Improving mitochondria and ER stability helps eliminate upper motor neuron degeneration that occurs due to mSOD1 toxicity and TDP‐43 pathology

Abstract Background Upper motor neurons (UMNs) are a key component of motor neuron circuitry. Their degeneration is a hallmark for diseases, such as hereditary spastic paraplegia (HSP), primary lateral sclerosis (PLS), and amyotrophic lateral sclerosis (ALS). Currently there are no preclinical assays investigating cellular responses of UMNs to compound treatment, even for diseases of the UMNs. The basis of UMN vulnerability is not fully understood, and no compound has yet been identified to improve the health of diseased UMNs: two major roadblocks for building effective treatment strategies. Methods Novel UMN reporter models, in which UMNs that are diseased because of misfolded superoxide dismutase protein (mSOD1) toxicity and TDP‐43 pathology are labeled with eGFP expression, allow direct assessment of UMN response to compound treatment. Electron microscopy reveals very precise aspects of endoplasmic reticulum (ER) and mitochondrial damage. Administration of NU‐9, a compound initially identified based on its ability to reduce mSOD1 toxicity, has profound impact on improving the health and stability of UMNs, as identified by detailed cellular and ultrastructural analyses. Results Problems with mitochondria and ER are conserved in diseased UMNs among different species. NU‐9 has drug‐like pharmacokinetic properties. It lacks toxicity and crosses the blood brain barrier. NU‐9 improves the structural integrity of mitochondria and ER, reduces levels of mSOD1, stabilizes degenerating UMN apical dendrites, improves motor behavior measured by the hanging wire test, and eliminates ongoing degeneration of UMNs that become diseased both because of mSOD1 toxicity and TDP‐43 pathology, two distinct and important overarching causes of motor neuron degeneration. Conclusions Mechanism‐focused and cell‐based drug discovery approaches not only addressed key cellular defects responsible for UMN loss, but also identified NU‐9, the first compound to improve the health of diseased UMNs, neurons that degenerate in ALS, HSP, PLS, and ALS/FTLD patients.


BACKGROUND
The corticospinal motor neurons (CSMN, a.k.a. upper motor neurons [UMNs]) have a unique role to collect, integrate, and transmit cerebral cortex's input to spinal cord targets, so that voluntary movement becomes the distilled end-results of cortical input. 1 UMNs are a clinically important and relevant neuron population, both within the context of injury and neurodegenerative diseases. [2][3][4] Longterm paralysis occur when UMNs degenerate in spinal cord injury patients. 5 In diseases such as hereditary spastic paraplegia (HSP), primary lateral sclerosis (PLS), and amyotrophic lateral sclerosis (ALS), the identification, characterization, and manifestation of disease are considered as a function of UMN degeneration over time. [6][7][8][9][10][11] Therefore, improving the health of degenerating UMNs will have broad implications both within the context of injury and neurodegeneration. The relevance of UMNs, especially with respect to ALS disease pathology, is becoming more evident, and the need to improve UMN health for developing effective and longterm treatment strategies is now considered a necessity. 12,13 However, to date, there has been no compound and no effective treatment strategy that targets the health of UMNs. Interestingly, even for motor neuron diseases characterized by the progressive loss of UMNs, none of the compounds that moved into clinical trials has ever been tested for their ability to improve UMN health. Current preclinical in vitro tests involve a different set of cell lines, at times not related to motor neuron biology. Most preclinical in vivo studies utilize mouse models that are generated by the mutation or pathology detected in patients, and that closely recapitulate many of the human condition. [14][15][16][17][18] So far, extension of lifespan in mouse models fails to translate to improved survival in patients [19][20][21] and calls for better and more informative preclinical assessments that translate to the human disease condition. Even though motor neuron diseases develop mostly because motor neurons degenerate, there has never been a study that investigates the health and betterment of diseased UMNs at a cellular level.
The only two drugs that have been approved by the FDA to treat ALS are riluzole, approved in 1995, and edaravone, approved in 2017; the latter works as a free radical scavenger and has been previously prescribed for stroke patients. [22][23][24][25][26][27] The ability of edaravone to improve UMN health has not been tested, and its efficacy has been studied only with the superoxide dismutase protein (SOD1), and not on the TDP-43, mouse model. Limited information is available on the cellular events that contribute to improved motor neuron survival. 22,28 Riluzole was approved prior to the development of hSOD1 G93A mice, and it works mainly on astrocytes to reduce astrogliosis-mediated toxicity. 29 Because riluzole failed to improve the longevity of the misfolded SOD1 (mSOD1) mouse model, 30,31 it probably would have failed preclinical testing had it been developed after the generation of hSOD1 G93A mice.
We and others find that UMNs in mice and UMNs in humans share many common features of motor neuron biology and display identical characteristics of neuropathology at the cellular level. 8,[32][33][34][35] For example, diseased UMNs in mice display degenerating apical dendrites, which is also observed in the UMNs of a broad spectrum of ALS patients, including sporadic ALS (sALS), familial ALS (fALS), and ALS/frontotemporal lobar degeneration (FTLD). 8,[35][36][37] Likewise, the UMNs that become diseased due to TDP-43 pathology have profound defects in their mitochondria and endoplasmic reticulum (ER), which are also observed in the UMNs of ALS patients with TDP-43 pathology. 34,38 This important translation at the cellular level further suggests that the emphasis needs to be on the neurons that degenerate and that a mechanismfocused and cell-based preclinical drug discovery platform would be informative and translational. [39][40][41] Furthermore, drug companies and the FDA now demand more information on the efficacy of compounds at the cellular level, which would expedite the success rate of clinical trials.
Misfolded SOD1 toxicity and TDP-43 pathology represent two distinct, and mostly nonoverlapping, causes of ALS. Recent studies have reported a positive effect of small molecules on SOD1 and TDP-43 models in vivo, but failed to investigate their impact on UMNs. [42][43][44][45] TDP-43 pathology is mostly excluded from the brains of patients with SOD1 mutations, and misfolded SOD1 is not observed in cases with TDP-43 pathology. [46][47][48][49][50][51] Therefore, being able to identify a compound that improves the health and stability of UMNs that become diseased due to these two different causes would have implications for a broad spectrum of patients.
Here we report that NU-9, a compound that was previously identified based on its ability to reduce mSOD1 aggregation in cell lines, to cross the blood brain barrier, have low toxicity and favorable drug-like properties, [52][53][54][55][56] has profound efficacy on stabilizing the cellular integrity of UMNs that degenerate due to mSOD1 toxicity and TDP-43 pathology. NU-9 treatment restored the structural integrity of mitochondria and ER, improved cytoarchitectural stability and integrity of UMN apical dendrite, eliminated the ongoing UMN degeneration that occurs due to two distinct underlying causes, and improved motor function that is related to UMN health. This is the first mechanismfocused and cell-based drug discovery study that lays the foundation for studies that will identify compounds based on their ability to restore neuron health, and also reports NU-9 as the first compound that eliminates UMN degeneration that occurs due to mSOD1 toxicity and TDP-43 pathology, an important step in drug discovery efforts for ALS, HSP, PLS, and ALS/FTLD patients.

