ANGIOGENESIS IS THE PROCESS by which new blood vessels are formed. This process occurs physiologically during wound healing, embryogenesis, and corpus luteum formation. More recently, angiogenesis has been recognized as a key element in the pathophysiology of tumor growth and metastasis (1). Once tumors grow beyond a diameter of 1–2 mm, passive diffusion is no longer sufficient to support the viability of malignant cells, and neovascularization becomes a necessity (2). Tumors that grow beyond the occult stage, therefore, are capable of activating the “angiogenic switch.” To this end, research has focused on the development of antiangiogenic and antivascular agents. Antiangiogenic drugs fall into two categories: analogs of endogenous angiogenesis inhibitors (e.g., endostatin and angiostatin), and inhibitors of proangiogenic factors (e.g., antivascular endothelial growth factor) (1). Antivascular agents are targeted against more mature vessels and, as such, target the vessel wall or are generally cytotoxic to endothelial cells.
This promising area of oncologic research has brought with it a demand for an accurate means of diagnosing tumor angiogenesis and monitoring treatment response aside from physical changes in the size of the tumor, which lag behind the physiological changes. Diagnostic candidates include blood levels of angiogenic markers (e.g., vascular endothelial growth factor (VEGF) (3)), circulating endothelial progenitor cells, biopsy specimens, and various imaging techniques. Currently, circulating angiogenic serum markers lack sufficient sensitivity and specificity but clearly hold great potential. New methods of measuring angiogenesis that include assessment of circulating endothelial progenitor cells are very promising and correlate with treatment response (4). Biopsy specimens are extremely information-rich but suffer from sampling bias and inherent invasiveness. The traditional “gold-standard” measure of angiogenesis is the histological estimate of microvascular density (MVD). The MVD quantifies the average number of microvessels within a selected microscopic field. However, this method is inherently invasive and cannot determine whether flow is present or whether the vessel is hyperpermeable. Additionally, tumors are heterogeneous. Angiogenesis is often maximal in the tumor periphery, and thus MVD measures will vary according to the location from which the biopsy is taken, and may under- or overestimate the degree of angiogenesis. Imaging modalities offer the ability to noninvasively sample the entire tumor. Imaging can provide a noninvasive means of detecting angiogenesis within and about the perimeter of the whole tumor, while giving functional information. The ability to accurately monitor angiogenesis response to therapy means that drug efficacy can be established at a very early stage of treatment, much before traditional criteria, such as size change, become apparent; likewise, nonresponders may be detected and management plans altered.
The modalities that have been proposed for imaging angiogenesis include MRI, computed tomography (CT), positron emission tomography (PET), single photon emission-computed tomography (SPECT), ultrasound (US), and optical imaging. CT, PET, and SPECT all expose the patient to ionizing radiation, which is particularly problematic when repeat scans are needed to monitor therapy. In the research environment, there are mandated quarterly and annual limits on the amount of radiation a research subject can receive for diagnostic purposes, which may be reached early in a study. CT has the virtue of producing images in which density measurements are directly proportional to the iodine concentration contained in the injected contrast media. This considerably simplifies image analysis; however, there is an inevitable burden of radiation exposure associated with repeatedly imaging over the same anatomical area and at repeated time points. Among PET studies, 15O water is excellent at estimating tumor perfusion noninvasively. However, such studies are cumbersome to perform because of the short half-life (approximately two minutes) of the agent, which necessitates an on-site cyclotron, and the poor imaging characteristics of dynamic PET images leading to low-resolution images. Fluoro-Deoxy-glucose (FDG)-PET alone has proven inadequate in the direct assessment of the tumor microvasculature, although it is possible to infer the state of the vasculature by the metabolic uptake of the tumor as depicted by the FDG-PET scan. PET and SPECT can measure picomolar concentrations of tracer molecules, but there are limited commercial radiotracers available for this task, and at best they offer low spatial resolution (5). There is a growing list of PET and SPECT agents that may be suitable for imaging angiogenesis, but none are routinely used in the clinic.
While it is safe to perform, US is highly operator-dependent. However, exciting progress is being made with the development of targeted microbubbles as contrast agents (6). The microbubbles are generally several microns in diameter, so they tend to remain confined within the vascular space. This property makes contrast-enhanced US (CEUS) a good measure of vessel perfusion and blood volume, rather than permeability. A major virtue of microbubbles is that they can be destroyed by increasing the applied US energy, thus permitting assessment of the refill rates, with good temporal resolution. Optical imaging is limited by its inability to penetrate more than a few centimeters below the skin surface, even with the best available agents (7). Even at this depth the dispersion of light renders spatially accurate imaging very challenging.
Among these modalities, MRI stands out by offering the greatest practical potential for imaging angiogenesis. It is widely available, is often used concurrently to assess tumor size and stage, does not involve radiation exposure, and the side effects of the gadolinium (Gd)-based contrast agents are well established and minimal. The installed base of MRI units is large and growing, and it is therefore accessible to the majority of patients in the developed world. MR is also an evolving field, with new scanning technologies and techniques continuing to emerge, allowing MRI to reflect not only anatomic abnormalities but also functional abnormalities. Meanwhile, MRI is moving from its historical role as a descriptive, qualitative technique to a more quantitative and reproducible technique. One of the keys to this will be the standardization of scanner types, pulse sequences, and analytic methods.
Broadly speaking, MRI techniques can be divided into two groups: 1) those that use extrinsic contrast agents (e.g., DCE-MRI and MRI with targeted agents) and 2) those that do not (e.g., arterial spin labeling (ASL) and blood oxygenation level-dependent (BOLD) imaging). In this review we will summarize the role of these MRI-based imaging techniques in monitoring angiogenesis, along with their respective advantages and disadvantages.
