|Home | About | Journals | Submit | Contact Us | Français|
Functional imaging with MRI contrast agents is an emerging experimental approach that can combine the specificity of cellular neural recording techniques with noninvasive whole-brain coverage. A variety of contrast agents sensitive to aspects of brain activity have recently been introduced. These include new probes for calcium and other metal ions that offer high sensitivity and membrane permeability, as well as imaging agents for high resolution pH and metabolic mapping in living animals. Genetically-encoded MRI contrast agents have also been described. Several of the new probes have been validated in the brain; in vivo use of other agents remains a challenge. This review outlines advantages and disadvantages of specific molecular imaging approaches and discusses current or potential applications in neurobiology.
As neuroscientists become increasingly brave in their efforts to study the functioning of neural systems in vivo, there is a growing need for measurement methods that can record comprehensive information about the functioning of living brains. Magnetic resonance imaging (MRI) is a special tool in this regard, because of its relatively high spatial resolution (~ 10 μm in high magnetic field scanners) and capacity to scan entire organisms noninvasively. Functional MRI (fMRI) with contrast dependent on cerebral hemo-dynamics provides an indirect readout of neural activity [1-3]. Although hemodynamic fMRI has had transformative impact in cognitive science, the techniques lack the specificity and temporal precision of electrophysiology and optical imaging, and have not been widely used in basic neurobiology experiments.
Another way to exploit the unique advantages of MRI for neuroscience is to perform the imaging in conjunction with molecular probes (contrast agents) sensitive to aspects of neuronal physiology . This approach is roughly analogous to performing optical neuroimaging with fluorescent dyes, but is currently far less well-developed. Most MRI contrast agents are paramagnetic chemicals that increase parameters called the T1 and T2 relaxation rates of water, as observed in tissue and solution; T1 or T2 relaxation enhancements produce image brightening or darkening, respectively. Additional classes of contrast agents work by a chemical exchange-based mechanism called CEST , or involve imaging nonstandard nuclei like 19F and 13C. The characteristics and physical mechanisms of different types of contrast agent are discussed at length in a number of book chapters and reviews [6-11], and are summarized in Figure 1. In general, for any agent to be used in functional imaging, either its ability to influence MRI contrast or its spatial distribution must be sensitized to neural activity in some way.
The past few years have seen significant advances in the design of new MRI contrast-based sensors and the introduction of protein contrast agents for brain imaging. These are nascent technologies—few of the efforts have progressed beyond an in vitro or proof-of-concept stage, but in several cases experiments using the new agents in animals can now be performed. The remainder of this review describes contrast agents suitable for functional imaging based on metal ions, pH, metabolic activity, and gene and protein expression. Prospects for future development and application of molecular fMRI methods are discussed.
Calcium ions are an important target for neuroimaging agents because neuronal calcium fluxes are dramatic and directly related to synaptic activity. Recent two-photon fluorescence imaging studies have demonstrated the power of calcium measurements to characterize neuronal population behavior in exposed regions of the brain [12,13]. MRI indicators for calcium can facilitate calcium imaging of deep tissue structures. Several relaxation-based contrast agents for calcium-dependent MRI have been introduced. An increasingly widespread approach was introduced by Koretsky and colleagues, who showed that Mn2+ functions as a paramagnetic Ca2+ mimetic and accumulates activity-dependently in neurons [14,15]. Because of its slow uptake and release kinetics (on the order of hours and days, respectively ), Mn2+ has proved useful as an “activity label” analogous to 2-deoxyglucose or c-Fos. The technique was recently used for 100 μm isotropic resolution T1-weighted mapping of auditory cortex in mice , and for a functional study of the antennal lobes of developing moths .
A contrast agent sensor designed for real-time calcium imaging was developed by Li et al. . The agent was formed by attaching the calcium chelator 1,2-bis-(O-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid (BAPTA) to two copies of the highly paramagnetic gadolinium complex Gd3+-1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (Gd-DOTA); the agent undergoes a change in T1 relaxivity (a measure of efficacy) near 1 μM Ca2+. Delivery of the sensor to neurons in sufficient quantity for functional imaging in vivo has not yet been reported. Partly in response to this situation, Atanasijevic et al.  have now described a new family of calcium sensors derived from extremely potent superparamagnetic iron oxide (SPIO) nanoparticle contrast agents. SPIOs were conjugated to “tunable” recombinant proteins that drove reversible particle aggregation around a midpoint of 0.8 μM Ca2+ and produced over 100% T2-weighted MRI signal changes in vitro. A key advantage of the these agents is that because of their high relaxivity, they may be used at concentrations (~ 1 nM) that are both easier to deliver and less disruptive to cellular calcium dynamics than Gd3+-based sensors (effective at 10−100 μM). A disadvantage of the SPIO calcium sensors is that they respond relatively slowly to calcium changes . As with the Gd3+-based calcium sensors, fMRI with SPIO sensors could be possible once effective intracellular delivery strategies are harnessed; SPIO uptake by brain cells in vivo has been demonstrated [22,23].
