Nature Medicine
9, 713 - 725 (2003)
doi:10.1038/nm0603-713
Imaging of angiogenesis: from microscope to clinicDonald M McDonald1
& Peter L Choyke21 Cardiovascular Research Institute, Comprehensive Cancer Center, and Department of Anatomy, University of California, 513 Parnassus Avenue, San Francisco, California 94143-0452, USA. dmcd@itsa.ucsf.edu 2 Imaging Sciences Program, National Institutes of Health, Bethesda, Maryland 20892-1182, USA. pchoyke@nih.gov Advances in imaging are transforming our understanding of angiogenesis and the evaluation of drugs that stimulate or inhibit angiogenesis in preclinical models and human disease. Vascular imaging makes it possible to quantify the number and spacing of blood vessels, measure blood flow and vascular permeability, and analyze cellular and molecular abnormalities in blood vessel walls. Microscopic methods ranging from fluorescence, confocal and multiphoton microscopy to electron microscopic imaging are particularly useful for elucidating structural and functional abnormalities of angiogenic blood vessels. Magnetic resonance imaging (MRI), computed tomography (CT), positron emission tomography (PET), ultrasonography and optical imaging provide noninvasive, functionally relevant images of angiogenesis in animals and humans. An ongoing dilemma is, however, that microscopic methods provide their highest resolution on preserved tissue specimens, whereas clinical methods give images of living tissues deep within the body but at much lower resolution and specificity and generally cannot resolve vessels of the microcirculation. Future challenges include developing new imaging methods that can bridge this resolution gap and specifically identify angiogenic vessels. Another goal is to determine which microscopic techniques are the best benchmarks for interpreting clinical images. The importance of angiogenesis in cancer, chronic inflammatory diseases, age-related macular degeneration and reversal of ischemic heart and limb disease provides incentive for meeting these challenges.Perspective
Imaging methods are proving to be indispensable for studying angiogenesis in the laboratory and clinic. They can pinpoint sites of angiogenesis, determine amount of blood vessel growth, characterize functional abnormalities of vessels, assess vascular heterogeneity and elucidate features that distinguish angiogenic blood vessels from normal blood vessels. Vascular imaging makes it possible to quantify the number and spacing of blood vessels, measure blood flow and vascular permeability, and analyze changes in blood vessel walls. Imaging methods also have potential utility in assessing the efficacy of angiogenesis inhibitors used in the treatment of cancer, age-related macular degeneration and other chronic diseases. This review describes contemporary methods for imaging angiogenesis in preclinical models and human disease, with a focus on the visualization of blood vessels in tumors.
Microscopic imaging methods are powerful tools for dissecting the cellular and molecular features of the microvasculature, and in an intravital setting can give real time functional readouts in preclinical models1. MRI, CT, PETand other noninvasive imaging methods make it possible to localize sites of angiogenesis and obtain functional data in animals and humans2,
3,
4,
5,
6.
Many methods for imaging angiogenesis in humans lack validation and have important limitations, however. Despite the power and utility of the methods, there is still a large discrepancy in spatial resolution between microscopy-based imaging methods and clinical methods. Compared with confocal microscopy and multiphoton microscopy, which have a resolution of  100 nm, and electron microscopy, with a resolution of a few nanometers, MRI and CT have resolutions in the range of 100 to 500  m, and PET, ultrasonography and optical imaging have resolutions of a few millimeters. In addition, some advanced clinical imaging methods are available only in the research setting or involve macromolecular intravascular contrast agents that have not yet been approved for use in humans. Also, few methods have been compared side by side in the same model of angiogenesis, and there is little consensus on the best standards for validation. Yet the clinical value and promise of imaging angiogenesis before and during therapy provides strong incentive for advancing the technology. Many of the approaches being developed for studying angiogenic blood vessels in tumors should also be useful for evaluating chronic inflammatory conditions and other disorders in which angiogenesis contributes to the pathophysiology.
Imaging angiogenic blood vessels
Vascular casts are a traditional but powerful method for examining the three-dimensional architecture of blood vessels and vascular networks (Fig. 1a and b). The vasculature can be made visible for light microscopic or scanning electron microscopic examination by intravascular injection of colored gelatin, latex or plastic casting material followed by tissue clearing or corrosion7,
8. This approach makes it possible to quantify vessel dimensions, intervascular distances, branching order and luminal surface features not readily assessed in tissue sections.
 | | Figure 1. Microscopic imaging of normal and angiogenic blood vessels. | | |  | (a) Scanning electron microscopic (SEM) imaging of polymer cast of normal microvasculature, showing simple, organized arrangement of arterioles, capillaries, and venules. Vasa vasorum of rat carotid sinus. (b) SEM image of cast of tumor microvasculature, showing disorganization and lack of conventional hierarchy of blood vessels. Arterioles, capillaries, and venules are not identifiable as such. Image of xenograft of human tumor in nude mouse provided by M. Konerding, University of Mainz, Germany (from ref. 8). (c,d) Fluorescence microscopic imaging showing effect of section thickness on appearance of tumor vessels in histological sections (c, 10-m section; d, 40-m section) of RIP-Tag2 mouse pancreatic islet cell tumor stained for CD31 immunoreactivity (Cy3 fluorescence). (e, f) Two images of same large tumor vessel showing endothelial sprouts (arrowheads) made visible by CD31 immunoreactivity (e, gold) but not by intravenous injection of FITC-labeled Lycopersicon esculentum lectin staining (f, green), because the sprouts lack a lumen accessible to the lectin. MCa-IV mouse breast carcinoma (from ref. 20). (g) Bright-field microscopic image of extravasated erythrocytes in blood lakes (red) visible in whole mount of RIP-Tag2 mouse tumor after intravascular blood was removed by systemic vascular perfusion of fixative. (h) Fluorescence microscopic image of same tumor showing functional blood vessels (arrowheads) labeled by intravenous injection of fluorescent cationic liposomes (red) and fluorescent lectin (green). Blood lakes (black) are not labeled by either fluorescent tracer and are thus not in continuity with the bloodstream. (i) Comparison of functional blood vessels (*) washed free of blood by vascular perfusion and erythrocyte-filled, stagnant blood lakes in histological section of RIP-Tag2 tumor stained with H&E. g−i: from ref. 20. Scale bar in i applies to all panels (50 m in a,b; 125 m in c−f; 150 m in g−i).
