Ultra-high resolution magnetic resonance microscopy of in situ gadolinium gold nanoparticle-labeled cells in the rat brain

Abstract

Mapping the distribution of cells within a tissue using MR imaging has remained a significant challenge for the field. Cellular MRI can trace cells within tissue, but typically does not achieve the resolution necessary to define a cell's precise anatomical location. To detect cells with ultra-high resolution MRI, a high r ₁ relaxivity intracellular contrast agent is required. Localizing this contrast within its biological context also necessitates an isotropic spatial resolution corresponding to the size of a cell's cytoplasm (∼20 μm) to place it within its biological context. We here demonstrate that gadolinium gold nanoparticles (GdAuNP) induce a high T₁-weighted cellular MRI contrast at ultra-high magnetic fields (9.4 T, and 11.7 T) that affords in situ labelled cell detection at very high resolutions (150, 100, 50, and 20 μm). A 20 μm 3D gradient-echo image (400 minutes scan) combined with MR image denoising robustly visualized the distribution of in situ labeled cells in the rat brain. Signal averaging (NA = 5) also consistently afforded the detection of labeled cells. Positive T₁-weighted contrast was confirmed to be caused by GdAuNP using histology. Immunohistochemistry confirmed the presence of GdAuNP almost entirely inside cells, primarily those of the neuronal lineage. Histology verified that the MR images accurately visualized individual cells' distribution within their anatomical context. Cellular resolution MRI of GdAuNP-labeled cells hence affords new avenues to investigate how individual cells contribute to the development, repair, and regeneration of tissues.

Fig. 1. Optimization of T₁ acquisition parameters. (A) To calculate the impact of GdAuNP on the T₁ signal in brain tissue, regions-of-interest (ROIs) corresponding to brain tissue and the signal produced by GdAuNP-labeled neural stem cells were measured on a T₁ map. (B) Scanning at both room temperature (22 °C) and physiological temperature (38 °C) afforded mapping of T₁ for individual voxels of brain tissue and GdAuNP-labelled cells under two common experimental conditions (i.e. ex vivo and in vivo). T₁ at 38 °C was marginally higher for both brain tissue (6.1%) and GdAuNP (4.6%). (C) Based on T₁ measurements of brain tissue and GdAuNP (22 °C) at both 9.4 T and 11.7 T, optimal parameters for T₁ acquisition were calculated using the Ernst equation for 4 different repetition times (TR, 50, 100, 500, 1000 ms). Optimal parameters were derived from maximum T₁w contrast rather than peak signal intensity for brain tissue or GdAuNP. With shorter TR, the optimal flip angle (FA) was lower, whereas with a TR of 1000 ms optimal flip angle was 90°. A disadvantage of longer TR is longer acquisition times. (D) To map the experimental space of TR and FA, contour maps were computed to illustrate their interaction and provide a means to choose optimal TR and FA to detect maximum contrast between GdAuNP and brain tissue. A TR of 1000 ms with a FA > 75° achieves maximum contrast (>22%), whereas a shorter TR of 500 ms achieves a lower contrast (17–22%), while affording twice the number of signal averages. (E) To illustrate how a GdAuNP decrease in T₁ affects contrast, a contour map of decreasing T₁ demonstrates that with a FA of 70°, T₁ contrast ranging from 10 to >46% can be achieved if sufficient contrast agent concentration is present.

