Hindlimb immobilization induces insulin resistance and elevates mitochondrial ROS production in the hippocampus of female rats

Abstract

Alzheimer’s disease (AD) is the fifth leading cause of death in older adults, and treatment options are severely lacking. Recent findings demonstrate a strong relationship between skeletal muscle and cognitive function, with evidence supporting that muscle quality and cognitive function are positively correlated in older adults. Conversely, decreased muscle function is associated with a threefold increased risk of cognitive decline. Based on these observations, the purpose of this study was to investigate the negative effects of muscle disuse [via a model of hindlimb immobilization (HLI)] on hippocampal insulin sensitivity and mitochondrial function and identify the potential mechanisms involved. HLI for 10 days in 4-mo-old female Wistar rats resulted in the following novel findings: 1) hippocampal insulin resistance and deficits in whole body glucose homeostasis, 2) dramatically increased mitochondrial reactive oxygen species (ROS) production in the hippocampus, 3) elevated markers for amyloidogenic cleavage of amyloid precursor protein (APP) and tau protein in the hippocampus, 4) and reduced brain-derived neurotrophic factor (BDNF) expression. These findings were associated with global changes in iron homeostasis, with muscle disuse producing muscle iron accumulation in association with decreased serum and whole brain iron levels. We report the novel finding that muscle disuse alters brain iron homeostasis and reveal a strong negative correlation between muscle and brain iron content. Overall, HLI-induced muscle disuse has robust negative effects on hippocampal insulin sensitivity and ROS production in association with altered brain iron homeostasis. This work provides potential novel mechanisms that may help explain how loss of muscle function contributes to cognitive decline and AD risk.

NEW & NOTEWORTHY Muscle disuse via hindlimb immobilization increased oxidative stress and insulin resistance in the hippocampus. These findings were in association with muscle iron overload in connection with iron dysregulation in the brain. Overall, our work identifies muscle disuse as a contributor to hippocampal dysfunction, potentially through an iron-based muscle-brain axis, highlighting iron dysregulation as a potential novel mechanism in the relationship between muscle health, cognitive function, and Alzheimer’s disease risk.

INTRODUCTION

Alzheimer’s disease (AD) is the fifth leading cause of death in adults ≥65 yr of age, affecting roughly 6.7 million Americans, and prevalence is expected to double over the next 20 years (1). Mechanisms of AD are incompletely understood, but age, genetics, comorbid medical conditions, environmental exposure, and lifestyle factors are known to contribute to risk (2). Physical activity (PA) is one lifestyle factor closely linked to AD risk (3, 4), with increased activity reducing risk of AD, whereas physical inactivity and sedentary behavior increase AD risk up to threefold (5). Despite this well-established link between physical inactivity and sedentary behavior as risk factors for AD, understanding of causation and molecular mechanisms remains incomplete. With the prevalence of physical inactivity rising to pandemic levels in contemporary societies (6), mirroring a notable increase in Alzheimer’s disease (AD) rates (7), understanding this intersection becomes increasingly crucial. Such investigations could prove pivotal in discovering approaches to prevent cognitive decline in older adults.

Skeletal muscle health (form and function) has been proposed as a plausible link between varying levels of PA and the risk of AD (8). Recent evidence has found that the detrimental impacts of reduced muscle health result in a nearly threefold greater risk for cognitive impairment, a 50% increase in AD risk, and an accelerated rate of cognitive decline (9–11). Several lines of evidence support a link between physical inactivity, muscle disuse, and cognitive impairment including 1) the profound effect of PA and exercise on muscle health (12), 2) substantial evidence supporting a positive relationship between muscle quality and cognitive function (13–15), and 3) negative effects of physical inactivity and disuse on muscle (16). In addition, there is evidence to suggest that muscle disuse exerts detrimental effects on brain function (17). For instance, preclinical hindlimb suspension studies in rats have found that 2 wk of muscle disuse reduces neurogenesis (18) and brain-derived neurotrophic factor (BDNF) levels in the hippocampus (19), a region of the brain responsible for learning and memory. A follow-up study demonstrated that dynamic foot stimulation during hindlimb suspension prevents deficits in neurogenesis (20), further supporting that muscle disuse drives this effect. In 2023, two studies evaluated the effect of muscle disuse via hindlimb suspension on spatial learning and memory, brain oxidative stress, glucose metabolism, and neurotrophin levels (21, 22). The first study found that 28 days of hindlimb suspension led to deficits in learning and memory, increased markers of oxidative stress, and loss of antioxidants in the hippocampus, as well as reduced BDNF protein levels in the hippocampus (21). The second study also used 28 days of hindlimb suspension and reported deficits in learning and memory based on cognitive behavior testing performance, increased markers of oxidative stress in the cortex, and signs of glucose metabolism disorder in the brain (22). In addition, iron, which is heavily involved in mitochondrial function (23), oxidative stress (24), and insulin resistance (25), has also been shown to accumulate in the muscle with aging and disuse (26, 27). Importantly, iron homeostasis in the brain is known to be dysregulated in many neurodegenerative diseases like AD (28, 29), but has not been assessed in the context of muscle disuse. Although the studies above provide evidence that muscle disuse affects hippocampal function, to our knowledge, there are currently no studies directly assessing mitochondrial function, insulin resistance, and dysregulated iron homeostasis as potential mechanisms responsible for the negative effect of muscle disuse on hippocampal function, which we aim to address in the current study.

We hypothesize that the relationship between muscle and brain function could be largely explained by skeletal muscle’s essential role in whole body metabolism, glucose homeostasis, and insulin sensitivity (30). Insulin resistance and diabetes mellitus (DM) are major risk factors for AD (31, 32), with over 80% of patients with AD having DM as a comorbidity (33). This has led to increased research interest in brain insulin resistance and its role in cognitive function and neurodegenerative disease risk, with AD recently being referred to as “type III diabetes mellitus” (34). Various studies have shown a strong bidirectional relationship between skeletal muscle metabolic health and DM (30, 35–37), and muscle disuse contributes to insulin resistance and increasing DM risk (38, 39). Interestingly, muscle iron accumulation is known to occur in response to muscle disuse (27), and elevated iron levels contribute to the development of insulin resistance (25), suggesting iron may be an underappreciated contributor to the occurrence of insulin resistance in muscle disuse. This also led us to determine whether muscle iron accumulation alters global iron homeostasis, with a particular focus on brain iron content, as a potential mediator for the negative effects of muscle disuse on hippocampal mitochondrial function and insulin sensitivity. Together, these findings have led us to hypothesize that muscle disuse negatively impacts hippocampal function, primarily by contributing to global iron dysregulation and insulin resistance as potential mechanisms driving the relationship between poor muscle health and cognitive dysfunction.