NON-CE IMAGING METHODS FOR ANGIOGENESIS
ASL is a noninvasive MRI technique for measuring tissue perfusion that uses radiofrequency (RF) pulses to invert the nuclear spin of inflowing arterial protons (8). This “magnetically” labels the blood before it flows into the region of interest (ROI), without the need for external contrast agents. This “tagged” blood then exchanges with water in the tissue of interest, changing its magnetization. The effect is small and cannot be measured directly; however, it is highly quantitative. To overcome the problem of small differences between the tagged and the untagged states, subtraction images are generated whereby ASL images are compared with non-ASL images in order to produce perfusion maps. The RF pulses can be applied either instantaneously, using a pulse with relatively low RF power, or continuously (9). Continuous ASL (CASL) employs a weaker RF field to repeatedly saturate blood water spins and, in theory, produces a greater signal-to-noise ratio (SNR). ASL may benefit from higher field strength MRI, such as 3.0 or 4.0 Tesla, where the increased spin magnetization and longer T1 relaxation could lead to an enhanced signal difference (10). However, higher field strengths will also result in higher specific absorption rate (SAR) values (the maximum value allowed is 4 W/Kg of body weight). This affects ASL more than DCE-MRI because there is the additional tagging pulse. One way to address this is to use a dual-coil, phase-encoded approach whereby a second smaller, localized coil is used for tagging. ASL has mainly been used for brain applications because it is relatively easy to “tag” the carotid vessels supplying the brain in the neck. Brain perfusion measured by ASL is reliable and reproducible, provided that the SNRs are not too low (11). A potential source of artifact with ASL is the length of transit time from the region of tagging to the ROI wherein the “tagged” spins lose their tagging due to T1 relaxation. This was first demonstrated in models of ischemic stroke, where the formation of collateral circulations was underestimated and was the primary source of vascular artifact in ASL, potentially causing errors in the calculation of perfusion values (12) (Fig. 1).
ASL can be used to generate regional maps of cerebral blood flow, and as a result initial research has centered on functional brain imaging. Increases in blood flow and volume have been proposed as markers of the degree of neovascularization within tumors, due to the increased size and number of vessels (and hence higher MVD). Thus, ASL may provide a noninvasive means of monitoring tumor angiogenesis indirectly by measuring blood flow. ASL has been used to calculate blood flow within brain tumors (13). Kimura et al (14) demonstrated ASL's ability to characterize blood flow in meningiomas. They demonstrated a significant correlation between the mean CASL signal intensity change and the histologically-derived microvessel area, thus showing the ability of ASL to assess angiogenesis and the microcirculation. Wolf et al (15) demonstrated a strong correlation between the grade of glioma and measurements of maximal tumor blood flow. CASL has also been used to measure blood flow as an outcome variable in 9L gliosarcoma rat models, with differences in VEGF-A expression (16). The VEGFs are a family of cytokines that stimulate the endothelium to proliferate, and are critical for angiogenesis. This group showed that under- (VEGF-A−) or overexpression of VEGF-A (VEGF-A+) in these tumors resulted in significant changes in tumor perfusion. Both VEGF-A− and VEGF-A+ showed elevated tumor blood flow, but only the VEGF-A+ group exhibited increased blood volume. This demonstrates that blood flow, as measured by ASL-MRI, is an important biomarker of angiogenesis, and that blood volume as determined by ASL is an independent variable. Interestingly, underexpression of VEGF-A led to an up-regulation of VEGF-D, emphasizing the fact that multiple biochemical pathways are involved in angiogenesis, and that tumors are highly adaptable, recruiting other angiogenic growth factors when one is blocked or inhibited (Fig. 2).
The issue of transit-time artifact has limited the application of ASL outside the brain. This is less of a problem for renal imaging, where the ROI is close to a main artery, but is potentially a greater challenge for more peripheral lesions or lesions in the liver, where dual blood supplies may be present. Methods to reduce this source of artifact include velocity-encoded ASL, where arterial spins are labeled throughout the field of view (FOV), including the ROI, but are distinguished according to their speed. For the control scans the same sequence is used, but a lower gradient is employed (17).
ASL has been applied to many areas of the body. For instance, Tempel and Neeman (18) used ASL to demonstrate vascular remodeling and angiogenesis within the ovary during the menstrual cycle. They mapped perfusion in the ovary and measured blood velocity within the ovarian artery, both before and shortly after ovulation. The results obtained were consistent with previous studies in showing increased preovulatory perfusion, due to the induction of follicular maturation, coupled with decreased postovulatory blood flow in the luteal phase. ASL has also been used to assess peripheral tumors, such as renal cell carcinoma (RCC) (19). Boss et al (20) showed ASL to be equivalent to CE T1 images in monitoring RCC recurrence following RF ablation therapy. ASL can be performed without exogenous contrast agents, which makes it available to patients with severe renal dysfunction or MR contrast-agent allergies.
ASL-MRI is a reproducible method for measuring tissue blood flow. Although initially used for functional brain imaging, it has the potential to quantify blood flow within tumors. Although it has been mainly limited to brain imaging, newer techniques may allow more accurate body imaging. ASL does not require external contrast agents, and thus may be suitable for performing repeated studies to monitor angiogenesis within tumors and response to treatment.
