Molecular imaging in cardiovascular disease: targets and opportunities

Abstract

The practice of clinical cardiology employs many imaging techniques for diagnosis, risk stratification and therapeutic monitoring. These imaging modalities largely provide anatomical or structural information, and only indirectly reflect underlying molecular events. Molecular imaging techniques report on entities or processes that can be defined at the molecular, as opposed to the anatomical, level. For example, molecular imaging can reveal the expression or activity of a specific protein, the fate or localization of a biomolecule, or the activity of a biological pathway. This Review highlights key processes and molecular targets that are currently being investigated experimentally by molecular imaging probes in vivo. Collectively, these targets have an important role in the development of atherosclerosis and acute plaque rupture, as well as in myocardial disease. Molecular imaging technology is now progressing towards clinical application in humans and has the potential to guide diagnosis, risk assessment and treatment response. Imaging-based surrogate end points could speed the development of new drugs, particularly those with novel mechanisms of action. More broadly, by noninvasively reporting on molecular processes in vivo, molecular imaging can reveal how specific proteins or pathways function in their native context, thus contributing to a systems-level understanding of cardiovascular disease biology.

Key Points

• The imaging techniques currently used in clinical cardiology predominantly provide structural or anatomical information

• Molecular imaging probes report on the expression or activity of proteins, biological pathways or cells, such as apoptosis, thrombus formation or stabilization, and cellular metabolism

• Molecular imaging targets can be engaged by antibodies, peptides, aptamers, or small molecules

• Existing molecular imaging probes have been widely tested in murine or large-animal models of atherosclerosis or myocardial disease; a small number of probes are being investigated in human trials

• Unbiased screens for novel imaging probes with no prespecified target could facilitate the discovery of novel imaging agents and targets

• Targeted molecular imaging probes can contribute to phenotyping of patients, drug development, translational studies, and a systems-level view of cardiovascular disease

Introduction

Imaging studies are widely used in clinical cardiology. A variety of imaging modalities—including standard radiography, fluoroscopy and angiography, ultrasonography, CT, MRI, single photon emission CT (SPECT), and PET—collectively help guide the management of patients. These techniques provide anatomical definition and functional information, such as vascular responsiveness and myocardial perfusion, viability, stiffness and contractility. However, currently available imaging techniques have limited ability to report on disease processes at the molecular level and are, therefore, unable to directly reveal important features of disease pathophysiology. Furthermore, risk stratification—with demographic risk factors, blood-derived markers, stress tests, or imaging studies—is largely performed at the population level, and this can lead to substantial variations in individual risk being obscured.

Molecular imaging techniques report on entities or processes defined at the molecular, as opposed to the anatomical, level. For example, molecular imaging can reveal the expression or activity of a specific protein, the fate or localization of a labeled biomolecule, or the activity of a specific biological pathway or cellular process. By reporting on molecular processes in vivo, molecular imaging provides insight into how proteins and cells function in their native context, subject to physiological regulation by interacting molecules, cells and organs. This information contributes to a holistic, systems-level understanding of biological processes that can inform studies of disease mechanism. In the clinical arena, molecular imaging could lead to the identification of novel imaging biomarkers or phenotypes for individual patients. Quantitative, imaging-based phenotypes that reflect specific pathophysiological processes could improve risk stratification of individual patients, define novel assessments of disease activity (for early diagnosis or monitoring response to therapy), clarify genotype–phenotype correlations for genetic studies, and enhance the development and evaluation of emerging therapies, particularly those with novel mechanisms.

In this Review, I will begin by summarizing a number of molecular imaging targets used in cardiovascular research, focusing on those that are widely used in vivo in animal models or human studies. I will also discuss emerging approaches to the discovery or application of new molecular imaging probes. For example, systematic screens based on imaging phenotypes could enable the unbiased discovery of new imaging probes in the absence of a prespecified protein target. In addition, imaging probes that reflect the output of biological pathways, or the activity of critical regulatory nodes, could reveal how diseases or treatments modulate the complex organization of proteins into pathways and networks.

Targets of molecular imaging probes

The pathogenesis of atherosclerosis involves a complex interplay of endothelial dysfunction, inflammation, monocyte adhesion and invasion, and uptake of oxidized lipoproteins into monocyte-derived macrophages, leading to the eventual formation of an atheromatous core. In this inflammatory milieu, proteases, such as matrix metalloproteinases (MMPs) and cathepsins, can erode the plaque's fibrous cap, leading to plaque rupture, exposure of prothrombotic factors, and formation of a platelet and fibrin clot.^1,2 Critical proteins and cells in this process have been targeted by several molecular imaging probes, which are summarized in two earlier reviews.^3,4 Some of these imaging probes—MMPs and probes for the presence and activity of macrophages, for example—can also report on myocardial damage and cardiac remodeling following myocardial infarction and inflammatory insults. In addition, probes for myocardial apoptosis, metabolism, and neurotransmitters can elucidate myocardial responses to injury and disease states. Here, I discuss several notable examples of imaging targets for vascular and myocardial disease. Most have been investigated using multiple imaging modalities, and the benefits and limitations of these different techniques have been reviewed.⁴