In this study, we used a model of hindlimb immobilization (HLI) in 4-mo-old female Wistar rats to determine whether muscle disuse for 10 days is sufficient to alter insulin sensitivity, mitochondrial function, and markers of AD and synaptic plasticity in the hippocampus of female rats. We began by analyzing how muscle disuse affected muscle mass, neuromuscular excitability, neuromuscular junction (NMJ) transmission, and muscle mitochondrial function and content. Next, we evaluated systemic glucose homeostasis and muscle and hippocampal insulin receptor surface expression as a marker of tissue insulin sensitivity. Then, we assessed hippocampal mitochondrial function and reactive oxygen species (ROS) production, markers of AD pathology, BDNF expression, and synaptic plasticity. Finally, we identified iron dysregulation in the brain, serum, and muscle in response to muscle disuse as a possible mechanism involved in the negative effects of muscle disuse on hippocampal function.

MATERIALS AND METHODS

Ethical Approval

All animal experiments herein were approved by the University of Missouri Animal Care and Use Committee (MU ACUC) (ACUC Protocol No. 35961). MU ACUC operates under the US Department of Agriculture (USDA) Animal Welfare Act along with the National Institutes of Health (NIH) Office of Laboratory Animal Welfare (OLAW) policy ensuring the humane care and use of laboratory animals. All steps have been taken to minimize any potential animal pain and suffering in all experiments. All investigators understand and acknowledge the ethical principles the journal operates under and have verified our work complies with the Animal Ethics Checklist.

Experimental Animals and Casting Procedure

Three-month-old female Wistar rats were ordered from Charles River Laboratories for this study. Before the start of the study, all rats were group housed until ∼15 wk of age when groups were chosen at random and casts applied. For veterinary cast application (BSN Medical Delta-Lite Plus Casting Tape) in the HLI group, rats were anesthetized using isoflurane, and casting material was applied to the extended hindlimbs with feet positioned in plantarflexion and knee at or near full extension. The Control Cast (Ctrl Cast) group underwent a similar process, except veterinary casting material was instead wrapped around the abdomen with no casting material on the hindlimbs to control for the casting process and mimic the variables associated with wearing a cast without producing muscle disuse. After cast application of Ctrl Cast and HLI groups, rats were dual housed within-group (Ctrl Cast or HLI) in large cages to maintain socialization and mobility. Rubber mats were placed on the bottom of the cages to further assist HLI rats in locomotion, and food was placed on the bottom of the cage for easier access ad libitum. Chewing rods and blocks were also provided for both the Ctrl Cast and HLI groups to act as environmental enrichment to reduce any perceived stress.

Due to the variety of outcomes measured and quantity of tissue needed for the performed techniques, this study consisted of five sets of experiments in five separate sets of female rats: Experiment 1: HLI rats (n = 8) underwent pre- and postelectrophysiological measurements to evaluate neuromuscular function before and after 10 days of HLI. Intraperitoneal (IP) injection of ketamine/xylazine was used in both instances for anesthetization before and after measurements and cast removal. Experiment 2: naïve control (Ctrl), Ctrl Cast, and HLI rats (n = 6/group) underwent 10 days of HLI, Ctrl Cast, or no intervention (Ctrl) and were given ketamine/xylazine IP injections on the 10th day for cast removal, subsequent euthanasia for blood collection [serum corticosterone measurement (Arbor Assays ELISA kit, K014-H5)], and tissue collection, with tissue being flash frozen in liquid nitrogen for molecular analysis and radioligand binding assays. Experiment 3: this set consisted of Ctrl Cast and HLI groups (based on no differences between Ctrl and Ctrl Cast Groups in experiment 2 with identical study design as experiment 2, except these groups underwent additional glucose tolerance testing on day 7 of HLI. These rats were also given ketamine/xylazine IP injections prior to cast removal and euthanasia, followed by whole hippocampus extraction for the isolation of mitochondria for high-resolution respirometry, whereas the remaining tissues were flash frozen in liquid nitrogen. Experiment 4: identical study design as experiment 3 but with whole brain, serum, and soleus muscle being collected for iron measurements using inductively coupled plasma optical emission spectroscopy (ICP-OES). Experiment 5: this experiment used Ctrl Cast and HLI rats (n = 8/group) that underwent casting for 5 days total and were then euthanized for molecular measurements of iron homeostasis in the soleus muscle. All rats (Experiments 1–5, 66 rats in total) were euthanized at ∼4 mo of age in all experiments.

Neuromuscular Electrophysiology Measurements of the Hindlimb Muscles Pre- and Post-HLI

Following a procedure similar to previous neuromuscular assessments performed in rats (40), 4-mo-old female Wistar rats had their neuromuscular connectivity with the hindlimb gastrocnemius muscle evaluated before and after 10 days of HLI. Before measurements, rats were anesthetized with a ketamine/xylazine cocktail. The hindlimb of the right leg was shaved and coated with electrode gel, then the active ring electrode was placed on the gastrocnemius muscle, whereas the reference ring electrode was placed over the midmetatarsal portion of the foot. Stimulation electrodes were then placed proximal to the hindlimb, and submaximal stimulation of the sciatic nerve was gradually increased until supramaximal levels were reached to determine maximal muscle excitability in the gastrocnemius [referred to as compound muscle action potential (CMAP)]. Next, CMAP is administered with repetitive nerve stimulation (RNS) at 50 Hz to evaluate neuromuscular junction (NMJ) transmission efficiency. Results were then obtained from the Sierra Summit program (Cadwell Industries, Kennewick, WA).