BOLD imaging takes advantage of the fact that the magnetic properties of hemoglobin differ depending on its oxygenation state. Deoxyhemoglobin is inherently paramagnetic, whereas oxyhemoglobin and tissues are diamagnetic. The presence of deoxyhemoglobin in the blood results in a mismatch between the magnetic susceptibility of tissue and blood that distorts the static magnetic field and shortens the T2* of the blood and its surroundings. When the level of deoxyhemoglobin in the blood decreases with increasing oxygenation, the mismatch is reduced and the T2* increases. Heavily T2*-weighted gradient-recalled echo (GRE) and echo-planar imaging (EPI) sequences can demonstrate this BOLD contrast effect, with a relatively higher signal produced by oxyhemoglobin compared to deoxyhemoglobin (21). Thus, oxyhemoglobin acts as an endogenous BOLD contrast agent in the presence of oxygenated tissue.
BOLD was originally developed as a technique for functional MRI (fMRI). The altered deoxy/oxyhemoglobin ratio during brain activity leads to a change in signal (22). The magnitude of the effects is relatively small (<5%) and requires statistically-based analytic programs to be reliably quantified. While fMRI has been extremely successful in brain activation studies, it has proven more difficult to apply BOLD methods to tumors. In order to induce a change in the deoxy/oxyhemoglobin ratio, cancer-imaging studies employing BOLD sequences use either inhaled 100% oxygen or carbogen (95% oxygen, 5% CO2) as a method of “challenging” tissue (23). BOLD signal relies on the degree of signal produced by oxyhemoglobin; however, this will in turn depend on a complex interaction between changes in blood volume, blood flow, and tissue oxygenation consumption (24). Most BOLD techniques compare changes following challenges with an inhaled gas. However, it should be noted that the BOLD response to carbogen is complex, and its relationship to angiogenesis is not always clear. The high oxygen level induces increased blood oxygenation saturation; however, the relative hypercapnia causes blood vessels (excluding pulmonary vessels) to dilate, hence increasing blood flow. Thus, there are several potential mechanisms for increasing the “BOLD” image intensity, e.g., increased oxygenation or decreased deoxyhemoglobin due to increased blood flow (21). Thus, the carbogen-BOLD effects reflect both changes in blood flow and oxygenation. Blood flow can be estimated because increased flow will lead to a decreased tissue extraction of oxygen; however, it is difficult to distinguish flow effects from reversible hypoxia. One such method is to use another imaging technique, such as Doppler US to demonstrate the blood flow contribution. Alternatively, T2* maps can be generated and kinetic modeling used to estimate the respective contributions of blood flow and deoxyhemoglobin levels to signal intensity change (21). (Fig. 3)
As with ASL-MRI, the ability of BOLD-MRI to noninvasively demonstrate blood flow has been investigated as a means of indirectly measuring angiogenesis within tumors. The technique is limited by susceptibility artifacts, particularly in the chest and abdomen, where air in the lungs and the bowel can lead to distortions in the image. Another problem is the poor spatial resolution of the technique due to the rapid acquisition of long-echo time (TE) images (25). BOLD imaging has also been shown to be influenced by body temperature, with a lower core temperature leading to reduced blood oxygenation and hence reduced signal (26). It can also be influenced by medications, including antihistamines, caffeine, and levodopa (27).
Gilead et al (28) used carbogen-BOLD imaging and histology to investigate angiogenesis in xenografted ovarian tumors at different stages of growth. Histologic staining was employed for alpha-SMA to help derive the measure of the “vascular maturation index,” which was shown to be significantly lower in the tumor. BOLD was shown to successfully map this heterogeneity of mature vessels. Initially new vessels were predominantly found in the periphery. Infiltration of alpha-SMA stroma cells into the tumor was associated with increased vascularization and tumor growth, implying that this is an important component of the angiogenic switch. Gilad et al (29) used BOLD imaging in normoxic (air) and hypercapnic (air/5% CO2) conditions to investigate the initiation of angiogenesis by either tumor cells or infiltrating stroma cells, and the role of different angiogenic factors. BOLD MRI was able to predict vascular maturation by measuring the degree of vasoreactivity to induced hypercapnia, and was validated by histological staining for alpha-SMA. In addition, in situ hybridization demonstrated that VEGF was expressed by the tumor cells, while the angiogenic factors angiopoietin-1 and -2 were expressed only by the infiltrating host stroma cells. Ludemann et al (30) used BOLD to estimate intratumoral blood volume by measuring the mean BOLD signal amplitude in patients with brain tumors while they performed tasks, and compared the results with those obtained from the contralateral hemisphere and control patients. They found a signal reduction within tumor regions, which correlated to the total intratumoral blood volume. Thus, the BOLD signal accurately reflected reduced tumor perfusion, which the authors attributed to the “steal phenomenon.”
Increasing tumor oxygenation is a means of radiosensitizing tumors prior to radiotherapy. BOLD has been shown to be a useful way to monitor this preprocedure sensitization in meningiomas (31) and rhabdomyosarcomas (32). The latter study showed that the addition of a perfluorocarbon emulsion to carbogen increased the BOLD signal by inducing greater hyperoxemia and further increasing the dissolved oxygen levels. BOLD was able to highlight differences in perfusion in prostate cancer between the center and the periphery of the tumor, and demonstrate intra- and intertumor heterogeneity, and was shown to be more rapidly reversible than DCE-MRI. This latter factor may make BOLD more suitable for accurately monitoring treatments that are designed to cause vascular disruption (33).
BOLD imaging has been more extensively investigated than ASL. It does not require exogenous contrast agents; however, studies often require the inhalation of a carbogen gas mixture, which can be cumbersome. For instance, carbogen is not a routine medical gas and must be especially ordered. Moreover, some patients exhibit extreme discomfort within the magnet due to the breathing apparatus and the feeling of “air hunger” induced by elevated carbon dioxide levels. Techniques can be applied in order to determine the relative contributions of blood flow and oxygenation to signal intensity and, indeed, demonstrate changes in tumor blood flow. The reproducibility of this method and its complexity are disadvantages that will delay its translation to the clinic. However, this methodology has excellent potential in the field of radiation oncology for delineating, and possibly manipulating, areas of oxygenation within tumors.
