α‐Motor neurons are spared from aging while their synaptic inputs degenerate in monkeys and mice

Summary Motor function deteriorates with advancing age, increasing the risk of adverse health outcomes. While it is well established that skeletal muscles and neuromuscular junctions (NMJs) degenerate with increasing age, the effect of aging on α‐motor neurons and their innervating synaptic inputs remains largely unknown. In this study, we examined the soma of α‐motor neurons and innervating synaptic inputs in the spinal cord of aged rhesus monkeys and mice, two species with vastly different lifespans. We found that, in both species, α‐motor neurons retain their soma size despite an accumulation of large amounts of cellular waste or lipofuscin. Interestingly, the lipofuscin profile varied considerably, indicating that α‐motor neurons age at different rates. Although the rate of aging varies, α‐motor neurons do not atrophy in old age. In fact, there is no difference in the number of motor axons populating ventral roots in old mice compared to adult mice. Moreover, the transcripts and proteins associated with α‐motor neurons do not decrease in the spinal cord of old mice. However, in aged rhesus monkeys and mice, there were fewer cholinergic and glutamatergic synaptic inputs directly abutting α‐motor neurons, evidence that aging causes α‐motor neurons to shed synaptic inputs. Thus, the loss of synaptic inputs may contribute to age‐related dysfunction of α‐motor neurons. These findings broaden our understanding of the degeneration of the somatic motor system that precipitates motor dysfunction with advancing age.

For example, we recently showed that IA/II proprioceptive sensory neuron soma and nerve ending at muscle spindles degenerate with increasing age and progression of amyotrophic lateral sclerosis (ALS) in mice (Vaughan, Kemp, Hatzipetros, Vieira & Valdez, 2015;Vaughan, Stanley & Valdez, 2016). These sensory neurons detect changes in the amount and rate of muscle contraction, and utilize this information to coordinate and modulate the activity of a-motor neurons. Hence, their degeneration would inevitably compromise motor function.
Although a-motor neurons receive, integrate, and relay all motor commands to skeletal muscles, the effect of aging on these neurons is still debated. Several studies, using histological assays to sample spinal cord sections, have reported that fewer a-motor neurons are present in aged humans and animals (Jacob, 1998;Tomlinson & Irving, 1977). However, these same studies show that the remaining amotor neurons enlarge rather than atrophy with age. Yet, other studies have found no difference in the number and size of a-motor neurons in old compared to adult mice (Chai, Vukovic, Dunlop, Grounds & Shavlakadze, 2011). Thus, the effect of aging on a-motor neurons is still an open question, and addressing it will contribute significantly to our understanding of NMJ and muscle fiber degeneration and accompanying age-related motor deficits.
There is even less known about the fate of synaptic inputs in aging spinal cords, which are responsible for initiating and modulating all voluntary movements. Synaptic inputs terminate throughout the soma and dendrites of a-motor neurons (Witts, Zagoraiou & Miles, 2014) and include excitatory (glutamatergic and cholinergic) and inhibitory (GABAergic and glycinergic) inputs originating from a variety of neurons. Glutamatergic inputs are largely responsible for activating a-motor neurons and emanate from neurons located throughout the central and peripheral nervous system. Cholinergic inputs, called C-boutons, modulate the output of a-motor neurons and arise from small interneurons located near the midline of the spinal cord. GABAergic and glycinergic inhibitory inputs originate from interneurons located in the spinal cord and are critical for finetuning and terminating the activity of a-motor neurons. Although these inputs have yet to be examined in the context of aging, they have been shown to degenerate in several diseases known to affect motor function, including ALS (Vaughan et al., 2015). Thus, age-associated dysfunction or loss of synaptic inputs terminating on a-motor neurons will also impair motor function.
In this study, we carried out histological and molecular analysis to determine the impact of aging on a-motor neurons and cholinergic inputs in the spinal cord of rhesus monkeys and mice. We examined these two species for several reasons: (i) Their vastly different average lifespans (25 years for rhesus monkeys and 2.12 years for C57/BL6 mice) make it possible to determine whether cellular changes are due to biological or chronological aging, (ii) rhesus monkeys are an important animal model for understanding human aging because they share approximately 93% of their genetic code with humans, and (iii) mice provide a tractable model system for interrogating age-related changes due to their small size and relatively short lifespan. In both species, we used light microscopy to examine the size of a-motor neuron somata, the presence of lipofuscin, the number of motor axons, and the glutamatergic and cholinergic inputs on a-motor neuron somata and throughout the ventral horn. In addition, molecular analyses were conducted for transcripts and proteins associated with motor neurons.

