intestinal adenomas in adenomatous polyposis coli (apc) mice,27,28
implanted MCa-IV mouse mammary carcinomas,20
and 11 normal organs. Immunoreactivity of 5ß1 integrin was co-localized with the endothelial cell marker CD31 (PECAM-1) to determine the uniformity of integrin expression in vascular networks.29
Fluorescence and confocal microscopic imaging, confirmed by fluorescence measurements, showed that 5ß1 integrin was uniformly expressed in blood vessels of all three tumor models, increased in expression during tumor progression, and, unlike the integrin in most normal organs, was rapidly accessible to antibodies in the bloodstream.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Three tumor models were studied. 1) Pancreatic islet tumors derived from ß-cells in islets of Langerhans were studied in RIP-Tag2 transgenic mice with a C57BL/6 background.26,30
Tumor-bearing RIP-Tag2 mice from our colony were identified by genotyping tail tip DNA and studied at 9 to 14 weeks of age.20,30
Tumors in individual RIP-Tag2 mice ranged from hyperplastic islets—little larger than normal islets—to carcinomas several millimeters in diameter. 2) Intestinal adenomas in adenomatous polyposis coli (apc) mice, which have 
35 adenomas by 16 weeks of age, vary in size but are arrested at about the same stage of progression.27,28
Adenomas in apc mice (C57BL/6J-ApcMin/J; Jackson Laboratory, Bar Harbor, MA) were studied at 16 to 21 weeks of age. 3) Cubes (2 mm) of MCa-IV mouse mammary carcinomas (gift from Rakesh Jain, Massachusetts General Hospital, Harvard University, Boston, MA) were implanted under the dorsal skin of male C3H mice (25 to 30 g) and studied 2 to 3 weeks later when the tumors were 5 to 9 mm in diameter.18,31
This highly reproducible tumor model was developed to visualize the tumor vasculature in a dorsal skinfold chamber in a setting not influenced by hormonal changes during the female estrus cycle.18
This approach was used in the present experiments so the data could be related to those obtained previously.20
Similar results were obtained when the tumors were implanted in mammary fat pads of female mice. Normal organs were studied in adult wild-type C57BL/6 mice. All experimental procedures were approved by the University of California, San Francisco, Institutional Animal Care and Use Committee.
Intravenous Injection of Antibodies
Mice, anesthetized by intramuscular injection of ketamine (87 mg/kg) and xylazine (13 mg/kg), received an injection of: 1) purified rat anti-mouse integrin 5 subunit antibody (50 µg per mouse or 2 mg/kg, anti-CD49e, clone 5H10-27, no. 553318; BD Biosciences PharMingen, San Diego, CA); 2) purified nonspecific IgG (50 µg, rat IgG2a,
monoclonal immunoglobulin isotype standard, clone R35-95, no. 553926; BD PharMingen); or 3) sterile saline (100 µl, 0.9% NaCl; Abbott Laboratories, North Chicago, IL) into a femoral or tail vein. To determine the extent of labeling of tumor vasculature, some mice received an injection of anti-5ß1 integrin antibody in combination with hamster monoclonal anti-CD31 antibody (50 µg, clone 2H8; Chemicon, Temecula, CA). Antibodies were diluted in sterile saline to final volume of 50 to 100 µl. Mice injected with saline were used for conventional immunohistochemistry.
The specificity of the antibody used to mark 5ß1 integrin is important to the meaningfulness of our results. This well-characterized rat anti-mouse monoclonal antibody was originally identified by T. Springer at Harvard University, where it was known as clone MRF5 (5H10) (http://cbr.med.harvard.edu/investigators/springer/lab/). The antibody is now available commercially as anti-CD49e antibody clone 5H10-27 from BD Biosciences Pharmingen (http://www.bdbiosciences.com), where specificity for 5 integrin has been tested by flow cytometry and cell binding assays. Because 5 integrin subunits pair solely with ß1 integrin subunits, we treated anti-5 integrin antibody as if it binds exclusively to 5ß1 integrin and designate it as such. The antibody has been reported to have blocked fibronectin binding in vitro, but function-blocking actions on endothelial cells or angiogenesis have not been demonstrated in vitro or in vivo.32,33
Vascular Perfusion Fixation and Immunohistochemistry
In most experiments antibodies, IgG, or saline circulated for 10 minutes before perfusion of fixative. In studies of the kinetics of accumulation in tumors, anti-5ß1 integrin antibody or IgG circulated for 10 minutes or 1, 6, or 24 hours. The chest was opened rapidly, and the vasculature was perfused for 3 minutes at a pressure of 120 mmHg with fixative [4% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.4; Sigma, St. Louis, MO] from an 18-gauge cannula inserted into the aorta via an incision in the left ventricle. The right atrium was incised to create an exit route. After the perfusion, tissues were immersed in fixative for 1 hour, rinsed with PBS, infiltrated overnight with 30% sucrose in PBS at 4°C, embedded in OCT (Sakura Finetek, Torrance, CA) and frozen at –80°C. Tissue sections were cut with a cryostat (Microm International GmbH, Walldorf, Germany) at a thickness of 100 µm, unless otherwise indicated, and dried on glass slides (Superfrost Plus; Fisher Scientific, Pittsburgh, PA). After removal of the OCT, sections were blocked and permeabilized by incubation in 5% normal goat serum containing 1% Triton X-100 in PBS for 3 hours.
