Cell Specificity of Manganese-enhanced MRI Signal in the Cerebellum

Abstract

Magnetic Resonance Imaging (MRI) resolution continues to improve, making it important to understand the cellular basis for different MRI contrast mechanisms. Manganese-enhanced MRI (MEMRI) produces layer-specific contrast throughout the brain enabling in vivo visualization of cellular cytoarchitecture, particularly in the cerebellum. Due to the unique geometry of the cerebellum, especially near the midline (vermis), 2D MEMRI images can be acquired from a relatively thick slice by averaging through areas of uniform morphology and cytoarchitecture to produce very high-resolution visualization of sagittal planes. In such images, MEMRI hyperintensity is uniform in thickness throughout the anterior-posterior axis of sagittal sections and is centrally located in the cerebellar cortex. These signal features suggested that the Purkinje cell layer, which houses the cell bodies of the Purkinje cells and the Bergmann glia, is the source of hyperintensity. Despite this circumstantial evidence, the cellular source of MRI contrast has been difficult to define. In this study, we quantified the effects of selective ablation of Purkinje cells or Bergmann glia on cerebellar MEMRI signal to determine whether signal could be assigned to one cell type. We found that the Purkinje cells, not the Bergmann glia, are the primary of source of the enhancement in the Purkinje cell layer. This cell-ablation strategy should be useful for determining the cell specificity of other MRI contrast mechanisms.

Introduction

A frontier for Magnetic Resonance Imaging (MRI) of the brain is to achieve cellular specificity and eventually cellular resolution. Assignment of signal or signal changes to cell types can help answer open questions in basic neuroscience such as how many cells of a cell type normally exist in an anatomical area and on what time course disease processes alter these cell types. Cell specific MRI techniques would be useful to guide the development of novel treatment strategies being attempted with cell labeling and MRI cell tracking techniques. General approaches in the field to achieve this goal involve imaging at high spatial resolution and introduction of high relaxivity contrast agents that have large effects on MRI signal. Examples include imaging various cell types using small particles of Iron oxide (Shapiro, Skrtic, and Koretsky 2005; Shapiro, Sharer, et al. 2006; Shapiro et al. 2004; Heyn et al. 2006; Liu et al. 2022; Kiru et al. 2022) or perfluorocarbons (Ahrens et al. 2005). In those studies, cell labeling was performed in vitro, cells injected into a host animal, and MRI performed in vivo. Very few studies use locally injected contrast agents to track populations of cells in vivo (Nieman et al. 2010; Pothayee et al. 2017; Shapiro, Gonzalez-Perez, et al. 2006). There is also evidence that when specific cells accumulate iron in the human brain, these can be measured with susceptibility-weighted imaging (Kirsch et al. 2009). This includes microglia in the motor cortex in patients with amyotrophic lateral sclerosis (Kwan et al. 2012) and microglia at the periphery of MS lesions (Bagnato et al. 2011). There are no generally applicable procedures to determine the cell specificity of MRI contrast and MRI contrast agents to assign signal to specific cell types.

Manganese (Mn²⁺)-enhanced MRI (MEMRI) has interesting tissue contrast but assignment of the cell types responsible for this contrast has not yet been determined. When administered systemically, Mn²⁺ has been shown to cross the blood-brain barrier and accumulate in non-uniformly in different areas of the brain (Lin and Koretsky 1997; Aoki et al. 2004; Watanabe, Frahm, and Michaelis 2013). The strong paramagnetic relaxation of water by Mn²⁺ has enabled MRI to detect laminar structures in the brain and to resolve changes in such tissue cytoarchitecture during dynamic processes such as development or disease progression (Aoki et al. 2004; Watanabe et al. 2002). Indeed, numerous previous studies of mouse cerebellar development and disease progression have been made possible by the high contrast images produced by MEMRI (Wadghiri et al. 2004; Szulc et al. 2013; 2015; Tan et al. 2018; Rallapalli, Tan, et al. 2020; Qiu et al. 2018; Rallapalli, Darwin, et al. 2020). In addition, proper application of Mn²⁺ ions have made MEMRI a useful tool for measuring neural activity (Yu et al. 2005; Lin and Koretsky 1997) and to enable MRI to detect anterograde tract tracing to dissect connectivity (Pautler, Silva, and Koretsky 1998). Despite the clear utility of MEMRI for a large variety of studies, the cell specificity of the contrast has not been established. For example, in studies of the cerebellum (CB), even though the CB is highly enhanced by systemic injection of Mn²⁺, the cellular source of Mn²⁺ uptake and hyperintensity in MEMRI images remains unclear. This is equally true of the hippocampus and olfactory bulb where MEMRI also detects cellular cytoarchitectural information.

