While changes in the proton signal intensity or relaxation times can be used to measure contrast agent uptake or tissue physiology, these methods are indirect measures of the contrast agent. An alternative method is direct detection of a signal originating from the agent itself, much as PET, SPECT and optical imaging directly map a nuclear or fluorescent tag incorporated onto the agent. There are a number of important nuclei that are visible by CMR techniques, such as 19F, 23Na, 31P, and 13C, however, the sensitivity of these elements tend to be quite low. Typical clinical MRI utilizes the proton (1H) signal, which represents a concentration of 110 M and a relative MRI sensitivity of 1.0. In comparison, the relative sensitivities of 19F, 23Na, 31P, and 13C are 0.83, 0.093, 0.066 and 0.016, respectively. Furthermore, the concentrations of these elements tend to be very low in biological samples, typically less than 10 mM, lowering their sensitivity by orders of magnitude compared to proton.
19F CMR of Perfluorocarbon Nanoparticles
Direct detection of 19
F has been explored in a number of research studies because it has a relatively high sensitivity, 83% compared to 1
H, and there is virtually no native background signal. Thus, fluorinated contrast agents can provide a definitive and quantitative CMR signature. For example, 19
F CMR of fibrin-targeted perfluorocarbon (PFC) nanoparticles could be used to map the formation of thrombi on ruptured atherosclerotic plaques and quantify the extent of ruptures in the fibrous cap. Human carotid endarterectomy samples were treated with fibrin-targeted paramagnetic nanoparticles and imaged at 4.7 T [74
H CMR showed high levels of signal enhancement along the luminal surface due to nanoparticle binding to fibrin deposits. A 19
F projection image of the artery, acquired in less than five minutes, displayed an asymmetric distribution of nanoparticles around the vessel wall corroborating the 1
H signal enhancement. Spectroscopic quantification of nanoparticle binding allowed calibration of the 19
F CMR signal intensity. Co-registration of the quantitative nanoparticle map with the 1
H image permitted visualization of both anatomical and pathological information in a single image (Figure ). Combining information from 1
H and 19
F CMR could allow prediction of subsequent occlusion or distal embolization from unstable or disrupted plaques, and aid clinical decision-making for acute invasive intervention vs. pharmaceutical therapies.
Figure 5 Direct quantitation of contrast agent binding utilizing 19F CMR and fibrin-targeted PFC nanoparticles. (a) Optical image ex vivo of a 5-mm cross section of a human carotid endarterectomy sample. This section showed moderate luminal narrowing as well as (more ...)
Further experiments performed on a 1.5 T clinical CMR system used rapid steady-state imaging to independently image and quantify two different populations of fibrin-targeted nanoparticles, perfluorooctylbromide (PFOB) or perfluoro-15-crown-5-ether (CE), based on their unique spectral signatures [75
]. Both imaging and spectroscopy could distinguish nanoparticles containing either PFOB or CE as the core material. The signal to noise for PFOB was lower than CE (10 vs. 25, respectively), presumably due to the single CE peak (20 equivalent fluorine atoms) compared to the multiple PFOB peaks (17 fluorine atoms distributed over 5 peaks). A clear linear relationship between the 19
F signal intensity and perfluorocarbon concentration was demonstrated for both PFOB and CE using both imaging and spectroscopy. From this demonstration on fibrin clots, it follows that multiple perfluorocarbon nanoparticle agents could be used to target different epitopes and achieve a noninvasive analogy to immunohistochemistry. For example, simultaneous quantification of angiogenesis in the vessel wall and fibrin deposition on the plaque cap could be used to evaluate the pathophysiological stage of an obstructive lesion.
Tissue inflammation is another biological marker of unstable atherosclerotic plaques, inducing the release of cytokines and upregulating the production of proteins like vascular cell adhesion molecule-1 (VCAM-1). PFC nanoparticles targeted to VCAM-1 were injected into genetically engineered ApoE-/-
mice, which mimic the clinical progression of atherosclerosis, to map inflammation [76
]. These mice display focal inflammation and macrophage infiltration in the kidneys [77
]. To definitively identify nanoparticle binding in the kidneys, 19
F CMR was performed on a 11.7 T research scanner 2 hours after injection of VCAM-1 targeted PFC nanoparticles (Figure ). The 19
F signal arising from the PFC core provided an unambiguous marker of particle accumulation, without the innate signal variations that can confound typical 1
H signal enhancement with traditional paramagnetic or superparamagnetic CMR contrast agents. VCAM-1-targeted nanoparticles accumulated in ApoE-/-
kidneys to a greater extent than non-targeted nanoparticles (3.7 billion particles per gram of tissue vs. 0.9 billion particles/gram). The uptake of targeted nanoparticles was also higher in the kidneys of ApoE-/-
mice compared to non-ApoE-/-
controls (3.7 vs. 1.6 billion particles/gram). Control animals also displayed no significant difference in the uptake of targeted versus non-targeted nanoparticles (1.6 vs. 1.5 billion particles/gram).