Ions of transition metals such as zinc and copper are also influenced by neural activity, and alterations of transition metal homeostasis have been associated with a number of neuropathologies. Gadolinium-based sensors for zinc  and copper , similar to the Li et al.  calcium sensor, have been synthesized and shown to produce T1 relaxation changes in vitro. Zhang et al.  recently introduced an MRI zinc sensor derived from the Mn3+ complex with 5,10,15,20-tetraphenylporphinetetrasulfonic acid (Mn-TPPS). Because their porphyrin framework and zinc-binding moieties are amphiphilic, these contrast agents are cell permeable. The authors demonstrated zinc-dependent MRI signal changes in cells incubated with the agent. Future studies will indicate to what extent this approach can be applied in intact animals.
The extracellular medium becomes slightly acidified during neural activity . Although these changes (in the range from pH 7.2−7.4) are not restricted to individual neurons, they could be monitored by pH-sensitive probes and used for functional fMRI. Both relaxation and CEST-based MRI contrast agents work by mechanisms that involve water or proton exchange (Figure 1A-C), which are inherently pH dependent and therefore easily compatible with pH sensing. In fact, a diverse set of pH indicators for MRI has been described (reviewed in ), though none of the indicators has yet been demonstrated to detect changes in neuronal activity. In one of relatively few in vivo studies, Garcia-Martin et al.  used a phosphonated Gd3+-based contrast agent to measure intravascular acidification in rat gliomas. Differences of the order of one pH unit could be distinguished, and absolute pH values were obtained using a calibration procedure . The contrast agent used in this study experiences changes in T1 relaxivity over the broad range from pH 6−8 ; the sensor could in principle be applied for functional brain imaging, following intracranial injection or blood-brain barrier disruption, but MRI signal changes would be expected to be less than one percent under realistic pH fluctuations (< 0.2 units) and agent concentrations (~ 100 μM). Synthesis of novel sensors optimized for sensitivity in the pH 7.2−7.4 range may therefore be critical to developing this approach for fMRI. An alternative is the use of intrinsic protein amide proton contrast for CEST-related pH imaging in the brain . Initial studies applied this method to detect focal ischemia in rats, involving pH changes on the order of 0.5 units.
Changes in metabolic activity are closely coupled to neural signaling, and MRI contrast agents sensitive to cellular respiration may be used for functional imaging. Given evidence that metabolic processes including the consumption of oxygen are locally regulated on a much faster timescale than changes in hemodynamics , direct monitoring of these variables could provide more precise information about brain function than hemodynamic fMRI techniques can. The best known oxygen sensitive contrast agent is the endogenous iron containing protein hemoglobin, which underlies blood oxygen level dependent (BOLD) contrast . Hemoglobin can also be used as an exogenous sensor in tissue , but it does not seem to provide enough sensitivity to detect rapid deoxygenation events (“initial dips”) associated with changes in neural activity. Cerebral metabolite uptake has traditionally been measured using radioactive glucose analogs, in conjunction with positron emission tomography (PET) or postmortem autoradiography. Golman et al.  have now introduced a sophisticated technique for performing similar experiments by MRI. The technique involves following the kinetics of a 13C-labeled metabolite, pyruvate, by 13C MRI (Figure 1D). Normally, 13C MRI is too insensitive to detect pyruvate at physiologically relevant concentrations, but here the authors used a method called dynamic nuclear polarization to boost the MRI signal from 13C1-pyruvate using a specialized device . Kinetics of pyruvate uptake and turnover to lactate and alanine were followed in the muscles and abdominal organs of rats and pigs. Whether this or related approaches can be useful for functional brain imaging is unclear at present. Principal difficulties involve the relatively rapid decay of MRI signal from the tracer (time constant 15−20 s in vivo), the fact that other metabolites (e.g. glucose) have much shorter decay times, and the need in fMRI applications for periodic or continuous supply  of polarized agents to the brain.
