2.1 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).

2.2 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 and are now available at Jackson Laboratory (stock no. 022476). Hemizygous UCHL1-eGFP females were bred to hemizygous prpTDP-43^A315T mice (procured from Jackson Laboratory, stock no. 010700) to generate prpTDP-43^A315T-UeGFP mice. In this study, only female mice were used for experiments and male mice were bred with WT females to generate more hSOD1^G93A-UeGFP or prpTDP-43^A315T mice. prpTDP-43^A315T mice were supplied with gel diet (DietGel 76A, ClearH₂O, ME, USA) to eliminate gastrointestinal (GI) complications. Transgenic mice were identified by PCR amplification of DNA extracted from their tail, as previously described.^{34, 57-59}

2.3 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-2-pyrrolidone (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.

In this study, WT-UeGFP (n = 10) and hSOD1^G93A-UeGFP mice (n = 6) were treated with vehicle, WT-UeGFP (n = 5) and hSOD1^G93A-UeGFP mice (n = 7) were treated with 20 mg/kg/day dosage of NU-9, and WT-UeGFP (n = 11), hSOD1^G93A-UeGFP (n = 9), and prpTDP-43^A315T-UeGFP mice (n = 4) were treated with 100 mg/kg/day dosage of NU-9. prpTDP-43^A315T-UeGFP mice (n = 3), which received no treatment, were used as controls for the TDP-43^A315T group (Table S2). Same animals were used for both immunofluorescence and EM analysis whenever possible.

2.4 Behavioral analyses

Behavior data were collected from hSOD1^G93A-UeGFP and WT-UeGFP mice (starting at P60 and every 7 days until P116) and from prpTDP-43^A315T-UeGFP mice (at P60, P90, and P120) with the following number of mice for each group: WT-UeGFP vehicle (n = 10), hSOD1^G93A-UeGFP vehicle (n = 6), WT-UeGFP 20 mg/kg/day NU-9 (n = 5), hSOD1^G93A-UeGFP 20 mg/kg/day NU-9 (n = 7), WT-UeGFP 100 mg/kg/day NU-9 (n = 11), hSOD1^G93A-UeGFP 100 mg/kg/day NU-9 (n = 9), prpTDP-43^A315T-UeGFP untreated (n ≥ 3), and prpTDP-43^A315T-UeGFP 100 mg/kg/day NU-9 (n = 4) (Table S3).

2.5 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.

2.6 Immunocytochemistry

The antibodies used are as follows: anti-GFP (1:1000, Invitrogen; or 1:1000, Abcam) and anti-misfolded SOD1 (B8H10, 1:250, Médimabs). Briefly, sections were treated with blocking solution (PBS, 0.05% BSA, 2% FBS, 1% Triton X-100, and 0.1% saponin) for 30 min at room temperature and incubated with primary antibody diluted in blocking solution overnight at 4°C. Appropriate secondary fluorescent antibodies (1:500, AlexaFluor-conjugated, Invitrogen) were added to the blocking solution at room temperature for 2 h in the dark.

2.7 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 50-µm 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.

2.8 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.

2.10 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% H₂O₂–10% 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 (OsO₄) (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-mm-wide × 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). Seventy-nanometer-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.

2.11 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.

2.12 Statistical analysis

All analyses were performed using Prism software (GraphPad 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. Two-way 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 multiple-comparison 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.

3.1 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).

Prior to processing for EM, mouse tissue sections were subjected to immunostaining for Ctip2, a marker for UMN^{1, 36, 57, 61} to distinguish UMN in layer 5 of the mouse motor cortex (Figure S1). UMN in human motor cortex were determined by their large soma size and deep layer 5 location, as previously described.^{8, 34} UMNs of normal control cases (n = 4) showed intact cell body with preserved cellular organelles (Figure 1A). Similarly, UMN of WT healthy mice showed well-preserved soma and nucleus (Figure 1C). However, UMNs of ALS patients (n = 9) displayed massive ultrastructural defects especially at the site of mitochondria and ER (Figure 1B). The UMNs of both hSOD1^G93A (Figure 1D) and prpTDP-43^A315T mice (Figure 1E) had similar problems.³⁴