 Full Figure and legend (181K) |
| |
Most microscopic studies of angiogenesis rely on histological tissue sections. Here, blood vessel detection is improved by immunohistochemical staining of endothelial cell markers, most commonly CD31, CD34, von Willebrand factor and Ulex europaeus lectin (Table 1). Endothelial cells can also be labeled by fluorescent or colored reporters expressed in transgenic mice9,
10,
11. As an alternative, blood vessels can be marked by antibodies to basement membrane proteins12,
13. Regardless of the marker, the appearance of blood vessels in histological sections is greatly influenced by section thickness. Only short segments of vessels are visible in thinner sections (Fig. 1c), but the vascular network can be seen in thicker sections (Fig. 1d).
| | |
A standard approach for quantifying angiogenesis in tissue sections is the measurement of microvascular density (maximal number of blood vessels per unit area of section), which provides an index of minimal intercapillary distance14. Used in hundreds of studies, microvascular density is a significant prognostic indicator in many cancers14,
15. But the approach has proven unreliable in some tumors16,
17 and in evaluating angiogenesis inhibitors when tumor mass decreases in parallel with the number of blood vessels15,
18.
Conventional approaches for identifying blood vessels in histological sections, even when labeled by reliable immunohistochemical markers, do not discern whether the vessels provide routes for blood flow. Functional blood vessels can be identified by intravital labeling with a tracer that can reach vascular targets only through the circulation19,
20. For example, a fluorescent lectin injected intravenously before tissue fixation labels functional vessels but not lumenless endothelial sprouts (Fig. 1e and f). Similarly, tumor cells supplied by functional blood vessels can be marked by injection of the fluorescent nuclear binding dye Hoechst 33342 (ref. 21). This approach can be used to monitor blood vessel regression after treatment or determine whether blood flows through channels lined by tumor cells at putative sites of vasculogenic mimicry22,
23. Collections of erythrocytes visible in histological sections are not sufficient evidence of blood flow because they may be static extravascular blood lakes or sites of hemorrhage (Fig. 1g−i)20.
Current angiographic methods of visualizing blood vessels in the clinical setting are excellent for evaluating larger arteries and veins but not the microvasculature of tumors or other sites of angiogenesis. An exception is retinal and choroidal angiography, where the eye has a natural window for visualizing angiogenesis. Here, blood vessels as small as 25 m, highlighted by intravascular fluorescence of sodium fluorescein or indocyanine green, can be viewed with a video camera or scanning laser ophthalmoscope24,
25. Blood vessels can thus be assessed in uveal melanomas (Fig. 2a), macular degeneration and diabetic retinopathy where neovascularization and leaky vessels are characteristic. Clinical angiography is also useful in evaluating therapies for ischemia of the heart or extremities but provides only an indirect measure of angiogenesis. X-ray images, acquired after rapid intra-arterial infusion of an iodinated contrast agent (Fig. 2b), have a resolution of 100 m and limited utility in visualizing the microvasculature3,
26. Also, the approach requires exposure to ionizing radiation and an iodinated contrast agent that is potentially nephrotoxic.
| | Figure 2. In vivo imaging of human tumors and structural basis of tumor vessel leakiness. | | |  | (a) Funduscopic image showing leakage of fluorescein into an eye tumor (uveal melanoma, arrowhead) near the optic nerve during the initial phase of an angiogram. Tumor has overall bright fluorescence, but most individual tumor vessels are not visible. Retinal arteries are bright (arrows); veins are dark. Courtesy of E. Chew, National Eye Institute, National Institutes of Health. (b) CT angiogram showing many renal artery branches (arrows) in a patient with bilateral renal tumors. Tumor microvasculature is not resolved. Inset, angiographic image of renal vasculature after latex injection showing larger vessels but not the microvasculature. (c−e) PET images of a patient with renal cancer revealing a metastatic tumor (arrowheads) in the pelvis. Each PET agent provides different functional information about the tumor: [18F]fluorodeoxyglucose primarily reflects tumor metabolism (c); [11C]carbon monoxide reflects tumor blood volume, which is greatest in the periphery (d); and [15O]water scan provides a quantitative measure of vascular perfusion, which is also greatest in the periphery (e). Courtesy of S. Bacharach, National Institutes of Health. (f−h) SEM images comparing luminal surface of normal blood vessel, which is smooth and has tight endothelial junctions (f, arrowheads, mouse mammary gland) and tumor blood vessel, which has widened intercellular spaces, overlapping endothelial cells, multiple cellular processes and other abnormalities (g, arrowheads, MCa-IV mouse mammary carcinoma). h, High magnification of a hole in the endothelium (arrows) showing the underlying basement membrane filaments (arrowheads). Scale bar in h applies to all panels (1 mm in a; 3 cm (inset 4 cm) in b; 5 cm in c−e; 5 m in f; 2 m in g; 0.5 m in h).