CE IMAGING METHODS FOR ANGIOGENESIS
The angiogenic process is heterogeneous within tumors, with some vessels demonstrating maturity and other vessels demonstrating incomplete layers with high permeability and fragility. As previously mentioned, angiogenic vessels have large gaps between the endothelial cells, the endothelium, and the basement membranes, as well as between the basement membrane and the pericytes, making the vessels hyperpermeable to many macromolecules. These properties can be exploited by DCE-MRI. MR contrast agents that leak slowly through the normal vasculature are able to pass more quickly through tumor vessels to produce differential enhancement. This results in a fast “wash-in” of contrast coupled with the rapid “wash-out,” and allows a functional analysis of the tumor microcirculation. DCE-MRI is performed with low-molecular-weight contrast media (LMCM) or macromolecular contrast media (MMCM). Gd-diethylenetriamine pentaacetic acid (Gd-DTPA) is the LMCM agent with the longest clinical track record and has a molecular weight (mW) of 567 Daltons (Da). MMCM range in mW but typically have mW > 30,000 Da (34). The majority of DCE-MRI studies rely on LMCM because of their clinical availability, but MMCM are being increasingly investigated.
DCE-MRI With LMCM
Over the last decade, LMCM DCE-MRI has been increasingly used to investigate angiogenesis within tumors, and in particular the response to antiangiogenic therapy. These agents include gadopentetate dimeglumine, gadoteridol, gadodiamide, and gadoterate meglumine. The longest experience is with gadopentetate dimeglumine (Gd-DTPA), which was approved in the United States in 1988. Gadobenate dimeglumine (Gd-BOPTA) was recently approved in the United States and has some protein-binding characteristics that make its kinetics somewhat more complex. Gadobutrol (or Gadovist) is available in some countries (35). This agent has been approved at 1.0-M concentration, which has theoretical advantages for perfusion imaging (35). However, except for those that demonstrate protein binding, these agents are thought to perform similarly with regard to their first-pass kinetics and clearance. After data acquisition, kinetic models can be applied in order to derive estimates of tissue perfusion and permeability based on the shape of the tumor wash-in and wash-out curves. Standardized terminology and analytic methodologies are required to achieve accurate comparability between centers. Tofts et al (35) proposed the terms Ktrans, kep, fpV, and ve as outcome parameters derived from a two-compartment general kinetic model, which is the most widely accepted model and which can readily be derived from first principles. Another semiquantitative measure that is often used is the initial area under the Gd concentration curve (AUGC). Ktrans represents the rate of contrast agent transfer from blood to interstitium, and kep is the reverse rate constant, representing backflow. The term fpV is the fraction of plasma volume, related to whole tissue volume, and ve is the fractional extravascular, extracellular leakage volume. These parameters can be depicted numerically or as color-encoded images. DCE-MRI parameters have been shown to correlate with vascular permeability, and hence angiogenesis, within tumor tissue (37).
DCE-MRI can be used to demonstrate the antiangiogenic effects of drugs early after their administration, and can predate traditional treatment response parameters such as changes in tumor size. This may be of benefit in selecting nonresponders at an early stage or in selecting ideal candidates for a particular drug. In this capacity, DCE-MRI was initially proven in a number of preclinical studies. Examples include the work by Marzola et al (38) using SU6668, a drug that targets VEGF, fibroblast growth factor (FGF), and PDGF tyrosine kinase receptors. They used an HT29 human colon carcinoma xenograft model in mice, and showed a 51% (tumor rim) and 26% (core) decrease in the average vessel permeability after only 24 hours of treatment. There was a 60% tumor growth inhibition after 14 days of treatment. DCE-MRI has been extensively studied in human clinical trials in which various antiangiogenesis drugs were investigated in patients with progressive tumors. Subsequently, a number of clinical trials have used DCE to prove drug efficacy. Wedam et al (39) recently demonstrated a reduction in Ktrans, kep, and ve in patients with locally invasive breast cancer after a single cycle of bevacizumab treatment (a monoclonal antibody to VEGF), which was correlated to a reduction in VEGF-receptor expression and increased apoptosis. PTK/ZK is an inhibitor of three separate VEGF-receptor tyrosine kinases that has been tested clinically. After administration of this drug, Lee et al (40) used DCE-MRI to demonstrate the significant reduction in AUGC values induced in patients with colon cancer patients and liver metastases, with results comparable to animal models. DCE-MRI can show response to PTK/ZK therapy as early as two days after therapy with significant reductions in AUGC (41) or permeability parameters (42), which also predicts subsequent response.
In addition, Xiong et al (43) treated patients with solid tumors with SU6668 and demonstrated efficacy via reductions in AUGC and maximum contrast concentration on DCE-MRI. LMCM DCE-MRI has also shown significant reductions in permeability values in patients treated with the antivascular agents AG-013736 (44) (an inhibitor of the VEGF, PDGF, c-Kit receptor tyrosine kinases), and SU5416, a selective inhibitor of VEGFR-2 tyrosine kinase activity (45). The latter study showed no significant change in normal liver over the two-month treatment course, despite a significant decrease in the AUC and maximum contrast concentration being observed within the tumors. (Fig. 4)
Consensus is still lacking on the exact kinetic model to be used in analyzing DCE-MRI studies. The differences among the various methods are often small, and the Cancer Imaging Program of the National Cancer Institute is attempting reach an international standard for kinetic modeling of DCE-MRI studies (46). Despite its simplicity, DCE-MRI is still not widely available, and protocols need to be standardized in order for studies to be comparable among institutions. However, DCE-MRI is rapidly emerging as the imaging technique of choice for monitoring clinical response in trials of new antiangiogenic and antivascular therapies.