Vascular cell adhesion molecule 1

Vascular cell adhesion molecule 1 (VCAM1) expression is upregulated in inflamed endothelium and this molecule mediates adhesion and transmigration of leukocytes. Hypomorphic mutation of VCAM1 in LDL-receptor knockout mice significantly reduces the development of atherosclerotic plaques.⁵ VCAM1 has been targeted in vivo by antibodies for contrast-enhanced ultrasonography⁶ and by antibodies and short peptides conjugated to magnetofluorescent iron oxide nanoparticles for magnetic resonance (MR) and optical imaging.^7,8 As proof-of-concept for therapeutic monitoring by a molecular imaging agent, Nahrendorf et al. used an apolipoprotein E knockout mouse model in which mice develop marked atherosclerosis on a high-cholesterol diet, and in which a magnetofluorescent nanoparticle bearing a VCAM1-targeting peptide colocalized to VCAM1-expressing plaques.⁷ After atorvastatin treatment a highly significant decrease in the MRI signal from the VCAM1 targeted nanoparticle was observed, which correlated well with decreased VCAM1 expression as demonstrated by immunohistochemistry⁷ (Figure 1).

Figure 1: A VCAM1-targeted magnetofluorescent nanoparticle reveals the effect of atorvastatin treatment in apolipoprotein E knockout mice on a high-cholesterol diet.

a | Short-axis MR image of the aortic root following nanoparticle injection at baseline. The VCAM-1 targeted nanoparticle causes a dramatic drop in aortic root wall signal intensity (red arrowheads). b | Corresponding MR image obtained after 8 weeks of atorvastatin, indicating decreased nanoparticle localization (red arrowheads). c | Color-coded fluorescence microscopy (showing nanoparticle localization) at same aortic level at baseline. d | Corresponding color-coded fluorescence microscopy after 8 weeks of atorvastatin treatment, showing decreased nanoparticle localization. e | Fluorescence reflectance imaging (FRI) of the aorta at baseline. f | Corresponding FRI following 8 weeks of atorvastatin treatment. Abbreviations: MR, magnetic resonance; VCAM1, vascular cell adhesion molecule 1. Permission obtained from M. Nahrendorf, Massachusetts General Hospital Center for Systems Biology and Center for Molecular Imaging Research, Harvard Medical School, Boston, MA, USA.

Macrophages, MSR and oxidized LDL

The class A macrophage scavenger receptor (MSR-A) is overexpressed by macrophages in atherosclerotic plaques and mediates uptake of modified lipoproteins, such as acetylated and oxidized LDL (oxLDL).⁹ Micelles containing paramagnetic gadolinium–diethylenetriamine pentaacetic acid (Gd–DTPA) complexes, and bearing anti-MSR-A monoclonal antibodies,¹⁰ showed significant MR enhancement of aortic plaques in vivo in apolipoprotein E knockout mice;^11,12 the MSR-A targeted micelles colocalize with plaque macrophages, as demonstrated by fluorescent microscopy.^11,12

Radiolabeled antibodies against epitopes found in oxLDL—such as MDA2, which binds to malondialdehyde (MDA)-lysine epitopes—localize to areas of atherosclerotic plaque, as demonstrated by autoradiography and scintigraphy in mouse and rabbit models of atherosclerosis.¹³ Furthermore, when the diet of LDL-receptor knockout mice is changed from one with high-fat and high-cholesterol to a standard chow diet, ¹²⁵I–MDA2 signal decreases in tandem with a decrease in oxidation-specific epitopes in the plaque.¹³ MDA2 and other, similar antibodies have also been incorporated into gadolinium-containing micelles for MRI of oxLDL uptake into macrophages in apolipoprotein E knockout mice.¹⁴

Although not explicitly targeted, dextran-coated, superparamagnetic iron oxide nanoparticles are avidly taken up by macrophages and have been widely used to image macrophages by MR in various in vivo animal models of atherosclerosis.^15,16 Nahrendorf et al. labeled a superparamagnetic and fluorescent nanoparticle with the PET tracer ⁶⁴Cu to create an imaging agent that was detectable optically and by PET and MR.¹⁷ In apolipoprotein E knockout mice, peak PET activity localized to areas of atherosclerosis, and macrophage-specific uptake was confirmed by flow cytometry and fluorescence microscopy. The enhanced sensitivity of PET enables a 10-fold decrease in the nanoparticle dose required, when compared with analogous MR probes.

