The triple transgenic (3xTgAD) mice used in this study have been previously described (Oddo et al., 2003a,b) and are maintained on a B6/129 background. This mouse model has been genetically engineered to contain homozygous mutations in three human genes associated with AD (PS1, APP, and Tau). The 3xTg-AD mice express the mutant human Tau protein, the human APP with the Swedish mutation and mutant presenilin. The mice were generated using two mutant transgenes (encoding APP_swe and Tau_P301L driven by the Thy1.2 promoter) microinjected into single-cell embryos harvested from PS1_M146VKI mice (Oddo et al., 2003b). The 3xTg-AD mice progressively develop Aβ and Tau pathology, with a temporal- and regional-specific profile that closely mimics their development in the human AD brain, including progressive impairments in learning starting as early as 3 to 5 months, spatial memory deficits at 6 months, and recognition memory deficits at 9 to 11 months of age. Despite equivalent expression of the human APP and human Tau transgenes, Aβ deposition develops prior to the tangle pathology. Intracellular Aβ immunoreactivity is first detected around 4–6 months of age; extracellular Aβ deposits start to manifest around 12 months of age and as the mice age, plaques can be easily detected throughout the brain (Oddo et al., 2003a,b, 2006, 2007). Accumulation of intracellular Aβ is the first step of the events followed by Tau protein hyperphosphorylation, extracellular deposition of Aβ protein, and appearance of paired intracellular neurofilaments from hyperphosphorylated Tau (Oddo et al., 2003a; Mastrangelo and Bowers, 2008). As the mice age, Tau phosphorylation becomes more severe and can be detected by phospho-specific markers, such as AT8 and AT180 (Oddo et al., 2003b,2007). All mice were caged in a pathogen-free environment with free access to standard chow and water and maintained on a 12 h light/dark cycle. Measurements were done using young (2 to 4 months) and old (17–20 months) mice unless otherwise stated. The Institutional Animal Care and Use Committee at Oklahoma Medical Research Foundation (Oklahoma City, OK, United States) approved all procedures.

Isometric contractile force generation was measured in situ in gastrocnemius (GTN) muscle based on the methods described previously (Larkin et al., 2011). Briefly, mice were anesthetized with isoflurane delivered by oxygen, and the whole GTN muscle was isolated and cleaned from surrounding muscles and connective tissues. After dissection, the isolated muscle was tied with a silk suture on the distal tendon, and the tendon was severed and mounted onto the force transducer (model 305B, Aurora Scientific). The mouse was placed on a temperature-controlled platform at 37°C and provided with continuous anesthesia. The electrode was placed on the surface of the GTN muscle directly, and the muscle optimal length was adjusted with single 0.2 ms stimulation pulses until a maximum twitch was reached. At the muscle optimal length, a series of 300 ms stimulus pulses were applied to achieve the maximum isometric tetanic force. After muscle stimulation, the electrode was moved from the muscle to the sciatic nerve, and the nerve filament was hooked firmly by the electrode. The same pulses of the tetanic stimulus were applied to the nerve to achieve the nerve-induced maximum isometric tetanic force. All of the above tetanic twitch protocols were repeated several times to confirm the reproducibility and reliability of the data. Comparing muscle and nerve stimulated contractile force generation allows us to measure the extent of denervation or loss of contractile force generation due to loss of intact NMJ innervation.

After all force measurements were completed, muscles were carefully removed and weighed, and the maximum tetanic force was normalized to the muscle cross-sectional area (CSA) calculated by the length and weight of the GTN muscle (dividing the muscle mass, mg, by the optimal length, mm, and the density of mammalian skeletal muscle, 1.06 g/cm³) to give the specific force (N/cm²). The NMJ function was presented by normalizing the nerve-induced force to the muscle-induced force as a percentage.