The ER in UMNs of normal controls and WT mice were well preserved with intact cisternae, adorned with ribosomes (Figure 1K, M arrows). Since smooth ER is very thin and difficult to identify accurately in tissue sections processed for immunoelectromicroscopic analysis, only rough ER with ribosomes were included in analyses. However, ER in UMNs of ALS patients were distended, swollen, and cisternae were damaged (Figure 1L arrowheads; normal control: 15% ± 3% ER cisternae/UMN/section; ALS: 88% ± 1% ER cisternae/UMN/section; p < .0001, Figure 1P). Same ultrastructural ER defects were also detected in the UMN of hSOD1^G93A (Figure 1N arrowheads; WT: 2% ± 1% cisternae/UMN/section; hSOD1^G93A: 44% ± 3% ER cisternae/UMN/section; adjusted p-value = .0001, Figure 1Q) and prpTDP-43^A315T mice (Figure 1O arrowheads; Q; WT: 2% ± 1% ER cisternae/UMN/section; prpTDP-43^A315T: 36% ± 2% ER cisternae/UMN/section; adjusted p-value = .0001). The average length of individual cisternae of ER was significantly shorter in UMNs of ALS cases compared to normal control (normal control: 1.7 ± 0.15 µm; ALS: 0.56 ± 0.03 µm; p < .001, Figure 1R). Similarly, as compared to WT, the average length of individual cisternae of ER was shorter in UMN of hSOD1^G93A (WT: 1.56 ± 0.07 µm; hSOD1^G93A: 0.50 ± 0.02 µm; adjusted p-value = .0001, Figure 1S), and prpTDP-43^A315T mice (WT: 1.56 ± 0.07 µm; prpTDP-43^A315T: 0.53% ± 0.01%; adjusted p-value = .0001, Figure 1S).

3.2 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^G93Ainduced protein aggregation. Hits from three families of compounds were filtered by a ligand-based computational approach, including substructure and similarity searching⁶² 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 cell-based assay.⁵² Multiple rounds of optimization were carried out, resulting in compound 2 (Figure 2A), which also had excellentin 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 properties⁵⁴ (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.

3.3 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 mice⁵⁸ 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 arrowhead), 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).

We next performed quantitative analyses to investigate whether these improvements were widely observed in diseased UMNs treated with NU-9 (Table S4). The total number of mitochondria in UMNs of hSOD^G93A-UeGFP mice significantly increased upon 100 mg/kg/day NU-9 treatment (21 ± 2 mitochondria/UMN/section; adjusted p-value = .001) when compared to UMNs treated with vehicle (9 ± 2 mitochondria/UMN/section). The total number of mitochondria after NU-9 treatment were comparable to WT-UeGFP mice (30 ± 1 mitochondria/UMN/section; adjusted p-value = .07). Furthermore, NU-9 treatment significantly increased the percentages of healthy mitochondria in UMNs of hSOD^G93A-UeGFP mice (WT-UeGFP: 93% ± 1%; vehicle: 10% ± 1%; NU-9 100 mg/kg/day: 86% ± 3%; adjusted p-value = .0001; Figure 2O,P).

NU-9 treatment also significantly improved the integrity of ER in diseased UMNs; there were significantly more healthy cisternae (WT-UeGFP: 31 ± 1; hSOD^G93A-UeGFP vehicle: 4 ± 1 ER cisternae/UMN/section; hSOD^G93A-UeGFP 100 mg/kg/day NU-9: 14 ± 2 ER cisternae/UMN/section, adjusted p-value = .008; Figure 2Q). In addition, the average length of ER cisternae of hSOD^G93A-UeGFP UMN treated with NU-9 (1.13 ± 0.1 µm) became comparable to the length of ER cisternae of WT-UeGFP UMN (1.56 ± 0.07 µm; adjusted p-value = .014), whereas diseased UMN had significantly shorter ER cisternae (0.50 ± 0.02 µm; adjusted p-value = .0001; Figure 2R; please refer to Table S4 for a complete list of total numbers of UMNs investigated and the total number of mice used for each quantitative analysis).

3.4 NU-9 treatment reduces mSOD1 in UMNs of hSOD1^G93A-UeGFP mice

NU-9 was identified on the basis of its ability to reduce misfolded mSOD1 in PC12 cells.^{55, 56} Since UMNs of hSOD1^G93A-UeGFP mice contained misfolded SOD1,⁶⁴ we next investigated whether NU-9 treatment would reduce levels of mSOD1 in diseased UMNs. To investigate the presence of misfolded SOD1 protein, we used the well-characterized B8H10 monoclonal antibody that can detect a wide spectrum of SOD1 mutants and metal-depleted WT SOD1 protein, but not intact WT SOD1.⁶⁵ UMNs in WT-UeGFP mice do not have misfolded SOD1, as expected (Figure 3A, Figures S2 and S3). However, UMNs of hSOD1^G93A-UeGFP mice and UMNs of hSOD1^G93A-UeGFP mice treated with vehicle have high levels of misfolded SOD1 (1.66 × 10⁵ ± 0.15 × 10⁵ arbitrary units [au]; Figure 3B,E). NU-9 treatment significantly reduces levels of mSOD1, especially in diseased UMNs (20 mg/kg/day: 1.33 × 10⁵ ± 0.11 × 10⁵ au; 100 mg/kg/day: 1.2 × 10⁵ ± 0.07 × 10⁵ au; adjusted p-value = .007; Figure 3C–E).