Fig. 2. Impact of field strength and sample temperature on T₁ acquisition. (A) Experimental verification of optimal acquisition parameters visually confirmed a robust detection of GdAuNP with a TR = 500 ms and FA > 70°, as well as TR = 1000 ms and FA = 90°. (B) Quantification of signal intensities for brain tissue and GdAuNP, as well as the contrast between both further illustrate how optimization improves T₁w contrast. (C) A direct comparison between optimized parameters of a TR of 500 and 1000 ms indicated a consistently greater (6–10%) contrast with the TR = 500 ms. A maximum of 29.5% contrast was achieved with TR = 500 ms and a flip angle of 90°. (D) To systematically investigate the impact of sample temperature and the Curie Point of gadolinium on T₁w contrast, sample temperature was controlled during scanning in a 11.7 T MRI scanner. T₁ of both brain tissue and GdAuNP signals increased by 16% in a linear fashion between 14 and 38 °C. The T₁ difference between both increased with rising temperatures >22 °C. Maximum T₁ difference was achieved with a physiological temperature of 37 °C, indicating that even scanning ex vivo samples at physiological temperature being preferable to room temperature (22 °C). (E) T₁ measurements of GdAuNP were increased by 14% (22 °C) and 18% (38 °C) at 11.7 T compared to 9.4 T. (F) However, T₁ contrast (i.e. the difference between brain tissue and GdAuNP) was only improved by 4% between 9.4 T and 11.7 T.

Fig. 3. Cellular resolution MRI. (A) Injections of GdAuNP into the lateral ventricle and the striatum resulted in an increase in T₁-weighted (T₁w) signal on MR images. At a 150 μm isotropic resolution, the signal was readily identified and the tissue signal (i.e. SNR) was high. However, the T₁w hyperintensity caused by GdAuNP was unfocussed due to partial volume effects. Increasing spatial resolution improved the localization of this hyperintensity of the GdAuNP, but SNR decreased. All scans used only 1 signal averaging. Although at 50 μm the GdAuNP hyperintensity was still visually identifiable, at the 20 μm cellular resolution tissue structure and the GdAuNP were no longer visually evident. (B) To mitigate the increase in noise at higher spatial resolution, denoising of MR images was implemented to recover the signal within these images. Denoising dramatically reduced the noise in tissue at all resolutions and provided a homogenous tissue signal. At higher resolutions (≤100 μm), especially the hyperintensity of the GdAuNP improved with a robust visualization of the lateral ventricle and its distribution in striatal tissue. The most dramatic result was evident for the 20 μm cellular resolution, for which denoising afforded a visualization of tissue structure (e.g. distinction between cortex, corpus callosum, striatum, ventricles), which was not feasible on the original image. Moreover, the hyperintensities induced by GdAuNP were readily distinguished from the surrounding tissues and afforded a very distinct localization. (C) Focusing on the injected hemisphere further illustrates the dramatic improvement of GdAuNP detection at cellular resolution (20 μm) using denoising to localize the signal in the lateral ventricle and the central region of the striatum.

Fig. 4. Improving signal averaging and denoising. (A) To further boost the signal at cellular resolution, signal averaging was investigated and revealed a consistent improvement in tissue signal and GdAuNP contrast. However, signal averaging comes at a substantial time penalty. Denoising significantly improved image quality even with a single average. (B) A combination of signal averaging with denoising provided the most dramatic improvements in GdAuNP detection to detail its tissue distribution. (C) Improvements in GdAuNP detection are very significantly impacted by a reduction in the noise component through signal averaging. However, a diminishing return is observed with more than 5 signal averages. Denoising overall reduces the noise component with improvements most noticeable in images with 5 signal averages. A reduction in signal variability is evident for both signal averaging and denoising, with more than 5 signal averages having diminishing returns. (D) An increase in SNR and CNR is achieved with both signal averaging and denoising. However, only minimal gains are achieved purely by using signal averaging. Denoising more significantly impacts both SNR and CNR, with an interaction of increasing signal averages evident. A strong denoising provided the most significant boost to SNR and CNR, but this came at a cost of signal sharpness (SDME). Beyond 5 signal averages, CNR revealed diminishing returns.