Mitochondrial function was measured in permeabilized muscle fibers. The permeabilization protocol has been previously described by our laboratory (Ahn et al., 2019). In general, a small piece of the red gastrocnemius muscle was excised from the body and finely dissected to separate the muscle fibers along their striations in cold buffer X containing (in mM): 7.23 K₂EGTA, 2.77 CaK₂EGTA, 20 imidazole, 0.5 DTT, 20 taurine, 5.7 ATP, 14.3 PCr, 6.56 MgCl₂ -6H₂O, and 50 K-MES (pH 7.1). A total of 30 μg/ml saponin was added to the fibers to induce permeabilization for 30 min, followed by 5 min washes for 3 times in washing buffer containing (in mM): 105 K-MES, 30 KCl, 10 K₂HPO₄, 5 MgCl₂ -6H₂O, 0.5 mg/ml BSA, 0.1 EGTA (pH 7.1). After washing, the permeabilized fibers were placed into the Oxygraphy-2K (O2k, OROBOROS INSTRUMENTS, Austria) following the protocols described before (Krumschnabel et al., 2015). The oxygen consumption rate (OCR) was determined using the oxygen probe, while the ROS production rate was measured by the O2K-Fluo LED2-Module Fluorescence-Sensor Green with Amplex UltraRed Reagent (Invitrogen, A36006). Substrates used for different complexes added sequentially as follows, leak state: 10 mM glutamate and 2 mM malate with no ADP; complex I: addition of 2.5 mM ADP; complex I + II: addition of 10 mM succinate; complex II: addition of 0.5 μM rotenone; complex IV: addition of 2 mM ascorbate and 0.5 mM TMPD added after 5 μM antimycin A. An H₂O₂ standard curve was measured each time before the actual experiments to calibrate the level of H₂O₂. All data generated from O2K was normalized to the muscle wet weights and analyzed using the official O2K software, Datalab Version 7.0.

Skeletal muscle contractile proteins (myosin heavy chain (MHC) and actin) composition was determined by SDS-PAGE gel. In this study, gastrocnemius muscle samples were homogenized and denatured as described previously (Xu et al., 2017, 2018). Then the denatured whole muscle sample was separated with a 10% SDS-PAGE gel. Then the gel underwent electrophoresis at 200 V at room temperature for 1 h, and immediately after running, Coomassie brilliant blue G250 was used to stain the gel and visualize the MHC bands. The images were collected by G:BOX Chemi (Syngene, United States), and the densitometry analysis was done for the ratio between MHC and actin, and also the proportion of MHC and action together to the total proteins. The densitometry analysis was conducted using ImageJ software (ImageJ, Fiji).

After the muscle was excised from the body, small muscle pieces were taken under the cold PBS in a petri dish along with the fiber directions with careful removal of fat and connective tissues. Muscle samples were transferred into a 24-well plate with 10% STUmol in deionized water (Poly Scientific R&D, #2832) for 1 h to fix the tissue by gently shaking. After the fixation, tissues were washed 3 times for 5 min in PBS at room temperature, then permeabilized tissue under the 2% Triton in PBS for 30 min on a shaker. After permeabilization, the tissues are placed into the blocking buffer containing 4% BSA, 1% Triton, and 5% serum, matching the host of secondary antibody diluted in PBS, blocking overnight in the cold room at 4°C. After blocking, we added primary antibodies, 1:50 SV2 (DSHB) for nerve terminals, and 1:50 2H3 (DSHB) for neurofilaments. The tissues were incubated overnight at 4°C, washed 6 times for 30 min in PBS at room temperature, and subsequently incubated with the secondary antibodies, 1:1,000 BTX-Alexa 488 (Invitrogen, #B13422), and 1:250 goat anti-mouse Cy3. After the incubation for secondary antibodies overnight at 4°C, all tissues were then washed 6 times for 30 min with PBS. Afterward, tissues were transferred onto slides and mounted with a mounting medium, and sealed with nail polish. Images of NMJ were taken by the Nikon confocal microscope under the magnification of 20×, and Z-stacks were taken to show the 3D structure of intact NMJ. The total thickness of optical sections is around 20 to 60 μm, and the stack interval was set at every 1 to 2 μm, so there are around 20 to 30 images per stack.

The NMJ area was analyzed as the area occupied by each individual labeled acetylcholine receptor (ACHR), and only the NMJs facing forward were analyzed. The fragmentation level was quantified by counting the pieces of each ACHR, and if there are five or more pieces per junction, the NMJ is considered a fragmented NMJ. The denervation score was quantified as follows: score 0: there is no denervation, the ACHRs fully overlap with nerve terminals; score 1: partial denervation, the ACHR partially overlaps with nerve terminals; score 2: complete denervation, there is little or no overlap of ACHR and nerve terminals. The statistical analysis for all images taken in this section was done by ImageJ software (ImageJ).

qRT-PCR was performed as described before (Bhaskaran et al., 2020). In brief, total RNA was isolated from 25 mg gastrocnemius muscle using Trizol, and cDNA was synthesized using the iScript cDNA kit (Bio-Rad, United States). qRT-PCR was performed using Power SYBR green PCR mix. The following PCR primers were used: TGFβ (forward- CCCTATATTTGGAGCCTGGA), TGFβ (reverse- CTTGCGACCCACGTAGTAGA), 18S rRNA (forward- GTGGAGCGATTTGTCTGGTT), 18S rRNA (reverse- CGCTGAGCCAGTCAGTGTAG). Calculations were performed by the comparative (2^ΔΔCt) method using 18S rRNA.