Tissue Collection

Experiment 2: blood was collected via heart stick 30+ min after administration of ketamine/xylazine. Next, the brains were rapidly removed following decapitation, placed in a brain matrix, and dentate gyrus (DG)-rich punches 3 mm in diameter were taken from 2-mm thick coronal brain slices. These punches were rapidly flash frozen in liquid nitrogen and stored at −80°C until tissue processing. In addition, hindlimb skeletal muscles were dissected out, weighed, then flash frozen in liquid nitrogen and stored at −80°C until tissue processing. Experiment 3: for high-resolution respirometry of the hippocampus, rats were decapitated 30+ min after ketamine/xylazine injections, and the brain was rapidly removed and placed in an ice-cold saline solution for 1 min. The hippocampus was then rapidly dissected and placed in ice-cold ATP buffer until the time of processing, which took place immediately after tissue collection. Hindlimb skeletal muscles were again weighed followed by flash freezing in liquid nitrogen. Experiment 4: rats were anesthetized using isoflurane, a heart stick was performed for blood collection, then rats were perfused with isotonic saline to clear blood from the tissues to prevent blood iron contamination of tissues before ICP-OES iron measurements. Tissues were then collected and rapidly flash frozen. Experiment 5: followed an identical process as experiment 2 for rat euthanasia and tissue collection.

RNA Isolation, cDNA Synthesis, and qRT-PCR

The DG was collected as described earlier. One DG-rich punch was used for RNA isolation, which consisted of bead homogenization in TRIzol, addition of chloroform, centrifugation, removal of the top phase of liquid following phase separation, then running the RNA through the Direct-zol RNA Miniprep + DNAse treatment kit (Zymo, R2051). Next, cDNA was synthesized using a High-Capacity Reverse Transcription Kit (Applied Biosystems, Carlsbad, CA). Gene-specific primers were generated using Primer3 (Table 1), then each sample was assayed in duplicate for target gene quantification using iTaq Universal SYBR green Supermix (Bio-Rad Laboratories, Hercules, CA). The 2^−ΔΔCt method was used to quantify mRNA expression levels, and all results for DG samples were normalized to Ppib. For the soleus muscle, it was first powdered, and then 10 mg of tissue was collected for RNA isolation identical to the process described earlier for the DG. In the case of the soleus, 18 s was used as the housekeeping gene.

Table 1.

Primers used for qRT-PCR

Gene	Forward	Reverse	Accession Number
Pgc1α	TTCAGGAGCTGGATGGCTTG	GGGCAGCACACTCTATGTCA	NM_031347
Sod1	AGCATTCCATCATTGGCCGTA	CAATCCCAATCACACCACAAGC	NM_017050
Bace1	CTATGTGGAGATGACCGTGGG	AGCCCCCACTGCAAAATTACT	NM_019204
Bdnf9	GTGACAGTATTAGCGAGTGGG	GGGTAGTTCGGCATTGC	NM_001270637.1
TNFa	AACACACGAGACGCTGAAGT	TCCAGTGAGTTCCGAAAGCC	NM_012675
Tnfrsfa	GTCCCCAGGGAAAGTATGCC	AGTAGGTTCCTTTGTGGCACTT	NM_013091
Tnfrsfb	ACAAGCCAGAACCTGGGAAC	ACACGGTGTCTGAAGTCTTGT	NM_130426
Slc40a1/Fpn1	TTTGCTGTTCTTTGCCTTAGTTGT	CCCCTTGTTTGTTCGGATGC	NM_133315
Nfe2l2	GACTTGGAATTGCCACCGC	ACACTTCTCGACTTACCCCAAG	NM_001399173
Mafbx	ACTACGATGTTGCAGCCAAGA	GCTTCCCCCAAAGTGCAGTA	NM_133521
MuRF1	GAGGGCCATTGACTTTGGGA	GTGTCCCTCTGTGGACACG	NM_080903
Ppib	CTCCGTGGCCAACGATAAGA	AGGTCACTCGTCCTACAGGT	NM_022536
18s	GCCGCTAGAGGTGAAATTCTTG	CATTCTTGGCAAATGCTTTCG	NR_046237

Western Blotting

One DG-rich punch and ∼20 mg of soleus tissue were used for protein isolation. Protein was isolated as described previously (41–43). In brief, tissue was homogenized in RIPA buffer, and protein concentrations were obtained using the bicinchoninic acid (BCA) assay method (Pierce Biotechnology, Rockford, IL). In the case of both DG and soleus, 25 µg of protein were loaded onto a 4%–15% SDS-PAGE gel along with Precision Plus Protein Dual Color Standards ladder (Bio-Rad) and ran for ∼50 min at 200 V. Proteins were transferred onto PVDF membranes using the TransBlot Turbo Transfer System (Bio-Rad Laboratories, Hercules, CA). Membranes were then incubated with NoStain Protein Labeling Reagent (Thermo Fisher, Waltham, MA) for total protein quantification for normalization. Blocker FL 10X Fluorescent Blocking Buffer (Thermo Fisher) was used to block nonspecific binding. The primary antibodies used in our studies are listed in Table 2. These primary antibodies were diluted in Tris-buffered saline + Tween 20 (TBS-T) with 5% BSA and incubated with blots overnight at 4°C. The next day, anti-mouse Alexa Fluor 633 or anti-rabbit Alexa Fluor 800 secondary antibodies (Thermo Fisher) were used at a dilution of 1:2,000 and incubated with blots for 1 h at room temperature. The iBright FL1500 Western blot imager (Thermo Fisher) was used, and the iBright analysis software (Thermo Fisher) was used to quantify band volume with rolling background subtraction, and values were normalized to total protein.

Table 2.

Primary antibodies for Western blot

Antibody	Dilution	Source	Molecular Weight, kDa
OXPHOS Cocktail	1:2,000	Thermo Fisher 45-8099	CV: 55, CIII: 48, CIV: 40, CII: 30, CI: 20
IGF1R	1:1,000	Cell Signaling 3027S	90
APP + AICD	1:1,000	Cell Signaling E4H1U	100–140, 9
Tau	1:1,000	Cell Signaling Tau46	50–80
pTau Thr181	1:1,000	Cell Signaling D9F4G	50–80
GluA1	1:1,000	Cell Signaling D4N9V	100
pGluA1 Ser831	1:1,000	Cell Signaling A502P	100
GluA2	1:1,000	Cell Signaling E1L8U	100
FTH1	1:1,000	Cell Signaling D1D4	21
TFR1	1:1,000	Thermo Fisher 13-8600	90
SLC40A1/FPN1	1:1,000	Novus Biologicals OX26	62
Akt	1:1,000	Cell Signaling C67E7	60
pAkt (Ser473)	1:2,000	Cell Signaling D9E	60
GPX4	1:1,000	Cell Signaling E5Y8K	20–22
Catalase	1:1,000	Cell Signaling D5N7V	60

C, complex; FPN1, ferroportin 1; FTH1, ferritin heavy chain 1; TFR1, transferrin receptor 1.