Fig. 5. Reproducibility of isotropic cellular resolution MRI. (A) Injection of GdAuNP into the striatum and lateral ventricle of 3 rats afford the visualization of its distribution using MRI at a 20 μm isotropic resolution and image denoising. (B) The isotropic resolution affords a visualization of the GdAuNP distribution at cellular resolution in the coronal, sagittal and axial plane. An isotropic resolution is essential to accurately localize individual voxels containing GdAuNP-labeled cells in their anatomical context. (C) Otsu thresholding of these images affords the selective visualization of the GdAuNP induced tissue contrast and facilitates their localization against the background T₁ signal of the brain. A 3D reconstruction of the GdAuNP reveals its distribution within the brain and could afford future overlays with other scans, such as T₂ or diffusion MRI for further interrogation of specific biological questions. (D) Individual denoised and thresholded T₁w slice images at 20 μm isotropic resolution reveal the patch-matrix of striatal tissue. Hyperintense regions correspond to the distribution of GdAuNP within the patch-matrix, whereas the region lacking T₁ signal corresponds to the patches. Minute anatomical changes become hence discernible on the MR images. LV – lateral ventricle.

Fig. 6. Verification of cellular resolution MRI by immunohistochemistry. (A) A macroscopic comparison of a thresholded T₁-weighted MR image with the corresponding histological slice revealing the fluorescence moiety of the GdAuNP. (B) An overlay of both MRI and histology reveals the overlap between both imaging modalities, again confirming that T₁ signal changes in individual voxels are caused by GdAuNP. (C) This is further highlighted in a zoomed-in region, whereas the delineation of the MRI is clearly defined by the presence of GdAuNP, hence verifying that the T₁ increase on MR images is due to GdAuNP.

Fig. 7. Intracellular localization of GdAuNP. (A) Macroscopic immunohistochemistry of the neuronal distribution and GdAuNP in the rat brain in comparison to the matching T₁-weighted MR slice. An overlay of both images verifies that the MRI signal increase is due to the presence of GdAuNP. (B) GdAuNP in the striatum and the SVZ are almost completely contained within individual cells. This is further highlighted by the distribution of the GdAuNP reflecting a cellular shape rather than a homogenous signal throughout the tissue, as would be expected from the contrast material being merely distributed through the extracellular space. (C) Within the SVZ, GdAuNP were only observed within neural stem/progenitor cells, but not the ependymal cell layer, highlighting a preferential uptake into cells of the neuronal lineage. Within adjacent striatal tissue, it is also evident that GdAuNP are contained within mature neurons, which endocytosed this agent, but there is almost no agent visible within the extracellular space. LV – lateral ventricle; SVZ – subventricular zone; CP – choroid plexus.

Fig. 8. Intra-neuronal GdAuNP. (A) Neurons within the striatum and neural stem/progenitor cells within the SVZ endocytosed a high level of GdAuNP, as evidenced by the intense red intracellular fluorescence. A very negligible amount of GdAuNP remains within the extracellular space. (B) Striatal neurons readily took up the GdAuNP C. 3D reconstruction of a single neuron reveals the peri-nuclear localization of GdAuNP. LV – lateral ventricle; CC – corpus callosum; CP – choroid plexus; SVZ – subventricular zone.

Fig. 9. Uptake of GdAuNP in non-neuronal cell types. (A) GdAuNP associated with astrocytes were a rarity. In cases of uptake, akin to neuronal cells, the GdAuNP were found to be localized around the nucleus rather than throughout the cytoplasm or cellular extensions. (B) There was some evidence of GdAuNP along blood vessels close to the injection sites. This could be due to GdAuNP in CSF, following a glymphatic drainage route. However, in a few cases, GdAuNP could also be seen within endothelial cells.

Fig. 10. Uptake of GdAuNP in microglia and macrophages. GdAuNP were also observed in phagocytic cells within the CNS (green arrows), notably resident microglia (CX3CR1+ cells) and peripheral macrophages (CD68+ cells) that invaded in response to tissue damage caused by the injection tract. However, most macrophages and microglial do not contain GdAuNP, even at the injection site (red arrows).

Scheme 1. Synthesis of small-molecule Gd(iii) agents (Gd-609 and Gd-828).

Scheme 2. Synthesis of single stranded DNA (ssDNA) agents.