The CB is a morphologically complex hindbrain structure that is critical for a range of behaviors including motor, social and cognitive brain functions (Figure 1). The CB, nevertheless, has a simple gross morphology and layered (laminar) cytoarchitecture. It consists of two hemispheres which are connected by the midline vermis, with distinct cortical layers and white matter boundaries. From outer to innermost layer of the three layers of the adult cerebellar cortex are the molecular layer (ML), the Purkinje cell layer (PCL), and the inner granule cell layer (IGL), which transitions into the white matter (WM). The ML is mostly composed of the parallel fibers (axons) of the granule cells, dendrites of the Purkinje cells (PCs), and processes of the Bergmann glia (BG). The PCL is home to a monolayer of PC bodies interspersed with BG soma, and thus the width of the layer is roughly defined by PC soma diameter. The PC soma are remarkably large (Herndon 1963) (~25 to 40 μm in diameter) compared to the smaller BG soma (~15 μm in diameter). BG outnumber PCs by approximately eight-fold (Korbo et al. 1993). The IGL is composed mainly of granule neurons and is the brain structure with the highest cell density. The WM is composed of axons from neurons outside the cerebellum that project to it (afferents) as well as from PCs which extend deep into the CB, synapsing on output neurons in the three bi-laterally symmetrical cerebellar nuclei located lateral to the vermis.

Figure 1:

Location, morphology, and lobules of the cerebellum as observed in our MEMRI experiments. (A) The cerebellum is the most dorsal-caudal structure in the brain (green) (B) it is composed of two hemispheres (H, red) lateral to the vermis (V, blue) in the midline. (C) a band of the vermis equidistant from midline (dotted lines) is relatively uniform in cytoarchitecture in the sagittal orientation. (D) sagittal MEMRI of the cerebellum, averaging through this band, reveals the 10 lobules and cortical layers within each lobule. (E) Schematic overview of the CB cytoarchitecture. A midsagittal section (i) through the CB vermis shows the lobules. Zooming in to lobule 9 (square) shows the laminar organization of the CB into the molecular layer (ML), Purkinje cell layer (PCL), inner granule layer (IGL), and white matter (WM). The PCL is composed of two cell types, the BG and the PCs.

Despite well-defined cytoarchitectural boundaries, there has been disagreement in MEMRI signal assignment to CB cortical layers. Over the years, precursor cells in the external granule cell layer (a transient germinal layer present at early postnatal stages) (Bartelle et al. 2013), IGL (Wadghiri et al. 2004; Aoki et al. 2004), and PCL (Watanabe, Frahm, and Michaelis 2013; Watanabe et al. 2002) have all been posited as putative sources of MEMRI enhancement. The disagreement in layer assignment of hyperintensity in CB MEMRI has arisen in part from the continuous improvement in imaging resolution and a better understanding of MRI pulse sequence parameter optimization for MEMRI.

Recent efforts have taken advantage of the uniquely patterned zones of homologous cytoarchitecture to produce high-resolution 2D MEMRI images of the cerebellar vermis, where morphology is conserved in the sagittal plane, using sagittal and axial sections (Watanabe, Frahm, and Michaelis 2013). It was made apparent from these images that the hyperintensity, presumably due to higher Mn²⁺ accumulation shortening T1, is centrally located in the CB cortex and uniform in thickness along the anterior-posterior axis. From this finding, the PCL was posited as the source. However, no validation experiments have been performed nor has the cellular source of this signal been determined.

Given the abundant body of evidence of Mn²⁺ uptake into cells via voltage gated Ca²⁺ ion channels and other transporters (Bartelle et al. 2013; Crossgrove and Yokel 2005; Lin and Koretsky 1997; Petrus et al. 2021), manipulating the relative populations of cerebellar cell types would have the potential to change signal distribution in CB MEMRI. To this end, genetic engineering approaches can be used to induce cell type-specific ablations. One method is to selectively express the diphtheria toxin receptor (DTR) in a specific cerebellar cell type and then inject diphtheria toxin (DT) at a specific stage to ablate the cells, since mice do not express a DTR (Buch et al. 2005; Bayin et al. 2018). For context, wild-type mice do not naturally express DTR but can be genetically engineered to do so in a cell-specific manner, using cell-type specific recombinases driven by marker gene promoters that in turn allows expression of DTR in the cells that express the marker gene of interest. In the developing cerebellum, the Pcp2 gene is specifically expressed in PCs and the Hopx gene in BG and some astrocytes (Bayin et al. 2021). Thus, PCs and BG can be selectively ablated via injection of DT into Pcp2-DTR (Bayin et al. 2018) and Hopx-DTR mice, respectively. Therefore, PCP2-DTR mice are susceptible to PC death and HOPX-DTR mice are susceptible to BG death.

In this study, we assigned hyperintensity in MEMRI to be maximal at the PCL in the CB by careful alignment with histology. Then, we used cell type-specific ablations to provide direct evidence that PCs are the primary cellular source of hyperintensity in the PCL for CB MEMRI. The strategy used should be useful for assign MEMRI contrast to cell types in other regions as well as for determining cell specificity to other types of MRI contrast.