Membrane Isolation and Insulin Receptor Radioligand Binding Assay

One DG-rich punch and ∼15 mg of soleus tissue were used for membrane isolation. Tissue was homogenized in 300 µL lysis buffer (25 mM Tris, 5 mM EDTA, 1 µg/µL aprotinin, 1 µg/µL leupeptin, and brought up to 20 mL volume with ddH₂O) for brain and 1 mL lysis buffer for muscle using a chilled bullet blender. Homogenized samples were centrifuged at 2,000 rpm for 5 min, and supernatant was transferred to a fresh tube for an additional spin at 17,500 rpm for 25 min. The resulting pellet was resuspended in binding buffer (75 mM Tris, 2 mM EDTA, 12.5 mM MgCl₂, 1 µg/µL aprotinin, 1 µg/µL leupeptin, 10% glycerol, and brought up to 5 mL volume with ddH₂O). Protein concentration was quantified using the Bradford protein assay. Next, both the total binding [50 µL binding buffer, 150 µL sample membrane, 50 µL hot ligand (1 nM ¹²⁵I-Insulin)] and nonspecific binding [50 µL BMS-536924 (10 µM), 150 µL sample membrane, 50 µL hot ligand (1 nM ¹²⁵I-Insulin)] reactions for the radioligand binding assay were prepared. The reactions listed earlier were set up in tubes and incubated at 37°C for 1 h in order for the reaction to reach equilibrium. Samples were then placed in harvester, and Whatman GC/F filter paper was soaked in 0.1% before placing on the harvester. Reaction was aspirated and washed five times with 5 mL cold wash buffer. Whatman filter circles from each reaction were taken and loaded into a tube for measurement on a gamma counter. Counts per million values were converted to molar values, specific binding was calculated by subtracting nonspecific binding from total binding, values were then normalized to protein content, and final data were expressed as fmol/mg protein.

Glucose Tolerance Testing

To evaluate whether glucose homeostasis was altered by skeletal muscle disuse, Ctrl Cast and HLI rats were administered a glucose tolerance test (GTT) 7 days after casts were applied. Rats were fasted for 4 h before taking basal blood glucose measurements via the tail vein. Following the basal measurement, rats were given an IP injection of glucose (1 g/kg body wt) with blood glucose levels measured 30, 60, and 120 min after glucose injection, as done previously (41).

High-Resolution Respirometry of the Hippocampus

Mitochondrial function in the hippocampus was evaluated using the Oroboros Oxygraph-2k (Oroboros Instruments, Innsbruck, Austria). Isolated hippocampi were placed in ice-cold ATP buffer until the time of processing approximately 1 h after tissue collection. Mitochondrial isolation was done as previously described (41). In brief, tissue was homogenized, centrifuged at 1,500 g for 5 min, then supernatant was placed in a Percoll gradient for centrifugation to produce the mitochondrial layer for extraction. Isolated mitochondria are given 30 min on ice to rest before loading in the Oroboros for assessment of H₂O₂ production and mitochondrial respiration. Before addition of the mitochondria, Amplex Red and horseradish peroxidase (HRP) were added to the Oroboros to detect H₂O₂ emissions, as Amplex Red in the presence of H₂O₂ is oxidized to the fluorescent molecule resorufin, which is ultimately catalyzed by HRP, allowing us to fluorescently detect H₂O₂ emissions. For mitochondrial respiration analysis, the following substrates were added after addition of mitochondria as previously described: malate (2 mM) and glutamate (5 mM) to stimulate Complex I activity (State 2 respiration), ADP (1,000 μM) to induce oxidative phosphorylation through complex I (state 3—complex I), succinate (10 mM) to activate Complex II and measure oxidative phosphorylation through both complex I + II (State 3—Complex I + II), FCCP (0.25 μM) for uncoupled maximal oxygen consumption (uncoupled respiration), and cytochrome C (5 μM) to confirm the integrity of the isolated mitochondria. The Oroboros DatLab system was used for analysis, and all results were normalized to protein content of the isolated mitochondria.

Inductively Coupled Plasma—Optical Emission Spectrometry

Soleus, serum, and brain samples were weighed out into a 50-mL digestion vessel and stored in the freezer before analysis. To each digestion vessel, 3 mL of nitric acid and 2 mL of perchloric acid were added followed by 0.1 mL of the 100 ppm yttrium internal standard. The digestion vessels were then vortexed for 5 s. The vessel lids were replaced by reflux caps, and the vessels were placed into a heating block at a temperature between 100°C and 120°C for 2 h. After heating digestion, the vessels were removed from the heating block and cooled to room temperature. Digestion vessels were again vortexed after the reflux caps were replaced, and the solution was transferred into a 15-mL Falcon tube. The vessels were washed with 3 mL of HPLC grade water, and the water in the digestion tubes was poured into 15 mL-Falcon tubes. The final volume of the solution in the 15 mL-Falcon tube was brought up to 10 mL with HPLC grade water. The Falcon tubes were vortexed and inverted before proceeding to the final ICP for analysis. The analysis of iron levels was completed at the Veterinary Medical Diagnostic Laboratory at the University of Missouri (Columbia, MO). The assessment was conducted by Thermo Scientific iCAP 7000 Plus Series inductively coupled plasma optical emission spectrometry (ICP-OES). Iron was detected at 259.940 nm with a linear calibration range from 0.1 to 100 µg/mL.

Statistical Analysis

Statistical analysis was performed in GraphPad Prism 10.1.2. For pre- and postassessments of body weight across groups, a repeated-measures two-way ANOVA was used to detect an interaction followed by Sidak’s post hoc comparisons. For all remaining comparisons between three groups at a single time point, a one-way ANOVA was used followed by Tukey’s post hoc analysis. For all comparisons between two groups, or for pre- and postmeasurements within the same group, a Student t test was used to detect any group differences. Grubbs’ test was used to detect and remove any significant outliers. Additionally, all data was assessed for normality using a Shapiro–Wilk normality test and found to have a normal distribution prior to running statistical analysis. P values <0.05 were considered significant, and all data are shown as means ± SE.

