Molecular MR Imaging Probes

Invited Paper

Magnetic resonance imaging (MRI) has been successfully applied to many of the applications of molecular imaging. This review discusses by example some of the advances in areas such as multimodality MR-optical agents, receptor imaging, apoptosis imaging, angiogenesis imaging, noninvasive cell tracking, and imaging of MR marker genes.

I. Introduction

The diverse array of rapidly growing imaging technologies detailed in this issue demonstrate the numerous approaches that are available to imaging molecular information in vivo. As the strengths and weaknesses of the individual methods are often complementary, no one modality will answer all biological questions. Here, magnetic resonance imaging (MRI) applications using exogenously applied imaging agents are reviewed. This review is not designed to be exhaustive but to show by example some of what is possible using MRI.

As a rule of thumb, with respect to many other whole body molecular imaging technologies, MRI has high-spatial resolution but relatively low sensitivity of detection. To improve the detectable signal changes, numerous agents including gadolinium-based small molecules [1], gadolinium-encapsulated liposomes [2], manganese-based small molecules [3], and iron oxide nanoparticles [4]-[6] have been employed. Typical detection levels of gadolinium small molecules are approximately 10^-4 M, while the iron oxide agents have detection threshold in the 10^-8 M range. The strength of clinical translatability for MRI-based molecular imaging approaches is seen in the fact that the imaging systems have been used hundreds of millions of times to obtain clinical images, and many of the newer molecularly selective imaging agents and techniques described here either could or have been tested in people.

II. Combination MR Optical Agents

MRI imaging agents that are observable by more than one imaging modality typically report as fluorescence with optical imaging as the second modality, while triple-labeled MR imaging agents may report with gamma rays with¹¹¹ In or^99m Tc imaging as the third detection method. For the data acquired using each modality to be useful, the two modalities should be complementary and report either upon different molecular or physiological parameters, take advantage of different spatial resolutions, or allow feedback with different time scales while minimizing instrument complexities.

Labeling cells ex vivo to determine where they migrate to under normal and pathological conditions is discussed in more detail below. A number of groups have labeled cells with MRI detectible agents, typically gadolinium-based [7]-[9] or iron oxide based [10]-[12], which are also fluorescently labeled, while others have simultaneously labeled cells with MR detectible and fluorescent agents to track their location [13]. The major advantage of dual labeling in these cases is that cellular migration may be visualized noninvasively over time by MRI, followed by histological confirmation of location at very high spatial resolution. The ability to stain these same cells for other markers including cellular differentiation markers and coregister this information with fluorescence reporting on labeled cell presence is another advantage.

Several investigators have used the fluorescence component of agents to understand at a microscopic scale the pharmacokinetics of contrast leak from the microvasculature in normal and tumoral tissues [14], [15]. In these cases, the fluorescence location and signal intensity correlate with agent delivery. An alternate paradigm has been to use the fluorescence component to report upon molecular activity and the MR component to report upon the location and concentration of probe in tissues. In particular, proteases, which are overexpressed in a variety of pathological states, can markedly increase the fluorescence of a class of “smart” optical agents [16] after cleavage of the imaging probes. This paradigm of multifold increase in fluorescence intensity after target interaction works as well with iron oxide nanoparticles as the carriers and quenchers of fluorescence [17], [18]. For example, Fig. 1 shows the draining axillary and brachial lymph nodes in a nude mouse [19]. The MRI scan shows the location of the agent in the lymph nodes as a T2 dark area, while the fluorescence intensity reflects protease activity within these nodes. Thus, the fluorescence signal in this case not only would help with intraoperative lymph node dissection after a preoperative MRI to see which nodes have normal or altered draining patterns [5], but additional in situ characterization based on fluorescence intensity reflecting protease activity [Fig. 1(b)] may allow increased specificity of metastatic status determination of the nodes.

Fig. 1.

(a) Axial and coronal MRI images top and near-infrared fluorescence image bottom right. MRI-optical multimodal particle was injected into the left forepaw 24 h prior to imaging. Draining (sentinel) lymph nodes are readily seen by MRI as well as transcutaneously by fluorescence imaging (from Bioconjug. Chem., vol. 13, no. 3, pp. 554-560, May-Jun. 2002, reprinted with permission). (b) Iron core partially quenches the fluorescence until the linker peptide is enzymatically cleaved by proteases and the fluorochrome is spatially separated from the core. Thus, the fluorescence signal intensity in (a) is proportional to enzyme (protease) activity within the lymph node, providing additional characterization.