Postmortem human brain samples
Postmortem human tissue collected according to protocols approved by an institutional review board was obtained from Northwestern University and University of Chicago. Clinical records were available for every subject. A neurologist examined all the patients and a neuropathologist had expertise in neurodegenerative disorders. Brains were fixed either in 10% neutral buffered formalin for 2 weeks or 4% paraformaldehyde (PFA) at 4 • C for 30 h. Areas of the primary motor cortex were retrieved by an expert neuropathologist, and 70-nm ultrathin sections were used for electron microscopy (EM) analysis, as previously described. 34 In this study, motor cortex isolated from normal control subjects with no neurologic disease (n = 4) and ALS patients (n = 9) were included (Table S1).

Mice
All animal procedures were approved by the Northwestern University Animal Care and Use committee and complied with the standards of the National Institutes of Health. All mice were on C57BL/6 background. Transgenic hemizygous males expressing a high copy number of the human SOD1 gene with a G93A mutation (B6SJL-Tg(SOD1*G93A)1Gur/J; The Jackson Laboratory) were bred to hemizygous UCHL1-eGFP(green fluorescent protein) females to generate hSOD1 G93A -UeGFP and wild type (WT)-UeGFP (control) mice. UCHL1-eGFP mice were generated in the Ozdinler Lab; they are reporter lines for UMNs, 57

NU-9 preparation and delivery
NU-9 was prepared as described previously. 54 Pharmacokinetic properties of NU-9 are listed in Table 1. For the formulation of 10 mg/ml concentration, 36.58 mg of test compound NU-9 was weighed, 0.274 ml of N-methyl-2pyrrolidone (NMP, Sigma-Aldrich) was added and vortexed, then 3.384 ml of olive oil was added and vortexed for ∼2 min until a clear yellow formulation was obtained. The 20 and 100 mg/ml doses were prepared separately and stored. hSOD1 G93A -UeGFP, prpTDP-43 A315T -UeGFP and WT-UeGFP mice were weighed and the required NU-9 dose per weight was calculated (20 or 100 mg/kg), which was administered once daily by oral gavage, starting at postnatal day (P)60 and continuing until P120. Animals from the untreated group received the vehicle (NMP and olive oil) only. For oral administration purposes, 20 ga × 38 mm plastic feeding tubes were used (Instech Laboratories, Inc). The gavage tip was inserted into the mouth directly over the tongue and into the pharynx. While observing the swallowing reflex, the tip was safely and smoothly slid into the esophagus. Once the administration was completed, the gavage tip was pulled straight out. Note. NU-9 was first identified for its ability to reduce mSOD1-mediated toxicity. 52 It has drug-like and favorable pharmacokinetic properties. 53,54 Abbreviations: ADME, absorption, distribution, metabolism, and excretion; NOAEL, no observed adverse effect level.

Rotarod test
Mice were placed on a rotating rod that accelerates linearly from 4 to 40 rpm (Rotarod, Ugo Basile), and the average time spent on the rotating rod for three consecutive trials was calculated for each mouse. Mice were allowed to run for a maximum of 5 min with a 5 min rest period between each run, and the latency to fall was recorded for three consecutive trials. The average of the three trials was taken as the data point for each mouse at the age tested, and the performance of different groups are reported as the mean ± SEM.

Hanging wire test
Mice were placed on a wire mesh, which was then inverted and suspended above the home cage; the time when the animal fell was recorded. This test was performed three times for a maximum of 60 s each session, with a 1-min rest period between each trial. The average performance for each session is presented as the average of the three trials.