Materials and Methods

Animals

All mouse experiments were performed according to protocols approved by the Animal Care and Use Committees at New York University School of Medicine, Memorial Sloan Kettering Cancer Center, and the National Institutes of Health. Both sexes were used in this study. For MEMRI experiments at increasing resolution, C57BL/6J animals (Jackson Laboratory Strain #000664) were used. To quantify the effects of PC ablation on MEMRI signal, two mouse models were engineered to express DTR in PCs, using two Pcp2^Cre mouse lines which express Cre in a relatively fewer (Saito et al. 2005) and greater (Barski, Dethleffsen, and Meyer 2000) number of PCs. These Pcp2^Cre mice were bred to R26^{loxSTOPlox-DTR/+} mice (Buch et al. 2005); Jackson Laboratory Strain #007900) to generate Pcp2^Cre; R26^{loxSTOPlox-DTR/+} mice that expressed DTR in fewer (“PC-X”) and greater (“PC-eX”) numbers of PCs after administration of diphtheria toxin. To quantify the effects of BG ablation on MEMRI signal, a Hopx^CreERT2 mouse line (Takeda et al. 2011; Jackson Laboratory Strain # 017606) was bred to the R26^{loxSTOPlox-DTR/+} mice to generate Hopx^CreERT2; R26^{loxSTOPlox-DTR/+} mice (“BG-X”). Hopx^CreERT2; R26^{loxSTOPlox-DTR/+} mice were given 200 μg/g Tamoxifen at postnatal day (P)0 to induce recombination and subsequent Cre expression. For cell-type specific ablations, diphtheria toxin (30 ng/g of mouse; List Biological Laboratories, Inc.) was injected intraperitoneally at P12 for PC-X and PC-eX animals and P16 for BG-X animals to ensure CB development was not impacted by the ablations. Across all models, littermates negative for the R26^{loxSTOPlox-DTR/+} transgene were considered controls.

Histological Analyses

Tissue preparation and histology

For histological analyses, anesthetized animals were transcardially perfused with 20 mL of ice-cold phosphate buffered saline (PBS) followed by 20 mL 4% paraformaldehyde (PFA). Brains were dissected and fixed 1-2 days more in 4% PFA at 4° C. Brains were incubated in 30% sucrose in PBS until they sank to the bottom of the tubes (1-2 days) and cryopreserved in optimal cutting temperature solution (Tissue-Tek) for cryosectioning. 14um thick frozen sagittal cryosections were obtained using a Leica cryostat (CM3050S) and stored at −20° C for future analysis.

Immunofluorescent staining and imaging

After warming to room temperature (RT), slides were washed once with PBS and incubated in blocking buffer (5% bovine serum albumin in PBS-T (PBS with 0.1% Triton-X)) at RT. Mouse anti-Calbindin1/CALB1 (1:1000, Swant, Cat no: 300) and goat anti-SOX2 (1:200, R&D, Cat no: AF2018 ) were used for staining and primary antibody incubation were performed at 4° C overnight. Slides were then washed three times with PBS-T (5 min each) followed by secondary antibody incubation with specific AlexaFluor conjugated secondary antibodies (1:500 in blocking buffer, Invitrogen) for 1 hour at RT. Counterstaining was performed using 1 μg/mL Hoechst 33342 (Invitrogen) and slides were mounted with Fluoro-Gel mounting media (Electron Microscopy Sciences).

Images were collected using a DM6000 Leica microscope and processed using ImageJ Software (NIH). For each quantification, three midline parasagittal sections/brain were analyzed, and data was averaged. Cells were counted using the Cell Counter plugin for ImageJ (NIH). Cell density was calculated by dividing the number of cells by length of PCL.

Multiplex Fluorescence Immunohistochemistry

Multiplex fluorescence immunohistochemistry (MIHC) was performed on 10 μm-thick 4% formalin-fixed mouse brain cryosections, as previously described (Maric et al. 2021). To prepare for the slides for MIHC staining, the sections were first treated with a 2-minute heat mediated antigen retrieval step in 10 mM Sodium Citrate buffer pH 6 using an 800W microwave set at 100% power. The sections were then sequentially incubated for 15 min at room temperature (RT) with Fc receptor blocking solution (Innovex Biosciences) to saturate endogenous Fc receptors, followed by Background Buster solution to minimize non-specific antibody binding. The sections were subsequently immunoreacted for 1 hour at RT using 1 μg/ml cocktail mixture of immunocompatible primary antibodies targeting microglia (IBA1), neurons (Calbindin, NeuN, MAP2, Neurofilament-Light Chain), astrocytes (GFAP, S100β), and oligodendrocytes (Myelin Basic Protein), as listed in Table 1. This step was followed by washing off excess primary antibodies in PBS supplemented with 1 mg/ml bovine serum albumin (BSA), followed by incubation of the sections using a 1 μg/ml cocktail mixture of the appropriately cross-adsorbed secondary antibodies (purchased from either Thermo Fisher, Jackson ImmunoResearch or Li-Cor Biosciences) conjugated to one of the following spectrally compatible fluorophores: Alexa Fluor 430, Alexa Fluor 488, Alexa Fluor 546, Alexa Fluor 594, Alexa Fluor 647, PerCP, IRDye 680 LT or IRDye 800CW. After washing off excess secondary antibodies, sections were counterstained using and 1 μg/ml DAPI (Thermo Fisher Scientific), a DNA labeling dye, for visualization of cell nuclei. Slides were then coverslipped using Immu-Mount medium (Thermo Fisher Scientific) and imaged using a multi-channel wide field epifluorescence microscope (see below).