RESULTS

HLI for 10 Days Induced Loss of Body Weight and Muscle Mass without Significantly Altering Stress Markers

The effect of muscle disuse, via HLI for 10 days, on hindlimb muscle mass was evaluated. HLI and Ctrl Cast groups had significant reductions in body weight relative to Ctrl (Fig. 1A) with the HLI group also having significantly lower bodyweights relative to the Ctrl Cast group [significant interaction (F_2,33 = 68.34; P < 0.0001), time (F_1,33 = 29.08; P < 0.0001), and group (F_1,33 = 5.842; P = 0.0067). Hindlimb muscle wet weights in the HLI rats (normalized to body weight) were significantly decreased relative to the Ctrl Cast group. Similarly, all hindlimb muscles in the HLI rats, except the extensor digitorum longus (EDL), were significantly decreased relative to the Ctrl group (significant main effect for all hindlimb muscles followed by Tukey’s post hoc analysis; Fig. 1, B–G). In the case of the plantaris, the Ctrl Cast group had a small, but significant, 8% increase in normalized muscle wet weight relative to the Ctrl group. Molecular markers of muscle atrophy, Mafbx and MuRF1, were also assessed and showed robust increases in the soleus and gastrocnemius muscles following 10 days HLI (Supplemental Fig. S1, C–F).

Figure 1.

Hindlimb immobilization (HLI) induces loss of body weight and severe muscle atrophy without altering stress markers. A: body weights before and after the 10-day HLI intervention in Ctrl (n = 12), Ctrl Cast (n = 12), and HLI (n = 12) groups. Muscle wet weight of the soleus (B), plantaris (C), gastrocnemius (D), quadriceps (Quad; E), tibialis anterior (TA; F), and extensor digitorum longus (EDL; G) normalized to body weight following HLI. H: adrenal wet weights normalized to body weight following HLI. I: serum corticosterone levels following HLI (n = 6/group). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

To evaluate stress as a potential factor in our experimental outcomes, markers of stress were measured but showed no differences between groups. Adrenal wet weights were collected and showed no differences between groups (F_2,33 = 0.5597; P = 0.5767; Fig. 1H). Similarly, serum corticosterone levels were measured with ELISA, and, despite a clear trend, no significant differences were present between groups (F_2,15 = 1.624; P = 0.2300; Fig. 1I). (For further assessments on the potential involvement of stress, see Supplemental Table S1 for Pearson correlations between adrenal weights, serum corticosterone, and molecular outcomes).

HLI for 10 Days Induced Neuromuscular Dysfunction Including Loss of Muscle Excitability and NMJ Transmission and Reduced Hindlimb Muscle Mitochondrial Electron Transport Chain Function

To assess the impact of muscle disuse on muscle health and function, we used electrophysiological methods to assess neuromuscular function in the gastrocnemius followed by molecular analysis of mitochondrial electron transport chain (ETC) proteins and biogenesis marker PGC1α levels in the soleus muscle. For neuromuscular electrophysiology assessment, measurements of muscle excitability (CMAP) and NMJ transmission (RNS) were performed pre- and post-HLI. When evaluating CMAP, we found a striking 70% reduction in muscle excitability in response to 10 days of muscle disuse (P < 0.0001; Fig. 2A). Next, we recorded RNS at 50 Hz to assess CMAP amplitude decrement to evaluate NMJ transmission efficiency. HLI for 10 days produced a significant reduction in NMJ transmission during RNS at 50 Hz (P = 0.0122), suggesting NMJ dysfunction as a striking feature of muscle disuse (Fig. 2B). Regarding skeletal muscle mitochondrial function, we measured ETC proteins in the soleus muscle (Fig. 2C). We found significant reductions in complex III (P = 0.036), complex IV (P = 0.001), and complex V (P = 0.043) with no further differences detected in Complex I or II (Fig. 2D). In addition, we measured the expression of Pgc1α, a transcription factor known to induce mitochondrial biogenesis (44), at the transcript level and found a roughly 60% reduction in our HLI group (P = 0.01; Fig. 2E).

Figure 2.

Muscle excitability, neuromuscular junction (NMJ) transmission, and mitochondrial function of hindlimb muscles are greatly reduced following 10 days of muscle disuse. Electrophysiological measurements of compound muscle action potential (CMAP; A) and repetitive nerve stimulation (RNS; B) pre- and post-hindlimb immobilization (HLI) (n = 8, RNS pre-HLI n = 5). C: Western blot images of the mitochondrial complexes in the soleus muscle of Ctrl Cast and HLI rats. D: quantification of the mitochondrial complexes expressed as fold change relative to Ctrl Cast (n = 6). E: Pgc1α levels in Ctrl Cast and HLI groups (n = 6). *P < 0.05; **P < 0.01; ****P < 0.0001.

HLI for 10 Days Reduced Glucose Uptake and Produced Signs of Brain Insulin Resistance

Based on the essential role of skeletal muscle in mediating glucose homeostasis and insulin sensitivity (30), we sought to determine whether 10 days of muscle disuse was sufficient to alter whole body glucose homeostasis and brain insulin sensitivity. We use a glucose tolerance test (GTT) on day 7 of HLI and found significantly higher blood glucose levels in the HLI group relative to Ctrl Cast at both 30 (P = 0.028) and 120 min (P = 0.034) after glucose administration (Fig. 3A) with an overall significant increase in area under the curve (AUC) blood glucose levels (Fig. 3B). Next, we isolated membrane fractions from the soleus and hippocampus to quantify insulin receptor surface levels via radioligand binding assay as a marker for insulin sensitivity in the skeletal muscle and brain in response to muscle disuse. We found a trending decrease (P = 0.059) in insulin receptor surface levels in the soleus of the HLI group (Fig. 3C). In the hippocampus, there was a significant decrease in insulin receptor surface expression for the HLI group (P = 0.042; Fig. 3D). To support this finding, we assessed downstream insulin signaling in the hippocampus by measuring Akt and its phosphorylation status (Ser 473) (Fig. 3E). Total Akt levels were unchanged while absolute Akt phosphorylation (P = 0.026) and Akt phosphorylation normalized to total Akt (P = 0.024) were significantly reduced (Fig. 3, F–H).

Figure 3.

Muscle disuse disrupts systemic glucose homeostasis and produces insulin resistance in the hippocampus. A: glucose tolerance test (GTT) on day 7 of hindlimb immobilization (HLI). B: AUC GTT between Ctrl Cast and HLI groups. C: detection of membrane insulin receptor expression in the soleus via radioligand binding following 10 days of HLI. D: measuring membrane insulin receptor expression in the hippocampus via radioligand binding following 10 days of HLI. E: Western blot images of Akt and p-Akt. F–H: quantification of Akt, p-Akt (Ser 473), and p-Akt:Akt ratio protein levels. *P < 0.05.