Full Figure and legend (98K) |
| |
Alternatives include computed tomographic angiography and magnetic resonance angiography. As with clinical angiography, an intravascular contrast agent is required, and resolution is insufficient to resolve the microvasculature (Fig. 2b, inset). Macromolecular MR contrast agents, not yet available for clinical use, provide a more sustained vascular enhancement phase that permits a spatial resolution approaching 200 m in animals27. Even then, most angiogenic blood vessels are not resolved. MicroMR and microCT have better spatial resolution (10 m in microMR) in preclinical models but have poor temporal resolution and large radiation exposure, which can be as high as 0.6 Gy per scan and reach 5% of the half-maximal lethal dose for mice6.
Color Doppler ultrasonography detects the frequency shift from moving blood and assigns color based on a scale proportional to velocity28,
29,
30. Resolution is limited to 200 m, and images are usually confined to superficial parts of the body. Microbubble contrast agents, often a perfluorohydrocarbon gas in 1 to 10-m hydrophobic vesicles, augment by <25 dB the acoustic signal from blood vessels and provide an index of tissue blood volume31. Microbubbles can be burst by high-intensity ultrasound, yielding a short but strong backscatter echo known as stimulated acoustic emission, which continuously reflects microvascular perfusion32.
Tumors labeled with green fluorescent protein, luciferase or another fluorescent or bioluminescent reporter can be noninvasively localized and measured in preclinical models by whole-body optical imaging6. The signal tends to be surface weighted, however, as the fluorescence of green fluorescent protein penetrates tissue poorly. Although the tumor vasculature is not fluorescent in these preparations, large vessels may be visible against a background of brightly fluorescent tumor cells33.
Imaging blood vessel function
Intravital microscopic imaging has revealed many bizarre features of tumor vessels in preclinical models. Tumors implanted in transparent chambers in skin and other sites can be imaged over time by fluorescence microscopy34 or multiphoton laser scanning microscopy, which reduces photon damage, increases contrast and expands the depth of imaging to hundreds of micrometers1. Rate of blood flow, red blood cell velocity, blood vessel diameter, vascular density, endothelial permeability, leukocyte−endothelial cell interactions, intravascular and interstitial fluid pressures, and interstitial diffusion and convection are among the variables measured by intravital microscopy coupled with fluorescent tracers and reporters, computer algorithms and mathematical models34. Changes in tissue pH and partial pressure of oxygen (pO2) and effects of vasoactive agents can also be measured35,
36. The approach has shown that blood flow in tumor vessels is variable, may transiently stop or reverse direction and does not deliver nutrients efficiently. Intravascular pressure in tumors is about normal, but interstitial pressure may be high enough to reduce the driving force for extravasation. Poor blood flow leads to tissue hypoxia, even in the presence of abundant angiogenesis.
Several approaches are available to measure blood flow and blood volume in a clinical setting. Cerebral blood volume can be determined by CT or MRI using low molecular weight (<1 kDa) contrast agents, iodinated contrast agents for CT and gadolinium chelates for MRI, provided the blood-brain barrier is intact. The area under the enhancement curve is directly related to blood volume, allowing for small amounts of recirculation37. Because of their low molecular weight and rapid extravasation, however, these agents do not accurately reflect blood flow and volume in the absence of the blood-brain barrier4,
38,
39.
Blood volume can be more accurately measured using macromolecular MR contrast agents that have a longer half-life in the circulation. Examples include superparamagnetic iron oxide particles40 and gadolinium-labeled albumin, dendrimers, protected graft copolymers or liposomes41,
42,
43. These agents, currently under development for imaging the vasculature, are beginning to be used clinically and hold great promise for estimating fractional plasma volume and perhaps vascular permeability in tumors42.
PET is another approach used to assess blood flow and blood volume in human tumors44. Fluorodeoxyglucose labeled with the positron emitter 18F can be used to localize primary tumors and metastases (Fig. 2c). Blood volume within tumors can be determined by inhaling carbon monoxide labeled with 11C or 15O, also positron emitters. When inhaled, trace amounts of labeled carbon monoxide irreversibly bind to red blood cells and distribute in accordance with vascular volume (Fig. 2d). One limitation is that PET images may be dominated by the signal from large vessels. The first pass of water labeled with 15O can be used to calculate blood flow within tumors because water is completely diffusible and redistributes quickly (Fig. 2e). The washout of the activity curve for a tumor is directly proportional to blood flow44. The half-life of 15O is only 2 min, so the procedure can be repeated at brief intervals. Spatial resolution is currently limited to 4 mm, and pixels may average multiple different types of tissue. Developments in detectors and scatter compensation may eventually provide resolutions of 1.5 mm (ref. 45).
Imaging blood vessel permeability
Defective endothelial barrier function is one of the best-documented abnormalities of tumor vessels46. Blood vessel leakiness is important because it contributes to the high interstitial pressure in tumors, may facilitate angiogenesis and enables antibodies and other macromolecular therapeutics to reach tumor cells from the bloodstream. Despite the barrier defect, tumor vessels do not leak as much as expected because the high interstitial pressure offsets the convective driving force for extravasation47. Imaging methods are helping to characterize vessel leakiness by determining the size, number, distribution and cellular basis of the endothelial defects, how increased endothelial permeability affects drug delivery to tumor cells, and whether vessel leakiness changes after treatment46. This information will help to design better drug delivery systems and to understand the effect of angiogenesis inhibitors on access of cancer therapeutics to tumor cells.