Dextran-coated, iron oxide nanoparticles have been tested in humans for their ability to detect macrophages in atherosclerosis in vivo. In a pilot study of symptomatic patients scheduled for carotid endarterectomy, histology and electron microscopy showed that ferumoxtran-10 nanoparticles (Sinerem^®, Guerbet, Roissy, France) accumulated in macrophages, particularly in ruptured or rupture-prone lesions. However, the sensitivity of MRI to detect uptake of ferumoxtran-10 into macrophages was decreased.¹⁸ The same ferumoxtran-10 nanoparticle (Combidex^®, AMAG Pharmaceuticals Inc., Lexington, MA) is being investigated for the detection of malignant lymph node metastases in patients with cancer.¹⁹

Activity-based probes

Matrix metalloproteinases

MMPs produced by macrophages, smooth muscle cells, and endothelial cells in atherosclerotic plaque proteolyze components of the extracellular matrix, and contribute to plaque instability.^20,21 MMPs also participate in cardiac remodeling after myocardial infarction. Coupling of small-molecule MMP inhibitors to ¹²³I, ^99mTc, ¹⁸F, or gadolinium has enabled localization of in vivo MMP expression by scintigraphy,²² SPECT-CT,²³ PET,²⁴ and MRI,²⁵ respectively, in animal models of atherosclerosis or myocardial infarction.

Some exciting molecular probes depend on the enzymatic activity, rather than simply the expression, of their targets offering the advantage of enzymatic amplification of imaging signals. One example is an MMP activity probe containing a peptide substrate for MMP-2, MMP-3, MMP-9, and MMP-13 conjugated to multiple cyanine 5.5 moieties for near-infrared fluorescence (NIRF). In the absence of MMP activity, fluorescence is quenched because of the proximity of the cyanine 5.5 molecules to one another. Cleavage of the peptide substrate by MMPs liberates cyanine 5.5 causing a 200-fold amplification in NIRF.²⁶ This probe has revealed MMP activity in murine models of myocardial infarction^26,27 and atherosclerosis.²⁸ Localization of probe activity closely reflects MMP enzymatic activity, as demonstrated by colocalization of NIRF signal with MMP expression on immunohistochemistry or in situ zymography; furthermore, treatment with an MMP inhibitor abrogates the NIRF signal.²⁸

Cathepsins

Cathepsins are cysteine proteases that act intracellularly in lysosomes, as well as extracellularly where they digest elastin, collagen, and other extracellular matrix components.²¹ Normal human arteries minimally express cathepsins, but macrophages and other cells in human atheroma express cathepsins S and K.²⁹ Murine genetic models of atherosclerosis, such as apolipoprotein E knockout or apolipoprotein E/nitric oxide synthase double knockouts, overexpress cathepsin B.³⁰ A first-generation cathepsin activity probe was originally developed to detect tumor-associated proteases, and consists of a poly-L-lysine backbone to which multiple side chains, containing methoxypolyethylene glycol and cyanine 5.5, are attached³¹ (Figure 2). Enzymatic cleavage of the poly-L-lysine backbone unquenches the cyanine 5.5. This probe has been used to demonstrate cathepsin B activity in atherosclerotic apolipoprotein E knockout mice using fluorescence tomography.³⁰ Subsequent probes incorporated specific cleavage sites for cathepsins and have demonstrated cathepsin activity in murine atheromas in vivo and in human carotid endarterectomy samples ex vivo.³² NIRF colocalizes in areas of disrupted elastin fibers and in close proximity to cathepsin-expressing macrophages.³² A probe activated by cysteine proteases (including cathepsin B) that uses a prototype intravascular catheter has been used to detect inflamed atherosclerotic plaques in hypercholesterolemic rabbits exposed to balloon injury.³³

Figure 2: Design of a cathepsin-activated imaging probe.

a | At baseline, fluorescence of the cyanine 5.5 moieties is quenched. Interaction with cathepsin results in cleavage of polypeptide backbone, with release and unquenching of cyanine 5.5. b | Top: near-infrared fluorescent image of a solution of cathepsin probe at baseline, and following enzyme activation. Note the dramatic increase in fluorescence after enzyme activation. Bottom: white light image of the same solutions. c | Chemical structure of the cathepsin probe, showing the protease cleavage site (green arrow), and quenching of cyanine 5.5 (red dots). Abbreviations: Cy, cyanine 5.5; MPEG, methoxy polyethylene glycol; PL, poly-L-lysine. Adapted from Weissleder, R. et al. Nat. Biotechnol. 17, 375–378 (1999).