Sections from anti-5ß1 integrin or IgG-injected mice were incubated for 12 to 15 hours in hamster anti-mouse CD31 antibody (monoclonal antibody 2H8; kind gift of Dr. William Muller, Department of Pathology, Cornell University Medical College, New York, NY or monoclonal 1398Z antibody, Chemicon). Sections used for conventional immunohistochemistry were incubated for 12 to 15 hours in CD31 antibody in combination with anti-5ß1 integrin antibody or nonspecific IgG used for intravenous injections. Primary antibodies (initial concentration, 1 mg/ml) were diluted 1:200 or 1:400 with PBS containing 5% normal goat serum and 1% Triton X-100.30
The identity of smooth muscle cells in intestinal villi was confirmed by staining with purified mouse monoclonal anti--smooth muscle actin antibody (Cy3-conjugated, no. C6198, Sigma).30
Sections were rinsed several times with PBS and then incubated for 6 hours in a combination of two secondary antibodies: Cy3-labeled anti-rat antibody for 5ß1 integrin and fluorescein isothiocyanate-labeled anti-hamster antibody for CD31 (Jackson ImmunoResearch, West Grove, PA). Secondary antibodies were diluted 1:200 or 1:400 with PBS containing 1% normal goat serum and 0.3% Triton X-100. Sections were mounted with Vectashield (Vector Laboratories, Burlingame, CA) for fluorescence microscopic imaging. Tissue sections from mice that had intravenous injections of both anti-5ß1 integrin and anti-CD31 antibodies were incubated for 6 hours in a mixture of Cy3-labeled anti-rat and fluorescein isothiocyanate-labeled anti-hamster secondary antibodies (Jackson ImmunoResearch), rinsed with PBS, and mounted with Vectashield.
Fluorescence Imaging
Tissue sections were examined with a Zeiss Axiophot fluorescence microscope equipped with single, dual, and triple fluorescence filters and a low-light, externally cooled, three-chip charge-coupled device (CCD) camera (480 x 640 pixel RGB color images, CoolCam; SciMeasure Analytical Systems, Atlanta, GA) and with a Zeiss LSM 510 confocal microscope (Carl Zeiss Micro-Imaging, Inc., Thornwood, NY) with argon, helium-neon, and UV lasers (512 x 512 or 1024 x 1024 pixel RGB color images). The amount of immunoreactivity in tissue sections prepared 10 minutes after intravenous injection of anti-5ß1 integrin antibody or nonspecific IgG was quantified in digital images of 100-µm-thick cryostat sections stained with Cy3-labeled anti-rat secondary antibody. For fluorescence measurements, RGB color images were captured with the CCD camera (x10 objective, x1 Optovar, tissue region 960 x 1280 µm), and then the red (Cy3) channel was converted to an 8-bit gray scale image (fluorescence intensity range, 0 < 255) for analysis using ImageJ v.1.29 (http://rsb.info.nih.gov/ij).
Immunofluorescence Measurements
Pilot studies revealed that, when section thickness, magnification, and microscope filters were held constant, subtle differences in immunohistochemical staining and CCD camera gain were the technical factors that contributed the most variability to the fluorescence intensity of images. Tissue staining was, therefore, standardized by simultaneously processing all slides in a particular data set (RIP-Tag2 tumors, apc tumors, or MCa-IV tumors). CCD camera gain was standardized as follows. First, after the brightest section (reference specimen) of a data set was identified, a digital image was acquired with the CCD camera. Next, the distribution of pixel fluorescence intensities of the image was displayed using the histogram function of Adobe Photoshop (Adobe Systems Inc., San Jose, CA). Camera gain was then adjusted, another image acquired, and the process repeated until the intensity histogram for the reference specimen image was well distributed, as judged by pixels throughout the entire intensity range with few or no pixels at maximal intensity of 255. When the camera gain for the reference specimen was optimized, the entire data set of images was acquired using this gain.
