Molecular fMRI of neurochemical signaling

Abstract

Magnetic resonance imaging (MRI) is the most widely applied technique for brain-wide measurement of neural function in humans and animals. In conventional functional MRI (fMRI), brain signaling is detected indirectly, via localized activity-dependent changes in regional blood flow, oxygenation, and volume, to which MRI contrast can be readily sensitized. Although such hemodynamic fMRI methods are powerful tools for analysis of brain activity, they lack specificity for the many molecules and cell types that play functionally distinct roles in neural processing. A suite of techniques collectively known to as "molecular fMRI," addresses this limitation by permitting MRI-based detection of specific molecular processes in deep brain tissue. This review discusses how molecular fMRI is coming to be used in the study of neurochemical dynamics that mediate intercellular communication in the brain. Neurochemical molecular fMRI is a potentially powerful approach for mechanistic analysis of brain-wide function, but the techniques are still in early stages of development. Here we provide an overview of the major advances and results that have been achieved to date, as well as directions for further development.

Keywords: Contrast agent; FMRI; Ion; Molecular imaging; Neurotransmitter; Sensor.

Fig. 1.

Mechanisms for molecular fMRI-based detection of neurochemicals. (A) Binding of a neurochemical analyte (magenta) to a T₁ contrast agent (gray) can cause changes in exposure of the agent’s paramagnetic center (green) to bulk water (blue background) that is directly detected by MRI. Longitudinal relaxation is enhanced (yellow halo) in the more solvent exposed state (right), brightening MRI signal in T₁-weighted scans. (B) To induce T₂ contrast changes, analytes (magenta) can induce clustering of MNPs (green). For the most commonly applied MNPs, clustering enhances T₂ relaxation (yellow halo) of the surrounding solvent (blue background), darkening signal in T₂-weighted scans (right). (C) Spectroscopic molecular MRI methods directly or indirectly detect specific atoms (cyan) that belong to a neurochemical probe or analyte of interest (magenta). Spectroscopic changes usually result from changes in the concentration of the analyte (e.g. increases, right), but changes in the analyte structure can also give rise to spectroscopic effects. The CEST effect (inset) exploits exchange between spectroscopically resolved protons with the bulk solvent (cyan arrows) as a means for amplifying detection of the analyte. Hyperpolarization involves transiently boosting the magnetic resonance signal associated with the resolved atom.

Fig. 2.

Molecular fMRI of dopamine. (A) Structure of the protein-based dopamine-sensitive MRI contrast agent BM3h-9D7, showing dopamine (DA, blue) binding near the paramagnetic heme group. This interaction blocks direct access of water molecules to the heme iron, required for effective T₁ relaxation. (B) Longitudinal relaxivity (r₁) of BM3h-9D7 in the absence vs. presence of excess dopamine. Error bars = standard deviations of three replicates each. Inset shows T₁-weighted MRI scans of corresponding samples in a microtiter plate. (C) Molecular fMRI maps of peak dopamine release concentration ([DA], top) and BOLD fMRI signals (bottom) observed in the medial striatum (bregma +1.5 mm) in response to rewarding stimulation of the medial forebrain bundle in rats. Dopamine signals were measured in the presence of BM3h-9D7 and the thin blue line shows the area filled with contrast agent; white lines indicate brain atlas features. Maps show a spatial dissociation of signals detected by dopamine and hemodynamic fMRI. (D) The full width at half-maximum (FWHM, left) and signal amplitude (%SC, right) of dopamine and BOLD responses in ventral striatum differ as a functional of stimulation frequency, providing further dissociation between neurochemical and hemodynamic signals. Error bars = standard errors from five animals. (E) Comparison of dopamine and conventional fMRI signals permits a neurochemically specific impulse response function (IRF) to be calculated over the mapping region for various time points (top, in seconds). The IRF indicates that the dominant effect of dopamine in ventral striatum is to boost the duration, as opposed to the amplitude of BOLD responses in this region. (F) By comparing striatal dopamine dynamics to hemodynamic fMRI signals observed throughout much of the brain (here, bregma –1.5 to 2.5 mm), hot spots of “dopamine tracking” can be identified in motor (M) and insular (I) cortex, as well as in the dorsal striatum (DS). Data from Brustad et al., 2012 and Li and Jasanoff, 2020.

