Molecular imaging is defined as the in vivo measurement of biological processes at the cellular and molecular levels.24-26 Molecular-imaging-based visualization of in vivo pathophysiologic processes could provide information regarding specific molecular alterations underlying the disease status of individual patients in real time.27-29 By providing molecular information unobtainable using conventional anatomy-based imaging modalities, molecular imaging would allow 1) earlier detection of diseases, 2) precise discrimination of the stable versus unstable disease status, and 3) both diagnostic and therapeutic quantitative monitoring of disease progression.30-32 The various modalities of molecular imaging (CT/MRI/fluorescent optical imaging/positron-emission tomography, PET/single-photon-emission CT/SPECT/ultrasound) have their own specific advantages and disadvantages that are related to how the images are generated.21,32

The application of fluorescent proteins or fluorochromes to the life sciences during the last 2 decades has considerably advanced basic and translational biological research, and these advances are now beginning to affect clinical practice.29,33,34 Fluorescence involves the absorption of light at characteristic wavelengths, and the emission of the stored energy at longer wavelengths.35 The advantages of fluorescence as a molecular imaging modality include picomolar molecular sensitivity, absence of ionizing radiation, the possibility of using it in many modalities with different scales, and relatively low cost. However, poor tissue penetration capability is a major obstacle to overcome.36,37 The attenuation of light by tissue is lowest in the near-infrared (NIR, 700-900 nm) region, and imaging in this region offers 1) markedly less photon absorption by blood hemoglobin, lipid, and water, allowing light to penetrate centimeters into the body; and 2) substantially reduced tissue autofluorescence, enabling higher sensitivity detection of targeted NIR fluorescent (NIRF) molecular imaging agents against a low background.31,36,38 This technology can potentially be combined with a non-invasive optical tomography system, intra-operative NIRF imaging system, or fluorescence-sensing catheter-based system.39 Jaffer et al. recently demonstrated that a NIRF-sensing catheter based on a clinical coronary artery guidewire could detect in vivo cathepsin B (CatB) protease activity in rabbit vessels the size of human coronary arteries in real time.

NIRF imaging requires a much smaller dose of fluorescence probes to detect molecules of interest-nanomoles of fluorochromes can be detected, compared to micromoles for MRI or millimoles for CT.31,32 The relatively high spatial resolution (typically less than 1 mm) of the catheter-based imaging system is another advantage of fluorescence imaging.39 Scatter reduces the spatial resolution of non-invasive fluorescence molecular tomography (FMT) relative to using an endoscopic imaging device (-1 mm), and is within the range of resolution provided by SPECT and PET.36

Studies have demonstrated that the formation of a thrombus due to rupture of unstable atherosclerotic plaques, followed by thrombotic or embolic occlusion of an artery, is the leading cause of stroke, accounting for up to 80% of cases of ischemic stroke in some autopsy series.10,40 There is a pressing need for tools to identify these vulnerable plaques, and thereby identify patients and lesions at high risk for vascular events, so that risk-altering treatments might be offered to improve clinical outcomes. According to the current practice guidelines and consensus, a carotid lesion is likely to cause ischemic stroke when stenosis of over 60 or 70% is detected by angiography.41 However, it has become clear that many strokes are attributable to plaques in the arteries with stenosis of 50% or less, highlighting the importance of plaque ruptures as a causative mechanism.42-44 Rupture-prone vulnerable plaques are not well identified by conventional measures of stenosis.45-47 Ultrasonic characterization of plaques as heterogeneous or of low echodensity on carotid duplex ultrasonography has been regarded by some as suitable for detecting unstable plaques, but firm conclusions await further studies.48,49 Plaque size and morphology (Fig. 1) are poor substitutes for the molecular events that shaped them, and the measurement of underlying molecular states provides the best hope for determining the propensity to cause complications.