Histology
Mice were deeply anesthetized using ketamine (90 mg/kg) with xylazine (10 mg/kg), and transcardially perfused with 4% PFA in PBS. The brains were removed intact and postfixed (4% PFA, overnight) and stored in PBS with sodium azide (0.01%) at 4 • C. Sections were cut in a coronal (50 µm) plane using a vibratome (Leica) and processed for immunocytochemical analyses.

UMN quantification
Since UMNs are genetically labeled with eGFP expression in the motor cortex of WT-UeGFP, hSOD1 G93A -UeGFP, and prpTDP-43 A315T -UeGFP mice, UMNs were identified based on the GFP expression. Quantitative analyses were performed on three matched rostrocaudal sections spanning the motor cortex. Three images per subject were taken to capture a 4× objective filed that encompass layer 5 of the motor cortex. An inverted epifluorescent Eclipse TE2000-E microscope (Nikon) with the same exposure time and intensity was used. UMNs were counted only if their soma and apical dendrite were both visualized in the same 50µm thick section. Images were analyzed in ImageJ (NIH) using the Find maxima processing command to determine the local maxima within a predetermined region of interest that circumscribes the layer 5 of the motor cortex. A universal noise tolerance, which ignores the local maxima corresponding to background and autofluorescence, was applied to all images.

Lower motor neuron (LMN) quantification
A 5-mm block of lumbar spinal cord was sectioned in 50-µm thick sections with a vibratome (Leica), and every other section was used for immunohistochemical analysis. LMNs were identified based on their location in the ventral horn and expression of the molecular marker ChAT. 10× Images were captured using an inverted epifluorescent Eclipse TE2000-E microscope (Nikon). GFP + /ChAT + and NeuN + /ChAT + LMNs were quantified per section. The average number of LMNs per section is reported per mouse with n = 5 mice per group.

2.9
Quantification of misfolded SOD1 50-µm thick sections of primary motor cortex were captured using the 20× objective on an inverted epifluorescent Eclipse TE2000-E microscope using the same settings and exposure time for all of the samples. eGFP positive (eGFP + ) UMNs in focal plane with sharp boundaries and a visible nucleus were traced using ImageJ (NIH). ROI were transferred to the misfolded SOD1 channel, and the integrated density was measured to determine the levels of misfolded SOD1 in GFP + UMNs only. Fifty to 90 UMNs were analyzed per mouse, and the average integrated density was reported per experimental group. For visualization purposes of misfolded SOD1 fluorescence intensity, nd2 files were opened in ImageJ and spectrum look-up table (LUT) was applied to the appropriate channel.

Electron microscopy
Mice were perfused with EM grade 4% PFA. One hemisphere of the brain was sectioned at 50 µm thickness coronally on a vibratome (Leica VT1000S, Leica Inc., Nussloch, Germany). The sections were postfixed in 2% PFA and 0.5% glutaraldehyde for 1 h, they were cryoprotected with glycerol-dimethylsulfoxide (DMSO) mixture followed by freeze-thaw at least four times, and treated with 1% sodium borohydride. They were then treated with 0.3% H 2 O 2 -10% methanol in TBS (100 mM Tris-HCl and 150 mM NaCl, pH 7.6) and 5% normal goat serum-1% bovine serum albumin in TBS to block nonspecific binding of primary antibody. This mixture was incubated overnight with rat anti-Ctip2 antibody (1:500, Thermo Fisher Scientific, Rockford, IL, USA). Biotinylated goat anti-rat IgG (1:500, Vector Laboratories, Burlingame, CA, USA) was used as the secondary antibody, and diaminobenzidine (DAB) was applied as the chromogen (ABC Elite kit, Vector Laboratories, Burlingame, CA, USA). Sections were then postfixed in buffered 2% osmium tetroxide (OsO 4 ) (Electron Microscopy Sciences, Hatfield, PA, USA), rinsed with distilled water, and stained in 1% uranyl acetate (Electron Microscopy Sciences, Hatfield, PA, USA), again rinsed with distilled water, dehydrated in ascending grades of ethanol with transition fluid propylene oxide (Electron Microscopy Sciences, Hatfield, PA, USA), embedded in the resin mixture with Embed 812 (Electron Microscopy Sciences, Hatfield, PA, USA), and cured in a 60 • C oven for 3 days. The sections in which primary motor cortex was present and visible under bright-field illumination on a dissecting scope were selected. Approximately, 5-mmwide × 7-mm-long pieces of the motor cortex from these sections were dissected under the microscope mounted on a resin block and were sectioned on a Leica Ultracut UC6 ultramicrotome (Leica Inc., Nussloch, Germany). The motor cortex was further trimmed, and only layer 5 was kept intact for preparing ultrathin sections. UMNs were identified based on the presence of Ctip2 immunohistochemistry (DAB) in their nuclei ( Figure S1). Seventynanometer-thin sections were collected on 200-mesh copper-palladium grids. Grids were counterstained with 8% radioactive depleted uranyl acetate and 0.2% lead citrate. Grids were examined on FEI Tecnai Spirit G2 TEM (FEI company, Hillsboro, OR, USA), and digital images were captured on a FEI Eagle camera.