Mitochondrial ROS Production Is Increased in the Hippocampus following HLI for 10 Days HLI

Based on the signs of hippocampal insulin resistance in response to muscle disuse, we evaluated whether this loss of insulin signaling was associated with decreased mitochondrial function and elevated ROS production in the hippocampus. Cognitive function depends on proper brain insulin signaling (45), glucose metabolism (46), and mitochondrial function (47). Brain insulin resistance and mitochondrial dysfunction are believed to be key initiating factors in neurodegenerative diseases like AD (48–51). Therefore, high-resolution respirometry was used to evaluate mitochondrial respiration and ROS production of mitochondria isolated from the hippocampus. We did not detect any differences across mitochondrial respiratory states between groups (Fig. 4A). However, when analyzing mitochondrial ROS production, we found that muscle disuse produced a threefold increase in basal respiration H₂O₂ emissions (P = 0.011; Fig. 4B) and a twofold increase in H₂O₂ production during State 2 respiration (P = 0.003; Fig. 4C). Next, we assessed additional markers of oxidative stress in the hippocampus. We found a significant increase in Sod1 (P = 0.001; Fig. 4D), an antioxidant that helps scavenge superoxide radicals, and Nrf2, a transcription factor that activates genes involved in the oxidative stress response (P = 0.024; Fig. 4E). We then measured the antioxidants glutathione peroxidase 4 (GPX4) and catalase at the protein level (Fig. 4F). GPX4 showed a trending increase in protein expression in HLI rats (P = 0.098) while catalase levels were unchanged (Fig. 4, G and H).

Figure 4.

Hippocampal mitochondrial reactive oxygen species (ROS) production is significantly increased following 10 days of muscle disuse. A: high-resolution respirometry across various states of respiration in mitochondria isolated from the hippocampus. B: H₂O₂ emissions under basal respiration conditions. C: H₂O₂ emissions during State 2 respiration. D and E: expression of Sod1 (D) and Nrf2 (E) in the DG region of the hippocampus. F: representative Western blot images of GPX4 and catalase. G and H: quantification of GPX4 (G) and catalase (H) protein levels. n = 6/group for all measurements. *P < 0.05; **P < 0.01.

HLI for 10 Days Increased Markers of AD Pathology and Reduced BDNF and Synaptic Plasticity in the Hippocampus

Based on the relationship between muscle health and AD risk (52), we next wanted to look at markers of AD pathology and synaptic plasticity in the hippocampus of HLI rats (Fig. 5A). We began by evaluating APP and its processing and cleavage products. Regarding APP processing, we found that Bace1, a protein that cleaves APP to form products like amyloid beta and APP intracellular C-terminal domain (AICD) (53), expression was increased to a significant degree at the transcript level in the DG of HLI rats (P = 0.024; Fig. 5B), whereas APP protein levels were not altered (Fig. 5C). Based on the difficulties of detecting amyloid beta in nontransgenic rats due to its low expression, we used an APP cleavage product, AICD, produced concomitantly with amyloid beta, as an indirect measure of the amyloidogenic cleavage of APP. We found a minor, but significant, increase in AICD protein levels in the HLI group (P = 0.047; Fig. 5D). In addition, levels of total tau protein were increased (P = 0.042; Fig. 5E), whereas phospho-tau levels were unchanged in HLI rats (Fig. 5F). Next, we looked at the transcript levels for the neuroplasticity marker, BDNF, whose expression is known to be indirectly regulated by factors secreted by muscle (54–56) (Fig. 5G). There was a highly significant reduction in Bdnf-IX transcript levels (P = 0.009), supporting that muscle disuse reduces BDNF expression in the hippocampus. AMPA receptors play a vital role in excitatory signaling and synaptic plasticity (57). In addition, dysregulated AMPA receptor expression and function have been shown to be an early feature of AD (58). Therefore, we analyzed the two most abundant AMPA receptor subunits: GluA1, its phosphorylation at S831, and GluA2. There was a significant increase in the protein expression of the GluA1 subunit (P = 0.022; Fig. 5H) in the HLI group, but this came without a subsequent increase in its phosphorylation status (Fig. 5I). However, the ratio of phosphorylated GluA1 relative to total GluA1 revealed a significant decrease in receptor phosphorylation for the HLI group (P = 0.009; Fig. 5J). Finally, we analyzed the protein levels of the GluA2 subunit of the AMPA receptor and saw no group difference (Fig. 5K), potentially suggesting a shift toward GluA1 calcium-permeable AMPA receptors in the DG.

Figure 5.

Hindlimb immobilization induced pathological APP cleavage, elevated Tau, reductions in brain-derived neurotrophic factor (BDNF), and dysregulated AMPA receptor function. A: Western blot images of Alzheimer’s disease (AD)-related proteins. B: mRNA expression for the APP-processing protein, BACE1. Protein levels of APP (C) and its cleavage product, AICD (D). Protein levels of Tau (E) and pTau (F). G: transcript levels of the coding exon for BDNF, exon IX. GluA1 protein levels (H) and its phosphorylation status at S831 (I). J: the ratio of total GluA1 to phosphorylated GluA1. K: GluA2 protein levels. n = 6/group for all measurements. *P < 0.05; **P < 0.01.

HLI for 10 Days Induced Iron Overload in the Soleus Muscle and Iron Dysregulation in the Circulation and Brain

Iron plays a vital role in mediating ROS production (24) and insulin sensitivity (25) and has been demonstrated to accumulate in aging and atrophied muscle (26, 27, 59–61). However, there is little to no evidence as to what effect muscle iron accumulation has on iron levels in peripheral tissues with high iron demands like the brain; therefore, we aimed to determine whether muscle disuse is sufficient to produce iron dysregulation in the brain. We began by measuring markers of iron storage, uptake, and export in the soleus (Fig. 6A). Storage of iron was increased in the soleus muscle by nearly 375% in HLI rats based on ferritin heavy chain 1 (FTH1) protein levels (P < 0.0001; Fig. 6B). Similarly, FTH1 protein levels were elevated 2.77-fold in the gastrocnemius muscle following 10 days HLI (P < 0.0001; Supplemental Fig. S2). Unsurprisingly, iron uptake in the soleus through the transferrin receptor 1 (TFR1) was reduced by ∼80% (P = 0.002; Fig. 6B), suggesting the soleus is experiencing iron overload. Next, we looked at ferroportin 1 (FPN1), a protein responsible for iron export out of the cell, at both the transcript and protein levels. Interestingly, FPN1 protein levels were reduced by ∼70% (P = 0.003; Fig. 6B) despite an over twofold increase in Fpn1 transcript levels (P = 0.002; Fig. 6C). To complement our molecular findings for iron content, we used ICP-OES to measure total iron in the soleus and serum and found a significant increase in iron concentration in the HLI soleus (P < 0.0001; Fig. 6D) along with a significant decrease in total iron serum concentrations (P = 0.0317; Fig. 6E).