Pancreatic islet tumors in RIP-Tag2 mice were classified by their diameters in tissue sections as small (< 500 µm), medium (500 < 1000 µm), or large (
1000 µm), which approximately correspond to stage of tumor progression.26
Digital images of three to four tumors of each size in one or more sections of pancreas from each mouse were acquired. Regions of interest (ROI), defined by tissue boundaries identified by their vascular pattern in specimens stained for 5ß1 integrin and CD31 immunoreactivities, were outlined with the freehand tool of ImageJ (Figures 1 and 2)
. The ROI of normal islets and small and medium size tumors was completely included in a single image. Large tumors, which exceeded the size of a single image, were sampled by acquiring three images at 4, 8, and 12 o’clock around the tumor, each image representing one ROI. Fluorescence intensities of the pixels within the ROI were then recorded with ImageJ. The extent of co-localization of anti-5ß1 integrin and anti-CD31 antibodies in RIP-Tag2 tumors was calculated using the co-localization function of ImageJ (Figure 2)
.
|
|
A histogram displaying the number of pixels at fluorescence intensities 0 to 255 in each ROI was prepared using the histogram function of ImageJ (Figure 2)
. Data for each mouse included intensity histograms for ROI in three to four tumors of each size in RIP-Tag2 mice, four ROI per mouse for apc adenomas and MCa-IV carcinomas, or four ROI for normal tissues in wild-type mice. Histogram data were exported to Microsoft Excel, and mean values for the number of pixels at each intensity were calculated from multiple ROI to obtain a single intensity histogram for each tumor or tumor size for each mouse. Weighted values for fluorescence intensity were then calculated by multiplying the mean number of pixels at each intensity in the ROI by their corresponding brightness value of 0 to 255 (Figure 2)
. This transformation discarded all black values, because the intensity of black (background) pixels was evaluated as 0. The mean intensity for each mouse was then calculated as the sum of the weighted intensities for all brightness values divided by the total number of pixels in the ROI. Mean intensity values for all mice in a group were averaged to obtain an overall mean intensity for the group. Three-dimensional plots of fluorescence intensities of tissue sections, in which brightness was displayed in the z axis, were prepared with the Surface Plot function of ImageJ.
Tumor Vascularity
An index of vascular area density (proportion of sectional area occupied by tumor vessels) was measured in fluorescence microscopic digital images of specimens from RIP-Tag2 mice injected with 5ß1 integrin antibody to relate vascularity (number of labeled vessels per unit area) to the intensity of accumulated antibody (mean brightness per labeled site) in tumors of different size.30
Area density measurements were made on three to four images of tumors of each of the three size groups per mouse (x10 objective, x1 Optovar, tissue region 960 x 1280 µm, n = 4 mice/group). Based on fluorescence intensities ranging from 0 to 255, specific vascular labeling by 5ß1 integrin antibody was distinguished from background by an empirically determined threshold value (intensity = 30) that included only blood vessels in specimens. Area density of vascular staining for 5ß1 integrin was calculated as the proportion of pixels having a fluorescence intensity equal to or greater than the corresponding threshold.30
Tumors from RIP-Tag2 mice injected with IgG were handled in a similar manner as a control.
Statistics
The significance of differences among groups was tested by analysis of variance followed by Fisher’s test for multiple comparisons or by Student’s t-test using JMP v.5.0.1.2 (SAS Institute, Inc., Cary, NC). P values <0.05 were considered significant. All groups consisted of four mice, with the exception of apc mice injected with nonspecific IgG, which consisted of three mice. Results are presented as mean ± SE.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
5ß1 Integrin in RIP-Tag2 Tumors
Islet cell tumors and exocrine pancreas both had strong 5ß1 integrin immunoreactivity in RIP-Tag2 mouse tissues stained by conventional immunohistochemistry. Most of the staining in tumors was associated with blood vessels, but most staining in exocrine pancreas was associated with ducts (Figure 1, A and B)
. When anti-5ß1 integrin antibody was injected intravenously, most of the immunoreactivity was associated with tumors, and little staining was evident in the exocrine pancreas (Figure 1C)
. Vessel immunoreactivity was stronger in tumors than in normal-size islets and increased in intensity with increasing tumor size (Figure 1C)
.
Measurement of Antibody Accumulation in Tumors
At 10 minutes after intravenous injection, 5ß1 integrin antibody labeled blood vessels in RIP-Tag2 tumors in a pattern that resembled the vascular network visible in tumor sections double-stained for CD31 by conventional on-section immunohistochemistry (Figure 2A)
. The amount of 5ß1 integrin antibody in tumors, assessed by analysis of the fluorescence intensity of pixels in digital images, was conspicuously greater than for nonspecific IgG (Figure 2B)
. The uniformity of 5ß1 integrin expression on tumor vessels was examined by comparing the distributions and amounts of anti-CD31 and anti-5ß1 integrin antibodies on tumor vessels after simultaneous intravenous injection of both antibodies into RIP-Tag2 mice. At 10 minutes after injection, the two antibodies had similar distributions in tumors, as reflected by 85% co-localization of 5ß1 integrin pixels with CD31 pixels (Figure 2; C to F)
. In addition, area density measurements showed that the amounts of accumulation of CD31 and 5ß1 integrin antibodies were essentially the same (Figure 2G)
. However, the two antibodies did not have identical staining patterns outside of tumors. Anti-5ß1 integrin antibody mainly labeled tumor vessels (Figure 2, C and D)
, but anti-CD31 stained the vasculature of the acinar pancreas (Figure 2E)
and other organs as well as tumor vessels.