The discovery of green fluorescent protein (GFP) and the development of genetically-encoded fluorescent indicators like cameleons  and synaptophluorins  are continuing to revolutionize the modern practice of neuroscience. Unlike fluorescent proteins, genetically encodable contrast agents (most of them paramagnetic metalloproteins) are plentiful in nature, but it is only in the past few years that any of these have been exploited as ectopically expressed markers for imaging. The iron storage protein ferritin (Ft) encloses a core of ferrihydrite with partially superparmagnetic properties, making Ft a close natural analog of SPIO contrast agents . In a 2005 paper, Genove et al.  demonstrated that viral-mediated overexpression of Ft in mouse brain led to clear changes in T2-weighted images (Figure 2A). Iron loading and relaxivity changes induced by Ft can be boosted by co-expressing transferrin receptor, another participant in endogenous iron metabolism . A recent report has now shown that Ft subunits expressed in transgenic mice can be detected in multiple tissue types and in utero without pathological side-effects , suggesting that Ft may find broad utility as a marker protein in MRI. This study showed that Ft expression even in relatively sparse endothelial cells led to detectable contrast changes (Figure 2B). Another protein contrast agent was cleverly designed by Gilad et al. , who boosted the concentration of exchangeable amine protons in transfected cells using an artificial lysine rich protein (LRP). Using the CEST MRI method (see Figure 1C), cells expressing the LRP could be distinguished from controls both in test tubes and in xenografted tumors (Figure 2C). Unlike Ft, which requires iron loading to induce contrast, LRP is a contrast agent as soon as it is translated; this may permit LRP expression changes to be detected on a shorter timescale than changes in Ft levels. On the other hand, MRI signal changes reported by Gilad et al.  were relatively subtle and required long imaging times to resolve (> 30 min.), and it is not yet known whether LRP expression-mediated contrast may be generalized easily to other contexts.
How could genetically encoded contrast agents be used for functional brain imaging? Although this has not been reported, a technically straightforward approach would be to express a protein contrast agent under control of a promoter known to be regulated by neural activity, like those of immediate early genes (IEGs) fos and arc. Because IEG protein induction generally persists for hours , a method like this would not be useful for functional imaging on the timescale of conventional fMRI or neurophysiology techniques, but it could be used in fairly simple (and potentially longitudinal) mapping studies in animals, somewhat like Mn2+ labeling technique discussed above. A more exciting direction from the perspective of systems neuroscience would be the engineering of MRI sensors for neural activity using protein contrast agents as building blocks. Key advantages of genetically encoded sensors over synthetic sensors include the possibility that they might be genetically targeted to specific cell types, the relative ease of delivering genes vs. imaging agents, and the fact that protein contrast agents may be cheaper to use and easier to modify than many synthetic contrast agents.
Genetic mechanisms can be used to direct MRI contrast due to exogenous agents; “semi-genetic” approaches to functional imaging might offer better sensitivity than protein contrast agent expression, particularly if high relaxivity agents or enzymatic amplification strategies are incorporated. Initial examples included detection of lacZ marker expression in developing frog embryos using a gadolinium-chelating β-galactosidase substrate , and monitoring of a transferrin receptor reporter gene in mice using transferrin-conjugated SPIOs . Semi-genetic contrast mechanisms based on a variety of marker proteins and receptors have now been reported (reviewed in [49-51]), and design of contrast agents targeting RNA transcripts has also been described [52,53]. Measurement of biological processes in the nervous system has not yet been convincingly demonstrated, however.
A number MRI contrast agents with potential utility for functional imaging have been discussed. Table 1 summarizes advantages and disadvantages of many of the approaches. Although some of the contrast agents have been applied in animals, only Mn2+ dependent labeling has so far been used for functional imaging of neural activity. For basic neuroscience studies, none of the new techniques is currently a surrogate for hemodynamic fMRI or invasive neural recording methods. Major progress has been achieved recently, however, with the development of new MRI probes for sensitive detection of brain-related physiological variables and the introduction of protein and genetically-controlled contrast agents. Several of these agents have been used to make measurements in vivo; applications to functional neuroimaging appear feasible in some cases, perhaps within the next five years. In addition to the persistent challenges of obtaining sensitivity and specificity for neural events, a hurdle in developing molecular fMRI techniques further will be the need to distinguish molecular signatures of activity from hemodynamic responses. Validation experiments in reduced preparations and in animals with suppressed BOLD responses may be valuable. Future work in this area will certainly focus on extending applications of the existing contrast agents in live animals, the development of more genetically-controlled probes, and the creation of MRI sensors for previously unexplored aspects of neural signaling, such as membrane potential and neurotransmitter release.
The author wishes to acknowledge support from the NIH (EB5723), the McKnight Endowment Fund for Neuroscience, and the Raymond and Beverly Sackler Foundation.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.