Table 1:

Primary antibodies used for MIHC of the mouse brain

Antibody	Vendor	Product #	Host Ig class	Clone	Conjugate
IBA1	Wako Chemicals	019-19741	Rabbit IgG	Polyclonal	Unconjugated
Calbindin	Cedarlane Labs	214004(SY)	Guinea Pig IgG	Polyclonal	Unconjugated
NeuN	Millipore Sigma	ABN91	Chicken IgY	Polyclonal	Unconjugated
MAP2	R&D Systems	MAB8304	Mouse IgG3	885232	Unconjugated
NF-L	BioLegend	846002	Mouse IgG1	NFL2	Unconjugated
GFAP	Thermo Fisher Sci	13-0300	Rat IgG2a	2.2B10	Unconjugated
S100b	Millipore Sigma	MAB079-1	Mouse IgG2a	15E2E2	Unconjugated
MBP	BioLegend	808402	Mouse IgG2b	SMI99	Unconjugated

Multispectral imaging of multiplex IHC stained sections

Images were acquired from whole specimen sections using the Axio Imager.Z2 slide scanning fluorescence microscope (Zeiss) equipped with a 20X/0.8 Plan-Apochromat (Phase-2) nonimmersion objective (Zeiss), a high resolution ORCA-Flash4.0 sCMOS digital camera (Hamamatsu), a 200W X-Cite 200DC broad band lamp source (Excelitas Technologies), and 9 customized filter sets (Semrock) optimized to detect the following fluorophores: DAPI, Alexa Fluor 430, Alexa Fluor 488, Alexa Fluor 546, Alexa Fluor 594, PerCP, Alexa Fluor 647, IRDye 680 LT and IRDye 800CW. Image tiles (600 x 600 μm viewing area) were individually captured at 0.325 micron/pixel spatial resolution, and the tiles seamlessly stitched into whole specimen images using the ZEN 2 image acquisition and analysis software program (Zeiss), with an appropriate color table having been applied to each image channel to either match its emission spectrum or to set a distinguishing color balance. Pseudocolored stitched images were then exported to Adobe Photoshop and overlaid as individual layers to create multicolored merged composites or exported as 8-bit grayscale BigTIFF files and imported in Fiji for image analysis.

MEMRI

Manganese chloride tetrahydrate in isotonic saline was injected intraperitoneally 24 hours before each imaging session (MnCl₂, 30mM, 60 mg/kg body weight). General anesthesia was maintained during imaging by 1.5% isoflurane in air.

For T1 (or R1) mapping experiments, MRI was performed at New York University on a Bruker 7 T micro-MRI system using a four-channel phased array cryogenically-cooled receive-only coil with a volume transmit coil (CryoProbe system, Bruker). Scan protocol included: low-resolution pilot; 50 μm in-plane (500 μm slice thickness) resolution RARE-VTR sequence. For multiple resolution MEMRI and registration to histology, MRI was performed at the NIH/NINDS on a 11.7 T animal MRI system (Magnex Scientific magnet with Bruker electronics and RRI 12 cm gradient/shim) using a CryoProbe system (Bruker). Scan protocol included: 1 min low-resolution pilot; FLASH images at 100, 80, and 50 μm isotropic resolution; and 30 μm in-plane (500 μm slice thickness) FLASH. The 50 μm in-plane images were generated by averaging together 10 midsagittal slices from the 50 μm isotropic resolution images.

T1 maps were generated using the Bruker Image Sequence Analysis tool and value inverted in Fiji to produce R1 maps. To quantify CB layer R1s, heavily T1 weighted images from the RARE-VTR protocol were manually segmented into ML, PCL, IGL, and WM ROIs. These segmentations were propagated to R1 maps for statistical analyses.

Image analysis

To show improvements in CB cortical layer definition at increasing MEMRI resolution, line intensity plots were generated in Fiji and rendered in GraphPad Prism. Line ROIs were drawn from CSF space to WM through lobule 3 at each resolution (n = 3 animals, n = 3 lines per animal).

Fifty micron isotropic resolution images were aligned to multiplexed immunohistology (MIHC) images using the landmark-based registration tool in Fiji. Point landmarks were drawn at the apices of each lobule and at the base of each sulcus between lobules - a total of 18 control points – on both the brightfield channel of MIHC and matching sagittal MEMRI slice. After registration, the MEMRI image in histological space was combined with the myelin basic protein (MBP), calbindin, DAPI channel images to create a composite image. Line ROIs were drawn on the composite image from CSF space to WM through various lobules (n = 10 lines). Intensities from each channel were normalized to the maximum intensity, then normalized channel intensities from each line were averaged together and rendered in GraphPad Prism.