Myeloperoxidase

Myeloperoxidase is produced by activated macrophages in human atherosclerotic plaques and in the inflammatory infiltrate after myocardial infarction. Myeloperoxidase generates hypochlorous acid (HOCl) and other oxygen free radical species. Under physiological conditions, HOCl acts as a powerful oxidizing agent with multiple deleterious vascular effects, including the oxidation of LDL. Myeloperoxidase-expressing macrophages have been implicated in the pathogenesis of complex atheroma and acute plaque rupture.^34,35 A myeloperoxidase-activity reporter that links Gd–DTPA to two phenol-containing 5-hydroxytryptamide moieties has been developed; myeloperoxidase oxidizes the phenol derivatives in this probe to form either cross-linked self-oligomers or covalent adducts with matrix proteins, yielding an increase in relaxivity³⁶ (Figure 3). This probe can visualize local myeloperoxidase activity in the inflammatory infiltrate in murine models of myocardial infarction³⁷ (Figure 3) and stroke.³⁸ In a complementary approach, a small-molecule sensor detects HOCl production by myeloperoxidase in physiological conditions, as demonstrated by strong fluorescence in an in vivo model of murine experimental peritonitis, and when incubated ex vivo with human carotid endarterectomy specimens.³⁹

Figure 3: Structure and activity of myeloperoxidase-activated probe.

a | Structure of gadolinium–diethylenetriamine pentaacetic acid-bis-5-hydroxytryptamide. b | Schematic of how phenol-containing molecules undergo oligomerization in the presence of oxidoreductases (such as myeloperoxidase) and H2O2, resulting in an increase in relaxivity. Parts c and d: in vivo MRI 1 day after murine myocardial infarction showing T1-weighted images acquired 1 h after Gd–DTPA-bis-5-hydroxytryptamide (MPO–Gd) injection. c | Mice with intact myeloperoxidase expression (MPO ^+/+) display dramatically increased enhancement in the infarcted left ventricular free wall (white arrows) after MPO–Gd injection. d | Decreased MR signal enhancement in myeloperoxidase knockout (MPO ^−/−) mice, confirming the specificity of the probe for MPO activity. Abbreviations: Gd–DTPA, gadolinium–diethylenetriamine pentaacetic acid; MPO, myeloperoxidase; MR, magnetic resonance. Permission for images in panels c and d obtained from M. Nahrendorf, Massachusetts General Hospital Center for Systems Biology and Center for Molecular Imaging Research, Harvard Medical School, Boston, MA, USA.

Fibrin and factor XIII in thrombi

A series of imaging agents have targeted fibrin in thrombi using antibodies or peptides. Specificity for fibrin helps direct the imaging probe (or a therapeutic) to localized fibrin in a clot, rather than to circulating fibrinogen. For example, antifibrin antibodies on lipid-encapsulated, Gd–DTPA-containing perfluorocarbon nanoparticles can localize to human thrombi in vitro⁴⁰ and in canine models⁴¹ when the contrast agent is preincubated in situ with a forming thrombus. These nanoparticles have also been used to deliver the plasminogen activator streptokinase to human thrombi in vitro.⁴² Similarly, an 11-amino-acid peptide that binds fibrin with micromolar affinity (and possesses 100-fold specificity for fibrin as opposed to fibrinogen) has been conjugated to four gadolinium–tetraazacyclododecane tetraacetic acid (Gd–DOTA) moieties.⁴³ This agent (EP-2104R, EPIX Pharmaceuticals, Lexington, MA) can detect carotid thrombi in vivo in a rabbit injury model,⁴⁴ and human thrombi that were formed ex vivo and delivered to the coronary or pulmonary circulation of pigs.⁴⁵ EP-2104R has also been tested in a pilot study of patients with documented intracardiac or arterial thrombi. Using a clinical 1.5T whole-body MRI system, thrombi were visualized with high-signal amplification and contrast:noise ratios relative to the surrounding soft tissue and blood pool⁴⁶ (Figure 4).

Figure 4: In vivo MRI of human thrombi using a fibrin-targeting peptide conjugated to gadolinium–tetraazacyclododecane tetraacetic acid (EP-2104R).

a | MRI before contrast enhancement (inversion recovery black-blood gradient-echo imaging). b | EP-2104R injection causes signal enhancement and visualization of left ventricular thrombus (arrow). c | In a different patient, the post-EP-2104R image shows signal enhancement corresponding to aortic thrombus (arrow). d | Corresponding contrast-enhanced CT shows atherosclerotic plaque with calcification (arrowhead). Abbreviations: LV, left ventricle; RV, right ventricle. With kind permission from Springer Science + Business Media: European Radiology, MR imaging of thrombi using EP-2104R, a fibrin-specific contrast agent: initial results in patients. 18, 2008, 1995–2005, Spuentrup, E., Botnar, R. M., Wiethoff, A. J., Ibrahim, T., Kelle, S., Katoh, M., Ozgun, M., Nagel, E., Vymazal, J., Graham, P. B., Günther, R. W. & Maintz, D. Figure 1 (part 2); Figure 3.

