Assessing neuraxial microstructural changes in a transgenic mouse model of early stage Amyotrophic Lateral Sclerosis by ultra‐high field MRI and diffusion tensor metrics

Abstract Objective Cell structural changes are one of the main features observed during the development of amyotrophic lateral sclerosis (ALS). In this work, we propose the use of diffusion tensor imaging (DTI) metrics to assess specific ultrastructural changes in the central nervous system during the early neurodegenerative stages of ALS. Methods Ultra‐high field MRI and DTI data at 17.6T were obtained from fixed, excised mouse brains, and spinal cords from ALS (G93A‐SOD1) mice. Results Changes in fractional anisotropy (FA) and linear, planar, and spherical anisotropy ratios (CL, CP, and CS, respectively) of the diffusion eigenvalues were measured in white matter (WM) and gray matter (GM) areas associated with early axonal degenerative processes (in both the brain and the spinal cord). Specifically, in WM structures (corpus callosum, corticospinal tract, and spinal cord funiculi) as the disease progressed, FA, CL, and CP values decreased, whereas CS values increased. In GM structures (prefrontal cortex, hippocampus, and central spinal cord) FA and CP decreased, whereas the CL and CS values were unchanged or slightly smaller. Histological studies of a fluorescent mice model (YFP, G93A‐SOD1 mouse) corroborated the early alterations in neuronal morphology and axonal connectivity measured by DTI. Conclusions Changes in diffusion tensor shape were observed in this animal model at the early, nonsymptomatic stages of ALS. Further studies of CL, CP, and CS as imaging biomarkers should be undertaken to refine this neuroimaging tool for future clinical use in the detection of the early stages of ALS.


| INTRODUC TI ON
The introduction of animal models has been one of the major steps forward towards a better understanding of the neuropathological processes occurring in humans. Based on their similarity to the human genome and easy availability, mammalian murine models have been one of the most commonly used representations of neurodegenerative diseases such as Amyotrophic Lateral Sclerosis (ALS). 1 As such, the phenotypical expression and histological discoveries parallel many of the symptomatology and neuropathological findings observed in patients with ALS. 2,3 During the last decades, the exponential growth of novel genetic tools has led to the detection of new mutations in the patient population, and thus to help in the development of new transgenic animal models for research and therapeutic discovery purposes. 4 From the original ALS mouse model representing the familiar form of ALS with the mutation of the superoxide dismutase 1 (G93A-SOD1) gene, 5 a growing number of rodent models that express different mutations, such as the fused in sarcoma (FUS), 6 the C9orf72 hexanucleotide repeat expansion mice, 7,8 and the transactive response DNA binding protein 43 kDa TDP-43 9,10 (among others) have been increasingly developed to address the effects of molecular changes on neuronal degeneration and death. However, such models represent less than 10% of the sporadic cases of ALS. Other models of sporadic ALS mutations, like the wobbler and VPS54 mice, 11,12 have been used aiming to understand the basic cellular mechanisms of motoneuron diseases, but these abnormalities are likely different from those occurring in ALS. 1 Despite the growing availability of murine animal models, the G93A-SOD1 mouse model is the original and longer tested mammalian model of ALS up to date. 13 In the last decades, the increasing development of MRI systems has been able to allow neuroscientists to analyze in real-time the neuropathological process occurring in physiological intact biological systems. 14,15 Also, the advances in computational power, hardware, and gradient strengths has been paramount in high-field resolution MRI particularly in small animal research. [16][17][18] Since its inception, diffusion tensor imaging (DTI) has been applied as an imaging technique to evaluate not only microstructure but also the integrity and connectivity of different CNS regions. 19 Diffusion tensor imaging provides a mathematical model of diffusion anisotropy and is widely used. Parameters, including fractional anisotropy (FA), mean diffusivity (MD), as well as parallel and perpendicular diffusivity, can be derived to provide sensitive, but nonspecific, measures of the altered tissue structure. 20 The popularity of DTI has been based on its relatively simple computational algorithms based on the directionality of the diffusion tensor's eigenvalues. 21 Therefore, this diffusion technique is one of the most widely utilized in the scientific and medical field to evaluate axonal injuries across different axonal tracts. 22,23 The diffusion tensor can be visualized using an ellipsoid where the principal axes correspond to the directions of the eigenvector system. 24,25 By applying the symmetric properties of this ellipsoid, the diffusion tensor can be decomposed into basic geometric measures to describe the shape of the diffusion tensor models. 26 In that regard, DTI derived parameters such as linear (C L ), planar (C P ), and spherical (C S ) anisotropy have been only applied to evaluate hippocampal structures on preclinical models of epilepsy. 27 Here, we extend this approach to characterize such DTI derived morphological parameters in the context of ALS. In previous work, we used ultra-high field diffusion MRI (UHFD-MRI) to asses microstructural changes in gray and white matter (GM & WM) applied to different murine models of neurodegenerative diseases. 28,29 Our results have shown that structural changes can be detected during earlier stages of the disease. 14,30,31 In line with such studies, we postulate that the intrinsic geometrical properties of monoexponential diffusion signals derived from DTI could add another level of specific information in relation to different neuraxial structures.
Hence, in this work we extended our previous investigations on DTI to analyze if parameters representing the DTI tensor contain additional information that: a) determine the specific geometrical characteristics of diffusion tensors across different neuronal tissue interrogated and b) to assess their value as additional biomarkers of disease in the context of neurodegenerative diseases (ALS mice). To test these hypotheses, we analyzed CNS tissues of an animal model of ALS (G93A-SOD1 mice) with UHF-MRI to determine the potential role of derived DTI anisotropic parameters in ALS.