| a-Motor neurons do not atrophy in old rhesus monkeys
We first examined the morphology of a-motor neurons in the spinal cord of rhesus monkeys and mice. To visualize a-motor neurons, 30lm spinal cord cross sections were stained with an antibody against the pan-neuronal marker, Neuronal Nuclei (NeuN) (Mullen, Buck & Smith, 1992), and an antibody against the vesicular acetylcholine transporter (VAChT). We identified a-motor neurons based on: (i) their location in the ventral horn of the spinal cord, (ii) the large size of a-motor neurons compared to other neurons in the spinal cord, and (iii) the presence of VAChT puncta, cholinergic inputs called Cboutons, throughout the perimeter of the soma and along the apical dendrite of a-motor neurons.
First, we compared the soma size between adult and old male rhesus monkeys. As depicted in Table 1, we analyzed the size of amotor neurons in the spinal cord of adult (6-to 17-year-old) and old (28-to 32-year-old) monkeys (Figure 1a-b). This analysis revealed no significant difference in the average soma size of old (1,388 AE 102.4 lm 2 ) compared to adult (1,231 AE 152.1 lm 2 ) rhesus monkeys ( Figure 1c). While these findings suggest that a-motor neurons do not atrophy in old rhesus monkeys, it remained possible that aging affects a subpopulation of a-motor neurons. For example, aging may preferentially affect slow, fast fatigue-resistant (FFR) or fast-fatigable (FF) a-motor neurons. These neurons vary in size (slow<FFR<FF), functional demands, and susceptibility to diseases (Stifani, 2014). If aging preferentially affects a subtype of a-motor neurons, the distribution of a-motor neurons soma size should be different between adult and old animals. However, a frequency histogram and a two-sample Kolmogorov-Smirnov (KS) test revealed no difference in the soma size distribution between adult and old rhesus monkeys (Figure 1d), indicating that, in rhesus monkeys, a-motor neurons retain their soma size throughout the aging process.

| a-Motor neurons do not atrophy in old mice
We next assessed the morphology of a-motor neurons in the spinal cord of C57BL/6 adult (2-to 3-month-old), old (21-month-old), and very old (28-month-old) male mice (N = 4 per age group for this and all other experiments involving mice). To this end, a-motor neurons residing in the lumbar region of the spinal cord were examined. To identify a-motor neurons, 30-lm spinal cord cross sections were stained with antibodies against NeuN and VAChT.
As in rhesus monkeys, the average size of a-motor neurons remains unchanged in old and very old mice compared to adult mice (adult = 626.9 AE 49.03 lm 2 ; old = 562.7 AE 61.95 lm 2 : pvalue = .4434; very old mice = 660.9 AE 47.27 lm 2 : p-value = .6916; Figure 1e). A frequency histogram and a two-sample KS test also showed that aging does not preferentially affect a subset of a-motor neurons, because the soma size distribution is similar between adult, old, and very old mice (adult vs. old: p = .2120; adult vs. very old: p = .1098; Figure 1f). We extended this analysis to a-motor neurons located in the thoracic and cervical regions of the spinal cord. As in the lumbar region, the size of a-motor neurons remains unchanged in the thoracic and cervical regions in old and very old mice ( Figure S1).
In addition, we counted the number of motor axons in L3 ventral roots in adult and very old mice ( Figure 2a). Axons were visualized in 15-lm cross sections using an antibody against neurofilament. The same tissue sections were also stained with an antibody against S100, a Schwann cell marker (Figure 2b,c). This analysis revealed no difference in the number of axons in L3 ventral roots between adult and very old mice (adult = 685.6 AE 7.232; very old = 644.4 AE 38.14: p-value = .2675; Figure 2d). Taken together, these data demonstrate that, in mice, a-motor neurons do not degenerate with age.
2.3 | Aging does not reduce genes preferentially associated with a-motor neurons in mice To more broadly assess the effect of aging on a-motor neurons, we analyzed levels of the homeobox gene 9 (Hb9), the gene encoding for the LIM homeodomain protein 1 (Isl-1), the choline acetyltransferase (ChAT), and the vesicular acetylcholine transporter (VAChT; Lu, Niu & Alaynick, 2015). These genes play important roles in the differentiation and function of a-motor neurons. Their expression patterns thus correlate with the number and functional status of a-motor neurons.
Using qPCR, we found mRNA levels for HB9, Isl-1, ChAT, and VAChT unchanged in the spinal cord of 27-month-old mice compared to 4month-old mice ( Figure S2a). We also examined protein levels for VAChT and Isl-1 by Western blotting. As expected, VAChT protein is present at similar levels in the spinal cord of very old and adult mice ( Figure S2b,c). We then used an antibody that detects Isl-1 and Isl-2 to assess changes in the aged mice. This antibody showed that Isl-1 and Isl-2 proteins significantly increase in very old mice ( Figure S2d, e). It also revealed that levels of Isl-1 mRNA and protein levels are uncorrelated in the spinal cord of very old mice. Additionally, we examined levels of NeuN to explore the effect of aging on most spinal cord resident neurons ( Figure S3). Like Isl-1, we found that, while aging did not alter the abundance of NeuN transcripts (Figure S3a), it significantly increased levels of NeuN proteins in the spinal cord of very old mice ( Figure S3b,c). Together, these findings show that genes associated with motor neurons are either unchanged or increased in very old age. These findings further indicate that the number of a-motor neurons is unchanged in aged animals.