An additional application of the preoperative MRI, intra-operative fluorescence imaging paradigm is in evaluating brain tumor margins [20]. Previously, it has been demonstrated that brain tumor delineation is enhanced with the use of long circulating iron oxide nanoparticles in both experimental systems [21] and in humans [22]. The multimodal nanoparticles report on the same preoperative information regarding brain tumor margins as the single modality particles [20]. However, as shown in Fig. 2, the sequestration of the particles allows relatively unencumbered near-infrared fluorescence detection instrumentation to aid the resection of tumors in the operating room, without the need for repeat intraoperative MR imaging.

Fig. 2.

MRI image (left), white light image (middle) and NIR fluorescence image (right) show a possible application of multimodal MRI-optical probes for preoperative MRI assessment and intraoperative NIR image assessment of brain tumor location and margins, allowing the radiologist and neurosurgeon to see the same lesion and same boundaries. Brain tumor is dark on T2 MRI images and is bright on NIR fluorescence images due to accumulation of probe (from Cancer Res., vol. 63, no. 23, pp. 8122-8125, Dec. 2003, reprinted with permission).

III. Receptor Imaging

Imaging of receptor overexpression seen in pathology is a useful paradigm that is applicable to most imaging technologies and has been used extensively clinically with SPECT agents. The disadvantage of MRI for receptor imaging is the lower sensitivity compared to the nuclear techniques and additional barriers to quantitation of receptor density; the advantage is the markedly higher spatial resolution and associated anatomic correlation that is automatically achieved. Initial studies, more than a decade ago, using surface-modified iron oxide nanoparticles and MR to detect receptor expression, focused on highly expressed asialoglycoprotein receptors present on normal hepatocytes but not expressed on malignant hepatocellular carcinomas [23]-[25], as a differential source of contrast.

More recently, imaging of HER-2/neu receptor expression in breast cancer lines in animal models was performed using a two component gadolinium-based system [26]. The authors of the study suggest that imaging HER-2/neu receptor expression would allow status determination noninvasively and could be used to select patients who would more likely benefit from antibody therapy. Another receptor target overexpressed in a number of cancers is the folate receptor. In ovarian cancer, the folate receptor is overexpressed in the majority of cases, and in contradistinction to HER-2/neu imaging where tumor characterization is the major goal, the goal of folate receptor imaging in ovarian cancer would be to improve detection of smaller metastatic foci and recurrence. As an example of this approach, tumor xenografts demonstrated MR contrast enhancement with gadolinium folate dendrimer administration, which was inhibited by the administration of free folic acid [27], [28].

Vascular surface targets have the advantage of fewer kinetic compartments that must be crossed for an intravenously administered contrast agent to reach its target. At sites of inflammation, endothelial adhesion molecules, such as intercellular adhesion molecule-1 and E-selectin, are upregulated. Diverse reporting moieties, including liposomes and iron oxide nanoparticles, have been used to report upon binding at sites of overexpression [29]-[32]. E-selectin overexpression has also been reported in prostate cancer epithelium itself and so may represent a more diverse target with imaging reflecting events beyond the endothelial surface [33]. Another vascular imaging target, alphaVbeta3-integrin, is upregulated in activated neovascular endothelium. Unlike the measures of angiogenesis below which reflect the vascular volume fraction and vascular leak, targets expressed in neovasculature reflect the differential rate of new vessel formation. Several groups have used gadolinium-based agents with moieties targeting alphaVbeta3 to image areas of neovasculature in tumors and atherosclerosis [34]-[36].