Leakage of Antibodies in RIP-Tag2 Tumors
Labeling of tumor vessels by 5ß1 integrin antibody was accompanied by irregular patches of immunoreactivity around some vessels (Figure 3, A and B)
. These patches were outside blood vessels and appeared to be sites of antibody extravasation. Injection of nonspecific IgG resulted in similar extravascular patches, but the staining was weak and tumor vessels were not labeled (Figure 3, C and D)
. The differences in uniform vascular labeling by anti-5ß1 integrin antibody and patchy leakage of IgG are illustrated schematically in Figure 3, E to G
. The dense vascularity characteristic of RIP-Tag2 tumors is shown by on-section staining for CD31 (Figure 3E)
. Uniform binding of 5ß1 integrin antibody to tumor vessels has essentially the same pattern as CD31 immunoreactivity (Figure 3F)
. By comparison, sites of IgG leakage are located in patchy regions between tumor vessels (Figure 3G)
.
|
5ß1 Integrin Distribution in RIP-Tag2 Tumors
Measurements of immunofluorescence in RIP-Tag2 tumors revealed significantly greater accumulation of anti-5ß1 integrin antibody than nonspecific IgG at all time points from 10 minutes to 24 hours after intravenous injection (Figure 3H)
. Values for anti-integrin antibody were already high at 10 minutes, increased further through 6 hours, and then decreased to the 10-minute level at 24 hours (Figure 3H)
. Extravasated nonspecific IgG was barely detectable at 10 minutes, increased a bit at 1 hour, but remained less than half the value for anti-integrin antibody throughout the 24-hour period of the study (Figure 3H)
.
Distribution and Accessibility of 5ß1 Integrin in Normal Islets
The vasculature of normal islets in wild-type pancreas had weak 5ß1 integrin immunoreactivity when stained by conventional immunohistochemistry (Figure 4, A and B)
. Ducts of normal acini had stronger staining (Figure 4B)
. Yet, when the antibody was injected intravenously, islet vasculature stained weakly, but ducts had little or no immunoreactivity (Figure 4, C and D)
. Pancreatic islets and ducts had little or no immunoreactivity after injection of nonspecific IgG (Figure 4, E and F)
.
|
5ß1 Integrin Expression during Progression of RIP-Tag2 Tumors
Expression of 5ß1 integrin increased with tumor size in RIP-Tag2 mice. More circulating anti-5ß1 integrin antibody accumulated in small tumors than in normal islets of wild-type mice (Figure 5; A to F)
. There was even greater accumulation of antibody in large tumors, which was evident in images and in surface plots of immunofluorescence (Figure 5, G and H)
. Measurements of fluorescence intensity showed the magnitude of these size-related differences and the differences between the accumulation of anti-5ß1 integrin antibody and nonspecific IgG after intravenous injection (Figure 5I)
. The mean intensity of 5ß1 integrin immunofluorescence in RIP-Tag2 tumors was three times the value for normal islets in wild-type mice and for nonspecific IgG in tumors.
|
5ß1 integrin labeling (Figure 5J)
, contributed to the increased accumulation of circulating 5ß1 integrin antibody. However, comparison of surface plots of tissue fluorescence for anti-5ß1 integrin antibody in tumors of different size (Figure 5; C, F, and H)
demonstrated that intensity of labeling (height of peaks = antibody binding per vessel) increased even more than the number of labeled sites (number of peaks = vascularity).
Distribution and Accessibility of 5ß1 Integrin in apc Intestinal Adenomas
Intestinal adenomas of apc mice had strong 5ß1 integrin immunoreactivity when stained by conventional immunohistochemistry (Figure 6A)
. The greatest staining was associated with blood vessels, stromal cells, and some tumor cells of the adenomas. Strong 5ß1 integrin immunoreactivity was also present in smooth muscle cells of the intestinal wall (Figure 6A)
and intestinal villi (Figure 6B)
,34
both in apc mice and in wild-type mice. The identity of smooth muscle cells was confirmed by co-localization with -smooth muscle actin immunoreactivity (data not shown). Immunoreac-tivity for 5ß1 integrin was not detected in the micro-vasculature or epithelium of normal intestinal villi in wild-type or apc mice (Figure 6B)
.