| Lipofuscin accumulation reveals that motor neurons age at different rates within and between animals
During this study, we found that lipofuscin progressively aggregates in the perinuclear region of aging a-motor neurons of mice and monkeys ( Figure S4). Lipofuscin, a form of cellular waste, preferentially accumulates in the cytosol of large neurons in animals and humans as they age (Gray & Woulfe, 2005;Keller et al., 2004). Lipofuscin is composed of lipids and highly oxidized and cross-linked proteins that accumulate within lysosomes. Because lipofuscin auto-fluoresces, it is easy to separate from proteins and subcellular structures labeled with specific fluorophores using light microscopy. In old rhesus monkeys, the number of a-motor neurons with aggregated lipofuscin was significantly higher compared to adult monkeys (adult = 44.57 AE 4.698%; old = 78.1 AE 5.021%; p < .0001; Fig-ure S4a-c). In mice, the incidence of a-motor neurons with aggregated lipofuscin increased with advancing age (adult = 0%; old = 73 AE 9.891%; very old = 87.33 AE 3.844%; Figure S4d). These findings were not surprising given that a-motor neurons are relatively large compared to other neurons, and lipofuscin has been primarily found in large neurons (Liang, Nelson, Yazdani, Pasbakhsh & German, 2004). As the soma size and number of a-motor neurons appear unchanged in old age, lipofuscin likely marks other well-documented aged-related afflictions known to affect a-motor neurons, such as degeneration of motor nerve endings at NMJs.
We made two additional and notable observations while analyzing lipofuscin in both rhesus monkeys and mice. First, we found aged a-motor neurons with vastly different lipofuscin profiles ( Figure S4a, b) often adjacent to each other and within the same spinal cord segment. In both aged monkeys and mice, lipofuscin aggregates were either absent, or, in small, or large amounts in these neurons. These varied lipofuscin profiles indicate that a-motor neurons age at different rates. Second, we found that lipofuscin begins to aggregate in amotor neurons of adult rhesus monkeys ( Figure S4c). In stark contrast, lipofuscin aggregates are absent from a-motor neurons of adult mice ( Figure S4d). We also found significantly more a-motor neurons with lipofuscin aggregates and occupying more of the cytosol in old rhesus monkeys ( Figure S4e) compared to old mice ( Figure S4f). This interspecies comparison suggests that the accumulation of lipofuscin Likewise, fewer a-motor neurons were found in aged rats (Hashizume, Kanda & Burke, 1988;Jacob, 1998). In stark contrast, other studies found no change in the number of a-motor neurons in aged cats and mice (Chai et al., 2011;Liu, Bertolotto, Engelhardt & Chase, 1996).
There are also contradictory findings on the effect of aging on the size of a-motor neurons. These discrepancies may be a consequence of the techniques utilized to analyze aged a-motor neurons. In several studies, a-motor neurons were identified following staining of nuclei or following the uptake of a fluorescence dye injected into the gastrocnemius muscle and analyzed using low-resolution microscopy.
In this study, we took a holistic approach to examine the effect of aging on a-motor neurons in two species with vastly different average and maximum lifespans. We used immunostaining to visual- Aging also reduces the total number of C-boutons in the ventral horn (d). Error bar = standard error. *p ≤ .05. **p ≤ .01. Scale bar = 20 lm in old age? In humans, the relationship between axonal loss and cause of death was not established, allowing for the possibility that diseases, rather than normal aging, cause the loss of motor axons. In mice, a previous study (Valdez et al., 2010) used hybrid transgenic mice expressing yellow fluorescence protein (YFP), which has been shown to contribute to degeneration of axons under stress (Bridge et al., 2009), whereas this study examined C57BL/6 mice.