| Theory
The tensor field data were diagonalized using the standard analytical methods and eigenvalues were obtained (λ 1 , λ 2 , and λ 3 ) to calculate Fractional Anisotropy (FA) as described in Eq. 1. Additionally, the tensor field was used to compute the DTI metrics, such as linear anisotropy (C L ), planar anisotropy (C P ), and spherical anisotropy (C S ) as described in Equation 2, 3, 4, respectively. 32,33 where; trace = (λ 2 + λ 2 + λ 3 ) FA and DTI metric maps were calculated, and a mean value of each parameter was extracted and calculated from each ROI ( Figure 1).

| Animals
All procedures used to obtain tissues followed an approved protocol from the animal care committee (ACC) at the University of Illinois at Chicago (UIC). In any situation of animal distress or pain, animals were sacrificed in carbon dioxide using standard protocols For MRI and histology imaging methods, ALS mice were obtained from the Jackson Laboratory (JAX#004435) and bred on a C57BJ6 background, overexpressing the SOD1 transgene with the G93A mutation. The G93A-SOD1 mice in this background have been previously characterized and they develop motor symptoms at approximately 110 days of age and die around 160 days. 34,35 We considered three groups of animals for this work: a wild-type control group (WT), a presymptomatic group at postnatal day 80 (P80) and a symptomatic group at postnatal day 120 (P120). For MRI studies, a total of 12 animals were used: wild-type (WT) control (n = 4) and ALS (G93A-SOD1) mice (n = 4 per group). Mice had easy access to food and water and were checked daily to assess their level of well-being and health.
Additional animals WT, P80, and P120 were used for further histological analysis, Specifically, we evaluated morphologic neuronal anomalies in the context of ALS, using additional mouse reporters expressing a yellow fluorescent protein (YFP) transgene specifically associated with a neuronal Thy1 promoter, were chosen. The first reporter group was chosen for the high YFP expression in axons located in spinal cord areas, the so-called YFP-J16 mice (JAX#003709).
The second group of YFP mice was chosen for its mild fluorescent Thy1 expression and higher background, making it ideal to study individual neuronal structural details in the cerebral cortex and hippocampus, so-called YFP-H mice (JAX#003782). Detailed molecular and neuronal population differences between both YFP reporters have been extensively described in previous work. 36,37