|
5ß1 integrin antibody. Blood vessels and smooth muscle cells had little or no 5ß1 integrin immunoreactivity in normal intestinal villi (Figure 6C)
, but blood vessels of apc adenomas had intense immunoreactivity (Figure 6, D and E)
. Vascular staining in adenomas had indistinct borders suggestive of antibody extravasation and binding to stromal cells (Figure 6, D and E)
. This stromal staining had a patchy distribution and was greatest near the apical surface of adenomas (Figure 6F)
. After intravenous injection of nonspecific IgG, patches of faint, diffuse immunoreactivity were scattered in adenomas of apc mice (Figure 6G)
. Measurements of immunofluorescence in apc adenomas showed that the mean value for anti-5ß1 integrin antibody after intravenous injection was twice the corresponding value for normal intestine and fourfold the value for apc adenomas after injection of nonspecific IgG (Figure 7)
.
|
5ß1 Integrin in MCa-IV Mammary Carcinomas
Blood vessels and scattered tumor cells had strong 5ß1 integrin immunoreactivity in implanted MCa-IV carcinomas stained by conventional immunohistochemistry (data not shown). Strong immunoreactivity was also present in these tumors after intravenous injection of the antibody, but here most of the staining was associated with blood vessels (Figure 8, A and B)
. Little or no staining was found after injection of nonspecific IgG (Figure 8, C and D)
. Fluorescence measurements showed that immunoreactivity of MCa-IV carcinomas was 7.5-fold greater after intravenous injection of anti-5ß1 integrin antibody than after injection of nonspecific IgG (Figure 7)
.
|
5ß1 Integrin in Normal Organs
To develop a perspective for interpreting the rapid, uniform labeling of tumor vessels by intravascular anti-5ß1 integrin antibody, immunohistochemically stained sections of several normal organs and tumors prepared 10 minutes after injection of anti-5ß1 integrin antibody or nonspecific IgG were imaged at standardized camera settings optimized for the Cy3 (yellow-orange) fluorophore (Figure 9
, Table 1
). The microvasculature of most normal organs had little or no detectable anti-5ß1 integrin antibody immunoreactivity (Table 1)
. Brain (Figure 9, A and B)
, lung, and testis were examples of organs with negligible staining. The absence of staining in these organs appeared to result from absence of 5ß1 integrin rather than absence of antibody leakage, because no immunoreactivity was found in brain vasculature stained for 5ß1 integrin by conventional immunohistochemistry, except for faint staining of large cerebral arteries (data not shown).
|
|
5ß1 integrin antibody (Figure 9C
, left; Table 1
). Liver of tumor-bearing mice and normal mice was similar in this regard. This staining was not the result of antibody accumulation around highly permeable hepatic sinusoids because strong 5ß1 integrin immunoreactivity was also found in liver stained for the integrin by conventional immunohistochemistry (Figure 9C
, right). Also, intravenous injection of nonspecific IgG, which would also leak, produced only faint staining (Figure 9D)
. Although the overall staining of hepatic sinusoids after injection of anti-5ß1 integrin antibody was even more intense than was found in the vessels of RIP-Tag2 tumors (Figure 9E)
, the higher vascular density of the liver limited the meaningfulness of this comparison. Faint staining of the liver after intravenous injection of nonspecific IgG was approximately the same as found in RIP-Tag2 tumors under the same conditions (Figure 9F
, Table 1
). High endothelial venules of lymph nodes were another site where strong immunoreactivity was present after injection of anti-5ß1 integrin antibody (Figure 9G
, Table 1
). These vessels had little or no staining after injection of nonspecific IgG (Figure 9H)
.
Assessment of Potential Toxicity of Injected Anti-5ß1 Integrin Antibody
No adverse effects of anti-5ß1 integrin antibody were detected in RIP-Tag2 mice during the 10-minute to 24-hour period of circulation in the bloodstream. None of the mice died or show signs of distress. Histological studies on tissues from these mice revealed no evidence of pathological changes in any of 11 organs, including the liver where sinusoidal endothelial cells were strongly labeled by the antibody.
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
5ß1 integrin in normal organs and tumors and to identify sites where the integrin was highly expressed and readily accessible to intravascular antibodies. We found that most normal blood vessels in mice had little or no 5ß1 integrin immunoreactivity. Although pancreatic ducts, intestinal smooth muscle, and some other normal tissues did have strong staining by conventional immunohistochemistry, this was not the case when the same antibody was injected into the bloodstream, indicating that anti-5ß1 integrin antibody had limited access to these sites during the 10-minute period of circulation. By comparison, 5ß1 integrin on blood vessels in three murine tumor models—islet cell tumors in RIP-Tag2 transgenic mice, intestinal adenomas in apc mice, and MCa-IV mammary carcinomas—was uniformly overexpressed and was rapidly accessible to intravascular antibody. Immunofluorescence measurements showed that most of the accumulation of anti-5ß1 integrin antibody in tumors resulted from direct labeling of tumor vessels rather than vessel leakiness. Overall accumulation of antibody in tumors, as reflected by immunofluorescence at 10 minutes to 24 hours after injection, was two to seven times the accumulation of extravasated nonspecific IgG. These findings indicate that 5ß1 integrin is overexpressed on tumor blood vessels and rapidly accessible to antibodies in the bloodstream.