Intravital imaging experiments have shown that the amount of leakage of intravascular tracers from tumor vessels correlates with molecular mass <160,000 kDa, suggestive of relatively large pores in the endothelium48. Microscopic imaging of particulate tracers indicates a pore cutoff size ranging from 380 to 2,000 nm, depending on tumor type and location49. Scanning electron microscopic images show clear differences in the endothelium of normal vessels (Fig. 2f) and leaky tumor vessels (Fig. 2g), some of which have separations or holes at intercellular junctions (Fig. 2h)20. Transmission electron microscopic images show similar holes20,
50 and also suggest the involvement of appreciably smaller (<100 nm) transcytotic channels or vesiculo-vacuolar organelles51.
Microscopic images of extravasated proteins illustrate the non-uniform distribution of leaky sites in tumors. IgG extravasates in a patchy manner (Fig. 3a), and endogenous fibrin has a similar distribution (Fig. 3b)52,
53. Particulate tracers also extravasate in scattered regions49. By comparison, antibody against  5 1 integrin labels tumor vessels uniformly because the integrin on endothelial cells is readily accessible (Fig. 3c). Access to 51 integrin expressed by nonvascular cells (Fig. 3d) is restricted by the endothelial barrier.
| | Figure 3. Imaging of leakage-based targeting and vascular targeting in tumors. | | |  | (a,b) Fluorescence microscopic images showing similar patchy distribution of extravasated IgG (a, red, arrowheads) 10 min after intravenous injection, and endogenous fibrin (b, red, arrowheads) in RIP-Tag2 tumors stained by immunohistochemistry. Blood vessels are marked by CD31 immunoreactivity (green). (c,d) Confocal microscopic images comparing the vessel-restricted distribution of 51 integrin (red) in RIP-Tag2 tumor 10 min after intravenous injection of antibody to 5 integrin (c) with the much broader distribution of the integrin evident by conventional immunohistochemistry (d). The subset of integrin that is accessible from the bloodstream is restricted to tumor blood vessels; little is present in normal-size pancreatic islets (c, arrowheads). By comparison, the overall distribution of the integrin in tissue sections includes conspicuous staining of normal pancreatic ducts around the tumor (d, arrowheads) and fainter staining of tumor vessels. Blood vessels are marked by CD31 immunoreactivity (green). (e,f) MRI of mouse comparing the appearance of a tumor after intravenous injection of a low molecular weight contrast agent (Gd-DTPA; e), which extravasates rapidly and makes the tumor uniformly bright, with a subsequent injection of 400 kD Gd-dendrimer contrast agent (f), which remains intravascular and highlights the larger tumor blood vessels (arrowheads). Courtesy of H. Kobayashi, National Cancer Institute, National Institutes of Health. (g,h) Immunohistochemical localization of RGD-4C peptide-expressing filamentous bacteriophage 6 min after intravenous injection into RIP-Tag2 mouse. Unbound phage in the bloodstream are removed by vascular perfusion. g, Phage are abundant in tumor vessels (gold, arrowheads) but are not present in the vasculature of normal pancreatic acini around the tumor. h, Confocal micrograph showing punctate appearance of phage (red, arrowheads) bound to tumor vessels. Vascular endothelial cells are marked by CD31 immunoreactivity (green). Scale bar in h applies to all panels (200 m in a; 65 m in b; 125 m in c,d; 5 mm in e,f; 150 m in g; 20 m in h).
Full Figure and legend (166K) |
| |
The leakiness of individual tumor vessels can be measured under direct microscopic visualization54, yet the complex architecture of the vasculature limits this approach to superficial regions of tumors. Methods dependent on the extravasation of Evans blue−labeled albumin or other macromolecular tracer are widely used, but the amount of leakage is influenced by vascular surface area and hydrostatic driving forces as well as endothelial permeability, and thus does not measure permeability per se.
MRI and CT are used to assess vascular permeability in humans and preclinical models. Low-molecular-weight contrast agents used in dynamic MRI extravasate rapidly at sites of angiogenesis and alter local relaxivity55. Because these agents also leak from normal vessels (Fig. 3e), however, it may be difficult to distinguish between normal and angiogenic vessels. Macromolecular intravascular MR contrast agents, such as iron oxide particles or gadolinium bound to larger molecules such as albumin or dendrimers (Fig. 3f), are better retained by normal vessels and more effectively test the barrier function of angiogenic vessels42,
56,
57. Such agents may soon be available in the clinic.
Dynamic contrast-enhanced MRI studies can provide a meaningful estimate of vascular permeability to the extent that a tumor is homogeneous. However, a major limitation is that each pixel reflects the signal from a relatively large volume (1  1 3 mm) and may represent multiple microenvironments in a tumor. MRI thus tends to average out variations in leakage from region to region. Indeed, dynamic contrast-enhanced MRI represents a complex summation of vascular permeability, blood flow, vascular surface area and interstitial pressure. Additional complexity stems from sensitivity to motion and measurement variability58. Because signal-to-noise ratios increase directly with magnetic field strength, new technologies that provide stronger magnets may improve both spatial and temporal resolution of MRI.
For imaging angiogenesis, dynamic contrast-enhanced CT is analogous to dynamic MRI. Contemporary CT scanners permit rapid repeated image acquisitions at low radiation doses. The linear relationship between signal attenuation and iodine concentration of the contrast agent provides CT with a distinct advantage over similar MRI techniques59. This advantage must be balanced against repeated exposure to ionizing radiation and potentially nephrotoxic contrast agents, however. To minimize these risks, dynamic CT studies are typically brief and use a low dose of contrast agent.