DCE-MRI With MMCM
MMCM generally range in mW from 5 kDa-90 kDa, and both Gd- and iron-oxide-containing agents are included in this category (34). MMCM were initially designed for prolonged intravascular retention for use with MR angiography (MRA). Unlike LMCM, they do not pass through normal endothelia, and thus are potentially more suitable for selective imaging of the tumor neovasculature, which tends to be highly permeable to larger molecules.
The increased size of MMCM means that they have different pharmacokinetic properties compared to LMCM. The extent to which Ktrans is dependent on flow and the permeability-surface area product varies. Where blood flow is low, the amount of contrast that leaks into the extracellular space will depend primarily on the rate at which it is supplied to the endothelium (i.e., flow), and Ktrans will reflect blood flow, a “flow-limited” state (47). If contrast delivery is not limited by perfusion, the “leakiness” of the endothelium predominates, and Ktrans reflects permeability to a greater extent in this “permeability-limited” state (48). Thus, high values of Ktrans indicate high permeability and probably high perfusion, while low values of Ktrans indicate either low permeability and/or low perfusion. Regardless of its exact meaning, a change in Ktrans likely represents a genuine pharmacodynamic effect of the drug. The size or “diffusibility” of the contrast agent is also important. LMCMs are freely diffusible tracers that yield a Ktrans proportional to the blood flow, whereas MMCMs are minimally diffusible tracers, where Ktrans is proportional to permeability. Thus, the increased size of MMCMs makes them less diffusible, and Ktrans values may more accurately reflect permeability within tumors. In addition they are excellent blood pool agents, so they can give more accurate estimates of tumor blood volume. The Ktrans of MMCMs is inversely related to their mW (and hence their diameter); however, this relationship is not linear. de Lussanet et al (49) showed this using monodispersed Gd-dendrimers of different generations (i.e., sizes) ranging from 0.7–51 kDa. A doubling of mW led to a 25% decrease in Ktrans. It must be noted that it is really the hydrodynamic diameter that is of importance, not necessarily the mW, since two molecules of equal mW can have radically different hydrodynamic diameters. MMCM have a delayed clearance from the body because they are not cleared exclusively through the kidneys. This may pose a problem for Gd-containing compounds, where prolonged retention of Gd could, in theory, lead to toxicity from dissociation of the Gd ion from its chelate. This is still a theoretical concern as toxicity studies have shown little cause for disquiet if the renal and hepatic function is normal or only mildly impaired. Iron-based compounds pose different problems. The body is able to naturally process the element; however, the length of time that iron is retained after contrast administration has led to concerns about iron overload. In addition, iron nanoparticles have been shown to have their own toxicities, usually related to anaphylaxis reactions; therefore, they must be injected slowly while the patient is monitored. (Fig. 5)
Albumin-(Gd-DTPA) is the prototype MMCM because of the ubiquity of albumin and its well known properties, but problems related to obtaining uncontaminated sources of human serum albumin have, until recently, limited enthusiasm for this method. Additionally, Gd-albumin tends to aggregate unless it is formulated in a way that inhibits this. Moreover, a major obstacle to clinical use was the difficulty in binding Gd ions to albumin with sufficient affinity to prevent free Gd ions from being liberated. This problem has been solved by new bifunctional chelates that link albumin to Gd with a high binding affinitiy. Gd-labeled albumin represents a very plausible MMMR contrast agent (50) but is not yet being commercially developed. A number of other macromolecular carrier compounds have been investigated, including MS-325 (an agent that reversibly binds albumin in vivo; generic name: gadofosveset trisodium), dextran compounds, viral particles, dendrimers, perfluorocarbon emulsions, liposomes, and iron-oxide compounds. Most MMCM studies have been performed only in animal models, although MS-325, dextrans, and iron-oxide compounds are being tested in human clinical trials. There are too many MMMR agents to cover in detail, but an overview of the most common MMCMs is presented below.
Albumin-(Gd-DTPA) is helpful in characterizing the microvessels of a wide range of tumors. Marzola et al (51) have shown that albumin-(Gd-DTPA) correlates with MVD and immunohistochemistry and is comparable to LMCM DCE-MRI following treatment with SU6668 and SU11248 (an inhibitor of the VEGF, PDGF, KIT, and FLT3 tyrosine kinase receptors) (44). Daldrup et al (52) showed that results from albumin-(Gd-DTPA)-enhanced MRI correlated with the histological grade of breast cancer, whereas poor correlations occurred with LMCMs. Another class of MMCM is liposomes, which are polydispersed spherical vesicles with one or more phospholipids bilayers and a hydrophilic interior. Liposomes can incorporate either Gd (53) or iron particles (54) as paramagnetic species, and are preferentially taken up by the liver and spleen, although coating them with polyethylene glycol reduces liver uptake (55). Although they have yet to be used in human trials, animal models have shown their potential for imaging metastatic lymph nodes (56). They have also been used for drug delivery and as targeted contrast agents (vide infra). Unfortunately, it is difficult to achieve uniformity of size during the synthesis of liposomes. Polydispersity is a major disadvantage as it complicates the analysis of permeability and clearance, making reliable DCE comparison between studies problematic. Dendrimers are another class of MMCM. They are highly branched, synthetically produced spherical polymers. In contrast to liposomes, these polymers can be produced in relatively uniform sizes as different “generations.” Two major types of dendrimers are commercially available: polyamidoamine (PAMAM) and diaminobutane core polypropylimine (DAB or PPI). After injection, the smaller dendrimer generations are excreted rapidly via the kidney and are too small to be considered MMCM, although they may be useful in assessing renal function (57). The G4 and G5 dendrimers (5–8 nm in diameter) are able to selectively leak through hyperpermeable tumor vessels. The larger dendrimer-based contrast agents demonstrate good vascular enhancement but minimal leakage from tumor vessels into surrounding tumor tissue (58). Gd-labeled dendrimers can enhance tumor vasculature in murine models (59) and have demonstrated the early effects of radiation therapy on tumor permeability (60), but they have yet to be tested in humans. Their multiple binding sites will potentially make them suitable for drug delivery and as targeting agents. Viral particles studies are at an early stage but hold promise. The virus' nucleic acids can be removed to leave an empty “protein cage” for the incorporation of paramagnetic MRI species and, indeed, therapeutic agents. They can be readily synthesized and are known to be biocompatible.