IV. Apoptosis

Apoptosis, programmed cell death, is a fundamental mechanism of tumor response to many types of chemotherapy. During this process, phosphatidylserine, typically present only on the inner surface of cell membranes, is exposed to the extracellular environment. The MR approaches to imaging this exposure are similar to the receptor targeting above, with phosphatidylserine as the imaging target, as schematized in Fig. 3. A major approach is to label annexin V, an endogenous human protein which recognizes this exposure. Originally, radiolabeled versions of the molecule were described [37], with subsequent testing in humans for evaluation of conditions as diverse as heart transplant rejection [38] and tumor chemotherapeutic response [39]. More recently, fluorescent versions [40], [41] have been used to evaluate apoptosis in tumors implanted in mice. Iron oxide nanoparticles coated with annexin V have been used in cell culture to image apoptosis, with increasing reduction of T2 MR signal changes correlating with increasing apoptotic fraction in cells [42]. Unlike the radio- and fluorescent-labeled annexin V, the iron oxide-based version is dominated in size by the nanoparticle; the altered blood half life and the ability to detect the particle by MRI after the local phagocytosis by host macrophages may provide a method for determining the apoptosis over a longer integral of time compared to shorter circulating detection molecules. Another approach has been to coat iron oxide nanoparticles with the C₂ domain of synaptotagmin I, which binds the membrane of apoptotic cells. Using this approach, the authors were successful in demonstrating increased binding of the conjugate, as measured by changes in MR T2 signal intensity, in tumor-bearing mice treated with chemotherapy [43]. Because of the importance of determining response to antineoplastic therapy, the approaches using MRI and other modalities will likely continue to increase.

Fig. 3.

Method of MR apoptosis imaging. When cells undergo apoptosis, phosphatidylserine (PS), typically present only on the inner surface of cell membranes, is exposed to the extracellular environment. Annexin V, a human protein with a high affinity for PS, is coupled to an MRI-detectable iron oxide nanoparticle to allow visualization.

V. Angiogenesis

Imaging of vascularity and neovascularity covers an enormous range of methods and applications in MR imaging. Numerous excellent reviews exist [44]-[47]; here, some broad themes are evaluated. A molecular-based approach to angiogenesis is to image endothelial markers overexpressed on new blood vessels using antibodies attached to MR detectable liposomes, as detailed above [34]-[36]. Characterization of lesions with respect to the vascular volume fraction or permeability of vessels provides alternate diagnostic information, and the change in these parameters is of marked importance in the early assessment of antiangiogenic therapy. The general approaches to these characterizations include the use of small molecular gadolinium chelates, larger long circulating gadolinium-based macromolecular agents, or the use of iron oxide nanoparticles.

Dynamic imaging of small molecular gadolinium chelates allows the calculation of several kinetic and related parameters including volume transfer constant and the fractional volume of extravascular extracellular space [48], [49]. Such imaging has been used extensively to evaluate the efficacy of anti-angiogenic therapy in murine models [50]-[52]. Moreover, given the millions of doses of low molecular weight gadolinium chelates that have been routinely administered to patients, the use of such agents for calculating in humans angiogenic status of tumors before and after therapy is an attractive approach [53]-[55].

The different kinetics of gadolinium macromolecular agents allow a differing broad range of imaging sequences to be employed compared with the smaller gadolinium chelates. This may allow multiple metastatic sites to be evaluated because rapid repeat imaging is usually not required. Using such an approach, imaging signal changes correlating with vascular endothelial growth factor (VEGF) expression were seen using a macromolecular gadolinium agent; these were not apparent on nondynamic Gd-diethylene triamine pentaacetate (DTPA) imaging [56]. Changes after administration of inhibitors of VEGF were also seen as early as one day after imaging, with the use of other macromolecular gadolinium agents [57], [58]. Macromolecular gadolinium agents which at early times after administration remain in the vasculature may be combined with the subsequent administration of smaller gadolinium chelates to separate the contributions from tumoral vascular volume fraction and interstitial volume fraction [59]. Iron oxide nanoparticles have also been used in both steady state [60] and dynamically [61] to assess vascular volume fraction and endothelial transfer coefficient, respectively.

VI. Cell Tracking

The ability to track cells over time is a major strength of noninvasive molecular imaging compared to histological methods which provide higher spatial resolution but allow analysis at only one time point. Numerous fundamental biological and therapeutic questions hinge on which cells migrate to various tissues in different pathological states and how this migration may be modulated. Most of the molecular imaging modalities have techniques to track cells, either through direct labeling of cells (for example, in the case of single photon nuclear imaging techniques) or the use of stable transfection of cells with maker genes which can then report on the location of cells with the administration of an imaging agent (for example using a PET reporter or a bioluminescence reporter). Cell tracking with MRI has many advantages compared to some of the other techniques. The methods, in general, do not change viability, function, or differentiation of cells at the labeling levels typically reported [62]. The extremely high spatial resolution that can be achieved deep inside tissue compared to other modalities and the associated anatomic information regarding adjacent structures is another benefit.