Another approach to molecular imaging of thrombi is based on factor XIII. Activated factor XIII has a transglutaminase activity that covalently cross-links fibrin polymers and incorporates plasmin inhibitors, such as α2-antiplasmin, thus rendering the thrombus more resistant to lysis. Probes that conjugate a factor XIII substrate peptide to near-infrared fluorochromes or ¹¹¹In-chelates have been synthesized.⁴⁷ Enzymatic cleavage of the substrate peptide enables unquenching of fluorophores and visualization of factor XIII activity in clotted human plasma in vitro, and in acute murine thrombi induced by FeCl₃.⁴⁸ Because the transglutaminase activity of factor XIII can also cross-link collagen fibers, the same imaging probes have been used to visualize factor XIII activity in the inflammatory response after myocardial infarction in murine models and in human myocardial samples.⁴⁹

Annexin V

Many studies have been based on the observation that apoptosis causes phosphatidylserine to become exposed on the outer leaflet of the plasma membrane. The endogenous anticoagulant protein annexin V binds negatively-charged phospholipids, such as phosphatidylserine, and has been used extensively to detect externalized phosphatidylserine as a phenotype of apoptosis by SPECT, SPECT-CT, MRI, PET, and optical imaging in a variety of animal models of atherosclerosis and myocardial infarction.^50,51,52 Annexin V probes have also been used in human studies to label cells in the infarct zone after myocardial infarction,⁵³ to preferentially label the culprit carotid artery in patients with transient ischemic attack,⁵⁴ and to label cells in cardiac transplant patients who have histological evidence of rejection and caspase 3 staining.⁵⁵ Findings from a pilot study of patients with nonischemic cardiomyopathy indicate that substantial ventricular uptake of ⁹⁹Tc–annexin V could predict future decline in left ventricular function.⁵⁶

Biomimetic analogs

Metabolite analogs

¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) PET imaging detects glucose uptake and phosphorylation in metabolically active cells, and is used clinically as a reporter of metabolic activity in cardiology, oncology, and other medical specialties. ¹⁸F-FDG PET has also been used to reveal the presence of macrophages in inflammatory atherosclerotic plaques, such as in vulnerable carotid plaques in patients scheduled to undergo carotid endarterectomy;^57,58 the mean ¹⁸F-FDG PET target-to-background ratio correlates well with macrophage staining in endarterectomy specimens.⁵⁸ This imaging technique has also been used to document changes in plaque activity following short-term drug treatment. Among patients with ¹⁸F-FDG uptake in the carotid arteries or thoracic aorta, the ¹⁸F-FDG PET signal decreased significantly in those who received a 3 month course of simvastatin, but not in patients treated with diet modification alone.⁵⁹

¹²³I-β-methyl-p-iodophenyl-pentadecanoic acid (¹²³I-BMIPP) is a radiolabeled, branched-chain fatty acid that reflects the uptake and early metabolism of fatty acids.^{60 123}I-BMIPP is taken up by myocytes and converted to ¹²³I-BMIPP-acyl-coenzyme A, which cannot diffuse back across the cell membrane and is only oxidized to a limited extent. ¹²³I-BMIPP SPECT can be used to document the shift from metabolism of fatty acids to that of glucose in the cardiac myocytes of patients with ischemia or cardiomyopathy.^60,61,62 Delayed imaging with ¹²³I-BMIPP SPECT shows persistent abnormalities up to 30 h after exercise-induced ischemia, which indicates that metabolic adaptation to ischemia (suppression of fatty acid metabolism) persists for some time after restoration of perfusion, and potentially allows the diagnosis of an acute ischemic episode long after symptom resolution.⁶² In a prospective study of patients undergoing long-term hemodialysis who did not have a history of myocardial infarction or revascularization, abnormal ¹²³I-BMIPP uptake on SPECT was associated with an increased risk of cardiac death over 3 years.⁶³ This finding indicates that marked abnormalities in fatty acid metabolism, perhaps related to persistent ischemia, could identify a subgroup of patients at increased risk of cardiac death.

