Mapping the microscale origins of magnetic resonance image contrast with subcellular diamond magnetometry

Abstract

Magnetic resonance imaging (MRI) is a widely used biomedical imaging modality that derives much of its contrast from microscale magnetic field patterns in tissues. However, the connection between these patterns and the appearance of macroscale MR images has not been the subject of direct experimental study due to a lack of methods to map microscopic fields in biological samples. Here, we optically probe magnetic fields in mammalian cells and tissues with submicron resolution and nanotesla sensitivity using nitrogen-vacancy diamond magnetometry, and combine these measurements with simulations of nuclear spin precession to predict the corresponding MRI contrast. We demonstrate the utility of this technology in an in vitro model of macrophage iron uptake and histological samples from a mouse model of hepatic iron overload. In addition, we follow magnetic particle endocytosis in live cells. This approach bridges a fundamental gap between an MRI voxel and its microscopic constituents.

Fig. 1

Subcellular mapping of magnetic fields in cells labeled for MRI. a Schematic of subvoxel magnetic field mapping using a NV magneto-microscope. b Illustration of a cell labeled with IONs and its expected magnetic field pattern. c Bright-field image of RAW 264.7 macrophage labeled with 200- nm IONs. White arrows point to internalized IONs. A bright-field imaging artifact also appears as black in the upper right corner of the cell. d Cartoon representation of each NV orientation and the corresponding representative spectra from fixed-cell experiments. The blue ball represents nitrogen and the red ball represents the adjacent lattice vacancy. Highlighted peaks in each relative fluorescence (RF) spectrum show the transition corresponding to each of the four orientations. e Magnetic field images of the field projections along each of the four NV axes of macrophages 2 h after initial exposure to 279 ng ml⁻¹ 200- nm IONs. f Images in e converted via Gram–Schmidt orthogonalization and tensor rotation to field maps along three Cartesian coordinates with the z axis defined perpendicular to the diamond surface and the x axis defined as the projection of the applied bias field onto the diamond surface plane. The y axis is defined to complete the orthogonal basis set. g Representative example of the procedure for dipole localization in cellular specimens. This procedure comprises three steps: first the local minima in the field map are identified and ranked; next, in decreasing order of magnitude, the neighborhood of each local minimum is fit to a point dipole equation and the resulting field is subtracted from the field map to reduce the fit-deleterious effect of overlapping dipole fields; and finally, the results of these fits are used as guess parameters for a global fit over the full field of view. The fit shown has a degree-of-freedom-adjusted R² of 0.97. Scale bars are 5 µm

Fig. 2

Predicted and experimental MRI behavior in cells. a Schematic of Monte Carlo modeling of spin relaxation using NV-mapped magnetic fields. A library of 11 cells mapped with vector magnetometry (three representative cells shown) in a 1:1 mix with unlabeled cells, was used to randomly fill a 108-cell FCC lattice with periodic boundary conditions and run a Monte Carlo simulation of spin-echo MRI to predict T₂ relaxation behavior. b Representative simulated MRI signal. c T₂-weighted MRI image of cell pellets containing a 1:1 mixture of supplemented and unsupplemented cells (+ IONs and –IONs, respectively) or 100% unlabeled cells (bottom). d Simulated and experimentally measured T₂ relaxation times for the 1:1 mixture. e Illustration of the same quantity of magnetic particles endocytosed or distributed in the extracellular space. f Simulated and experimentally measured relaxivity for endocytosed and extracellular distributions of IONs. Measurements and simulations have N = 5 replicates. All error bars represent ± SEM

Fig. 3

Magnetometry of histological samples. a Diagram of mouse model of iron overload, prepared by injecting 10 mg kg⁻¹ of 900 nm iron oxide nanoparticles into the tail vein. b 7T T₂-weighted MR image of fixed, excised mouse livers from mice injected with IONs or saline. c NV magnetic field maps of 10 µm liver sections obtained from the mice in b. d Fluorescence images of the tissue samples in c. Fluorescence images were taken with autogain to reduce the necessary exposure time, resulting in the visibility of the autofluorescence of the tissue in the saline control. Magnetometry scans were taken with a fixed gain. This experiment was repeated a total of three times, with data from two additional experiments shown in Supplementary Fig. 6. Scale bars in b and c–d are 5 mm and 10 µm, respectively

Fig. 4

Dynamic magnetic microscopy in live mammalian cells. a Cartoon showing the typical progression of endocytotic uptake of IONs. b Bright field and series of time-lapse magnetic field images of RAW macrophages over 10 h. Three additional replicates are shown in Supplementary Fig. 7. c Bright field and series of time-lapse magnetic field images of a RAW macrophage with 10 min between magnetic field images. Two additional replicates of this experiment are shown in Supplementary Fig. 7. Scale bars are 5 µm