Quantification of mitochondria and ER
EM images of Betz cells of normal control and patients, Ctip2-positive (Ctip2 + ) UMN of WT, and hSOD1 G93A -UeGFP and prpTDP-43 A315T -UeGFP mice ( Figure S1) were acquired from 70-nm ultrathin sections of the motor cortex, and EM images were taken on FEI Tecnai Spirit G2 TEM using FEI Eagle camera. Mitochondria were quantified, as described previously. 34 Mitochondria with intact outer and inner membrane and with intact cristae were considered healthy. Mitochondria with morphological defects, such as broken outer and/or inner membrane and cristae, were counted individually and were reported as percentage of defective mitochondria. ER that are disintegrated and that display profound expansion were considered unhealthy. Length of ER cisternae was measured using line tool of ImageJ software (NIH) upon calibration for pixels/micrometer using set scale function. Individual ER cisternae were traced using freehand line tool of ImageJ. The length of each traced cisternae per UMN per section was recorded. On an average, 8-10 cells/group were used for measurement of ER cisternae. The total number of mitochondria and ER, and the total number of UMNs investigated by EM-based quantification are presented in Table S4.

Statistical analysis
All analyses were performed using Prism software (Graph-Pad Software). The D'Agostino and Pearson normality test was performed on all datasets. Statistical differences between two groups were determined using either a parametric (Student's t-test) or a nonparametric test (Mann-Whitney t-test), when appropriate. For pairwise comparison of two groups, an unpaired t-test with Welch's correction was used. Statistical differences between more than two groups were determined by one-way ANOVA, followed by the Tukey's post hoc multiple-comparison test. Twoway ANOVA with Tukey's post hoc multiple-comparison test was used for comparison of behavior data of SOD1 mice, and mixed-effects analysis with Tukey's multiplecomparison test was used for TDP-43 mice. Statistically significant differences were taken at p < .05. Please refer to Tables S3 and S5 for results of all statistical tests performed.

Diseased UMNs in patients and UMNs of disease models share common cellular defects
UMN loss is a defining characteristic of ALS, and diseased UMNs display pathology at a cellular level (Figure 1). 8,34,35 Likewise, the UMNs of hSOD1 G93A and the prpTDP-43 A315T mice, which were developed to mimic mSOD1 toxicity and TDP-43 pathology mediated motor neuron degeneration in patients, 58,59 develop progressive UMN loss 34,57,60 and display similar defects at a cellular level. EM revealed such striking similarities between UMNs, albeit in different species (Figure 1).

Identification of NU-9
A high-throughput screen of >50,000 drug-like compounds was carried out using a SOD1 G93A -expressing PC12 cell-based assay 55,56 to identify neuron-protection scaffolds that mitigated protein aggregation and cytotoxicity. 52 The cytotoxicity protection assay identified compounds that protected cells from the toxicity of aggregated SOD1 G93A , and the protein aggregation assay targeted compounds that inhibited SOD1 G93A induced protein aggregation. Hits from three families of compounds were filtered by a ligandbased computational approach, including substructure and similarity searching 62 and a clustering technique. 63 A hit compound 1 (Figure 2A) was selected from >50 analogs of the cyclohexane-1,3-dione family of compounds, based on its ability to reduce mSOD1-mediated toxicity and to inhibit mSOD1-induced protein aggregation in a PC12 cellbased assay. 52 Multiple rounds of optimization were carried out, resulting in compound 2 (Figure 2A), which also had excellent in vitro absorption, distribution, metabolism, and excretion (ADME) properties, but did not penetrate into neurons. 53 Further modifications of compound 2, led to the generation of NU-9 (Figure 2A), 54 which penetrated cortical neurons, crossed the blood brain barrier, and had favorable pharmacokinetic properties 54 (Table 1). In summary, NU-9 had good in vitro potency (300 nM), microsome stability (t 1/2 74 min), plasma stability (t 1/2 >2 h), little inhibition of cytochrome P450s and the hERG channel, good brain permeability (8 µM), and extended the lifespan of the ALS mouse by 13% at 20 mg/kg. In vivo pharmacokinetics in mice showed t 1/2 = 2.7 h and 94% oral bioavailability. A 7-day repeat toxicology study in mice gave a no observed adverse effect level (NOAEL) of 100 mg/kg, indicating negligible levels of toxicity.