Other tracers for myocardial metabolism include endogenous biochemical substrates labeled with ¹¹C, which can be imaged using PET as they undergo uptake, metabolism, and clearance from cardiomyocytes. These molecules include ¹¹C-palmitate, which traces long-chain fatty acid metabolism, ¹¹C-glucose, which can be used to follow glucose uptake and oxidation, and ¹¹C-acetate, which is used to report on flux through the Krebs cycle, and on overall oxygen consumption.⁶⁰

HDL-mimetic nanoparticles

HDL mitigates the development of atherosclerosis by facilitating reverse transport of cholesterol from plaque macrophages to the liver.⁶⁴ Fayad and colleagues have developed synthetic HDL-mimetic nanoparticles that can mediate cholesterol efflux from macrophages and are also suitable for use with a variety of imaging modalities. These nanoparticles incorporate either lipoproteins (such as apolipoprotein A-I) extracted from human HDL,⁶⁵ or amphipathic helical peptides (such as L-37pA) that can promote ATP-binding cassette (ABC)A1-dependent and, to a lesser extent, ABCA1-independent lipid efflux in vitro.⁶⁶ The lipoproteins or amphipathic peptides are reconstituted with phospholipids, gadolinium-chelates and fluorescent moieties in the final nanoparticle (Figure 5). These HDL mimetics can mediate efflux of tritiated cholesterol from macrophages in vitro,⁶⁷ and localize to intimal macrophages when injected into apolipoprotein E knockout mice.^65,67 Variants of these nanoparticles that incorporate nanocrystal cores of quantum dots, gold, and iron oxide have also been synthesized for fluorescence imaging, CT, and MRI, respectively.⁶⁸

Figure 5: Synthetic HDL-mimetic nanoparticles.

a | Schematic of synthetic HDL-mimetic nanoparticle containing the amphipathic peptide 37pA and various phospholipids. b | Transverse MRI of the abdominal aorta (arrow) in apolipoprotein E knockout mice, before and 24 h after injection of the synthetic HDL-mimetic depicted in part a. Note the heterogenous signal enhancement in the aortic wall in the postinjection image. c | Confocal microscopy of aortic tissue following HDL nanoparticle injection. i | DAPI (4′,6-diamidino-2-phenylindole) nuclear stain; ii | rhodamine fluorescence (from nanoparticle); iii | anti-CD68 antibody conjugated to Alexa 647 (labeling macrophages); iv | merged image, showing colocalization of rhodamine (nanoparticle) and Alexa (macrophage) fluorescence. Abbreviations: DMPC, dimyristoyl phosphatidylcholine; DMPE, 1,2-dimyristoyl-sn-glycero-3-phosphoethanolamine; Gd–DTPA, gadolinium–diethylenetriamine pentaacetic acid; Rhod-PE, 1,2-dimyristoyl-sn-glycero-3-phosphoethanolamine-N-(lissamine rhodamine B sulfonyl). Cormode, D. P., Briley-Saebo, K. C., Mulder, W. J., Aguinaldo, J. G., Barazza, A., Ma, Y., Fisher, E. A. & Fayad, Z. A. An ApoA-I mimetic peptide high-density-lipoprotein-based MRI contrast agent for atherosclerotic plaque composition detection. Small (2008) 4, 1437–1444. © Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission.

Neurotransmitter analogs

Neurotransmitters have a critical role in the regulation of myocardial function. For example, heart failure is accompanied by a derangement in normal sympathetic regulation, with increased norepinephrine levels and downregulation and uncoupling of cardiac β-adrenergic receptors. Modulation of the sympathetic axis by β-blocker therapy can improve outcomes in chronic heart failure.⁶⁹ These lines of evidence suggest that imaging of the sympathetic axis could convey useful prognostic or pathophysiologic information. ¹²³I-metaiodobenzylguanidine (mIBG) is a norepinephrine analog that cannot be hydrolyzed by monoamine oxidase and can report on the degree of sympathetic innervation and tissue norepinephrine content by SPECT. ¹²³I-mIBG uptake on SPECT is a strong predictor of sudden cardiac death in patients with heart failure,⁷⁰ and of recurrent arrhythmic events in patients with a history of ventricular tachycardia or fibrillation.⁷¹

Emerging approaches

Phenotypic screens for novel imaging agents

The vast majority of imaging agents described in this Review were developed using a target-based approach; a protein is first nominated as an imaging target because of its experimentally demonstrated role in a disease process or its expression in cells of interest. Antibodies, peptides, aptamers, or small molecules are then screened against the target and conjugated to a radioisotope or nanoparticle platform to create the final imaging agent. Phenotype-based screens of libraries of potential imaging agents represent a complementary approach to the discovery of imaging probes. This approach does not require a prespecified imaging target. Instead, imaging probes are identified in an unbiased fashion by screening potential agents for their preferential binding or uptake into a particular cell type, typically using an in vitro cell-based assay (Figure 6). This phenotypic approach allows the discovery of imaging probes for cellular states that are defined on the basis of their biological or medical significance—such as mutant versus wild-type cells or healthy versus diseased vessels—rather than on the basis of a single target protein. This approach could lead to the discovery of unexpected mechanisms underlying the particular cell state of interest. In this sense, phenotypic imaging screens are logically analogous to forward genetic screens in genetically tractable model organisms—such as the mutant screens that identified key genes involved in the determination of body pattern and polarity in Drosophila, and in cell-cycle regulation in yeast. Although phenotypic screens require a separate process to determine the target of the imaging probe, affinity purification combined with quantitative proteomics can identify interacting target proteins much more rapidly than in the past.⁷²

Figure 6

Schematic comparison between a | target-based and b | phenotype-based approaches to imaging probe discovery.