Figure 6.

Hindlimb immobilization (HLI) for 10 days induced signs of iron overload in the soleus and iron dysregulation in the serum and brain. A: Western blot images for proteins involved in iron storage [ferritin heavy chain 1 (FTH1)], uptake [transferrin receptor 1 (TFR1)], and export [ferroportin 1 (FPN1)] in the soleus. B: fold change protein levels of FTH1, TFR1, and FPN1 in the soleus (n = 6/group). C: fold change transcript levels of Fpn1 in the soleus (n = 6/group). D and E: total iron concentrations using inductively coupled plasma optical emission spectroscopy (ICP-OES) in the soleus and serum, respectively (n = 8/group). F: Western blot images for FTH1, TFR1, and FPN1 proteins in the dentate gyrus (DG). G: fold change protein levels of FTH1, TFR1, and FPN1 in the DG (n = 6/group). H: fold change transcript levels of Fpn1 in the DG (n = 6/group). I: whole brain total iron concentrations using ICP-OES (n = 8/group). J: correlation between soleus and brain iron concentrations. *P < 0.05; **P < 0.01; ****P < 0.0001.

Next, we measured markers of iron homeostasis in DG-rich punches of the hippocampus (Fig. 6F). There was a minor but significant increase in cellular iron storage based on FTH1 protein levels (P = 0.035; Fig. 6G). Interestingly, we also saw a highly significant increase in TFR1 (P = 0.001; Fig. 6G), a surprising finding based on the already elevated FTH1 levels. Then, looking at the transcript and protein levels of FPN1, we saw no change at the protein level (Fig. 6G) but a significant increase in Fpn1 transcript levels (P = 0.038; Fig. 6H), partially paralleling what we found in the soleus muscle for FPN1. Next, we used ICP-OES to look at total iron in the whole brain and found that total brain iron concentration was significantly decreased in the HLI group (P = 0.0052; Fig. 6I), contrasting our molecular findings suggesting elevated iron content in the DG. Finally, when analyzing correlations between soleus and brain iron concentrations, we found a strong negative correlation between the iron concentrations of the two tissues (r = −0.675, P = 0.004; Fig. 6J). These findings suggest that muscle disuse produces iron accumulation in the soleus leading to iron dysregulation in the blood and brain.

HLI for 5 Days Induced Signs of Iron Overload in the Soleus Prior to the Occurrence of Significant Muscle Atrophy

Building on the above findings demonstrating global changes in iron homeostasis, we next aimed to determine whether muscle iron dysregulation rapidly occurs in response to muscle disuse. Therefore, in this portion of experiments, a subset of rats underwent HLI for 5 days, rather than 10, followed by molecular analysis of iron markers in the soleus. In this case, 5 days of HLI did not result in significant reductions in either body weight (P = 0.5261; Fig. 7A) or soleus wet weight (P = 0.2410; Fig. 7B) despite elevations in molecular markers of muscle atrophy [Mafbx and MuRF1 (Supplemental Fig. S1, A and B)], demonstrating this time point predates the onset of significant muscle atrophy. Interestingly, when analyzing iron storage, uptake, and export proteins (Fig. 7C), FTH1 protein levels were significantly increased (P < 0.0001), whereas TFR1 (P < 0.0001) and FPN1 (P < 0.0001) protein levels were significantly decreased following 5 days of HLI. These findings support the hypothesis that muscle iron overload is induced by muscle disuse rather than atrophy.

Figure 7.

Muscle disuse for 5 days induced iron accumulation in the soleus prior to significant muscle atrophy. A and B: body weight and soleus wet weight following 5 days of hindlimb immobilization (HLI) (n = 8/group). C: representative Western blot images of FTH1, TFR1, and FPN1 protein expression in the soleus. D: fold change protein levels of ferritin heavy chain 1 (FTH1), transferrin receptor 1 (TFR1), and ferroportin 1 (FPN1) in the soleus in response to 5 days of HLI (n = 8/group). ****P < 0.0001.

DISCUSSION

The primary objective of this study was to assess whether muscle disuse, induced by hindlimb immobilization (HLI), had a detrimental impact on hippocampal insulin sensitivity and mitochondrial function. Our findings revealed that a 10-day period of muscle disuse resulted in a significant reduction of systemic glucose uptake and insulin receptor surface expression in the hippocampus by ∼55%. These findings indicate that the loss of muscle metabolic function, during disuse, is associated with disruption of overall glucose homeostasis and diminished brain insulin sensitivity. Given the well-established link between insulin signaling in the brain and cognitive function, these findings suggest a potential mechanism for increased risk of Alzheimer’s disease (AD) associated with muscle disuse and physical inactivity.

Our investigation also found a drastic elevation in hippocampal mitochondrial reactive oxygen species (ROS) production during muscle disuse. Increased H₂O₂ production was accompanied by elevated Sod1 and Nrf2 transcript levels, which we hypothesize represent a compensatory response to combat oxidative stress in the hippocampus of HLI rats. Surprisingly, GPX4 and catalase levels were mostly unaffected, suggesting that elevated H₂O₂ levels are likely due to changes in the fenton reaction process, based on the observed iron dysregulation, rather than deficits in antioxidative capacity. The bidirectional link between ROS and insulin resistance supports the hypothesis that these factors are primary drivers of how muscle disuse negatively affects the hippocampus. Elevated ROS and insulin resistance in the brain are known to impair cognitive function and contribute to AD development (62, 63). Supporting this, markers of amyloidogenic processing of APP and tau protein levels in the hippocampus, indicative of AD pathology, increased with muscle disuse. This finding is supported by a prior study demonstrating that 14 days of HLI in an AD mouse model accelerated AD onset and progression (64), further demonstrating a relationship between muscle disuse and AD pathology.