Imaging cellular and molecular abnormalities of blood vessels
Imaging studies have identified multiple cellular and molecular abnormalities that distinguish tumor vessels from their normal counterparts and help to explain their unusual appearance, disturbed blood flow and leakiness. Most tumor vessels do not fit into the conventional hierarchy of arterioles, capillaries and venules. Abnormalities involve all components of the vessel wall: endothelial cells, pericytes (mural cells) and basement membrane60. Despite the abnormalities, most tumor vessels do have endothelial cells and pericytes60. Evidence suggesting that endothelial cells are absent in focal regions or even entire blood vessels in tumors22,
61 deserves further validation to exclude local heterogeneity of marker expression or vessel-like structures not functionally connected to the bloodstream.
Membrane proteins selectively expressed by angiogenic blood vessels provide potential markers for imaging tumors. Reported examples include v3, v5, and 51 integrins62,
63, CD105 (endoglin)64,
65, complexes of vascular endothelial growth factor (VEGF) and its receptor VEGFR-2 (ref. 66), CD36 (thrombospondin-1 receptor)67,
68, Thy-1 (ref. 69), prostate-specific membrane antigen70 and tumor endothelial markers71.
In vivo phage display is a powerful approach for identifying membrane proteins for targeting tumor vessels72. Filamentous bacteriophages are genetically engineered to express small peptides encoded by a piece of DNA inserted in the phage genome. Phage that express the cyclic peptide CDCRGDCFC (termed RGD-4C) target v3 and v5 integrins on tumor vessels (Fig. 3g). The phage peptide, which mimics the RGD (Arg-Gly-Asp) motif of vitronectin, binds cognate receptor sites on integrins73,
74. RGD-4C phage imaged by confocal microscopy appear as tiny dots on endothelial cells in tumors (Fig. 3h).
Integrins on tumor vessels are used as targets for imaging agents in preclinical models. Novel PET tracers using 18F-labeled glycopeptides containing RGD sequences have been developed to target v3 and v5 integrins on angiogenic endothelial cells75. Related agents have been developed for targeting gadolinium-labeled liposomes for MR imaging76 and microbubbles for ultrasonography77. The ED-B domain of fibronectin78, transforming growth factor- receptor endoglin65 and anionic phospholipids79 are additional targets on tumor vessels potentially useful for diagnostic imaging agents. None of these markers is expressed by all tumor vessels, however, and some may be expressed by normal cells as well.
Another feature of endothelial cells of tumor vessels that can be exploited for imaging is their avid binding and internalization of cationic liposomes19,
80. After binding, cationic liposomes enter endothelial cells by endocytosis19, and few extravasate. The opposite is true for anionic, neutral and stealth liposomes19,
49. Targeting of endothelial cells by cationic liposomes can be used to deliver diagnostics or therapeutics to tumor vessels81.
As a key cellular component of the microvasculature, pericytes are a potential target for vascular imaging agents, but the cells are located outside the endothelial barrier. Reports that pericytes are an inconsistent component of tumor vessels and that their absence is a sign of vessel immaturity raise additional questions82,
83,
84. Pericytes are identified by their distinctive location within the vascular basement membrane or by selective markers. Immunohistochemical markers of pericytes include -smooth muscle actin (-SMA)85, platelet-derived growth factor receptor (PDGFR)-86, desmin87, nestin88, smooth muscle myosin89, tropomyosin90, high-molecular-weight melanoma-associated antigen (NG2)91, aminopeptidase A92, aminopeptidase N (CD13)88, matrix metalloproteinase-9 (gelatinase B)93, sulphatide94, VEGFR-1 (ref. 95) and regulator of G-protein signaling-5 (ref. 96). In addition, there is a transgenic mouse with pericytes that express LacZ86.
Unfortunately, none of these markers is unique to pericytes. Moreover, pericytes are heterogeneous and may not express all markers85. The most commonly used marker, -SMA, is expressed by the mural cells of most arterioles and venules, but not capillaries, and has variable expression in tumors and other sites of angiogenesis85,
87,
93,
97. Therefore, the absence of immunoreactivity for -SMA does not indicate absence of pericytes. Desmin, NG2 and PDGFR- are expressed more widely by pericytes but are not restricted to these cells. In some tumors, CD13 is also expressed by endothelial cells98. Nevertheless, when multiple markers are used for identification, pericytes are consistently found on tumor vessels99 and growing blood vessels100 at a stage when they do not express -SMA.
Pericytes on normal blood vessels are intimately associated with endothelial cells (Fig. 4a), but those on tumor vessels are loosely attached, have multiple layers and extend cytoplasmic processes away from the vessel wall86,
99 (Fig. 4b). In transgenic and implanted tumors in mice, pericytes labeled with -SMA and desmin are present on >97% of vessels and cover more than half of the vessel surface99. Pericytes are also associated with most endothelial sprouts in these tumors99. Because of vessel leakiness and the endothelial support function of pericytes, these cells on tumor vessels may be reasonable targets after all (see note added in proof).
| | Figure 4. Imaging of tumor vessel pericytes, basement membrane and perivascular sleeves. | | |  | (a,b) Confocal microscopic images showing -SMA-immunoreactive pericytes (red, arrowheads) that tightly envelop endothelial cells (b, green, CD31 immunoreactivity) of a normal vessel (a) but are loosely attached to blood vessels in Lewis lung carcinoma. (c) Confocal micrograph showing coverage of blood vessels in Lewis lung carcinoma by an abnormally loose, fragmented and multilayered basement membrane (red, arrowheads) stained for type IV collagen immunoreactivity. Endothelial cells are marked by CD31 immunofluorescence (green). (d) Transmission electron microscopic image showing multiple layers of basement membrane (arrowheads) on blood vessel in RIP-Tag2 tumor. RBC, red blood cell.(e) Scanning electron microscopic view of cut surface of MCa-IV mouse mammary carcinoma showing 100-m thick sleeves of viable tumor cells (arrowheads) around blood vessels (*), interspersed by regions of stroma and necrosis. Blood was removed by vascular perfusion of fixative. Scale bar in e applies to all panels (25 m in a,b; 35 m in c; 0.5 m in d; 200 m in e).