NU-9 treatment improves the integrity of mitochondria and ER
We recently generated a reporter line for UMNs, UCHL1-eGFP mice, in which UMNs are genetically labeled with eGFP expression that is stable and long lasting, 57 so that their cellular responses to compound treatment can be quantitatively assessed both in vitro and in vivo. 34,57 In an effort to visualize diseased UMNs and to assess their cellular response to compound treatment, hSOD1 G93A59 and the TDP-43 A315T mice 58 were crossed with UCHL1-eGFP to generate UMN reporter disease models, hSOD1 G93A -UeGFP and prpTDP-43 A315T -UeGFP mice, in which UMNs with mSOD1 toxicity and TDP-43 pathology were labeled with eGFP expression. 57 NU-9 ( Figure 2A) was delivered to both hSOD1 G93A -UeGFP and WT-UeGFP mice ( Figure 2B, Table S2) at two different doses (20 and 100 mg/kg/day) daily via oral gavage starting at P60, when mice begin to show symptoms and UMNs display cellular defects. 60 All mice were sacrificed at P120, which is considered end stage and about 60% of UMNs in the motor cortex are lost while the remaining UMNs have smaller soma size and vacuolated and disintegrated apical dendrites. 57 EM allowed cell type-specific analyses of UMNs and their key organelles with high precision at the ultrastructural level ( Figure 2C-F). At P120, UMNs of vehicle treated hSOD G93A -UeGFP mice lost most of their cytoplasmic integrity. There were very few intact organelles in the soma ( Figure 2G). However, the presence of disintegrated mitochondria ( Figure 2H,I arrowheads) and ER ( Figure 2J arrowheads) were strikingly evident. Healthy mitochondria were defined by the presence of double membranes, and intact cristae structure. Mitochondria mostly lost the integrity of their inner membrane ( Figure 2I  UeGFP mice with NU-9 treatment. **p < .002, One-way ANOVA followed by Tukey's post hoc multiple-comparison test aggregated, enlarged, or began to disintegrate. The ER also displayed broken and dispersed cisternae ( Figure 2J arrowheads). Such profound ultrastructural defects at an organelle level begin to reveal the cellular problems diseased UMNs face in hSOD G93A -UeGFP mice at P120, and it is thus of great importance to investigate whether NU-9 treatment would have an impact.
NU-9 treatment displayed profound improvements in both the structure and integrity of mitochondria and ER of diseased UMNs (100 mg/kg day dose, the only dose investigated at the EM level; Figure 2K-N). Upon treatment, the overall picture of the soma was dramatically improved with the presence of an intact nuclear membrane, which was devoid of any invaginations or protrusions, and detection of numerous organelles that were proper in size, location, and interactions among each other ( Figure 2K). The mitochondrial inner membrane was intact with proper cristae ( Figure 2L,M arrows), which were in close contact with the ER (Figure 2N arrows). The integrity and structure of the ER were maintained, and the expansion of the lumen was eliminated ( Figure 2N arrows).

NU-9 treatment improves cytoarchitectural integrity of UMN apical dendrite
Because the apical dendrite is the site of cortical integration and its stability is the key to proper UMN modulation, function, and health, we next investigated whether NU-9 treatment would also improve the cytoarchitectural integrity of UMN apical dendrite and reduce the extent of its vacuolization and disintegration. WT-UeGFP mice treated with vehicle have mostly healthy apical dendrites that extend toward the top layers, and only a small percentage had vacuoles (28% ± 12%; Figure 4A,E). WT-UeGFP mice treated with 100 mg/kg/day of NU-9 also have healthy apical dendrites with few vacuoles (16% ± 5%; adjusted p-value = .9235). On the other hand, most of hSOD1 G93A -UeGFP UMNs treated with vehicle continued to have vacuolated and disintegrating apical dendrites (76% ± 11%; Figure 4B,F), and the difference between WT-UeGFP controls was highly significant (adjusted pvalue = .0133; Figure 4G). However, NU-9 treatment significantly improved the integrity of disintegrating apical dendrites in hSOD1 G93A -UeGFP mice in a dose-dependent manner ( Figure 4C,D). Upon 20 mg/kg/day treatment ( Figure 4C), the percentage of UMNs with vacuolated apical dendrites was reduced to 44% ± 7%, and this is further reduced to 23% ± 10% when hSOD1 G93A -UeGFP mice are treated with 100 mg/kg/day of NU-9 (Figure 4C,D,G; adjusted p-value = .0019), which was comparable to the F I G U R E 3 NU-9 treatment reduces misfolded SOD1 levels in upper motor neurons (UMNs) of hSOD1 G93A -UeGFP mice. (A) Representative images of UMNs and B8H10 antibody staining that recognizes misfolded SOD1 protein in the motor cortex of WT-UeGFP or (B) hSOD1 G93A -UeGFP mice treated with vehicle, (C) 20 mg/kg/day NU-9, or (D) 100 mg/kg/day NU-9. Scale bars, 20 µm; n ≥ 3 biological replicates. (E) Average integrated density of misfolded SOD1 fluorescence in UMNs in the motor cortex of WT-UeGFP or hSOD1 G93A -UeGFP mice treated with vehicle, 20 mg/kg/day NU-9, or 100 mg/kg/day NU-9; mean, SEM, and individual data points shown for n ≥ 3 biological replicates. **p < .01, ****p < .0001, One-way ANOVA followed by Tukey's post hoc multiple-comparison test integrity of apical dendrites in the control WT-UeGFP mice (adjusted p-value = .9956).

NU-9 treatment significantly improves UMN retention in the motor cortex of hSOD1 G93A mice
The timing and the extent of UMN loss in hSOD1 G93A -UeGFP mice 57 is comparable to that of UMN loss in hSOD1 G93A mice. 60 Vehicle treatment did not have an impact on UMN numbers in WT-UeGFP mice (59 ± 3 UMNs; Figure 5A) or hSOD1 G93A -UeGFP mice (5 ± 2 UMNs; Figure 5B). UMN loss with respect to disease progression remained significant (adjusted p-value < .0001) with vehicle treatment. NU-9 treatment even at the 100 mg/kg/day dose did not have any adverse effects on UMN numbers in WT-UeGFP mice (66 ± 1 UMNs; adjusted p-value = .5368). However, when hSOD1 G93A -UeGFP mice were gavage treated daily with NU-9, more UMN cell bodies were detected in the motor cortex (Figure 5C,D). There were 21 ± 5 UMN after 20 mg/kg/day of NU-9 treatment (adjusted p-value = .0446; Figure 5E), and upon treatment with a 100 mg/kg/day dose, there was a dramatic increase in the numbers of UMNs retained in the motor cortex (46 ± 3 UMN), which was highly significant when compared to vehicle-treated mice (adjusted p-value < .0001) and mice treated with 20 mg/kg/day (adjusted p-value < .0001; Figure 5E). Importantly, the average number of UMN present in the motor cortex of hSOD1 G93A -UeGFP mice treated 60 days with 100 mg/kg/day of NU-9 became almost comparable to that of UMN numbers present in healthy WT-UeGFP mice (adjusted p-value = .0306).