Iron homeostasis was examined as muscle atrophy and aging are known to produce muscle iron overload (26, 27, 59–61), and iron plays a vital role in mitochondrial respiration, ROS, and insulin sensitivity (24, 25). We hypothesized that muscle iron overload may impact iron homeostasis in other tissues with high iron demands, like the brain (65). In support of our hypothesis, we observed that muscle disuse led to increased muscle iron content while reducing iron content in the circulation and whole brain. This alteration in iron distribution provides a potential mechanism for how muscle disuse influences hippocampal insulin sensitivity and mitochondrial function. Interestingly, analysis of whole brain iron content conflicted with our molecular assessments of iron content markers in the hippocampus. We suspect this could partially be attributed to a protective mechanism by which the hippocampus rapidly upregulates TFR1 expression in an attempt to prevent iron loss and maintain redox balance and mitochondrial respiratory function, resulting in a period of elevated intracellular iron storage. Additionally, this disconnect could be influenced by the presence, or absence, of blood in the collected samples. Whole brain samples for mass spectrometry were from perfused rats, minimizing the influence of iron in the blood, while hippocampus samples were not perfused. Therefore, the elevated iron markers in the hippocampus could be influenced by circulating factors and cell types. We also add important findings on the timeline of iron accumulation in muscle atrophy by providing evidence that iron overload is an early event, already present at 5 days of muscle disuse and precedes the occurrence of significant muscle atrophy. We suggest that iron overload and subsequent iron-induced oxidative stress play a significant role in disuse-induced muscle atrophy. These novel findings shed light on potential mechanisms underlying the relationship between physical inactivity, muscle disuse, and the heightened risk of AD. Further investigation uncovered significant deficits in neuromuscular function during muscle disuse, emphasizing the impact on muscle excitability. This finding is supported by clinical studies reporting that altered neural function explains 48% of strength loss in response to muscle disuse (66). We propose that improvements in neuromuscular function could be a targeted approach for treating muscle disorders such as sarcopenia. In addition, the reduction in soleus mitochondrial electron transport chain (ETC) proteins suggests a connection between muscle iron dysregulation and decreased ETC function, potentially contributing to increased free iron levels.

Assessing whole body metabolism and insulin sensitivity, we observed a substantial decrease in glucose uptake after 7 days of muscle disuse. The reduction in soleus insulin receptor surface expression was only a trend, possibly underestimated as insulin receptor expression is fiber type dependent (67). However, a robust decrease in insulin receptor surface expression in the hippocampus was evident after 10 days of muscle disuse, suggesting a potential contribution to brain insulin resistance. Our findings add a potential upstream mechanism for prior studies showing that hindlimb suspension reduced glucose metabolism in the cortex (22) as a loss of insulin signaling is a likely explanation. Overall, this finding is particularly significant given the crucial role of insulin in brain function and its association with AD risk (31).

Insulin resistance also contributes to impaired mitochondrial function and oxidative stress (68, 69), which is drastically elevated in the hippocampus of our HLI rats. A large body of work has demonstrated the importance of brain mitochondrial function in either producing resiliency or predisposing the brain to neurodegenerative diseases like AD (70–72). Brain mitochondrial dysfunction and ROS are known to contribute to AD onset and progression (73), and our data herein demonstrate that muscle disuse can elevate ROS levels in the hippocampus of otherwise healthy rats. We add important context to previous findings that muscle disuse increases markers of oxidative stress and reduces antioxidant levels in the brain (21, 22) by implicating elevated mitochondrial ROS production as a likely mechanism. In addition, brain iron dysregulation could play an exacerbating role in the elevated hippocampal ROS levels as iron readily converts H₂O₂ into highly reactive hydroxyl radicals (known as the Fenton reaction) that can damage the structure and function of various molecules in the cell (74). Importantly, these findings implicate altered insulin sensitivity, mitochondrial function, and oxidative stress as potential mechanisms explaining the relationship between muscle and cognitive function.

The involvement of stress is a limitation worth addressing in this study. HLI is a form of chronic mild stress despite all efforts being made to minimize stress as much as possible (added environmental enrichment, dual-housing, improved food and water access, added rubber mats to improve mobility). Therefore, a Ctrl Cast group was incorporated to roughly approximate the chronic mild stress associated with wearing a cast, although the added stress of reduced mobility and hindlimb immobilization cannot be fully controlled for. There was a clear trend for increased serum corticosterone in HLI rats although this effect did not reach statistical significance. Overall, we cannot rule out the influence of stress as a contributing factor in our study. However, it is worth noting that there was a lack of correlation between adrenal weights, serum corticosterone levels, and our primary molecular outcomes, suggesting that stress was not a primary driver of the observed phenotype.

Our study provides compelling evidence that muscle disuse, even over a short duration, induces maladaptive molecular changes in the hippocampus that may enhance susceptibility to neurodegeneration. These alterations are associated with major shifts in iron homeostasis, implicating iron dysregulation as a potential contributor to the observed negative effects of muscle disuse on the brain. The intricate interplay between muscle and brain function elucidated in this research offers a molecular explanation for the relationship between muscle health and increased AD risk. Future work should further investigate muscle as a contributor to sporadic AD and explore the role of iron dysregulation as a driver of these effects.

DATA AVAILABILITY

Data will be made available upon reasonable request.

SUPPLEMENTAL MATERIAL

Supplemental Table S1 and Supplemental Figs. S1 and S2: https://doi.org/10.6084/m9.figshare.26005237.

Supplemental Figs. S2–S7: https://doi.org/10.6084/m9.figshare.26005195.

GRANTS

This work was financially supported by Frank W. Booth, PhD and was supported in part with resources and the use of facilities at the Harry S. Truman Memorial Veterans Hospital in Columbia, Missouri.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

N.R.K. and F.W.B. conceived and designed research; N.R.K., C.W.M., C.-H.C., J.M.B., T.J.K., T.E.C., W.D.A., L.A.G., and X.D. performed experiments; N.R.K., C.W.M., C.-H.C., W.D.A., L.A.G., and X.D. analyzed data; N.R.K., R.S.R., W.D.A., L.A.G., and F.W.B. interpreted results of experiments; N.R.K. prepared figures; N.R.K. drafted manuscript; N.R.K., C.-H.C., T.J.K., T.E.C., R.S.R., W.D.A., L.A.G., X.D., and F.W.B. edited and revised manuscript; N.R.K., C.W.M., C.-H.C., J.M.B., T.J.K., T.E.C., R.S.R., W.D.A., L.A.G., X.D., and F.W.B. approved final version of manuscript.