MS-325 is an agent that reversibly binds albumin and was the first MMCM to be used in human trials (61). Only the albumin-bound form can be considered an MMCM, but the reversible nature of the Gd binding allows Gd to be more rapidly cleared from the body. MS-325 has been used in clinical trials as an MRA agent (62), but has yet to be investigated for angiogenesis assessment in humans. Turetschek et al (63) used a murine breast tumor model to compare MS-325 to albumin-(Gd-DTPA) in one of the few studies of MS-325 for angiogenesis imaging. MS-325 showed no correlation to MVD or tumor grade, although it should be noted that the pharmacodynamic properties of MS-325 vary greatly among different species due to the differing compositions of albumin and its binding affinity for MS-325 across species.
Iron-Based MMCM Agents
Iron-oxide nanoparticles are often used as MMCM. It is difficult to avoid oxidation of iron molecules; hence, most particles are iron oxides. Moreover, since iron oxides have a tendency to aggregate, the particles must be coated in order to reduce this. Dextran provides an effective coating for iron-oxide particles without interfering with the superparamagnetic properties of the nanoparticle. Dextran has an excellent safety profile and has been used in human studies, primarily for MRA (64) and cardiac perfusion studies (65). A number of different iron-oxide compounds have been developed as MMCM. Iron oxides are superparamagnetic and they predominantly act to shorten T2* relaxation time to produce a “negative” enhancement. They vary in size, which alters their pharmacokinetics. Superparamagnetic iron oxide (SPIO) particles are usually 50–150 nm in diameter. They have a polycrystalline core coated with either dextran or silica, and are mainly taken up by phagocytic cells within the reticuloendothelial (RES) and lymphatic systems. Tumor cells within these areas contain fewer phagocytic cells and therefore less iron particles. This makes them appear brighter than surrounding tissue, and helps in the diagnosis of metastases. Ultrasmall SPIO (USPIO) are 10–50 nm in diameter. Their smaller size means that they are taken up more slowly by the RES (66). USPIOs have also been used to image the angiogenic processes in murine breast cancer models, with results for Ktrans shown to correlate with both tumor grade (67) and histological MVD (68). Very small SPIO particles (VSOP) have been developed more recently, and have a diameter of 2–10 nm. They may prove more useful in the field of MRA than in angiogenesis imaging (69).
Stem Cell Imaging
Another approach to the imaging of angiogenesis is to label endothelial progenitor stem cells and track their behavior in vivo. These cells naturally track to regions of increased endothelial proliferation and areas of neovascularization, and hence provide a method of imaging angiogenesis. Iron particles are preferable to Gd ions due to their better toxicity profile and the fact that even single cells loaded with iron particles are readily detected by MRI (70). In theory, the harsher environment of the lysosome or endosome, with its lower pH, could lead to dechelation of the Gd and release free, highly toxic Gd ions. SPIO particles can be incorporated into endosomes within the stem cell cytoplasm using protamine sulfate (71). Another method of incorporating iron into cells, termed “magneto-electroporation,” uses electroporation techniques adapted from those used to transfect cells with viruses or DNA (72). Once incorporated into stem cells, the iron particles allow in vivo tracking of the cells with MRI. Progenitor cells can be detected within tumors as early as three days post-intravenous injection (73). Preliminary research has shown that the iron particles are not toxic to mesenchymal stem cells labeled with feruxomide-protamine sulfate and do not affect their ability to function, proliferate, or differentiate, and that the cells can continue to be imaged during their lifespan (74, 75). However, it should be noted that groups using other transfection agents, such as poly-L-lysine, have demonstrated an inhibition of stem cell differentiation (76). Iron is essential for a number of physiological processes in the body, but is not without potential side effects. Intracellularly, iron is required in its free state (Fe2+ and Fe3+) for cell growth (77). However, increased levels of unbound iron can lead to formation of reactive oxygen species, which may also lead to cell death (78). In theory, the introduction of ferumoxide complexes into cells may increase the formation of reactive oxygen species and hydroxyl-free radicals; nevertheless, in preliminary studies cell survival did not appear to be affected (71). While it is unlikely at the doses administered, there remains the theoretical risk of iron overload, where free iron in the bloodstream can convert hydrogen peroxide to free-radical ions that attack cellular membranes, protein, and DNA. Rapid intravenous injection of iron-containing compounds has been associated with transient back pain and hypotension; therefore, the infusion should be performed slowly and the patients carefully observed during this period.