Statistics

Statistical analyses were performed in R (v 4.0.5) and detailed in Supplementary information: Statistical Analyses (R Core Team 2021). Statistical significance was set at p<0.05.

Results

High-resolution MEMRI of the mouse cerebellum defines the position and width of the cortical layers

To localize enhancement on MEMRI to the CB layers, we imaged at increasing resolutions (Figure 2). At 100 μm isotropic resolution, the position and shape of the lowest enhanced region were correlated with that of the internal WM. Higher signal intensity extended radially outward from the outer edge of the WM to the cerebrospinal fluid and skull space. This hyperintensity correlated with the IGL, PCL, and ML of the cerebellar cortex. The distribution of intensities in CB MEMRI remained consistent across imaging resolutions. The highest signal band at 100 μm isotropic resolution became more stratified at higher resolutions. At 80 μm isotropic resolution, three distinct bands of signal were clearly apparent in the CB cortex: an inner medium signal intensity band with a variable width immediately adjacent to the white matter, a high signal intensity band with uniform width adjacent to the inner medium signal band, and an outer medium signal intensity band with variable width extending from the high signal intensity band to the CSF space. We hypothesized that the high intensity signal band between the medium intensity bands correlated to the PCL due to its position in the cortex and uniform width. By extension, we hypothesized that the inner and outer signal bands correlated with the IGL and ML, respectively. At 50 μm in-plane, 50 μm isotropic, and 30 μm in-plane resolutions, the hyperintense band in the cerebellar cortex reduced in width uniformly over the entire CB, while the inner and outer medium intensity bands relatively increased in width indicating that the low resolution had significant partial volume of laminar contrast. The total width of the cerebellar cortex (~ 400 μm) in lobules 1/2 did not change between 80 μm and 50 μm isotropic resolution images. However, the high intensity band continued to decrease in thickness even with the 30 μm in-plane resolution. This suggested that even at this resolution, the hyperintense band has not been properly sampled and is still partial volumed, likely due to the thicker slice used and the fact that PCs are about 30 μm in diameter.

Figure 2:

MEMRI of the CB at increasing resolution (top row). As resolution increases, the ability to resolve the cortical layers (ML, PCL, and IGL) and WM improves. This effect is quantified using line intensity analysis (dotted lines) through lobules 1/2 of the cortex (bottom row).

These results were quantified in Table 2 by measurement of peak intensity position and full width at half-maximum (FWHM) intensity in the cortex of lobule 2. At lower resolutions, peak MEMRI intensity in the cerebellar cortex varied in position. At higher resolutions, the peak intensity position was relatively fixed. This indicates that partial volume effects on signal localization are mostly mitigated by 50 μm in-plane resolution imaging. As expected, the FWHM intensity of the hyperintense cortical layer generally decreased as resolution increased to 30 μm in-plane.

Table 2:

Location and spread of peak MEMRI intensity in the CB cortex

MEMRI resolution	Peak position from cortical surface (μm)	FWHM (μm)
100 μm isotropic	199	210
80 μm isotropic	213	230
50 μm isotropic	182	166
50 μm in-plane	182	210
30 μm in-plane	180	80

Alignment of MEMRI to histology associates hyperintensity with the Purkinje cell layer

Next, we registered 50 μm isotropic resolution MEMRI images to matched immunohistochemistry images (Supplementary Figure 1, Figure 3). Calbindin staining was used to mark the PC dendrites in the ML, soma in the PCL, and axons in the WM. DAPI staining marked cell nuclei, with signal being dominated by granule cells in the IGL at low power. Layer-specific MEMRI signal was not altered due to voxel interpolation to IHC resolution before landmark-based registration was used. However, some widening of the hyperintense signal band was observed and could be explained by the bicubic resampling blur (Supplementary Figure 1). Overlaying the MEMRI on the IHC to create a composite image at low power revealed hyperintensity spanning the boundary between high DAPI and Calbindin staining. Increasing magnification showed MEMRI colocalization with PC soma in the PCL, proximal dendrites in the ML, and a small number of granule cells closest to the PCL in the IGL. Line intensity plots corroborated this colocalization pattern. MEMRI signal intensity reached a maximum at the peak Calbindin fluorescence intensity, indicating that the highest MEMRI enhancement was localized to the PCL.

Figure 3:

Registration of MEMRI to matched multiplexed immunohistology (MIHC) implicates the PCL as the primary source of hyperintensity in MEMRI of the CB. (A) MEMRI are registered into the MIHC space. The MBP (yellow) and Calbindin (red) channels from MIHC are shown to highlight the WM and PCL, respectively, whereas low and high DAPI (blue) indicates the ML and IGL respectively. (B) Composite of the channels + MEMRI aligns the MEMRI hyperintensity mainly to the PCL. (C) Zoomed in view of lobule 2 reinforces this assignment. (D) Intensity analysis of lines drawn on the composite images validates the coincidence of Calbindin (PC soma) and MEMRI signal peaks.