For asymptomatic patients with significant carotid stenosis, the number needed to treat to prevent one stroke from any cause at two years is 67, which could increase to 111 to prevent one large-artery stroke at 2 years, since not all strokes originate from a narrowed internal carotid artery.50 For symptomatic patients, endarterectomy is beneficial in the long term even in the presence of a contralateral occlusion and increased perioperative risk.51,52 However, from a patient's standpoint, opting for a procedure with a 3-6% immediate perioperative risk to reduce a future stroke chance by 6-15% at 5 years in both asymptomatic and symptomatic cases can be daunting.53 This illustrates the difficulty of weighing the risks and benefits in treating an individual patient.51

Characteristic features of vulnerable atherosclerotic plaques are accumulation of inflammatory cells inside the plaque (Fig. 2), formation of a large lipid core, and thinning of the overlying fibrous cap.5 Heavily clustered macrophages in and around the shoulder region of the cap would secrete matrix-disorganizing proteolytic enzymes such as cathepsins and matrix metalloproteinases (MMPs), which could render atherosclerotic plaques prone to rupture.6,54 Thus, inflammatory protease activity could be regarded as a hallmark of the plaque vulnerability. In fact, using the Affymetrix gene chip to profile genes expressed in stable and unstable atherosclerotic plaques revealed that the expressions of CatB, cathepsin S, and MMP-9 were up-regulated in unstable atherosclerotic plaques.55 Taken together, these data suggest that imaging protease activity in vivo could allow identification of vulnerable plaques.

The Weissleder group and others have developed protease-sensing NIRF molecular imaging agents.56-60 Unlike "always-on" probes such as nonspecific NIRF probes [e.g., indocyanine green (ICG)] or targeted NIRF probes relying on affinity ligands, the protease-sensing NIRF probes are "optically silent at injection" because of intra-molecular autoquenching between closely spaced fluorochromes. Enzyme-specific protease-mediated cleavage spatially separates fluorochromes so that they become dequenched and brightly fluorescent. Multiple NIR fluorochromes positioned on a polylysine backbone can be activated by CatB.34 The gelatinase substrate sequence GGPRQITAG has been used to synthesize a NIRF probe that can be activated by MMP-2/9.58 Animal studies using these activatable probes have shown that in vivo and ex vivo NIRF imaging could visualize the protease activity in the aortic or carotid atheromata of ApoE knock-out mice (Figs. 3 and 4).58,59,61,62 Immunohistochemical analyses demonstrated that the NIRF signal precisely reflected the spatial distribution of the inflammatory protease activities. In addition, the protease immunoreactivity co-localized with Mac-3 or CD-68 positive macrophages.

Despite the high expectations for and rapidly increasing interest in molecular imaging techniques, there are significant challenges in translating the advances from the laboratory to the atherosclerosis clinic. As a step toward clinical translation, we recently demonstrated that CatB NIRF imaging could reliably reflect the anti-atherosclerotic effect of atorvastatin and glucosamine as well as the pro-atherosclerotic effect of a high cholesterol diet (Fig. 4). We also revealed that plaque populations were heterogeneous within individuals, with some plaques showing high and others lower CatB-related signals.61 Notably, these differences in signal intensity could not be predicted by visual inspection of the plaques. In every case where imaging predicted an inflammatory component in the plaque, histological studies confirmed it to be the case.61 Regarding translation steps from animals to humans, catheter-based fluorescence imaging systems are likely to be available in the atherosclerosis clinic. In addition, it should be remembered that since NIR photons can potentially penetrate >5 centimeters into the body, noninvasive FMT systems may eventually detectNIRF signals from human carotid atheromata.36,37 Moreover, the protease-sensing probe is expected to enter clinical trials in 2010.31 Other molecular imaging biomarkers that reflect plaque vulnerability include annexin for apoptosis,63,64 integrins for angiogenesis,65 and inflammatory cells such as macrophages.66-68

To summarize, molecular optical imaging could contribute to stroke prevention by detecting protease activity in atherosclerotic plaques in vivo; it would thereby serve as a powerful tool for evaluating vascular inflammation, for determining individualized therapeutic strategies including pre-operative planning, and for monitoring the effects of therapeutic interventions.