Most of the MR cell tracking methods have relied on cell labeling with iron oxide nanoparticles. A major reason for this, as stated in Section I, is the lower concentration compared to small molecular paramagnetic agents needed for detection. The majority of cell labeling is performed ex vivo and the cells readministered either systemically or focally at the target tissue. Often, simple coincubation with nonspecific particles has been sufficient for detection, especially for cells such as macrophages that have increased phagocytic capability [63] and for cells expected to be in high concentration at a target site. The surface modification of the dextran coat of the iron oxide nanoparticles with the HIV tat peptide sequence markedly increases the uptake of the particles. This increase was up to 100-fold, compared with unmodified nanoparticles in lymphocytes, even for early preparations [10], [11] which have been subsequently improved with respect to uptake [64]. Some investigators have used a variety of transfection agents to increase the uptake of unmodified dextran-coated iron oxide nanoparticles [65], [66], while others have used larger particles to achieve their detection threshold goals [67]. Detection thresholds in cell suspension have achieved the single cell level, initially at higher field strengths [68] and more recently at clinically standard field strengths [69]. An application-specific in vivo approach for cell labeling which utilized the relatively high uptake by macrophages was used to track macrophages at sites of inflammation in experimental autoimmune encephalomyelitis [70].

Cell tracking applications have included the temporal evaluation of neural or embryonic stem cell migration into regions of cerebral infarction [7], [71]. Other neurologically focused applications have included the tracking of oligodendrocyte progenitor cell migration in the spinal cord of rats [72]. Such methods should allow a more rapid rate of therapeutic optimization coupled to such cell-based treatments. Labeling of mesenchymal stem cells has likewise been utilized to track their delivery and migration in areas of myocardial infarction [73], [74]. A utility of high spatial resolution tracking over a time of T cells is the ability to determine where within a tumor systemically administered cells are delivered. As seen in Fig. 4, over the course of 36 h, cytotoxic T cells preferentially home to the tumor which expresses an antigen that they have been primed against (compared to the contralateral tumor which does not express the antigen) and, importantly, the distribution within the tumor is quite heterogeneous [75]. Additionally, because the cells were labeled with a known amount of probe, the changes in signal intensity could be correlated with cell number per voxel within the tumor.

Fig. 4.

T2 weighted images (top row), and false color T2 maps (bottom row) from left to right, before, and 12, 16, and 36 h after adoptive transfer of T cells primed against an antigen expressed on the tumor in the right flank but not expressed in the tumor on the left flank of a mouse. Dark areas represent location of T cells, quantified per voxel in the bottom panel. Prior to intravenous injection, cells were labeled with an iron oxide nanoparticle, allowing their noninvasive visualization over time by MRI (from Cancer Res. vol. 63, no. 20, pp. 6838-6846, Oct. 2003, reprinted with permission).

VII. MR Marker Genes

Each of the major molecular imaging modalities has one or more marker gene strategies available which have been tested in vivo. The goal of marker genes is to create a method of determining when and where a gene of interest is expressed, without having a significant therapeutic effect. By coupling such a marker gene with a second gene expression that has a desired therapeutic effect, a generalizable scheme is created. A major application for such an approach is to monitor the efficacy and spatial distribution of gene therapy agent delivery. As with other applications, the spatial resolution of MRI is advantageous for observing heterogeneity of vector delivery and gene expression. To overcome the decreased sensitivity compared to isotopic approaches, a two-step procedure was developed to allow improved detection.

The transferrin receptor is a cycling receptor that is expressed by cells in conditions of low cellular iron. The transferrin receptor gene was engineered to remove the iron-regulatory region and mRNA destabilizing motifs. Thus, constitutive expression of the cell surface receptor occurs regardless of cellular iron status. However, the two atoms of serum iron attached to transferrin and which are internalized into the cell with each receptor cycle are insufficient by themselves to allow visualization of the location of engineered transferrin receptor expression noninvasively by MRI in vivo. A second step, coupling transferrin to iron oxide nanoparticles which have several thousand atoms of iron which are in a superparamagnetic state and which result in further changes in signal intensity when brought into proximity to one another in endosomes, allows detectability [76]. As seen in Fig. 5, two tumors, one of which was stably transfected to express the engineered transferrin receptor and the other, the identical wild type control that expressed the native transferrin receptor gene, show imagable differences in R2* relaxivity, confirmed to be secondary to differences in iron oxide nanoparticle accumulation in the stably transfected tumor. Further studies [77] have confirmed that the internalization of the nanoparticles was secondary to receptor-mediated endocytosis from the engineered transferrin receptor. An important step for the translation of this paradigm to act as an MR marker gene was to show that the marker gene could be delivered with a viral vector delivery system and that MR marker gene expression correlated with therapeutic gene expression delivered in the same amplicon [78]. Further improvements in the transferrin-iron oxide moiety [79] have also resulted in the ability to image native levels of transferrin receptor expression in cell culture. Thus, another potential use of the imaging agent might be to provide a degree of in situ characterization based on transferrin receptor expression.