Mouse models of ablation of Purkinje cells and Bergmann glia produce significant shifts in Purkinje cell or Bergmann glia density

Three genetically engineered mouse models were generated to study the effects of BG (BG-X) or PC (PC-X, PC-eX) ablation on MEMRI signal. In order to confirm ablation and assess the reduction in cell numbers, we performed immunostaining with a PC marker, CALB1, and BG marker, SOX2, on midsagittal CB sections. Cell density was quantified ~10-14 days after ablation of PCs or BG at P12 and P16, respectively, by injection of DT (Figure 4). As expected, we observed a significant reduction in the PC and BG densities in all models. In the PC-X model, PC density (CALB1+ soma) in the PCL was significantly reduced compared to their NoDTR littermate controls (percent reduction = 28.7% ±8.71%, *p=0.030), whereas the PC-eX model showed a larger reduction in the PC density upon ablation compared to control littermates (percent reduction = 54.2% ±4.37%, ****p<0.001). Additionally, in the BG-X model, the density of SOX2-positive cells in the PCL was also significantly reduced upon ablation (percent reduction = 53.6% ±12.9%, **p=0.004). Collectively, these results demonstrate that DT-mediated cell ablation can reduce PC and BG densities in the PCL, allowing us to test which cell type is responsible for hypersensitivity in MEMRI.

Figure 4:

Three mouse models of cell ablations in the PCL: BG-X, PC-X, and PC-eX. (A) Timeline of DT injections and histological analyses. (B) Immunohistochemical comparisons of the densities of the two cell types in the PCL in the three models between experimental and NoDTR conditions. Across all models, experimental animals appear to have reduced cell counts in the PCL compared to NoDTR. SOX2 (green) marks the BG soma in the PCL and astrocytes elsewhere. CALB1 (Calbindin1, red) marks PC soma in the PCL and dendrites in the ML. Hoechst (blue) marks all nuclei. (C) Quantitative analysis of cell numbers shows statistically significant reductions in the PCs and BG in the PCL across in the relevant models. Percent reduction in cell density and p-values for each model: BG-X: 53.6%,**p=0.004; PC-X: 28.7%, *p=0.030; PC-eX: 54.2%, ****p<0.001.

MEMRI of Purkinje cell ablation mouse models reveals the Purkinje cells, not the Bergmann glia, as the source of MEMRI hyperintensity in the Purkinje cell layer

R1 mapping was performed and layers of the CB were manually segmented. Because the R1 values obtained include both endogenous tissue and Mn²⁺ effects, we first estimated average R1 in mice that did not receive Mn²⁺. In the cortex, including the ML, PCL, and IGL, the average measured R1 was 0.62 (std. dev. = 0.082; n=10). In the white matter, the measured R1 was 0.66 (std. dev. = 0.079; n=10). These values were subtracted from the raw R1 measurements made in the ML, PCL, IGL, and WM of BG-X, PC-X, and PC-eX animals (ΔR1).

R1 mapping was subsequently performed in transgenic animals with PCL cell ablations that also received Mn²⁺ injection (Figure 5). Layer-wise signal intensity differences were analyzed between NoDTR and DTR animals (Supplementary information: Statistical Analyses).

Figure 5:

MEMRI R1 mapping of BG-X, PC-X, and PC-eX animals indicates PCs, but not BGs, are the primary source of Mn²⁺ uptake and hyperintensity in the PCL. (A) MEMRI images of the CB were manually segmented into lamina: ML (green), PCL (red), IGL (blue), WM (yellow). These segmentations were propagated onto R1 maps to obtain layer-specific R1 measurements. (B) Representative R1 maps from BG-X, PC-X, and PC-eX animals. Few apparent differences are visible in CB lamina between NoDTR and DTR conditions of BG-X animals. However, PC-X and PC-eX animals exhibit noticeable reductions in R1 in the PCL in the DTR condition compared to NoDTR. (C) ΔR1 values of each CB layer. Significant differences in ΔR1 between DTR and NoDTR conditions were apparent in the PCL of the PC-X (18.46%, *p=0.036) and PC-eX (27.22%, **p=0.004), but not the BG-X model. No other significant differences were apparent in any other layer in any of the models.

We performed comparisons between NoDTR and DTR animals of the same background using the ΔR1 in each tissue layer. For the BG-X model, no significant differences in ΔR1 were observed in any cortical layer nor WM between NoDTR and DTR conditions. For the PC-X model, a significant reduction in ΔR1 was observed in the PCL (estimated difference = 0.073 s⁻¹, 18.46%, std. error = 0.026, *p=0.036). No significant differences in ΔR1 were observed in the ML, IGL, nor WM. For the PC-eX model, as in the PC-X model, a significant reduction in ΔR1 was observed in the PCL (estimated difference = 0.096 s⁻¹, 27.22%, std. error = 0.026, **p=0.004). No significant differences in ΔR1 were observed in the IGL, ML, nor WM. Note that a similar analysis based on raw R1 values (without subtraction of estimated endogenous tissue R1) is shown in Supplementary Figure 2.