3.5 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 p-value = .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 integrity of apical dendrites in the control WT-UeGFP mice (adjusted p-value = .9956).

3.6 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 mice⁵⁷ 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).

3.7 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 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 distinct and different disease models. To visualize diseased UMNs and to assess their cellular response to compound treatment, TDP-43^A315T mice⁵⁸ were crossed with UCHL1-eGFP mice⁵⁷ 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-UeGFP mice (Figure S3), and there is no TDP-43 pathology reported in SOD1 mouse models or patients with SOD1 mutations,^46-51 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).

prpTDP-43^A315T-UeGFP mice (n = 4) were treated with a 100 mg/kg/day dose of NU-9, and sex- and age-matched untreated prpTDP-43^A315T-UeGFP mice (n = 3) were used as the negative control. As both hSOD1^G93A and prpTDP-43^A315T mice were mated with the same WT-UeGFP mouse colony to generate hSOD1^G93A-UeGFP and prpTDP-43^A315T-UeGFP mice, same WT-UeGFP cohort was used as healthy control for both groups.

NU-9 treatment resulted in profound improvements in both mitochondria and ER of UMNs that become diseased as a result of TDP-43 pathology (Figure 6D–F). Mitochondria, especially the inner membranes of mitochondria, became intact (Figure 6E arrow), and there were no signs of mitoautophagy³⁸ or mitophagy. The ER retained its structure with ribosomes attached, and there was no enlargement or disintegration of cisternae (Figure 6F arrows). Quantitative analysis confirmed significant increase in the numbers of total mitochondria per UMN per section after 100 mg/kg/day NU-9 treatment (32 ± 3 mitochondria/UMN/section), when compared to mitochondria in diseased UMN (prpTDP-43^A315T-UeGFP: 20 ± 1 mitochondria/UMN/section; p < .033; Figure 6G). Interestingly, the number of mitochondria after NU-9 treatment in prpTDP-43^A315T-UeGFP mice became comparable to that of WT mice (30 ± 1 mitochondria/UMN/section; adjusted p-value = .76). The average percentage of healthy mitochondria also significantly increased with NU-9 treatment (87% ± 1% mitochondria/UMN/section) when compared to diseased UMNs (prpTDP-43^A315T-UeGFP: 22% ± 2% mitochondria/UMN/section; adjusted p-value = .0001; Figure 6H) and were comparable to the average percentage of healthy mitochondria in WT-UeGFP mice (93 ± 1 mitochondria/UMN/section).

NU-9 treatment also increased the average number of intact ER cisternae in UMNs that become diseased as a result of TDP-43 pathology (WT-UeGFP: 31 ± 1 ER cisternae/UMN/section; prpTDP-43^A315T-UeGFP: 12 ± 2 ER cisternae/UMN/section; prpTDP-43^A315T-UeGFP [100 mg/kg/day NU-9]: 36 ± 2 ER cisternae/UMN/section, n = 4 mice, adjusted p-value = .0002; Figure 6I). The average length of ER cisternae became also significantly longer in the UMN treated with NU-9 (WT-UeGFP: 1.56 ± 0.07 µm; prpTDP-43^A315T-UeGFP: 0.53% ± 0.01%; adjusted p-value = .0001; prpTDP-43^A315T-UeGFP treated with NU-9: 1.37% ± 0.05%; adjusted p-value = .0001; Figure 6J).

3.8 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 pathologyin vivo.

Health and integrity of apical dendrites in prpTDP-43^A315T-UeGFP mice displayed profound improvement with NU-9 treatment as the percentage of UMNs with vacuolated primary apical dendrites was significantly reduced (untreated: 81% ± 5%; NU-9 [100 mg/kg/day]: 11% ± 1%; adjusted p-value = .0005; Figure 7A–E). Most interestingly, the average number of UMNs in the motor cortex of prpTDP-43^A315T-UeGFP mice treated with NU-9 experienced a profound increase compared to that of diseased untreated prpTDP-43^A315T–UeGFP mice (untreated: 37 ± 4 UMNs; NU-9 [100 mg/kg/day]: 59 ± 3 UMNs; adjusted p-value = .0121; Figure 7F–H). Same WT-UeGFP cohort was used as healthy control for both hSOD1^G93A-UeGFP and prpTDP-43^A315T-UeGFP mice, as previously mentioned. The UMN numbers with NU-9 treatment were comparable and almost identical to those of the healthy control mice treated with vehicle (WT-UeGFP mice 59 ± 4 UMNs; adjusted p-value > .9999), revealing the ability of NU-9 to eliminate the ongoing degeneration of UMNs that become diseased due to TDP-43 pathology.