Full Figure and legend (64K) |
| |
The vascular basement membrane (basal lamina) is a self-assembled layer of proteins, glycoproteins and proteoglycans formed by endothelial cells, pericytes and smooth muscle cells. Type IV collagen, laminin, fibronectin, nidogen (entactin) and the heparan sulfate proteoglycan perlecan are among the main components101. Basement membrane of normal blood vessels tightly envelops pericytes and endothelial cells, but that of tumor vessels is loosely associated with these cells (Fig. 4c), varies in thickness and has multiple layers in some regions (Fig. 4d). Basement membrane also covers most endothelial sprouts and pericyte processes. Redundant layers of vascular basement membrane, visible under conditions where blood vessels undergo regression and regeneration, may provide a scaffold for vascular regrowth102,
103.
Hypoxia is a common feature of tumors, a stimulus for angiogenesis and a potential target for imaging. Under conditions of limited nutrient delivery, tumor cells may grow as sleeves around tumor vessels (Fig. 4e). Hypoxic regions in tumors can be detected by MR imaging. Deoxyhemoglobin shortens the time required for water protons to completely dephase, thus causing a small change in signal intensity referred to as the blood oxygen level−dependent (BOLD) effect104. By having the subject breathe room air, followed by a mixture of 95% oxygen and 5% CO2 (carbogen), and then room air again, deoxyhemoglobin/oxyhemoglobin ratios and the MR signal undergo changes (Fig. 5a and b) that reflect vascular reactivity to hypercapnia and reversal of hypoxia4.
| | Figure 5. Imaging of tumor hypoxia and effects of angiogenesis inhibitors. | | |  | (a,b) MR imaging of vascular changes in a tumor implanted in a mouse that breathed room air, followed by carbogen (95% O2 + 5% CO2), followed by room air. Changes in signal intensity (a) and the color-encoded map of the MR image (b) reflect carbogen-induced changes in blood flow and oxygenation within the leg tumor, with the largest changes occurring in the brightest regions (b, yellow). Courtesy of Y. Zhang, National Institutes of Health. (c) Fluorescence microscopic imaging of VEGFR-2 immunoreactivity (red) of vascular endothelial cells in RIP-Tag2 tumor of an untreated mouse (left) and a mouse treated for 7 d with VEGF-Trap, a soluble receptor inhibitor of VEGF signaling (right)140. VEGF inhibition not only decreased the number of VEGFR-2-immunoreactive blood vessels, but also decreased the intensity of immunofluorescence of individual vessels. (d) Surface plot showing the magnitude of difference in VEGFR-2 immunofluorescence of tumor vessels in c, reflecting decreased VEGFR-2 expression after treatment. (e) Confocal microscopic imaging showing relatively normal-appearing PDGFR--immunoreactive pericytes (arrowheads) closely associated with blood vessels of RIP-Tag2 tumor after treatment with VEGF-Trap for 7 d (compare with untreated tumor in Figure 4b). Arrow marks aggregate of PDGFR--immunoreactive cells not associated with individual tumor vessels. (f) Confocal image of RIP-Tag2 tumor treated with VEGF-Trap for 7 d, showing strands of basement membrane (arrowheads, type IV collagen immunoreactivity, red) that superficially resemble blood vessels but do not have endothelial cells (CD31 immunoreactivity) or blood flow. Scale bar in f applies to all panels (5 mm in b; 60 m in c; 40 m in e; 50 m in f).
Full Figure and legend (112K) |
| |
Direct measurements of tumor pO2 are possible with electron paramagnetic resonance, Overhauser-enhanced MRI and [19F]perfluorocarbon MRI105,
106. Electron paramagnetic resonance uses free-radical contrast agents that signal differences in redox and oxygen status between tumor and normal tissues106. Overhauser-enhanced MRI uses a free-radical contrast agent that requires very low magnetic fields and measures tissue oxygen concentration105. The width of the resonance peak of these agents varies inversely with pO2 and thus provides an in vivo measure of tissue oxygenation. Because the MR signal from [19F]perfluorocarbon is linearly dependent on tissue pO2, pO2 maps can be generated when this agent is injected into the tumor and the tumor is scanned at the resonance frequency of fluorine107. None of these techniques is currently approved for human use.
Imaging to assess efficacy of angiogenesis inhibitors
The identification of imaging readouts that reflect the efficacy of angiogenesis inhibitors has great potential but is at an early stage, and the results thus far are not straightforward. Part of the difficulty stems from incomplete knowledge of how these agents affect blood vessels. Strictly speaking, angiogenesis inhibitors block the growth of new blood vessels from existing vessels. However, to the extent these agents also eliminate certain existing vessels, such as tumor vessels, they are antivascular agents with selective specificities. Blood vessel regression has received much less attention than angiogenesis and has mainly been studied in the context of vascular networks that naturally regress during development108,
109 or cyclic changes in the reproductive system110.