To further analyze the relationship of R1 to PC ablation suggested by the ΔR1 results, we plotted the ΔR1 measurements made in the PCL of PC-X and PC-eX NoDTR and DTR animals against the estimated PC density reduction estimated for each condition (Figure 6). After pooling the NoDTR cases, linear regression, and extrapolation to 0% PC density, we found that the PC ablation was highly predictive of ΔR1 (R² = 0.989). However, the intercept derived from the fit was non-zero (ΔR1 = 0.125 s⁻¹). This analysis suggests that the intercept ΔR1 represents a residual amount of uptake of Mn²⁺, likely into the remaining BG soma in the PCL. Based on this extrapolation the PCs are responsible for approximately two-thirds of the total Mn²⁺ uptake in the PCL. Indeed, the extrapolated ΔR1 of the PCL to 0% density of PCs is very close to the ML and IGL ΔR1 measurements. This indicates that the entirety of the increased laminar contrast (i.e., between the PCL and other cortical layers) is due to PCs.

Figure 6:

Correlation analysis of MEMRI ΔR1 changes to cell density reduction measured by histology shows linear relationship between PC density and ΔR1 (y = 0.002x + 0.125, R² = 0.99). The strong linear correlation between PC density and ΔR1 suggests that PCs are the source of hyperintensity in CB MEMRI. The offset term in the regression may represent R1 changes due to nonspecific uptake of Mn²⁺ in the brain.

Discussion and Conclusions

In this study, we assigned hyperintensity in CB MEMRI to the PCL. Using cell-type ablations, we were also able to identify the cellular source of most of this hyperintensity as the PCs and not the BG. First, we performed CB MEMRI at increasing resolution. Next, we cross-checked our high-resolution MEMRI results by registering to multiplexed immunohistochemical (MIHC) treated sagittal cerebellar sections at matched mediolateral levels to show that the PCL was the likely source of hyperintensity. Because there are only two major cell types in the PCL, the PCs and BGs, we performed MEMRI relaxometry on BG-X, PC-X, and PC-eX animals that underwent selective cell ablation of either one of the cell populations and demonstrated that the PCs were the primary source of hyperintensity in CB MEMRI and solely responsible for enhancement of the PCL above the ML and IGL.

MEMRI has clear utility in the analysis of anatomy, function, and connectivity in vivo (Silva et al. 2008). There are many advantages to using in vivo MEMRI over in situ or ex vivo methods, including fluorescence activated cell sorting (FACS) and inductively coupled plasma mass spectrometry (ICPMS), for Mn²⁺ localization. Imaging in vivo mitigates mislocalization of Mn²⁺ as it has been shown to disperse immediately postmortem (Sato et al. 2018). It is also challenging to recover viable or intact neurons from dissociated brain tissue, which is particularly true for isolating PCs (Baptista, Blazeski, Hatten, et al. 1994). ICPMS requires relatively large sample dry weights (~10 – 500 mg) to obtain reasonable trace metal detection limits³. So, even if PC isolation were feasible, one would have to keep the Mn²⁺ in place after sacrificing one animal, sort intact PCs and BG using FACS after dissociation, and pool together cells from multiple animals to have enough tissue to perform ICPMS for Mn²⁺ (Radke, Ensley, and Hansen 2020). We have demonstrated that MEMRI of genetically-engineered mouse models that have cell ablation can measure Mn²⁺ concentrations in the μM range (Maddage, Marques, and Gruetter 2014), in vivo in a single animal, at resolutions as high as 30 μm in-plane.

Our ability to resolve neuroanatomical structures on MRI is primarily dependent on sensitivity, resolution, and contrast (Duyn 2012). The presented R1 mapping experiments were performed at 7 T, leveraging the improved SNR made possible using a CryoProbe. At this field strength, we were able to resolve layers of the CB but not individual cell bodies within a cortical layer. It is likely that imaging resolutions on the order of BG diameter will be required to resolve both BG and PC soma. The presented experimental results make clear MEMRI should enable cellular contrast if sufficient resolution can be achieved with next generation 17-20 T small animal MRI systems (Leftin et al. 2015; Lopez-Kolkovsky, Mériaux, and Boumezbeur 2016). MEMRI at 11.7 T has been used to detect individual glomeruli within the glomerular layer in the olfactory bulb (Chuang, Belluscio, and Koretsky 2010). It will be interesting to investigate whether specific cell types can be assigned to signal enhancement in other brain areas such as hippocampus and olfactory bulb using similar strategies as used in our study.