| MRI and DTI protocol
Animals were rendered unconscious with CO 2 inhalation, then transcardially perfused with a PBS and 4% paraformaldehyde (PFA) solution. After the skull was opened, mouse brains were extracted intact and immersed in PFA (>48 hours). Before scanning, brains were soaked overnight in phosphate-buffered saline (PBS) (Corning cellgro, catalog #21-040-CV, lot#12417001) to remove free fixative. Three brains were stacked in a 10 mm inner diameter glass tube (Fisher Scientific, cat#14-961-26) and surrounded with fluorocarbon oil (Fluorinert®, 3M, Maplewood, MN). Images from brains were acquired with a 17.6T vertical-bore Avance II scanner using a 25 mm RF coil. Spinal cords (SC) were placed on a 5 mm NMR tubes (New Era #NE-MQ5-7, 300-400 MHz) and scanned in a 5 mm RF coils as described in previous studies. 38 F I G U R E 1 Anatomical maps and areas of segmentation in the ALS mouse. A, T2 map anatomical representation of a mouse brain showing different white (WM) and gray matter (GM) segmented regions used for diffusion tensor imaging (DTI) metrics analysis. B, MRI diffusion map at b0 from a mouse spinal cord (SC). SC segmentation was centered in the lumbar WM and GM regions F I G U R E 2 MRI Diffusion tensor derivatives in different tissues of a wild-type mouse. A, Diagram representing the different parameters used to assess tensors geometry, such as linear, planar, and spherical anisotropy (CL, CP, and CS, respectively) and its spatial relationship with eigenvalues (e). B, Representation of CL, CP, and CS in Bayesian coordinate plots across different central nervous system (CNS) region in a wild-type rodent (WT). Note that different gray and white matter structures yield tensors with a unique & different geometry. Animals (n = 4), *P < .05, **P < .01. Abbreviations: WM, white matter; GM, gray matter; CC, corpus callosum; CST, corticospinal tract; CCX, cortex; Hipp, hippocampus; SC wm , spinal cord white matter; SC gm , spinal cord gray matter Two imaging sessions of six mouse brains (N = 12), with a total of 170 MRI slices were acquired, coronally centered and oriented along the rostral-caudal axis of each brain. Additionally, a group of six spinal cords was imaged (16 axial slices per SC) using an acquisition centered in the lumbar region. For all the scans, diffusion-weighted images were acquired using a spin-echo sequence with TR = 10 000 ms and TE = 20 ms, interleaved 0.  39 ROIs from each central nervous system structure were manually segmented following anatomical landmarks described in the standard stereotaxic coordinate mouse brain and spinal cord atlases and data extracted using ITK-SNAP. 40
*P < .05 from control; **P < .01 from control. selected matching stereotaxic coordinates. 41,42 Fluorescent confocal microscopy images were obtained with a 534 nm laser channel using standard techniques previously described 29,37 and images processed using ImageJ software. 43,44

| Statistical analysis
Quantitative data were tabulated and analyzed using statistical   Figure 2). Overall, the C L was greater in WM than GM regions, the C P was slightly greater in WM structures, and the C S was substantially greater in the GM. Interestingly, SC wm had a larger C L and C S values when compared to other WM regions ( Table 1). As such, tensor metric differences across each WM tracts can be accounted for their specific eigenvalues, marking the proportion of crossing fibers on each WM structure (Figure 3).  Figure 5A, B = 10 microns

| D ISCUSS I ON
The use of transgenic animal models imposes a significant advantage in neurobiology in the detailed analysis of biomarkers. 45 Despite the recent introduction of additional animal models of ALS, 46,47 the vast majority of preclinical mammalian studies are based on rodents. 48 The possibility of neuronal labeling fusing specific fluorescent tags has allowed neuroscientists to use the tagged fusion expressed in the animals not only to improve the visualization of axonal degeneration in diverse experimental conditions but also as a ground  [49][50][51] In recent years, DTI metrics have been used in several neuro-oncological, 52-55 as well as neuroinfectious [56][57][58][59] clinical scenarios (Table 2). In oncological conditions, a decrease in C P and C L was associated with an increase in C S , probably due to a significant increase in tissue cellularity. In neurological and extra neurological 60 inflammatory pathologies, the C P and C L component  [71][72][73][74][75] ; it is also possible that a relative reduction of the neuraxial volume could determine a relative increase in cellular density as described in our previous histological finding. 29,38 Previous DTI work in canine animal models determined that the distribution of data for the WM internal capsule differed markedly from the WM centrum semiovale region. 33 Furthermore, data for the internal capsule were distributed in a relatively tight cluster, possibly reflecting the compact and parallel nature of its fibers, whereas data for the centrum semiovale were more widely distributed, consistent with the less compact and often crossing pattern of its fibers. This indicates that the tensor shape plot technique can depict data in similar WM regions as being alike, adding more specificity to these parameters. Results presented in our studies (Table 2)  The validity of animal models for the study of human diseases such as ALS has been often criticized particularly due to their high phenotypical dependence on their genetic background. 83 Lastly, the interrogation of alterations observed in complex brain microstructures (superficial and deep GM regions) could also be limited by the exclusive use of Gaussian diffusion models. Thus, additional implementation of multi-exponential diffusion models 28,88 will be required for more accurate representation to improve monitoring, as well as testing new therapeutic strategies in ALS. DMR-1157490 and the State of Florida.

CO N FLI C T O F I NTE R E S T
None.