3.9 Effect of NU-9 on lower motor neurons (LMNs)

In an effort to determine whether NU-9 treatment also improves the health and survival of lower motor neurons (LMNs), we investigated the lumbar spinal cords of both hSOD1^G93A-UeGFP and prpTDP-43^A315T-UeGFP mice (Figure 8). As indicated by previous reports,^{58, 66} there was no prominent LMN loss in the spinal cords of prpTDP-43^A315T-UeGFP mice, even at P120 (Figure 8A,B), and thus investigation of NU-9 treatment on the survival of LMNs was not possible. However, as extensively reported in the field,^{59, 67-69} there was a dramatic reduction in the numbers of LMNs in hSOD1^G93A-UeGFP mice (17.1 ± 1.3 LMN) when compared to the healthy controls (52.8 ± 3 LMN; adjusted p-value < .0001; Figure 8C,D). We quantitatively assessed the changes in the numbers of LMNs in the lumbar spinal cord of mice that are treated with vehicle, NU-9 (20 or 100 mg/kg/day), and control healthy mice (Figure 8E,F). NU-9 treatment, regardless of dose, was not sufficient to eliminate ongoing LMN degeneration in hSOD1^G93A-UeGFP mice. There was no difference in LMN numbers between vehicle and treated (NU-9 [20 mg/kg/day]: 18.6 ± 0.4 LMN; vs. vehicle adjusted p-value = .9526; NU-9 [100 mg/kg/day]: 14.1 ± 1.3 LMN; vs. vehicle adjusted p-value = .7255) and untreated mice at P120 (Figure 8E,F; Table S5).

As the UCHL1-eGFP reporter selectively labels small diameter alpha and gamma motor neurons that are not vulnerable to degeneration in ALS,^{57, 68-71} we investigated both the NeuN⁺/ChAT⁺ LMN subpopulation vulnerable to degeneration and GFP⁺/ChAT⁺ LMN subpopulation resistant to degeneration. There was a dramatic reduction in the numbers of NeuN⁺/ChAT⁺ LMNs in the diseased mice (10.3 ± 0.9 LMN) when compared to the healthy controls (46.8 ± 2.6 LMN; adjusted p-value < .0001; Figure 8F). However, NU-9 treatment, regardless of dose, was not sufficient to eliminate LMN degeneration in hSOD1^G93A-UeGFP mice; there was no statistical difference in NeuN⁺/ChAT⁺ LMN numbers between treated (NU-9 [20 mg/kg/day]: 12.1 ± 0.4 LMN; vs. vehicle: 10.3 ± 0.9; adjusted p-value = .8524; NU-9 [100 mg/kg/day]: 9.7 ± 0.9 LMN; vs. vehicle: 10.3 ± 0.9; adjusted p-value = .995; Figure 8H) and untreated mice at P120. The numbers of GFP⁺ and ChAT⁺ LMN were comparable, regardless of genotype or treatment (Figure 8I, Table S5), revealing neuroprotective effects of NU-9 to be selective for UMNs.

3.10 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).

prpTDP-43^A315T mice performed worse than WT littermates on rotarod and hanging wire tests.^{34, 75, 76} However, when treated with a 100 mg/kg/day dose of NU-9, they performed significantly better on the hanging wire test, so much so that they became comparable to healthy WT mice at P120 (untreated prpTDP-43^A315T-UeGFP: 46 ± 7 s; prpTDP-43^A315T treated with NU-9: 59.5 ± 0.3 s; adjusted p-value < .0001; WT-UeGFP [vehicle]: 60 ± 0 s; vs. untreated prpTDP-43^A315T-UeGFP adjusted p-value < .0001 WT vehicle vs. prpTDP-43^A315T-UeGFP treated with NU-9 [100 mg/kg/day] adjusted p-value = .9688; Figure 9E, Table S3).

Different from hSOD1^G93A mouse model, even the rotarod test revealed significant improvement in TDP-43 model only after 30 days of NU-9 treatment (untreated prpTDP-43^A315T-UeGFP: 153.1 ± 4.8 s; prpTDP-43^A315T-UeGFP treated with NU-9 [100 mg/kg/day]: 232.6 ± 18.1 s; adjusted p-value = .0482; WT-UeGFP [vehicle]: 283.9 ± 8.9 s; vs. untreated prpTDP-43^A315T-UeGFP; adjusted p-value < .0001 WT vehicle vs. prpTDP-43^A315T-UeGFP treated with NU-9 [100 mg/kg/day] adjusted p-value = .0482; Figure 9D; please refer to Table S3 for all values and statistical analysis at each time point).