Fibrinolytic therapy, which is the only FDA-approved treatment for acute cerebral infarction, inevitably imposes a hemorrhagic risk.10,13,14 Delayed or intra-arterial thrombolysis is associated with higher complication rates, and enormous effort has been expended to empirically define the risks and benefits of thrombolysis at different times after vascular occlusion.14 Predicting the thrombolytic resistance and hemorrhagic risk in individual patients would be extremely helpful when deciding whether or not to perform thrombolytic therapy in individual patients.

As a first step toward imaging thrombolytic resistance in vivo, we and others chose to devise and characterize a molecular imaging agent to probe for the activity of factor XIII (FXIII) enzyme, a thrombin activated transglutaminase.69-71 The FXIII probe contains NIR fluorochromes attached to the peptide sequence NQEQVSPLTLLK, which is specifically recognized by the coagulation enzyme as its substrate. The normal function of activated FXIII (FXIIIa) enzyme is to stabilize a thrombus by cross-linking fibrin fibers-thus increasing the tensile strength-so as to covalently bind molecules that impede plasmin activity, such as α2-antiplasmin.72

Imaging this coagulation enzyme activity may be useful to staging the thrombotic disease when determining the appropriate therapy by clarifying the acute-versus-chronic state of the clot, which is difficult or impossible to achieve with anatomic imaging only. When the probe was injected intravenously into mice with ferric-chloride-induced thrombi in the femoral artery, the FXIIIa-related NIRF signal intensity was proportional to the age of the thrombi.71

The activity of FXIIIa enzyme is expected to change far more rapidly than the clot volume, and may serve as an early predictor of the success or failure of a therapeutic regimen.73 In this context, we investigated whether FXIIIa imaging could reflect treatment effects. To study the effect of heparin in a mouse model of cerebral venous thrombosis, heparin was administered immediately before thrombus induction and the FXIII probe was injected 3 hours later.70 The FXIIIa-related NIRF signal, captured through a cranial window using an intravital fluorescence microscope (Fig. 5), was significantly weaker in the heparin-pretreated group than in the untreated group. Notably, the therapeutic response to the anticoagulation treatment could be monitored in real time. We are currently studying if FXIIIa imaging can predict thrombolytic resistance in an arterial stroke model. Thrombolytic resistance is likely multifactorial in origin, and hence other imaging or plasma biomarkers of thrombolytic resistance merit further investigations, such as plasminogen activator inhibitor-1, α2-antiplasmin, and factor V.74,75

The future combination of molecular optical imaging with fluorescence-sensing intra-arterial catheter systems is expected to be suitable for clinical application, with molecular diagnosis guiding therapy with fibrinolytic agents or thrombin inhibitors and their successor drugs in acute ischemic stroke. This approach may allow discrimination of thrombi that would be resistant to chemical thrombolytic agents. Early triage to mechanical recanalization and clot retrieval methods would be rational when chemical thrombolysis is expected to fail, if patients could be triaged appropriately to these treatments (Fig. 6).

Intra-operative molecular imaging may assist neurosurgeons who are performing procedures such as post-stroke decompression surgery or revascularization operations to visualize cerebral perfusion and penumbra in real time. Woitzik and colleagues76 introduced intraoperative cerebral videoangiography, an imaging system integrated into the surgical microscope, which allowed 1) real-time imaging of the cortical blood flow in patients undergoing decompressive craniectomy and 2) cortical perfusion areas to be stratified into ischemic core and penumbral areas. They used ICG, which binds tightly to plasma proteins and becomes confined to the vascular system; its short half-life and emission at a NIR wavelength (835 nm) make it an advantageous contrast agent for in vivo angiography. A cerebral blood flow index (the ratio of the difference in fluorescence intensity and rise time) was calculated to determine flow maps that were used to assign a significant portion of tissue (15-37%) to the penumbra,76 which is a potential target to maximize benefits from the decompressive craniectomy, which in turn reduces intracranial pressure and improves cerebral blood flow.77