As proof-of-concept, phenotypic screens have been performed for magnetofluorescent iron oxide nanoparticles bearing various small molecules on their surfaces. Weissleder and colleagues screened a library of nanoparticles in 384-well microtiter plates for preferential binding to specific cell types.⁷³ Although the unmodified nanoparticle binds primarily to activated and resting macrophages with the same degree of affinity, conjugation of glycine to the surface of the nanoparticle results in preferential uptake into macrophages activated by oxLDL and other substances rather than into resting macrophages.⁷³ In vivo, these glycine-modified nanoparticles preferentially label 'M2-like' tumor-associated macrophages, which express high levels of F4/80 protein, vascular endothelial growth factor, and Tie-2.⁷⁴ Thus, glycine modification of magnetofluorescent iron oxide nanoparticles yields a novel probe for a functionally distinct subset of macrophages in vivo. Phenotypic screening in primary human cell isolates has also led to the identification of a small-molecule-modified nanoparticle that demonstrates enhanced endothelial binding in vitro;⁷⁵ endothelial targeting was confirmed in human carotid endarterectomy samples ex vivo, as well as in murine vessels in vivo.⁷⁵

Phenotyping of patients and drug development

Carotid artery intima–media thickness, which is measured by ultrasonography, is a prime example of an imaging phenotype that is widely accepted as a surrogate end point for cardiovascular disease. That is, carotid intima–media thickness correlates well with the presence of coronary artery disease and the risk of cardiovascular events; therapies that reduce risk of cardiovascular events also reduce the rate of progression of carotid intima–media thickness.^76,77 Other noninvasive and invasive imaging techniques have also been used to provide surrogate end points in clinical trials, including CT, MRI, brachial and intravascular ultrasonography, and optical coherence tomography.^77,78

Molecular imaging phenotypes could complement existing imaging surrogate end points, enhancing care of patients and drug development in several ways.⁷⁹ By reporting on specific molecular events in the arterial wall that are associated with rapid lesion progression, thrombotic complications, or plaque rupture, molecular imaging could lead to the identification of novel imaging phenotypes for diagnosis, assessment of disease status, individualized risk stratification, and treatment response. Importantly, molecular imaging phenotypes may report on cellular processes that are closer to the mechanism of a therapeutic intervention than other biomarkers or clinical end points and, therefore, facilitate the in vivo evaluation of drugs with novel mechanisms in small, rapid trials.

Molecular imaging is a powerful approach to testing therapeutic hypotheses in vivo. Imaging phenotypes can reveal if a pharmacological intervention causes the effect that is expected based on the proposed mechanism of action, which is distinct from whether the intervention affects the likelihood of a clinical end point. If a drug fails to cause a mechanistically expected imaging phenotype, the drug is less likely to cause the desired therapeutic effect. This scenario could potentially provide a rationale for timely decisions on whether or not to proceed with drug development. However, given that most imaging phenotypes will not be used for drug registration studies, clinical end points will ultimately still be required for FDA approval of new therapeutics.

Molecular imaging phenotypes could facilitate the development of drugs with novel mechanisms of action. For example, the enzyme lipoprotein-associated phospholipase A₂ (Lp-PLA₂) hydrolyzes oxidized phospholipids in LDL to generate proinflammatory and proatherogenic lysophosphatidylcholine and oxidized nonesterified fatty acids. These molecules stimulate homing of inflammatory cells to the developing plaque, generate further proinflammatory mediators locally, and increase apoptosis of multiple cell types, including macrophages.⁸⁰ Darapladib—an inhibitor of Lp-PLA₂—has been investigated in swine and human atherosclerosis, where it reduced the growth of plaque necrotic core volume and the incidence of plaque instability.^81,82,83 Imaging agents directed at several of the targets discussed above (macrophages, cathepsins, MMPs, phosphatidylserine) could directly and noninvasively address the therapeutic hypothesis that darapladib decreases plaque inflammation and macrophage infiltration in vivo. Similarly, the HDL-mimetic nanoparticles discussed in this Review^65,67 could lead to the development of imaging probes that inform the development of therapies directed at HDL metabolism.