NU-9 improves integrity of mitochondria and ER of UMNs with TDP-43 pathology
We recently discovered that cellular defects, which occurred in UMNs of Betz cells of ALS patients with TDP-43 pathology, were fully recapitulated in the UMNs of TDP-43 A135T mice. 34 EM helped visualize and reveal intracellular defects that occur in UMNs ( Figure 6A). Mitochondria were severely affected, their inner membrane was broken, and mitochondria disintegrated ( Figure 6B arrowheads). Likewise, ER lost the integrity of their architecture; cisternae were broken ( Figure 6C arrowheads), and pieces of ER with ribosomes attached were detected in the soma.
Since NU-9 improved ultrastructural integrity of both mitochondria and ER of UMN diseased due to mSOD1 mean, SEM, and individual data points shown for n ≥ 5 biological replicates. *p < .05, ***p < .001, ****p < .0001, One-way ANOVA followed by Tukey's post hoc multiple-comparison test toxicity, and these were indeed the prominent defects detected in the UMNs with TDP-43 pathology, we reasoned that a mechanism-based treatment strategy would suggest NU-9 to improve the mitochondrial and ER defects observed in UMNs that become diseased by TDP-43 pathology as well, even though these are two dis-tinct and different disease models. To visualize diseased UMNs and to assess their cellular response to compound treatment, TDP-43 A315T mice 58 were crossed with UCHL1-eGFP mice 57 to generate an UMN reporter disease model prpTDP-43 A315T -UeGFP mice. 34 There is no misfolded SOD1 detected in the UMN of prpTDP-43 A315T -  (Figure S3), and there is no TDP-43 pathology reported in SOD1 mouse models or patients with SOD1 mutations, [46][47][48][49][50][51] in fact mSOD1 toxicity and TDP-43 pathology are accepted to be distinct causes of motor neuron degeneration. Therefore, UMN degeneration in TDP model cannot be explained by mSOD1 toxicity. We decided to take the leap and investigated whether NU-9 treatment would also be effective with respect to TDP-43 pathology ( Figure 6A-C).

NU-9 treatment eliminates degeneration of UMNs with TDP-43 pathology in vivo
We previously reported problems with mitochondria and ER of UMNs with TDP-43 pathology, 34 and because NU-9 treatment enhanced the integrity of mitochondria and the ER of diseased neurons at the ultrastructural level ( Figure 6), we next investigated whether NU-9 treatment would also support cellular integrity and survival of UMN with TDP-43 pathology in vivo.

NU-9 treatment improves upper motor neuron function
Even though most behavioral assays fail to properly assess UMN health and connectivity, the hanging wire test is reported to be more specific to UMN integration, as evidenced by Fezf2 −/− mice that are born without UMN and CST axons, which perform very poorly in this test. 72 Therefore, we performed both the well-studied rotarod 73,74 and the hanging wire test, which reveals the ability of the mouse to use its fingers and digits to hold on to the wire (Figure 9).
There was a distinction between healthy WT-UeGFP and diseased hSOD1 G93A -UeGFP mice on the hanging wire test, as hSOD1 G93A -UeGFP mice failed to grab and hold on to the inverted wire as disease progressed. The difference became significant by P95 and continued to be significant throughout (P95: WT-UeGFP: 60 s; hSOD1 G93A -UeGFP: 49.5 ± 3.7 s, adjusted p-value = .0328; Figure 9C). Untreated hSOD1 G93A -UeGFP mice were not able to stay on the hanging wire and their performance continued to decline with age ( Figure 9C, Table S3). On the contrary, hSOD1 G93A -UeGFP mice treated with a 100 mg/kg/day dose of NU-9 performed significantly better than hSOD1 G93A -UeGFP mice treated with vehicle by P102 (hSOD1 G93A -UeGFP [vehicle]: 42.3 ± 3.8 s; hSOD1 G93A -UeGFP [100 mg/kg/day NU-9]: 53.3 ± 2.9 s; adjusted p-value = .0254), and this performance was comparable to that of healthy mice at that age (WT-UeGFP [vehicle]: 60 s; adjusted p-value = .2529). Unlike hanging wire test, NU-9 treatment did not result in significant improvement in rotarod performance, regardless of dose ( Figure 9B, Table S3).