Because destruction of selected blood vessels is implicit in the concept of antivascular therapy, it seems that measurements of microvascular density would provide a useful index of efficacy. Indeed, when tumor vessels regress more rapidly or aggressively than tumor cells, microvascular density can decrease substantially (Fig. 5c), as has been reported in many preclinical models111,
112,
113,
114. This is not always the case, however15,
18. If changes occur gradually or tumor cell death is tightly coupled to blood vessel regression, microvascular density may not change, or may even increase, despite an overall reduction in tumor size and number of vessels15,
18. Moreover, high microvascular density is not necessarily predictive of favorable response to angiogenesis inhibitors15.
Intravital microscopic studies provide an instructive framework for thinking about blood vessel regression in tumors. In androgen-dependent tumors viewed in a windowed chamber in mice, vascular diameter, tortuosity, permeability and leukocyte rolling all decrease progressively after castration115. Within 24 h, some endothelial cells become TUNEL-positive, indicative of apoptosis. Apoptosis of tumor cells begins a day or more later. Blood vessel density can decrease by 80% within 2 weeks. Antibodies to VEGF114,
116 and some other inhibitors117 have similar effects. Blood vessels that survive treatment are structurally and functionally different from those eliminated and appear more normal118.
Microscopic imaging has shown that the number of endothelial cells undergoing apoptosis, as indicated by TUNEL staining, increases as microvascular density decreases82,
112,
113,
115. The number of apoptotic endothelial cells present at any time is small in comparison to the total number of endothelial cells, however, and other mechanisms of cell death may also be involved. The number of proliferating tumor endothelial cells, identified by BrdU or Ki67 labeling, does not correspond to microvascular density or rate of tumor cell proliferation119,
120,
121.
Microscopic readouts of expression of receptors and other key signaling molecules offer additional indicators of the response of endothelial cells to angiogenesis inhibitors114. After treatment, VEGFR-2 immunoreactivity may decrease by two mechanisms. The number of VEGFR-2-positive tumor vessels decreases as the vasculature regresses (Fig. 5c), and the amount of immunoreactivity per vessel also decreases (Fig. 5d), suggesting reduced VEGF responsiveness of surviving vessels.
Few imaging studies have addressed the fate of pericytes after endothelial cells are destroyed by an angiogenesis inhibitor. Do pericytes degenerate, remain in place, migrate away, change their expression of molecular markers or transform into other cells? The number of pericytes seems not to decrease as much as endothelial cells after treatment, and pericyte coverage of surviving vessels may increase122. After treatment, some pericytes are not associated with endothelial cells, while others are closely associated with endothelial cells of surviving vessels, consistent with normalization of blood vessel phenotype (Fig. 5e). The proportion of tumor vessels with -SMA-positive pericytes increases during blood vessel regression after VEGF withdrawal in tumors that conditionally express VEGF82. Blood vessels in human prostate cancer have the same feature after castration of the patient82. It is unclear whether tumor vessels with -SMA-negative pericytes are selectively eliminated or whether the phenotype of pericytes changes from -SMA-negative to -SMA-positive.
Microscopic studies of blood vessel regression in a variety of natural conditions show that the vascular basement membrane does not disappear as rapidly as endothelial cells degenerate102,
103. In the context of tumors, immunoreactivity for vascular basement membrane normally matches the pattern of blood vessels, but after treatment the basement membrane does not regress in parallel with the endothelial cells. In treated tumors, thin strands of basement membrane resembling blood vessels interconnect the surviving vessels as telltale signs of vessels that were once present (Fig. 5f). These remnants of tumor vessels may provide a path for blood vessel regrowth and may also be a useful morphological index of drug efficacy.
The introduction of angiogenesis inhibitors as cancer therapies has challenged traditional methods of imaging tumors in patients123. Measures such as time to progression or survival are well-accepted markers of clinical efficacy. Inhibitors of angiogenesis pose special problems, however, because the vascularity of a tumor can change without a corresponding change in tumor size and vice-versa. In animals, tumors can be imaged with MRI, PET or CT at frequent intervals after drug administration, tissue can be obtained from specific regions based on imaging features and regional intratumoral responses can be followed. This level of monitoring is usually not possible in patients.
The utility of PET and MRI studies is under evaluation in clinical trials of angiogenesis inhibitors (Fig. 6a and b). One approach is to assess glucose metabolism with fluorodeoxyglucose (FDG) imaged by PET, as tumor glucose metabolism would be expected to decrease as nutrient vessels regress during antivascular therapy. A decrease in glucose metabolism, monitored by FDG uptake, would presumably indicate a response. Yet, uptake of FDG by tumors may increase in tissues rendered hypoxic by the angiogenic inhibitor, because expression of the Glut1 glucose transporter is increased by hypoxia. Anaerobic metabolism uses more glucose than aerobic respiration, increasing FDG accumulation. Thus, FDG uptake may paradoxically rise initially after antiangiogenic therapy, confounding the interpretation of the studies123.
| | |
Initial results from trials of angiogenesis inhibitors monitored with dynamic MRI, dynamic CT and FDG-PET imaging suggest that changes in vascular permeability, vascular volume fraction and metabolism can be detected in tumors soon after therapy begins123,
124. Some angiogenesis inhibitors produce significant but unpredicted changes. An agent may decrease the total number of blood vessels and increase vessel caliber without changing vascular permeability125. These changes do not always predict clinical efficacy as indicated by improved survival, however. Also, evidence of decreased tumor vascularity in imaging studies may not be accompanied by corresponding histological changes. Therefore, more effective surrogate markers of angiogenic inhibition are needed.