One disadvantage of this method is that iron is a negative contrast agent. Once the target has turned dark, the image cannot become darker; hence it is difficult to obtain a dynamic range of intensities that could quantify the number of stem cells, or, indeed, differentiate negative contrast from areas of signal void due to tissue absence or artifact (79). Some groups have attempted to differentiate iron “signal” from signal void by using a series of spin-echo images to quantify the fast-decaying T2* relaxation of tissue containing highly concentrated iron-labeled cells (80). Iron-containing tissue has a very short T2* (<2 msec), and ultrashort T2* maps can be created to distinguish the presence of iron (80). Some groups have tried to circumvent the problem of “negative” contrast by using imaging sequences such as gradient-recalled echo acquisition for superparamagnetic particles (GRASP) to create a “positive” iron contrast. These scans can be used concomitantly to distinguish between iron signal and artifact (81). Other techniques to produce positive iron signals are also being developed experimentally (82, 83). (Fig. 6)
Anderson et al (84) injected iron-loaded endothelial precursor cells into a mouse bearing a glioma, and demonstrated the trafficking of these cells to the angiogenic portions of the tumor. MRI demonstrated cell migration toward and incorporation into the tumor neovasculature. Histology confirmed the presence of these differentiated, iron-labeled cells around the tumor rim, the area where angiogenesis is maximal. Arbab et al (73) also demonstrated the ability of iron-labeled AC133 cells to migrate to xenografted glioma tumors and be present intratumorally three days post-injection in preformed tumors. However, in this experimental situation low signal intensity could not be detected until the tumor had reached a size of 1 cm. Confocal microscopy showed the cells' incorporation into the neovasculature, and immunohistochemistry confirmed the transformation of the administered cells into endothelial cells.
Although it is at a relatively early stage, this research offers the potential to image angiogenesis and monitor its progression during the cell's lifespan.
Targeted Imaging Agents
The ideal MR contrast agent would specifically target molecular markers that are present on angiogenic blood vessels, thus allowing highly specific angiogenesis detection. Conveniently, such endothelial cell surface markers are already known, having previously been discovered and used with immunohistochemical staining for the ex vivo diagnosis of angiogenesis. Some of these markers are now being used as molecular targets to aid in vivo diagnosis by MR, PET, SPECT, or optical imaging. One advantage is that these epitopes are located on the vessel walls and thus are highly accessible for intravenous contrast agents. Examples of angiogenic molecular targets include CD31 (PECAM-1), CD34 (hematopoietic progenitor cell antigen-1), E-selectin (CD62E), endoglin (CD105), endosialin (CD248), VEGFR-2 (Flk-1), αvβ1 integrin, and αvβ3 integrin (LM609) (85). An inherent problem with molecular imaging of angiogenesis using MRI is the relatively low number of these targets in comparison to background signal from unbound agent within the vessels (86). Moreover, vessels comprise only a small percentage of the total mass of a typical tumor. Thus binding to these epitopes must be not only highly specific but also highly efficient, while wash-out of unbound conjugates must be rapid. The imaging techniques used to detect these conjugates must be sensitive enough to detect very low concentrations of the agent. PET and SPECT are generally more suited to this type of imaging, but lack sufficient spatial resolution. MRI would be ideal were it not for its relatively poor sensitivity (on a molar basis) for detecting vascular receptors. This can be overcome by producing MRI compounds that utilize monoclonal antibodies (mAb) to improve specificity, and incorporating high concentrations of magnetic agents to increase sensitivity. However, such macromolecules, due to their increased size, will take many days to clear from the blood pool.
A number of specific agents targeted to activated endothelial targets have been tested with MRI. E-selectin is overexpressed in proliferating endothelial cells, and as such can be used as a marker of angiogenesis (87). Kang et al (88) used MRI and an anti-E-selectin mAb (H18/7) linked to iron-oxide nanoparticles to demonstrate binding to human endothelial umbilical vein cells (HUVEC) in vitro. Il-8 was used to up-regulate HUVEC expression of E-selectin. The antibody conjugate retains its specificity for E-selectin, as confirmed by the 100–200-fold higher binding of conjugate in up-regulated vs. control cells. MRI of the cells was able to reveal a significant T2-weighted signal decrease for cells in the antibody targeted group. The same group recently used this method to image HUVEC cells in a mouse model in vivo (89). Il-1β was used to up-regulate E-selectin expression on the implanted cells, with the iron-oxide nanoparticle conjugates producing a three times greater MR signal reduction on T2* images in comparison to controls.
VEGF is a circulating cytokine that is known to play a vital role in angiogenesis. It mediates its effect on endothelial cells mainly via the VEGFR-2 (Flk-1) tyrosine kinase receptor. Tumor neoangiogenic endothelial cells have greater numbers of VEGFR-2 than the normal endothelium (90). In addition, some tumor cells can directly express VEGFR-2. These characteristics have been exploited to develop a number of antiangiogenic inhibitors (vide supra), but they also make VEGFR-2 an ideal target for molecular-based angiogenesis imaging (91). Backer et al (92) emphasized this potential, although they used near infrared imaging as their method of detection. They conjugated VEGF to a fifth-generation dendrimer and demonstrated significantly increased binding to VEGFR-2 in xenografted breast tumors as compared to VEGFR-2-receptor-blocked control mice. To date, little research into VEGFR-2 has involved targeted imaging; however, given the extensive work on VEGFR-2 for drug targeting, receptor imaging developments are likely to follow.