Fig. 3.

Detection of amino acid neurotransmitters. (A) A T₁ agent designed to sense amino acid transmitters glutamate (Glu), GABA, and glycine (Gly) incorporates a crown ether moeity chosen to bind each ligand’s amine (blue) and a gadolinium chelator moiety capable of coordinating the analyte’s carboxylate group (red). Coordination of Gd³⁺ by the amino acid carboxylate is expected to displace an inner-sphere water molecule (cyan), decreasing effective relaxivity of the sensor and lowering T₁-weighted MRI signal. Adapted from Oukhatar et al., 2015b. (B) Magnetic resonance spectrum from a volume of interest (VOI) in the human brain showing the spectroscopic signature of glutamate (label) on a parts-per-million (ppm) frequency scale. Inset at right shows location of a spectroscopic VOI (dashed square) with respect to brain regions activated by a visual stimulus (color). Inset at left shows mean glutamate levels from 12 subjects, measured by functional spectroscopy before, during, and after periods of visual stimulation (cyan shading). Error bars denote Cramér-Rao bounds on glutamate concentration estimates at each time point. Data reproduced with permission from Bednařik et al., 2015.

Fig. 4.

Design of a nanosensor for MRI of acetylcholine. (A) A polymer scaffold is conjugated to the enzyme butyrylcholinesterase (BuChE) and a pH sensitive MRI contrast agent. Catalytic hydrolysis of acetylcholine by the BuChE releases acetate in the vicinity of the sensor, lowering the local pH and shifting the pH-sensitive agent from its high-pH state (blue) to a low-pH state (magenta). (B) At neutral pH, the pH-sensitive agent contains a deprotonated nitrophenol oxygen (blue) that acts as a Gd³⁺ ligand. At low pH, this group becomes protonated (magenta), freeing up an additional metal coordination site for interaction with water molecules and increasing the probe’s T₁ relaxivity. Adapted from Luo et al., 2018.

Fig. 5.

Sensing extracellular calcium ions in the brain. (A) Structure of a gadolinium-based calcium sensor in its calcium-free low-r₁ state (left) and calcium-bound high-r₁ state (right). (B) A coronal T₁-weighted MRI scan showing a hyperintense region injected with the contrast agent of panel A (label), overlapping somatosensory cortex (red outline). (C) Time courses of molecular fMRI signal in five rats before, during, and after ischemia induced by medial cerebral arterial occlusion (MCAo, start and stop time ranges indicated in cyan), showing transient suppression of the MRI signal during the ischemic period. (D) Equivalent data to panel C, obtained using a control calcium-insensitive contrast agent and demonstrating absence of the transient MCAo-induced signal decrease observed in the presence of the calcium sensor. Panels A-D adapted from Savić et al., 2020. (E) Design of a magnetic calcium-responsive nanoparticle (MaCaReNa) MRI sensor, actuated by extracellular calcium-dependent clustering of the calcium binding domain C2AB (red) with magnetic nanoparticles coated by a mix of phosphatidylcholine (PC) and phosphatidylserine (PS) lipids. (F) Rat brain regions injected with MaCaReNa (right) or calcium-insensitive lipid-coated nanoparticles (LCIO, left), showing MRI signal changes (color overlay) elicited by ipsilateral electrical brain stimulation of neuronal fibers that project to the injected brain area. (G) Time courses of responses as in panel F observed in presence of MaCaReNas under ipsilateral (red) or control contralateral (purple) stimulation, or in the presence of calcium insensitive LCIO particles (gray). Cyan vertical bars denote stimulus timing and shading around each trace denotes standard errors over 3–4 animals each. Panels E-G adapted from Okada et al., 2018.