Fig. 5.

Imaging of MR marker gene expression is shown as false color R2* relaxation maps superimposed upon two tumors in this mouse. Marker gene is an engineered transferrin receptor which is constitutively expressed in the tumor on the left and not expressed in the tumor on the right. Gene expression is revealed after the administration of an iron oxide nanoparticle surface modified with transferrin. Internalization of the particle increases R2* (from Nature Med. vol. 6, no. 3, pp. 351-355, Mar. 2000, reprinted with permission).

Another approach to MR marker gene imaging has been use of “smart agents.” This class of agents changes physically detectible properties after target interaction. Different smart probes have been developed for near-infrared optical imaging [16] and for MR imaging. One of the MR methods relies on changes in T1 relaxation as the imaging agent, EgadMe, is modified by beta-galactosidase, a common optical marker gene. In this method, the beta-galactosidase acts as an MR marker gene as well as an optical marker gene by cleaving a blocking group that shields from adjacent water the gadolinium at the center of the developed molecule [80]. This approach allowed the noninvasive observation of marker gene expression in Xenopus embryos. Other MR smart agents include magnetic relaxation switches [81], [82] that change their T2 relaxation properties in the presence of the target (DNA, protein, enzyme activity), by reversibly altering self-assembly of disperse magnetic particles into stable nanoassemblies. A third approach for MR smart agents takes advantage of changes in T1 relaxation times based upon changes in rotational correlation time of gadolinium as a monomer is polymerized. In this case, the MR marker is peroxidase, which results in the oligomerization [83]. Thus, a variety of marker gene and smart probes has been developed.

The noninvasive approaches that have been detailed here show some of the potential for answering biological questions in vivo and for determining clinical paradigms in the future. The broad diverse applications for MRI molecular imaging will continue to grow in the future as molecular biological approaches are further combined with novel imaging agents and new imaging methods.

Biographies

Umar Mahmood received the B.S. degree in chemistry from Caltech and the M.D. and Ph.D. degrees from Cornell University, Ithaca, NY. His doctoral and postdoctoral work in biophysics was performed at Memorial Sloan Kettering Cancer Center and focused on tumor physiology studies using ³¹P NMR spectroscopy. His medical residency training was in radiology at the Massachusetts General Hospital (MGH), Boston.

He is currently an Assistant Professor of Radiology, Harvard Medical School, Cambridge, MA, and Director of Small Animal Imaging at the MGH Center for Molecular Imaging Research. He enjoys using different parts of the electromagnetic spectrum, from radiowaves, to gamma rays, to image molecular events in vivo.

Lee Josephson received the B.S. degree in chemistry from the University of Wisconsin, Madison, and the Ph.D. degree in biochemistry from the State University of New York, Stony Brook.

During a postdoctoral fellowship in the laboratory of Dr. G. Guidotti, Harvard University, Cambridge, MA, he collaborated with Dr. L. Cantley to discover the inhibitory activity of vanadate on ATPases and phosphatases. He was a Cofounder and Chief Scientific Officer of Advanced Magnetics, Cambridge, MA, between 1982 and 1997, a company that pioneered the development of specialty magnetic materials for applications in biology and medicine. An early worker in the field of nanotechnology, he collaborated with Dr. Weissleder to demonstrate that nanoparticle-enhanced MRI could determine the metastatic status of lymph nodes in an animal model. He is currently an Associate Professor at Harvard Medical School, Cambridge. His research interests include the design and uses of magneto/optical nanoparticles, relaxation mechanisms of nanoparticle-based MR contrast agents, and imaging the effects of chemotherapy. He is a coauthor of over 60 publications and a coinventor on over 30 U.S. patents.