Role of 5ß1 Integrin in Angiogenesis
The integrin 5ß1 and its principal ligand fibronectin are essential for normal embryonic development. Targeted deletion of either the 5 integrin subunit or fibronectin results in major disturbances in the vasculature and embryonic lethality.35-37
Mouse embryos lacking 5 integrin have distended blood vessels and abnormal vascular patterning.37
In the adult, 5ß1 integrin promotes angiogenesis and is essential for the growth of new blood vessels under some conditions.14
Both 5ß1 integrin and fibronectin are up-regulated in angiogenesis in tumors13,24
and certain other diseases.38
Integrin 5ß1 on blood vessels is up-regulated by basic fibroblast growth factor (bFGF, FGF2), tumor necrosis factor-, or interleukin-8 but not by vascular endothelial growth factor (VEGF).13
Expression of 5ß1 integrin is regulated by the homeobox transcription factor Hox D3, which induces expression by binding directly to the promoter of the integrin 5 subunit.39
Integrin 5ß1 promotes angiogenesis by supporting endothelial cell migration and survival.14,40
This process is mediated in part by suppression of an apoptotic program dependent on protein kinase A activity.14
Integrin 5ß1 may also play a role in angiogenesis by serving as a binding site for fibrinogen41,42
and endostatin.43
Inhibition of 5ß1 integrin signaling may lead to caspase-3- and caspase-8-dependent endothelial cell death via apoptosis.13,14
This mechanism has been exploited in preclinical models in which tumor growth has been slowed by inhibition of 5ß1 integrin.13,14,24
In this regard, 5ß1 integrin, like vß3 integrin, is a potential target for inhibiting angiogenesis.12,44-46
Tissue Distribution 5ß1 Integrin
The present study showed that 5ß1 integrin 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, which are also sites of expression in humans.47,48
However, most normal blood vessels in adult mice had little or no expression of 5ß1 integrin detectable by conventional immunohistochemistry. Two exceptions were hepatic sinusoids and high-endothelial venules of lymph nodes. Unlike 5ß1 integrin, fibronectin is nearly ubiquitous in basement membranes of normal adult tissues.49
In contrast to their normal counterparts, blood vessels in RIP-Tag2 tumors, apc adenomas, and MCa-IV carcinomas had uniformly strong expression of 5ß1 integrin. The finding of strong immunoreactivity of blood vessels in these tumors is consistent with the documented increased expression of 5ß1 integrin on tumor vessels.13,24,30,50,51
In RIP-Tag2 tumors and MCa-IV carcinomas, most 5ß1 integrin expression was associated with vascular endothelial cells, whereas in apc adenomas, the integrin was expressed not only by blood vessels but also by stromal cells and to a lesser extent tumors cells.
Increasing 5ß1 Integrin Expression during Tumor Progression
The use of the RIP-Tag2 model made it possible to follow changes in 5ß1 integrin expression during tumor progression. The changes were assessed by comparing the accumulation of anti-5ß1 integrin antibody. Faint immunoreactivity was detected on blood vessels of normal pancreatic islets, but labeling was significantly greater in blood vessels of even the smallest tumors, and the amount of anti-5ß1 integrin antibody that accumulated increased with tumor size. The change in antibody binding appeared to result both from increased expression of the integrin on tumor vessels and from increased tumor vascularity. Fluorescence measurements and surface plots of immunofluorescence showed increases in both the height of the fluorescence intensity peaks, reflecting amount of antibody binding per vessel, as well as the number of peaks, reflecting the density of the vasculature.30
Accessibility of 5ß1 Integrin
One objective of the present study was to test the accessibility of sites of 5ß1 integrin expression in tumors to antibodies in the bloodstream. To address this issue we injected anti-5ß1 integrin antibody into the circulation and then compared the amount and distribution of labeling at 10 minutes with that found by conventional immunohistochemistry. The brief circulation time was part of the experimental design to identify sites of greatest accessibility to the antibody. At 10 minutes, labeling was restricted to tumor vessels and scattered focal sites of leakage nearby.
Intravascular anti-5ß1 integrin antibody prominently labeled blood vessels in all three tumor models. In RIP-Tag2 mice, preferential labeling of tumor vessels sharply highlighted pancreatic tumors against the remainder of the pancreas. Strikingly different results were obtained when the antibody was injected versus when it was applied directly to tissue sections. Conventional immunohistochemistry showed that 5ß1 integrin was expressed by multiple cell types both in tumors and in some normal organs.
Kinetic studies of anti-5ß1 integrin antibody injected into RIP-Tag2 mice revealed that the labeling of tumor vessels was already strong at 10 minutes. Thereafter, the amount of labeling increased somewhat throughout 6 hours and remained high for the entire 24-hour period of the study. The amount of labeling by the anti-integrin antibody significantly exceeded that produced by extravasated nonspecific IgG at all time points. Accumulation of nonspecific IgG in tumors and of antibodies was manifested by patches of leakage rather than vascular labeling. No intravascular IgG remained after the blood was removed by vascular perfusion of fixative.