Dual-function imaging techniques are emerging. Contrast agents targeted to proliferating or activated endothelial cells offer the possibility of not only imaging sites of angiogenesis but also targeting drugs directly to those sites126. Technetium-labeled endostatin has been used to mark tumors in preclinical models127. Liposomal nanoparticles that bind v3 integrin can deliver a mutant Raf gene that induces endothelial cell apoptosis in a preclinical model128. Such nanoparticles could also be labeled with imaging agents, permitting simultaneous imaging and treatment of tumors126,
129. This approach raises the possibility of closer links between imaging and treatment in the future.
Conclusions
Imaging technologies have a central role in understanding the mechanisms and functional implications of angiogenesis. This began in the laboratory with microscopic methods, and now offers great promise in the clinic. The combination of microscopic and clinical approaches is providing a remarkably detailed story about the peculiarities of angiogenic blood vessels and beginning to yield some clues of the action of angiogenesis inhibitors. An ongoing dilemma is, however, that microscopic methods provide their highest resolution on preserved tissue specimens, whereas clinical methods can image living tissue deep within a tumor but at a resolution too low to see vessels of the microvasculature. Questions remain on how to bridge the resolution gap, and which microscopic criteria are the best benchmarks for interpreting clinical imaging data. Noninvasive imaging methods that assess blood flow, blood volume or vascular permeability are still relatively insensitive and may not detect functionally significant changes in structure or gene expression in endothelial cells or pericytes. Nor can many of the methods used in preclinical studies be used in patients. Although it would be desirable to have reliable surrogate markers of angiogenesis in blood or urine, the links between current surrogate markers and drug efficacy are largely indirect and are only beginning to be evaluated. There is an urgent need for a more detailed understanding of changes in tumor blood vessels after treatment. There is also a need for more sensitive indicators of the efficacy of angiogenesis inhibitors. Macromolecular contrast agents and molecular imaging probes that specifically target activated or apoptotic endothelial cells will greatly improve the utility of clinical imaging. Ultimately, markers should make it possible to pinpoint and monitor specific molecular and cellular actions of angiogenesis inhibitors. Imaging will be involved in making this happen.
Note added in proof: For more information on the feasibility of using both endothelial cells and pericytes of tumor vessels as targets in antiangiogenic therapy, see Bergers, G. et al., J. Clin. Invest.
111, 1287−1295 (2003).
REFERENCES
- Brown, E.B. et al. In vivo measurement of gene expression, angiogenesis and physiological function in tumors using multiphoton laser scanning microscopy. Nat. Med. 7, 864-868 (2001). | Article | PubMed | ChemPort |
- Weissleder, R. Scaling down imaging: molecular mapping of cancer in mice. Nat. Rev. Cancer 2, 11-18 (2002). | Article | PubMed | ChemPort |
- Pearlman, J.D., Laham, R.J., Post, M., Leiner, T. & Simons, M. Medical imaging techniques in the evaluation of strategies for therapeutic angiogenesis. Curr. Pharm. Des. 8, 1467-1496 (2002). | PubMed | ChemPort |
- Neeman, M. Functional and molecular MR imaging of angiogenesis: seeing the target, seeing it work. J. Cell. Biochem. Suppl. 39, 11-17 (2002). | PubMed |
- Costouros, N.G., Diehn, F.E. & Libutti, S.K. Molecular imaging of tumor angiogenesis. J. Cell. Biochem. Suppl. 39, 72-78 (2002). | PubMed |
- Weissleder, R. & Ntziachristos, V. Shedding light onto live molecular targets. Nat. Med. 9, 123-128 (2003). | Article | PubMed | ChemPort |
- Less, J.R., Skalak, T.C., Sevick, E.M. & Jain, R.K. Microvascular architecture in a mammary carcinoma: branching patterns and vessel dimensions. Cancer Res. 51, 265-273 (1991). | PubMed | ChemPort |
- Konerding, M.A., Miodonski, A.J. & Lametschwandtner, A. Microvascular corrosion casting in the study of tumor vascularity: a review. Scanning Microsc. 9, 1233-1244 (1995). | PubMed | ChemPort |
- Schlaeger, T.M. et al. Uniform vascular-endothelial-cell-specific gene expression in both embryonic and adult transgenic mice. Proc. Natl. Acad. Sci. USA 94, 3058-3063 (1997). | Article | PubMed | ChemPort |
- Motoike, T. et al. Universal GFP reporter for the study of vascular development. Genesis 28, 75-81 (2000). | Article | PubMed | ChemPort |
- Gale, N.W. et al. Ephrin-B2 selectively marks arterial vessels and neovascularization sites in the adult, with expression in both endothelial and smooth-muscle cells. Dev. Biol. 230, 151-160 (2001). | Article | PubMed | ChemPort |
- Miettinen, M., Lindenmayer, A.E. & Chaubal, A. Endothelial cell markers CD31, CD34, and BNH9 antibody to H- and Y-antigens--evaluation of their specificity and sensitivity in the diagnosis of vascular tumors and comparison with von Willebrand factor. Mod. Pathol. 7, 82-90 (1994). | PubMed | ChemPort |
- Schmidt, D. & von Hochstetter, A.R. The use of CD31 and collagen IV as vascular markers. A study of 56 vascular lesions. Pathol. Res. Pract. 191, 410-414 (1995). | PubMed | ChemPort |
- Weidner, N. Intratumor microvessel density as a prognostic factor in cancer. Am. J. Pathol. 147, 9-19 (1995). | PubMed | ChemPort |
- Hlatky, L., Hahnfeldt, P. & Folkman, J. Clinical application of antiangiogenic therapy: microvessel density, what it does and doesn't tell us. J. Natl. Cancer Inst. 94, 883-893 (2002). |
|