Figure 6 Mapping tissue uptake of VCAM-1 targeted PFC nanoparticles with 19F CMR. Multinuclear imaging of kidneys from atherosclerotic ApoE-/- (top) and wild-type control (bottom) mice imaged at 11.7T. (A) Proton MR of kidney anatomy. (B) 19F CMR for direct detection (more ...)
Typically, targeted CMR contrast agents generate image enhancement as a result of both nonspecific blood pool signal as well as the specific binding of the agent to the biomarker of interest. Separating out these two contributions can be very difficult to achieve in vivo. One method to suppress the nonspecific signal utilizes diffusion weighted 19
F spectroscopy to null the signal arising from moving particles, representing the unbound fraction in the blood pool, while retaining the signal from stationary particles that are specifically bound to the target epitope [78
]. A genetically engineered mouse model of squamous cell cancer, derived by incorporating human papilloma virus into the mouse genome [79
], was studied with a 11.7 T research scanner. Transgenic and age matched control mice were injected with 1 ml/kg αν
-targeted PFC nanoparticles (corresponding to a PFC dose of 0.2 ml/kg) and scanned 90 minutes later to assess angiogenesis in these precancerous lesions. Both the transgenic and control mice displayed decreased 19
F signal with increasing b-values. However, 60-100% of the 19
F signal remained at b-values near 60,000 s/mm2
in the transgenic animals, while no detectable 19
F signal was observed in control mice at b-values of 1500 s/mm2
. The calculated apparent diffusion coefficient (ADC) of PFC nanoparticles was 33.1 m2
/s in the transgenic mice, significantly lower than the 19,563 m2
/s ADC in the controls.
Angiogenesis is also a prominent feature in the progression of aortic valve stenosis. Cholesterol-fed rabbits develop aortic valve sclerosis, characterized by gross thickening, macrophage infiltration, calcification, and eventual bone formation that mimics the clinical presentation of the disease [80
]. Cholesterol-fed rabbits underwent 19
F CMR after injection of αν
-targeted nanoparticles to quantify angiogenesis in the aortic valve leaflets [84
]. The cholesterol feeding caused gross thickening of the aortic valves accompanied by extensive foam cell infiltration, non-calcified bone deposition, activation of myofibroblasts, abnormal microvascular proliferation and upregulation of αν
-integrin expression. None of these abnormalities were observed in the normal valve tissue from control animals. Rabbits received IV injections of 2.2 ml/kg αν
-targeted nanoparticles, nontargeted nanoparticles or in vivo competitive inhibition of αν
-integrin binding via pretreatment with αν
-targeted safflower oil nanoparticles. Two hours after nanoparticle injection, the aortic valve leaflets were excised for 19
F MR spectroscopy at 11.7T.
The crown ether (CE) signal arising from the nanoparticle contrast agent was readily detected and distinguished from a perfluorooctylbromide (PFOB) quantification reference based on the chemical shifts of these perfluorocarbon species (Figure ), allowing quantification of the total volume of bound nanoparticles (Figure ). The volume of targeted nanoparticles bound to the valves was 19.5 nL, which was more than three times higher than the amount of nontargeted nanoparticles (5.6 nL). Competitive inhibition of ανβ3-integrin binding reduced the amount of nanoparticles in the valves by about half (10.3 nL). Valves from healthy rabbits treated with targeted nanoparticles contained almost nine times fewer nanoparticles (2.3 nL) than the valves from cholesterol-fed rabbits. These techniques may be useful for assessing atherosclerotic components of preclinical aortic valve disease in patients and could assist in defining efficacy of medical therapies. The sensitivity of this approach for molecular detection of sparse quantities of inflammatory epitopes in very thin structures at high field strengths establishes a basis for future efforts to develop localized spectroscopic methods at clinical field strengths that could be useful for detecting disease and monitoring therapies.
Figure 7 Quantitative 19F spectroscopy of angiogenesis in aortic valve disease. The 19F signal was utilized to quantify binding of nanoparticles to the valve leaflets from (A) a rabbit treated with ανβ3-targeted nanoparticles and (B) a (more ...)
Figure 8 Quantitative comparison of nanoparticle binding in valve leaflets. The volume of nanoparticles (in nanoliters) bound to the valves was calculated from the 19F signal. The valves treated with ανβ3-targeted particles displayed three (more ...)
The ability of 19F CMR to directly detect and quantitate the binding of specifically targeted nanoparticles is a significant advantage over the use of paramagnetic or superparamagnetic contrast agents, which are only visible based on their effects on the water signal. However, a lingering disadvantage of these techniques is the limited sensitivity of CMR to any signal other than the bulk water. These studies usually overcome this inherent limitation by using some combination of high magnetic field strength, employing MR spectroscopy, and/or very long scan times.