The αvβ3 is a membrane integrin and a well-recognized molecular marker of angiogenesis. It has been shown to regulate the production of VEGF in tumor cells that express it (93). αvβ3 is only weakly expressed on mature, quiescent endothelial cells, but it is strongly expressed on the activated endothelial cells of angiogenic vessels (94). The αvβ3 binds proteins that present the triple amino acid sequence arginine-glycine-aspartic acid (RGD) (95). Contrast agents can potentially target the αvβ3 integrin by expressing this peptide sequence, as demonstrated by radionuclide imaging (96). In addition, RGD multimers can increase the affinity to αvβ3, and increase the tumor-to-background ratio (97). Other molecules that can be used to target the αvβ3 integrin include NC100692 peptide (98) and the antagonist SU015 (99). LM609 is a monoclonal antibody that binds to αvβ3, and can offer a way to target the integrin more specifically. (Fig. 7) (100)
Winter et al (101) selectively targeted the αvβ3 integrin with perfluorocarbon, Gd-containing nanoparticles covalently coupled to an αvβ3 integrin peptidomimetic antagonist. The resulting complexes were relatively large, containing ∼90,000 Gd ions each, and had an approximate diameter of 270 nm. Using a rabbit tumor model two hours post-injection, the targeted conjugate increased MRI signal within tumors by 126%, predominantly in the periphery. Histology confirmed the presence of neoangiogenesis within these areas. This group used the same method to image melanoma xenografts in murine models (102). Contrast enhancement of neovascularity in animals that received the αvβ3-targeted complex increased 173% after two hours. Sipkins et al (103) used polymerized vesicles with a Gd payload to target αvβ3 via the LM609 mAb. They were able to detect angiogenic “hot spots” within rabbit tumors that were not seen with conventional MRI. Mulder et al (104) have produced Gd-liposomes with ∼ 700 RGD multimer peptides attached. These liposomes also contained fluorescein-PE to allow fluorescence imaging. In vivo T1 MR images showed enhancement, and ex vivo fluorescent microscopy demonstrated the association of the liposomes with activated tumor epithelium. Research is beginning to take place to target dendrimers to αvβ3. For instance, Shukla et al (105) have synthesized a PAMAM-G5 dendrimer conjugated to a RGD-4C ligand. Flow cytometry demonstrated the uptake of this complex by HUVEC cells expressing with αvβ3 integrins. The αvβ3 integrin has great potential and appears to be the most likely candidate as a molecular angiogenesis target. Thus, although none of these agents have been tried clinically, preclinical data supports the potential for targeted MRI of angiogenic vessels. (Fig. 8)
A problem for MRI targeted imaging is its relatively low sensitivity compared to other modalities. MRI is at least 10−6 less sensitive than PET, but this does not necessarily mean it cannot be used to detect angiogenesis markers. Assuming that the vessels make up less than 5% of a tumor, and that not every endothelial cell can be labeled, it may be difficult to achieve sufficient contrast agent concentration within the abnormal vessels to allow imaging. Fortunately, larger and more potent macromolecular contrast agents increase signal dramatically and higher field strength magnets combined with specially designed coils may allow for the detection of angiogenic vessels even at these low concentrations.
As macromolecular contrast agents continue to be developed and eventually gain FDA approval, they will be increasingly used to target the above angiogenic markers and other molecular epitopes.
The development of antiangiogenic drugs has brought a need for an accurate, noninvasive means of assessing the angiogenic process. MRI is one of a number of imaging modalities that can be employed for this purpose. MRI is widely clinically available, offers a good spatial resolution, and does not require ionizing radiation exposure. As such, it is emerging as the optimal candidate for the diagnosis and monitoring of angiogenesis, and much research has taken place to improve the technique and develop new types of contrast agent.
Many MR techniques are available but they can be divided into two main groups: those that require external contrast agents, and those that do not. ASL and BOLD are imaging methods that do not require intravenous contrast agent administration, which in itself is an advantage, and can provide indirect measures of angiogenesis. ASL directly measures blood perfusion quantitatively but is limited in its applicability. BOLD depends on both tissue oxygenation and blood flow, and is difficult to quantify with regard to flow. Its use is still experimental for this application.
DCE-MRI with low-mW Gd-DTPA exploits the hyperpermeable nature of neoangiogenic tumor vessels. A kinetic model can then be applied in order to derive parameters of permeability. This has been the most investigated method for imaging angiogenesis, and correlation with histological evidence of angiogenesis and response to treatment is well documented. Gd-DTPA DCE-MRI with LMCM is currently the most widespread and accurate means of imaging angiogenesis, and new methods will continue to be compared to this standard. There is, however, a pressing need to standardize the acquisition and analysis of DCE-MRI data to make the findings comparable from site to site and from study to study.
The use of macromolecular agents for DCE-MRI may ultimately prove more specific than Gd-DTPA for blood flow measurements. MMCMs are slowly beginning to gain clinical approval and have the added potential to incorporate drugs for concurrent tumor treatment and also utilize targeting vectors to increase specificity. A number of potential epitopes exist for the targeting of contrast agents to areas of angiogenesis. Integrins appear to offer the greatest potential, due to their relative selective expression on the “active” endothelium undergoing angiogenesis. Although it is at a relatively early stage in development, research in this field continues at a rapid rate. Closely related to this field is the use of labeled stem cells, which can be “trained” to target activated endothelium. Iron labeling is the most common form employed, although this has the disadvantage of being a negative contrast agent and is thus prone to artifact.
Thus, many new MRI-based approaches to the study of angiogenesis already exist and are being refined for broader clinical use. In the meantime DCE-MRI serves as the most practical tool for evaluating the microvasculature in vivo.
This research was supported by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research. Thank you to Drs R. Wolf, D. Alsop, A. Heerschap, S. Anderson, J. Frank, G. Lanza and W. Mulder for their kind permission to reprint images.