Gastrocnemius muscles were homogenized in RIPA buffer containing 50 mM Tris (pH 7.4), 150 mM NaCl, and protease inhibitors. Then the total protein was quantified with the Bio-Rad protein assay kit (Sigma-Aldrich, Poole, United Kingdom), and the same amount of protein was loaded and separated with SDS-PAGE gels at certain percentages, i.e., 10 or 12%. The gel was then run at 200 V for 1 h and wet-transferred onto 0.45 μm nitrocellulose membranes (Bio-Rad) with the conditions of 100 V, 30 min at 4°C, same as described before (Murphy et al., 2012). After the transfer of the gel, total proteins in each lane were quantified using ponceau staining (Sigma, #P3504), then the membrane was washed with ddH₂O to remove the ponceau staining and blocked with 1% BSA solution in TBST for at least 1 h at room temperature. Soon after blocking, primary antibodies were added onto the membrane and incubated overnight at 4°C. After the incubation of the primary antibody, the membrane was washed with blocking buffer and then exposed to the secondary antibody for 30–60 min. After the secondary antibody, the membrane was washed with TBST for the last time to clean the background. Protein bands were visualized and quantified using the Gene tool system (SynGene-Frederick, MD, United States). The relative content of each protein measured using western blot analysis was normalized to sample total protein content measured using ponceau stain and densitometry of total ponceau in that sample lane. Primary antibody information, rabbit anti-Amyloid Precursor Protein (APP) (Abcam, ab32136), rabbit anti Tau (Abcam, ab76128), chicken anti-TGF-β1/1.2 (R&D systems, AF-101-NA), rabbit anti-Smad 2/3 (Cell Signaling, #3102), rabbit anti-phospho-Smad2/3 (Cell Signaling, #8828), Smad2/3 pure protein (cell extract) (Cell Signaling, control: #26725, phosphorylated: 47986), mouse anti MuRF1 (Santa Cruz, sc-398608 HRP).

Samples were obtained from the brain, gastrocnemius muscle, and spinal cord and mechanically homogenized in 7.5 volumes of ice-cold TBS with phosphatase and protease inhibitors. Following centrifugation at 55,000 rpm for 1 h at 4°C, the supernatant (TBS extract) was collected. The pellet was then resuspended and mechanically homogenized in 7.5 volumes of TBS with 1% Triton X-100 containing protease and phosphatase inhibitors. Following centrifugation at 55,000 rpm for 1 h @ 4°C, the supernatant (TBS-T extract) was collected. The pellet was then resuspended in 2 volumes of 88% formic acid and sonicated at 4°C for 1 min and gently agitated on a shaking platform at 4°C overnight. The formic acid extract is sonicated at 4°C for 1 min, clarified by spinning for 10 min at 1,800g at 4°C, and neutralized by adding non-buffered Tris 1 M, pH11.

Aβ42 and Aβ40 concentrations in TBS, TBS-T, and formic acid extracts were measured by multi-plex electrochemiluminescence immunoassay using Aβ peptide panel 1 (6E10) kit (Cat # K15200E-2) according to manufacturer’s instructions (Mesoscale Discovery). The kit was supplied with peptides (Aβ42 and Aβ40), diluent 35 and 100, SULFO-TAG Anti-Aβ 6E10 antibody, read buffer T, and MSD plates. The samples were thawed on ice and diluted 2 times in diluent 35. Eight standards were prepared by dissolving 10 μl of each peptide in 370 μl of diluent 35 and four-fold serial dilutions. Diluent 35 (150 μl) was added to each well of the MSD plate and incubated in a platform shaker at 700 rpm and at room temperature for 1 h. All subsequent incubation steps were carried out similarly. The antibody solution was prepared with 60 μl of SULFO-TAG Anti-Aβ 6E10 antibody in 2,940 μl of Diluent 100. After incubation, wells were washed thrice with PBS-0.05% tween solution followed by the addition of 25 μl of antibody solution. A 25 μl of prepared samples and standards were added to the wells and incubated for 2 h. Read buffer was prepared by diluting read buffer T 2 times in deionized water. Wells were washed thrice with PBS-0.05% tween solution followed by adding 150 μl of the read buffer. The plates were read immediately after the addition of reading buffer with MSD Quickplex SQ 120.