Mapping Tissue Oxygenation
As with proton imaging, 19
F contrast agents have been developed for mapping tissue physiology, most notably oxygenation. Oxygen is essential for tissue viability and without an adequate supply, cellular dysfunction and death rapidly occurs. Noninvasive monitoring of tissue oxygenation could be utilized for diagnostic and therapeutic applications in a range of common diseases, including myocardial ischemia, cancer, stroke and peripheral vascular disease. The most common CMR technique to monitor tissue oxygenation is Blood Oxygen Level Dependant (BOLD) imaging. BOLD imaging is sensitive to the ratio of oxyhemoglobin and deoxyhemoglobin and can be utilized for high spatial and temporal mapping of the dynamic changes in brain oxygenation during functional CMR studies [85
]. However, the BOLD CMR signal is not quantitatively related to tissue oxygenation, because it depends strongly on a number of other factors including blood volumes, blood flow, hematocrit and pH [86
]. Other CMR oximetry techniques use contrast agents that are sensitive to the local oxygen tension (pO2
). A number of these agents are based on perfluorocarbon molecules, which display a linear dependence of the 19
F spin lattice relaxation rate on pO2
]. In order to accurately measure pO2, the T1 relaxation time of the fluorinated compound must be quantified in order to avoid the influence of contrast agent concentration, T2 and other factors on the CMR signal. Molecular oxygen is paramagnetic and the solubility of oxygen in perfluorocarbons is three to ten times higher than in water [88
]. Oximetry based on 19
F CMR capitalizes on a number of strengths: it is a spin 1/2 nucleus, the sensitivity is approximately 83% compared to 1
H, it is 100% abundant, and endogenous fluorine in biological samples only occurs at very low levels and is typically undetectable because of very short T2 relaxation times. Due to the lack of a background 19
F signal in the body, PFC agents can be definitively identified and quantified by CMR.
PFCs generating multiple spectroscopic resonances can be used provide multiple estimates of pO2
or to measure pO2
and another physiological parameter, such as temperature, that affects the 19
F R1 value by solving simultaneous equations [90
]. For imaging, however, the multiple 19
F resonances can generate chemical shift artifacts or reduce the overall signal-to-noise [91
]. PFCs with a single resonance, such as perfluoro-15-crown-5-ether or hexafluorobenzene offer high pO2
sensitivity and minimal temperature sensitivity [93
]. For example, hexafluorobenzene has been investigated for mapping tumor oxygenation during hyperoxic interventions using echo planar imaging (EPI) to maximize the temporal resolution [95
Quantification of Metabolic Flux
Another important physiological marker of disease is metabolic flux rates. For example, the creatine kinase (CK) reaction is responsible for replenishing adenosine triphosphate (ATP) by using phosphocreatine (PCr). Quantitative 31
P spectroscopy studies have shown significant reductions in cardiac PCr and ATP concentrations in MI patients compared to healthy controls [96
]. However, these measurements do not distinguish between cell death, reduced substrate availability or impaired enzyme activity. MR spectroscopy can be used to measure the rate of metabolic reactions by tracing the exchange of saturated spins from one molecule to another. The pseudo-first-order CK rate constant, k, reflects the intracellular CK reaction kinetics and is independent of myocyte number, while CK flux is defined as the product of [PCr] and k. The value of k can be interpreted as the fraction of the PCr pool used to create ATP via the CK reaction each second, which is a measure of intracellular metabolic function. Therefore, k depends only on the surviving cells that contribute to the 31
P MRS signal and is not confounded by myocyte loss. On the other hand, reduced myocardial CK flux can be due to a loss of total enzyme activity, altered intracellular substrate levels, or allosteric modifications of the enzyme.
In a clinical study of myocardial infarction patients, the CK kinetics were measured noninvasively using 31
P spectroscopy. The tissue concentrations of ATP and PCR were measured and CK kinetics (k and CK flux) were measured by magnetization transfer on a 1.5T clinical scanner [97
]. Myocardial [ATP] and [PCr] were 39% to 44% lower in MI patients compared to healthy controls, however the myocardial CK rate constant, k, was normal in these patients. As a result of the lower tissue PCr levels, the CK flux was reduced by 50% in the MI patient population. These results demonstrate that ATP loss following MI is a direct result of PCr depletion, most likely due to myocyte loss. The maintenance of normal k values indicates that intracellular CK metabolism is maintained in the surviving myocytes. These results reinforce the use of therapies for MI patients that combat substrate loss or reduce energy demand, rather than those that increase workload in the surviving tissue. For example, beta blockers are routinely prescribed for MI patients because they reduce the heart rate and myocardial oxygen consumption. Using similar 31
P spectroscopy techniques, a 50% reduction in CK flux has been measured in patients with non-ischemic dilated cardiomyopathy and mild-to-moderate chronic heart failure (CHF) [98
] and a 65% decrease in CK flux has been reported in patients with pressure-overload left ventricular hypertrophy and CHF [99
Although this study demonstrates 31P CMR on clinical patients, there are a number of technical hurdles that limit the wide-spread use of this method in the clinic, including the instrumentation required, limited resolution and long scan times. 31P spectroscopy requires specialty instrumentation that is not found on standard clinical CMR scanners, such as dedicated cardiac 31P coils and data processing packages. Also, the 31P data was localized in only one dimension, yielding a relatively poor resolution of roughly 6.5 by 6.5 by 1 cm. This would limit the application to patients with relatively large and well defined MIs. Furthermore, completion of the spectroscopy exam took about 70 minutes. As a result, the other CMR exams, including cine, tagging and delayed myocardial enhancement, were performed during a separate imaging session.