Co-localization of 5ß1 integrin with sites of CD31 immunoreactivity confirmed that endothelial cells were the main cellular location of rapidly accessible integrin in tumors. Injection of anti-5ß1 integrin and anti-CD31 antibodies together gave very similar results in tumors, where both antibodies labeled most tumor vessels and most 5ß1 integrin co-localized with CD31. However, the two antibodies gave very different results elsewhere because of the generalized vascular labeling by the CD31 antibody.
It was not surprising that anti-CD31 antibody rapidly labeled tumor vessels, but why were tumor vessels preferentially labeled by intravascular anti-5ß1 integrin antibody? Access of antibodies in the bloodstream to 5ß1 integrin in tumors is determined by the anatomical distribution of sites of integrin expression, macromolecular permeability of the vasculature at those sites, convective driving forces, and endothelial surface area for antibody extravasation.19
The well documented leakiness of blood vessels in tumors may be a factor.52
Indeed, the vasculature of MCa-IV tumors has an unusually large (2000 nm) pore size.18,20
The presence of focal, patchy accumulations of nonspecific IgG and anti-5ß1 integrin antibody in the interstitium confirmed the leakiness of tumor vessels to macromolecules in all three models we studied. Patchy regions of antibody extravasation were identifiable because they co-localized with fibrin in the stroma53
and with components of basement membrane and extracellular matrix (unpublished data).54
However, most of the intense labeling of tumor vessels by anti-5ß1 integrin antibody is unlikely to simply reflect vascular leakiness. Most anti-5ß1 integrin antibody that accumulated in RIP-Tag2 tumors co-localized with CD31 immunoreactivity, which is known to be expressed by most blood vessels in these tumors and to match the distribution of intravenously injected fluorescent Lycopersicon esculentum lectin and four other endothelial cell markers (CD105, VEGFR-2, VEGFR-3, and 5ß1 integrin).30
Unlike the intense, uniform labeling of tumor vessels by the 5ß1 integrin antibody, accumulations of nonspecific IgG were patchy, weakly fluorescent, and primarily extravascular. Labeling of tumor vessels by anti-5ß1 integrin antibody is consistent with a uniform vascular expression of the integrin, whereas patchy, extravascular labeling is consistent with the heterogeneous leakiness of tumor vessels.19,21,22
Luminal Expression of 5ß1 Integrin on Endothelial Cells
Did the anti-5ß1 integrin antibody bind to the luminal or abluminal surface of tumor vessels? Binding to 5ß1 integrin on the abluminal surface of endothelial cells or on stromal cells would require intravascular antibodies to cross the endothelial barrier. However, if 5ß1 integrin is expressed on the luminal surface of tumor vessels, binding sites would be immediately accessible to antibodies in the bloodstream. In pilot studies, we attempted to determine the location of the labeling by using confocal microscopy and fluorescence microscopy with filters for viewing two fluorophores simultaneously to examine tissue sections of various thicknesses of RIP-Tag2 tumors stained for both 5ß1 integrin and CD31 immunoreactivities. Although the approach seemed reasonable, it did not convincingly show whether 5ß1 integrin was on the luminal surface, abluminal surface, or both surfaces of endothelial cells (unpublished data). The problem appears to concern the thinness of the endothelial cells compared to the resolution of fluorescence and confocal microscopy. Immunogold labeling for electron microscopy may work, but even that approach will be tricky because the 5ß1 integrin antibody leaks into the tumor at some locations. Accumulations at sites of leakage will have to be distinguished from sites of binding.
The location of anti-5ß1 integrin antibody binding was explored further in a companion study by comparing the kinetics of accumulation of antibodies to 5ß1 integrin to those of antibodies to three other targets in RIP-Tag2 tumors.53
During the 24-hour period of that study, the 5ß1 integrin antibody maintained its strong, uniform vascular association and also accumulated throughout time at sites of leakage in tumors. By comparison, antibodies to fibronectin, type IV collagen, and fibrin accumulated mainly at sites of leakage. Accumulation of these antibodies was minimal at 10 minutes, peaked at 6 hours, and then plateaued. The half-life of labeling by 5ß1 integrin antibody was less than 10 minutes, but the half-life of labeling by the other antibodies was a few hours. Also, the targets were different, with the 5ß1 integrin antibody mainly targeting tumor vessels and the others accumulating at sites of leakage. These striking differences in kinetics and distribution add further evidence that 5ß1 integrin is expressed on the luminal surface of tumor vessels and fibronectin, type IV collagen, and fibrin are not. Endothelial cells in culture express integrins on both apical and basolateral surfaces.55
Bacteriophage displaying peptides containing RGD (Arg-Gly-Asp) sequences on their surface rapidly home to tumors by binding to integrins on blood vessels.56
Similarly, RGD-modified liposomes can target the endothelium of tumor vessels and deliver therapeutic payloads.16
Functional Effects of Anti-5ß1 Integrin Antibodies
The functional significance of the rapid accumulation on tumor vessels of 5ß1 integrin antibody in the bloodstream cannot be directly tested in mice because of the lack of availability of function-blocking antibodies with documented functional efficacy in vivo. The 5H10-27 anti-5 integrin antibody used in the present study inhibits fibronectin binding and cellular migration in vitro in systems not involving endothelial cells,32,33
but has not been shown to affect endothelial cells or angiogenesis in vivo and has not been used to inhibit integrin ligation in tumor models.