All results are presented as mean values ± standard error of the mean (SEM), and comparisons among different groups were performed with one-way ANOVA and Tukey’s multiple comparison test. The statistical analysis was completed using GraphPad Prism 8, and the statistical significance was set at p values less than 0.05.

To characterize the expression of APP and Tau transgenes in neural and muscle tissue of young and old control and 3xTgAD female mice we measured APP and Tau protein in homogenates from the brain, spinal cord, sciatic nerve, and gastrocnemius muscle by western blotting (Supplementary Figure 1). While APP expression was elevated in brain tissue of both young and old 3xTgAD compared to controls, APP expression in the spinal cord, sciatic nerve, and gastrocnemius muscle was increased only in old 3xTgAD mice compared to controls (Supplementary Figures 1B,D,F,H). Tau expression was elevated in the brain, spinal cord, and sciatic nerve tissue from both young and old 3xTgAD female mice compared to wild-type mice (Supplementary Figures 1A,C,E). Tau expression was not different between groups in gastrocnemius muscle tissue. We also measured APP and Tau expression in male 3xTgAD mice and age-matched control mice, and although we had a small sample size (n = 3) we did not observe a significant increase in the amount of these two proteins in male 3xTgAD mice, in agreement with other reports (Mendell et al., 2020).

To determine whether soluble and insoluble forms of Aβ accumulate in neural and muscle tissue of 3xTgAD mice, we performed multiplex immunoassays (meso-scale discovery) to measure Aβ40 and Aβ42 from TBS, TBS-T, and formic acid-soluble fractions from the brain, spinal cord, and gastrocnemius muscle of old and young 3xTgAD mice (and old wildtype as control) (Figure 1). As expected, no human Aβ40 and Aβ42 were detected in any of the fractions of wild-type mice. In the 3xTgAD mouse brain, Aβ40 and Aβ42 both increased with age, particularly insoluble Aβ found in formic acid-soluble fractions. Strikingly, Aβ40 and Aβ42 were detectable in the TBS soluble fraction of both spinal cord and gastrocnemius muscle of old female 3xTgAD mice, and in gastrocnemius muscle their amounts were significantly elevated compared to young female 3xTgAD mice (Figure 1). The large variation observed in the Aβ amount might be due to normal biologic variability.

We measured the mass of the gastrocnemius and quadriceps muscle (normalized to body weight) in the hindlimb of young and old wildtype and 3xTgAD mice. Body mass is increased with age in both genotypes but was not different between wildtype and 3xTgAD mice (22.4 ± 0.6 vs. 19.4 ± 0.4 g for young female wildtype vs. 3xTgAD mice and 31.8 ± 3.6 vs. 40.0 ± 1.3 g in the old control and 3xTgAD mice). The normalized muscle weight was significantly decreased in gastrocnemius muscles from both young (15%) and old (26%) female 3xTgAD mice compared to age-matched wildtype control mice (Figure 2A). Quadriceps muscle mass was not reduced in the older wildtype mice but was significantly lower in older 3xTgAD mice (Figure 2B).

To determine whether the reduction in mass was also accompanied by a loss in function, we measured the muscle maximum specific force and the function of the neuromuscular junction (NMJ) in young and old female mice using in situ electrical stimulation of the sciatic nerve. Maximum specific force production was not different between wildtype and 3xTgAD female mice (Figure 3A). The integrity of NMJ function was determined by normalizing the nerve-induced force when stimulated to the muscle-induced force as a proportion (an intact normal NMJ should be able to trigger a nerve-induced force close to 100% of the muscle-induced force). The NMJ function in old 3xTgAD mice was significantly reduced by 17% of the nerve-induced force when compared with wild-type mice (Figure 3B). NMJ function in young 3xTgAD mice was preserved with no significant reduction in nerve-induced force production (Figure 3B). Together these data indicate that muscle strength was not affected by AD phenotypes, but NMJ function was impaired in older female 3xTgAD mice compared to wild-type mice.