DISCUSSION
The cellular events that give rise to selective neuronal vulnerability leading to neurodegenerative diseases are now better understood than a decade ago, 77-84 and many more compounds are generated and characterized with druglike properties. [85][86][87][88] Yet, there has been no effective cure for any of the motor neuron diseases, especially for the diseases of the UMNs. One of the major limitations has been the lack of proper tools and drug discovery platforms that would utilize UMN response as the readout. In their absence, preclinical assays rely heavily on extension of mouse lifespan as outcome measure. 14,89-93 However, the lack of translation from mouse models to humans has resulted in numerous failed clinical trials and compounded frustrations. The need to develop better preclinical assays that provide information about the survival needs of vulnerable and degenerating neurons in patients has become evident. 40,41,94 Even though UMNs are a critical component of motor neuron circuitry, the idea that their degeneration is secondary to LMN loss and is a byproduct of an ongoing degeneration, previously diminished their importance as a potential cellular target for therapeutic interventions. 67,95,96 However, mounting experimental data now reveal that the cellular pathology of UMNs becomes evident much earlier than symptom onset. 32,36,37,57 Spine loss and apical dendrite degeneration occurs prior to neuronal loss, 97-100 and cortical hyperexcitation is even used as an early detection marker of ALS. 35,[101][102][103] In a rat model of mSOD1, the reduction of G93A mutation levels only in the motor cortex improved the health and integrity of global motor neuron circuitry, 104 further supporting the idea that UMNs are feasible targets. Recent studies also confirmed the importance and relevance of UMNs in ALS pathology. When UMNs were ablated from the SOD1 mouse models by crossing with the fezf2 null, the results revealed that UMNs indeed play an important role for initiating and modulating disease pathology and that their degeneration is not by mere consequence. 72,105 In addition, floxed hSOD1 G37R mice recapitulate UMN loss. When mutant hSOD1 G37R is removed from UMN of floxed mice by crossing them with the CrymCreER T2 mice, UMN loss is prevented suggesting that UMN degeneration relies on cell autonomous mechanism. 106 These recent findings establish UMNs as important contributors to disease pathology in ALS and further suggest that their survival needs to be considered for building effective treatment strategies.
The second reason that has diminished interest in UMNs has been the assumption that treatments that benefit LMNs should also be beneficial to UMNs, and that since they are both motor neuron populations, they do not need to be investigated separately. This could be the reason why none of the compounds that have been in clinical trials for ALS have ever been tested for their ability to improve UMN health. It has been assumed that if a compound improves the health and integrity of LMNs, it should also improve the health of UMNs, and therefore no special or additional emphasis has been given to the UMNs. The assumption that these two neuron populations are similar is unfounded because they are born from different progenitors, and their differentiation, maturation, target recognition, and integration to circuitry occur at different time points, at different sites, and via different molecular mechanisms. 107 Their gene expression profiles and neuronal identities also are very different. Thus, it is not reasonable to think that they should respond similarly to treatment. In fact, their requirements for survival could indeed be very different. [108][109][110] By focusing our attention on the needs of diseased neurons, and building effective treatment strategies that take their survival requirements into account, we could set the stage for treatments that are translational. 41,94 We have developed a novel platform in which the responses of UMNs to compound treatment can be readily assessed at a cellular level with precision and clarity that was not previously possible. Our study revealed that NU-9 treatment improved the ultrastructural integrity of mitochondria and the ER, both of which are exceptionally important organelles for motor neuron health. 34,38 Mitochondria are responsible for the generation of ATP. 111 They also play a key role in the initiation of innate immunity. 112 Therefore, the health and integrity of mitochondria are crucial for motor neurons that have high levels of energy demand and must control neuroimmune reactions for improved health. [113][114][115] Mitochondrial problems occur very early and selectively in UMNs, which develop the disease because of mSOD1 toxicity, lack of Alsin function, and Profilin mutations. 34,37,38,57,110,116,117 Misfolded SOD1 selectively binds to mitochondria and affects their shape and function. [118][119][120][121][122][123][124][125] TDP-43 also binds to mitochondria, and inhibiting TDP-43 binding to mitochondria improves motor neuron function. [126][127][128] Therefore, reducing the levels of misfolded SOD1 and TDP-43 pathology potentially improves mitochondrial function. Likewise, improving mitochondrial function also reduces levels of misfolded SOD1 in motor neurons. 125 Therefore, further studies are required to reveal the details of cause and effect. 129 The ER is the site of protein production and initial folding, where defects result in ER stress, one of the converging pathologies shared among many different neurodegenerative diseases. 130 We previously reported that diseased UMNs are especially prone to ER stress, and increased ER stress contributes to their early vulnerability, while other cortical neurons remain healthy. 36 Electron microscopy studies reveal the presence of ER stress as a result of the enlargement of lumen, followed by disintegration of the ER. 34 The ability of NU-9 to improve the integrity of both mitochondria and the ER is exceptionally significant, because, even though the underlying causes of the disease are heterogeneous, many of the pathways converge on the health and function of mitochondria and the ER. [131][132][133] Disruption of intracellular membrane organelles, such as the Golgi apparatus, has been suggested as a possible cause for ALS 134 and is proposed to be upstream of the ER dysfunction. 135 Maintenance of the ultrastructural integrity of these two key organelles, especially in the neurons that display primary vulnerability, would enable them to perform their much required function. This may explain why NU-9 treatment improves UMN cytoarchitecture and eliminates their progressive degeneration in both hSOD1 G93A and TDP-43 A315T mice.
The underlying pathologies of the UMNs in wellcharacterized mouse models and UMNs in patients with motor neuron diseases are almost identical at the cellular level. For example, the apical dendrite degeneration observed in UMNs was also recapitulated in the Betz cells of fALS, sALS, and ALS with FTLD patients. 8 The apical dendrite is exceptionally important for the function of UMNs. We previously identified apical dendrite defects in UMNs that become diseased by mSOD1 toxicity, 37,57 lack of Alsin function, 116 Profilin defects, 117 and TDP-43 pathology, 32 and these cellular defects were fully recapitulated in the Betz cells-the UMNs in humans-of a broad spectrum of ALS patients, including sALS, fALS, as well as ALS/FTLD. 8 UMNs of patients had vacuolated apical dendrites, which display massive disintegration. 8,32,36,37 This is the site where other cortical neurons connect and communicate with the UMNs. 8,35,136,137 When apical dendrite disintegrates, it is not possible for UMNs to be properly modulated and thus they fail to convey cerebral cortex's signal to the spinal cord targets. We and others find that UMN apical dendrite degeneration is an early event in ALS and contributes to the dysregulation of cortical hyperexcitation and hypoexcitation even prior to symptom onset. 35 Therefore, being able to reverse the ongoing apical dendrite degeneration would have significant outcomes for neuronal health and connectivity. We find that NU-9 treatment enhances the cytoarchitectural stability of apical dendrite, such that they become comparable to WT healthy controls. These findings would have important implications for the health and connectivity of UMNs. One of the outcomes of improved neuronal connectivity in vivo is improved motor function. Even though detection of UMN involvement in motor function of mice is challenging, recent evidence suggests a unique utility for the hanging wire test to interrogate UMN involvement. We find that NU-9 treatment improves the ability of both prpTDP-43 A315T and hSOD1 G93A mice to do better in the hanging wire test, further suggesting that the observation of a cellular function has a direct consequence in motor behavior. NU-9 treatment not only improves the health of UMNs that become diseased as a result of mSOD1 toxicity but also because of TDP-43 pathology, which represent two important, yet independent, causes of motor neuron degeneration. Mutations in the SOD1 gene were identified in ALS patients, 138 and the disease mouse models were based on the mutations detected in patients 59 ; these models mimicked many aspects of human pathology, including progressive UMN loss. 57,110 It is important to note that misfolded SOD1 inclusions are also detected in patients with mutations in C9orf72 and other ALS/FTLD associated genes. 139,140 In addition, ER stress leads to accumulation of even the WT SOD1 aggregates. 141 Misfolded WT SOD1 thus can propagate in a prion-like fashion and seed cytotoxic misfolding of WT SOD1. 142,143 Therefore, it is important to identify compounds that can reduce misfolded SOD1mediated toxicity.
TDP-43 pathology develops regardless of a mutation in the TARDP gene, 144,145 and it is observed in the brains of about 95% of ALS patients. 146 Most patients with SOD1 mutations do not display TDP-43 pathology in their brains, 46,50 and TDP-43 accumulations are not detected, [46][47][48]50 albeit interactions between mSOD1 and TDP-43 have been suggested. [147][148][149] In SOD1 mouse models with G93A, G37R or G85R mutations, there is no mislocalization of TDP-43 to the cytoplasm in motor neurons of mutant SOD1 transgenic mice, nor association of TDP-43 with ubiquitinated inclusions. 50 In addition, abnormally phosphorylated or truncated TDP-43 species were not detected in fractionated ALS mouse spinal cord or brain. 51 DNA strand breaks and TDP-43 mislocalization are absent in the murine hSOD1 G93A model of ALS both in vivo and in vitro. 49 Therefore, identification of a compound that improves the health of UMNs that become diseased by these two prominent and distinct causes is rather significant. It also suggests that NU-9 would have an impact in a broad spectrum of patients, including patients with ALS, HSP, PLS, and ALS/FTLD.