We found no adverse effects from 10 minutes to 24 hours after injection of the anti-5ß1 integrin antibody, and observed no pathological changes in histological studies of liver and other sites of antibody binding. The absence of functional changes could be a manifestation of absence of blocking action, brief treatment period, dose, or insensitivity to integrin inactivation despite the presence of binding sites for 5ß1 integrin antibody. However, the toxicity issue was directly addressed for a potent function-blocking anti-human 5ß1 integrin antibody with established anti-angiogenic activity, and the antibody was found to have minimal toxicity in primates.57
Preclinical studies performed on cynomolgus monkeys show no significant toxicity with doses as large as 50 mg/kg i.v. (unpublished data).57
Similarly, the antibody has minimal side effects in humans given doses up to 15 mg/kg i.v. in a phase I clinical trial.25
Antibody 5H10-27 was used in a dose of 2 mg/kg in the present studies.
Integrins as Vascular Targets
Multiple molecular markers have been identified as vascular targets in tumors. Integrin vß3, which is up-regulated in tumor vessels, is one of the most extensively studied targets for inhibiting and imaging angiogenesis in tumors.15,44
Tumstatin, a cleavage fragment of the 3 chain of type IV collagen, appears to inhibit angiogenesis by binding vß3 integrin, and part of the anti-angiogenic action of endostatin may be mediated by binding 5ß1 integrin.43
Integrin targeting by RGD peptides increases efficacy of doxorubicin and tumor necrosis factor- in preclinical tumor models.10,16,58
The uniform distribution and rapid accessibility of 5ß1 integrin on tumor vessels suggest that it would be an effective target for delivery of agents preferentially to blood vessels in tumors. The results of studies using inhibitors of 5ß1 integrin on preclinical tumor models support this function.13,24
The distinctive anatomical distribution and high level of expression of 5ß1 integrin warrant further functional studies to explore specificity and efficacy as a target for imaging and therapeutics in cancer.
In conclusion, the results of these experiments taken together show that 5ß1 integrin on tumor blood vessels is overexpressed and rapidly accessible to antibodies in the bloodstream. The strong, uniform labeling of tumor vessels by anti-5ß1 integrin antibody circulating in the bloodstream contrasts with weak labeling of most normal blood vessels. The rapid accessibility of the integrin on tumor vessels distinguishes it from most sites of integrin expression outside the vasculature. Because the labeling of 5ß1 integrin by intravascular antibody far exceeds the accumulation of nonspecific IgG, abluminal overexpression of integrin combined with vessel leakiness is unlikely to explain the rapid accumulation of anti-integrin antibody in tumors. Overexpression of the integrin on the luminal surface of tumor vessels appears to be a more important factor. Because of these properties, 5ß1 integrin has a promising role as a vascular target for antibodies in tumors.
|
Acknowledgements |
|---|
5ß1 integrin expression in endothelial cells; Barbara Sennino and Tsutomu Nakahara for help with some of the experiments; and Shunichi Morikawa, formerly at University of California, San Francisco, now at Tokyo Women’s Medical University, Tokyo, for help with the immunohistochemical staining of tumor vessels. |
Footnotes |
|---|
Supported in part by the University of California (BioSTAR grants S98-50 and 99-10067), the National Institutes of Health (grants HL-24136 and HL-59157 from the National Heart, Lung, and Blood Institute; P50-CA90270 from the National Cancer Institute), the AngelWorks Foundation, the Vascular Mapping Project (to D.M.M.), and by Microgravity Science, NASA Glenn Research Center (to P.P.-W.).
P.P.-W. and I.M.K. contributed equally to this work.
Present address of P.P.-W.: Microgravity Science Division, NASA Glenn Research Center, Cleveland, OH 44135; I.M.K.: Genentech, Inc., South San Francisco, CA 94080; S.Z. and A.R.: Department of Oncological Sciences, University of Turin, Italy; C.F.: Integra LifeSciences, San Diego, CA 92121; U.J.: Chiron Corp., Emeryville, CA 94608.
Accepted for publication March 10, 2005.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
P < 0.05 for large tumors compared to small tumors. Scale bar,150 µm.