To directly assess the morphology of the NMJ, neurofilaments (red) and acetylcholine receptors (ACHRs, green) were stained and examined by confocal microscopy, and NMJ area, degree of fragmentation, and denervation score were measured. NMJ morphology was altered, with significant denervation and fragmentation, in the old female 3xTgAD mice compared to age-matched controls (Figure 4A). The percentage of fragmented NMJs is 30% in old female 3xTgAD mice, which is dramatically higher than its age-matched wildtype (10%) (Figure 4C), with the majority of NMJ being partially or completely denervated (Figure 4D). In contrast, the fragmentation and denervation were similar in wild-type mice and the young 3xTgAD mice (Figures 4C,D). In addition, the measurement of NMJ area did not show a difference between old wildtype and 3xTgAD (Figure 4B), however, in the young mice, this area was reduced in 3xTgAD mice (602 μm²) compared to wildtype mice (700 μm²) (Figure 4B). These data show that the NMJ dysfunction in the old female 3xTgAD muscles is related to NMJ morphological alterations. To determine whether the NMJ phenotype is related to changes in motor neuron decline in the spinal cord, we measured the number of alpha-motor neurons in the lumbar region of the spinal cord sections using the NeuN antibody (Figure 4E). There was no change in the number of alpha motor neurons (Figure 4F).

It is well known that loss of innervation and NMJ dysfunction induce elevated mitochondrial dysfunction and generation of mitochondrial peroxides in muscle (Muller et al., 2007; Arthur et al., 2008; Pharaoh et al., 2020). To determine whether the changes in NMJ function and morphology observed in old female 3xTgAD mice altered muscle mitochondrial function, we measured oxygen consumption rate (OCR) and ROS generation as peroxide emission in permeabilized muscle fibers from gastrocnemius muscle from old female wildtype and 3xTgAD mice. Using respiratory substrates to stimulate electron flow through different electron transport chain complexes, we found a decrease in OCR in transgenic compared to wildtype mice using glutamate/malate, complex I linked substrates but not in response to complex II using succinate rotenone (Figure 5A). There was no difference in State 1 (leak state, mitochondria respiring without the addition of external substrate) or complex I (mitochondria respiring with the addition of glutamate and malate) ROS production rate (Figure 5B). Overall, these data suggest that mitochondrial function may be mildly impaired in the old female 3xTgAD mice but not at a level that induces elevated mitochondrial ROS generation.

To determine whether there are differences in the contractile protein composition of the muscle, we measured the abundance of contractile proteins, myosin and actin, relative to total muscle protein in whole muscle homogenates from gastrocnemius muscle by using SDS-PAGE gel (Figure 6A). When normalized to total protein, the relative proportion of contractile proteins was not different in young or old 3xTgAD female mice compared to their age-matched wildtype mice (Figure 6B). Moreover, there is no difference between the ratio of actin to myosin, in young or old 3xTgAD female mice compared to age-matched wild-type mice (Figure 6C). These data show that the basic contractile components of skeletal muscle were not affected even in the old female 3xTgAD mice.

It has previously been reported that the amount of TGF- β is elevated in neural tissues with AD (Kinney et al., 2018). In skeletal muscle, TGF-β downstream proteins such as Smad and MuRF1 are closely associated with muscle atrophy (Sartori et al., 2009; Bollinger et al., 2014; Kinney et al., 2018). In order to determine if the TGF-β associated muscle atrophy pathway is altered in 3xTgAD mice, we measured the level of TGF-β transcription in gastrocnemius muscle from young and old female wildtype and 3xTgAD mice. TGF-β mRNA is elevated in old female 3xTgAD mice compared to age-matched control and young mice (Figure 7A). Consistent with mRNA level, a similar increase in TGF-β protein levels (∼30%) was observed in muscle from old female 3xTgAD vs. wildtype mice, with no change detected in young mice (Figure 7B).

Smad 2/3 is a protein involved in TGF-β signaling that is activated by phosphorylation and is associated with muscle atrophy (Sartori et al., 2009). Phosphorylation of Smad 2/3 in gastrocnemius muscle was undetectable by western blotting in old female wildtype and both wildtype and 3xTgAD in young mice but was apparent in 3xTgAD old mice (Figure 7C). MuRF1 is a downstream effector protein of Smad 2/3 (Bollinger et al., 2014). MuRF1 levels were significantly increased in old female 3xTgAD mice (∼25%) but were unchanged in young female 3xTgAD mice compared to controls (Figure 7D). Together, these data confirmed that the TGF- β associated muscle atrophy pathway is activated in old female 3xTgAD muscle, and is a potential effector of the decreased muscle mass seen in the gastrocnemius muscle in old female 3xTgAD mice. We also measured the TGF-β signaling pathway in male 3xTgAD and control mice, and both mRNA and protein levels of TGF-β are not different between 3xTgAD and wildtype mice, and Smad 2/3 phosphorylation is not elevated. The level of MurF1 protein is also the same in male 3xTgAD and wild-type mice (Supplementary Figure 4).