Other illustrative examples can be found in the oncology literature. As in the field of cardiology, the substantial period of time before the occurrence of a 'gold standard' clinical end point—such as a radiographic decrease in tumor size—can delay treatment decisions and the evaluation of novel therapies. If molecular imaging phenotypes more closely reflect pathophysiological or therapeutic mechanisms, then they might be useful as early surrogates for later clinical end points. For example, in patients with Hodgkin lymphoma, or aggressive forms of non-Hodgkin lymphoma, early ¹⁸F-FDG PET imaging provides prognostic information comparable with that provided by traditional radiologic studies performed after completion of therapy.⁸⁴ Furthermore, ¹⁸F-FDG PET imaging could be useful as a specific probe of metabolic pathways that functionally interact with glucose uptake, rather than merely as a general indicator of metabolic activity. For example, inhibitors of the mammalian target of rapamycin (mTOR)—a critical integrator of nutrient response—and phosphoinositide-3 (PI3) kinase—an upstream regulator of the serine-threonine kinase Akt, and a key mediator of insulin and other growth factor signaling—cause a dramatic, early ¹⁸F-FDG PET response in murine cancer models that is out of proportion to radiographic decreases in tumor mass^85,86 (Figure 7). As mTOR, PI3 kinase, and Akt have been associated with a variety of cardiovascular processes, such as cardiac hypertrophy and contractility, and vascular smooth muscle cell proliferation, ¹⁸F-FDG PET could prove to be a useful indicator of pathway activity in the evaluation of therapies directed at cardiomyocyte or vascular smooth muscle metabolism.

Figure 7: Use of a metabolic imaging probe (¹⁸F-FDG PET) to visualize the effect of therapies targeting metabolic pathways.

A proof-of-concept example from the oncology literature. Mice with lung adenocarcinoma (caused by inducible expression of an activating allele of PI3 kinase) were treated with an experimental inhibitor of PI3 kinase and mTOR. a | CT scans before and 4 days after treatment show decrease in tumor size (arrows), but persistent tumor mass. b | ¹⁸F-FDG PET-CT image showing intense activity in the tumor before treatment, and abrogation of the PET signal after treatment at a time when persistent tumor mass is still present. Abbreviations: ¹⁸F-FDG, ¹⁸F-fluorodeoxyglucose; H, heart; mTOR, mammalian target of rapamycin; Pl3, phosphoinositide-3 kinase. Reproduced from Engelman, J. A. et al. Nat. Med. 14, 1351–1356 (2008).

Future opportunities

Molecular imaging will soon be poised to contribute systematically to the phenotypic evaluation of disease in animal models and in humans. Studies of human disease genetics, including genome-wide association studies, are providing a comprehensive 'parts list' of genes that interact with environmental and behavioral factors to influence an individual's risk of disease. However, our understanding of the functional importance of individual genes, how multiple genes connect in pathways and networks, and how genetic variants implicated in disease perturb these networks, is incomplete. Targeted imaging agents can noninvasively provide quantitative information about the activity of biological pathways subject to endogenous physiological regulation. In particular, the effects of perturbations with small-molecule drugs or nutrients can be monitored in real time with temporal and spatial resolution. The development and systematic application of molecular imaging agents that monitor the activity of key pathway 'nodes' can powerfully complement existing approaches to human disease studies, such as invasive physiological measurements, gene expression, proteomics, and metabolomics.

Realizing the potential benefits of molecular imaging in clinical medicine or drug development is a long-term challenge. Even with compelling proof-of-concept studies in small and large animal models, widespread application in humans requires extensive clinical investigation into the relationships between molecular imaging phenotypes and other biomarkers and clinical end points. Issues of measurement standardization and reproducibility, safety of nanomaterial imaging agents,^87,88 financial cost, and potential radiation exposure will also need to be resolved. One potential model for how these challenges might be addressed is the Alzheimer's Disease Neuroimaging Initiative,⁷⁹ a consortia of academic, government, pharmaceutical and device company scientists. This collaborative effort seeks to validate imaging approaches, serum and other fluid biomarkers, and neuropsychiatric tests, with the aim of developing a robust, widely available suite of assessment tools for both care of patients and drug development.

Conclusions

Collectively, the vignettes discussed in this Review outline how a repertoire of molecular imaging agents can elucidate not only the presence or activity of a particular imaging target, but also provide a broad range of biological and clinical insights. Molecular imaging holds great promise to enrich our understanding of fundamental disease mechanisms, as well as translate those insights into new approaches to the diagnosis and treatment of cardiovascular disease.

Review criteria

The PubMed database was searched for English language papers, with no constraint on the year of publication. Searches were not limited to full-text articles, but wherever possible references with full-text papers available through institutional library subscriptions were used. The broadest searches used the terms “molecular imaging” or “in vivo imaging”, combined with either “cardiac” or “cardiovascular”. Papers on individual target proteins were searched for by combining the name of the protein or target, and the term “imaging”. Imaging targets were selected based on novelty and potential impact. Citations were chosen to recognize both early contributions and more recent studies.