Clinically, in vivo retinal angiography with ICG has long been used to visualize neovascularization and hemorrhage in retinal and choroidal tissues.78 In a mouse model with a cranial window, we showed that cerebral ICG-based angiography could reveal permeability of the blood-brain barrier by sensing ICG leakage from cerebral vessels due to thrombosis-induced venous hypertension in real time in vivo.70 The technique was recently used to evaluate cortical microvascularization to compensate for a reduced cerebral blood flow in Moyamoya disease.79 ICG-based angiography demonstrated that the microvascular density and diameter were significantly elevated in patients with Moyamoya disease.

In addition to the cerebral perfusion status, the penumbral area could be defined by various physiological parameters including decreased pH due to increased intracellular lactate accumulation.80-82 pH-Activatable molecular optical imaging probes conjugated to a cancer-targeting monoclonal antibody were developed recently.83 The probes consist of a boron-dipyrromethene (BODIPY) as a fluorophore and a dialkylated aniline that renders the fluorophore fluorescent at different acidic pHs. BODIPY dye with a nonprotonated form of dialkylated aniline is nonfluorescent, whilst the dye with a protonated form of aniline is highly fluorescent. Conjugation of the probe to a cancer-specific monoclonal antibody, trastuzumab, leads to recognition and internalization by HER2-positive cancer cells via the endosomal-lysosomal degradation pathway. An acidic pH in the lysosome then turns on the fluorophore, and hence the probe offers tumor specificity with a minimal background signal. It should be noted that the imaging agent is not an "always-on" probe like conventional targeted probes. Instead, it becomes fluorescent under specific conditions, which improves the specificity and sensitivity as in the case of protease activatable probes. More importantly, since the proton pump is required to maintain the acidic condition in the lysosome, only viable cancer cells can be visualized by the probe. This kind of pH-sensing probe may be useful in delineating cerebral penumbra and infarcted tissues.

Molecular imaging researchers are actively pursuing 1) the development of novel molecular imaging strategies, particularly multi-modal molecular imaging, 2) innovations in molecular imaging-based theranostics, and 3) bench-to-bedside translation of associated cutting-edge technologies. Cardiovascular and neurovascular molecular optical imaging researchers pursue similar aims. Multi-modal molecular imaging probes are detectable by optical, MRI, and nuclear approaches simultaneously, which provides both high molecular sensitivity and high anatomical resolution as shown by Kircher et al.84 combined MR and optical imaging. Nahrendorf et al. labeled dextranated and DTPA-modified magnetofluorescent nanoparticles with ⁶⁴Cu to yield a macrophage-targeted trimodality reporter for combined MRI, PET, and optical imaging.67

Combining diagnostic molecular imaging with therapeutic (theranostic) approaches would allow diseases to be detected at an early stage, drug delivery to desired targets to be monitored, quantification of the therapeutic efficacy in an individual patient in vivo, and the treatment for the patient to be tailored based on the imaging results. The development and utilization of protease-mediated photodynamic agents could be regarded as a theranostic approach.85 The agents are nontoxic in their native state, but they become fluorescent and produce singlet oxygen on protease conversion; after being delivered and activated in and around the macrophages infiltrating vulnerable plaques, intravenously injected theranostic agents are expected to not only visualize rupture-prone plaques but also stabilize them as a photodynamic therapy while avoiding systemic phototoxicity.

Healthcare in the 21^st century is evolving to be personalized, predictive, and preventive. Molecular imaging provides a fundamentally new type of information that allows a timely and precise assessment of pathophysiologic states at the cellular and molecular levels, thereby complementing that available from anatomically based imaging modalities. Molecular optical imaging will provide the information needed to provide personalized stroke care. Most importantly, well-designed translational animal studies and human trials will be key to establishing the use of molecular imaging tools in clinical stroke management.86