Careful alignment of MEMRI to MIHC treated sections validated the assignment of peak hyperintensity in CB MEMRI to the PCL. This strategy could be readily extended to assign hyperintensity to cell types in other enhancing areas, such as the olfactory bulb and hippocampus. The histological markers selected for layer assignment were guided by a well-established literature (Wassef et al. 1985; Boggs 2006; Wechsler-Reya and Scott 1999). In the future, registration of whole brain MEMRI to MIHC of whole brain sections can yield unbiased assignment of signal to cell type or layer (Maric et al. 2021).The fact that MEMRI extended beyond the cell bodies of the PCL could be due to partial volume effects, or because there is some PC volume in the other CB cortical layers (e.g., the dendrites of the PCs are found in the ML), or because of Mn²⁺ accumulation by other cell type populations not manipulated in our study.

Our results from imaging the PC-X and PC-eX mouse models showed that increasing PC cell depletion also decreased ΔR1, further pointing to the PCs as the major source of hyperintensity. Our hypothesis was that selective Mn²⁺ uptake into cells produced MEMRI contrast, therefore ΔR1 would decrease linearly with cell ablation. When changes in ΔR1 across PC ablation models were analyzed, we were able to extrapolate a 0% PC density R1 for the cells remaining in the PCL. This intercept value was remarkably similar to the other to CB cortical layers, indicating that the hyperintensity of the PCL compared to the IGL and ML is almost entirely due to PCs. Uptake by BGs is the likely explanation for this residual ΔR1 (Aschner, Gannon, and Kimelberg 1992; Erikson and Aschner 2006). Care must be taken to reassess cell contributions under conditions where cell types with different Mn²⁺ uptake may be infiltrating an area or whose cell phenotype may be changing which may have affected our measurements. For example, in Niemann Pick Type-C (NPC) mice (Rallapalli, Darwin, et al. 2020), even after significant PC loss, the CB cortex exhibited MEMRI enhancement in late-stage NPC animals compared to controls. These animals also exhibited increased astrocyte activation and likely increased numbers of astrocytes. Although the models used in the current study were chosen to minimize any inflammatory effects on ΔR1 and images were acquired soon after ablation, inflammation and necrosis may have also contributed to the residual enhancement in the PCL not related to PCs (Waghorn et al. 2011; Kawai et al. 2010).

Mn²⁺ is likely to enter all cell types to some extent and to interact with different proteins in different cells, which can alter relaxivity. It is also likely that some fraction of Mn²⁺ is taken up and then sequestered by proteins restricting access to water protons. This Mn²⁺ fraction would be inactive for generating MRI contrast and therefore unable to be localized using MEMRI or relaxometry. Overall R1 in the mouse brain increases with administration of MnCl₂, even though this increase is heterogeneous in a regional specific way. Cytoarchitectural information is readily observed in the CB, hippocampus, and olfactory bulb at lower doses, but higher levels are required to see cerebral cortical layers (Aoki et al. 2004; Silva and Bock 2008). Thus, the cell type specificity may be changing as a function of Mn²⁺ dose. The ΔR1 of the ML and IGL reflect a lower level of Mn²⁺ accumulation at the dose used in the current study, and it is likely that the BG and other cell types that may enter the PCL after PC ablation accumulate Mn²⁺ in the PCL. Our finding of a non-zero intercept in Figure 6 represents an estimate of this residual uptake in the PCL.

Our estimates of the baseline endogenous R1 for the CB in these experiments were limited. It should be noted that the PCL ablation models were of mixed background, so any strain-dependent variation in laminar R1 adds uncertainty to these results. R1 measurements of the CB WM and CB cortex were made without MnCl₂ administration. It was not possible to segment the hyperintense layer without the addition of Mn²⁺ at 7 T. For this reason, we assumed that the R1 values in the ML, PCL, and IGL were equal prior to Mn²⁺ administration to the average cortical estimate for computing ΔR1. The fact that there is little contrast prior to Mn²⁺ addition makes it likely that this is a good assumption, but it may influence the quantitative interpretation of the ΔR1 magnitudes.

Taken together, the results of our study represent an important advancement in MEMRI by identifying the cerebellar cell type responsible for signal enhancement. The strategy used in our study demonstrates the power of pharmacogenetic ablation approaches to assign MRI signal to specific cell-types in the brain and should be useful for other cell types and other MRI contrast mechanisms. To date, no other studies have assigned signal to a specific brain cell type in the context of systemically delivered contrast agents. It is very likely that Mn²⁺ uptake and sequestration mechanisms are not the same for all brain areas and cell types in those areas may behave differently. The strategy taken here has potential to provide significant insight for interpretation of MEMRI as well as for other types of MRI contrast (Nyarko-Danquah et al. 2020; Wadghiri et al. 2004; Silva et al. 2008). Additional studies investigating other influences and pathologies will expand interpretability. For instance, influences of CSF circulation of Mn²⁺, cell density, and differential expression of metal transporters will likely all contribute to MEMRI signal levels. Assignment of Mn²⁺ signal to cell type as performed here provides a key first step to more biologically significant interpretations of preclinical MRI data at the level of specific cell types.

Highlights.

• The Purkinje cell layer is the source of high signal in Manganese-enhanced MRI (MEMRI) of the cerebellum

• Purkinje cells, not Bergmann glia, are responsible for this high signal.

• MEMRI signal is cell type specific.