CONCLUSIONS
Today, drug discovery efforts are more focused on reversing the disease-causing cellular mechanisms and improving the health of neurons that display selective and progressive degeneration. However, numerous challenges still exist, especially for the diseases of the UMNs. For example, the underlying causes of neuronal vulnerability are not well defined, and there has been no preclinical assay that assesses cellular responses of UMNs to compound treatment. None of the compounds that are in clinical trials for motor neuron diseases have ever been tested on diseased UMNs. Here, we first find that the mitochondrial defects and problems with ER are observed both in the UMNs of ALS patients and in the UMNs of mouse models that are developed to mimic patients with mSOD1 toxicity and TDP-43 pathology. There is translation at a cellular level, and even though the UMNs are in different species, the underlying causes of UMN vulnerability are the same. Mitochondrial defects and problems with the ER, therefore, offer a target for intervention. We find that NU-9, a compound that was previously characterized to reduce mSOD1 aggregates in cell lines and a compound that crosses the blood brain barrier with favorable pharmacokinetic properties, has the unique ability to improve the structure and the integrity of both mitochondria and ER. This unique ability results in enhancing the cytoarchitectural integrity of degenerating UMNs and, most importantly, stopping the progressive degeneration of UMNs that become diseased as a result of mSOD1 toxicity and TDP-43 pathology, two independent and overarching causes of neurodegeneration. Our findings mark the identification of the first compound that improves the health of diseased UMNs, and lay the foundation for future mechanismfocused and cell-based drug discovery studies.

C O N F L I C T O F I N T E R E S T
The authors declare that there is no conflict of interest.