| The rate of aging varies among a-motor neurons in the same anatomical region
We found that a-motor neurons retain their soma size in the presence of substantial amounts of lipofuscin, demonstrating that lipofuscin is a marker and not a driver of neuronal aging. Interestingly, we identified a-motor neurons with vastly different lipofuscin profiles. This finding strongly suggests that a-motor neurons age at different rates, a possibility supported by published data showing that motor axon nerve endings degenerate at different rates, even within the same muscle (Valdez et al., 2012). As the soma size and number of a-motor neurons appear unchanged in old animals, the aggregation of lipofuscin must correlate with age-related subcellular changes in a-motor neurons that include altered neurotransmission (Arasaki, Iwamoto & Tamaki, 1995), deregulated expression of proteins critical for the assembly, and early and progressive degeneration of presynaptic sites at the NMJ (Valdez et al., 2010).

| Aging affects cholinergic and glutamatergic synaptic inputs that innervate a-motor neurons
We show that excitatory inputs, both cholinergic and glutamatergic, degenerate in aged spinal cords despite the lack of atrophy of amotor neurons. These findings support published data indicating that synaptic degeneration occurs before connecting cells atrophy in aged animals. In the neocortex of old rhesus monkeys, fewer synapses are found on pyramidal neurons (Morrison & Baxter, 2012). Synapses also degenerate in the hippocampus, cortical regions, and olfactory bulb of aged rodents prior to degeneration of the neuronal soma (Azpurua & Eaton, 2015;Barnes, 1994;Grillo, 2016;Hof & Morrison, 2004;Punga & Ruegg, 2012). The relationship between synaptic and soma changes with aging has been best documented at the NMJ, pri-

| Immunohistochemistry
Mouse and monkey spinal cord sections were blocked using 5% lamb serum, 3% bovine serum albumin, and 0.1% Triton X-100 in F I G U R E 6 In mice, VGLUT1-positive synaptic inputs innervating motor neurons decrease with advancing age. Immunostaining for VGLUT1 (Green) and the motor soma with anti-NeuN (Blue) in the lumbar region of the spinal cord (a, b). There are significantly fewer VGLUT1 puncta present on the motor soma of very old mice compared to adult mice (c). However, the total number of VGLUT1 puncta does not change in the ventral horn of mice (d). Error bar = standard error. *p ≤ .05. Scale bar = 20 lm not be seen through it, were also measured in ImageJ. Synaptic inputs were counted using ImageJ analysis software. For specifically counting C-boutons and VGLUT1-positive inputs on the soma, any puncta contacting the cell body of an a-motor neuron were counted by hand. For counting the overall number of VAChT puncta in the ventral horn, the cell counter tool was used in the ImageJ software. The red (VAChT) color channel was separated, and a threshold of about 2% of the maximum was applied. Any puncta between 1 and 40 lm 2 were counted.
Anomalies were then removed by hand from the image with all channels visible using the ROI manager feature in ImageJ.

| Imaging
All images were taken with a Zeiss LSM 700 or Zeiss LSM 710 con-  Table 3.

| Statistics
Significance between animals of different ages was determined using unpaired t test and one-way ANOVA with post hoc Bonferroni test.
The Kolmogorov-Smirnov test was used to compare the distribution of a-motor neuron soma size between groups. Data are expressed as the mean AE SE (standard error), and p ≤ .05 was considered statistically significant.

ACKNOWLEDG MENT
We thank members of the Valdez laboratory for constructive comments on and edits to the manuscript.

CONFLI CT OF INTEREST
None.

AUTHOR CONTRI BUTIONS
N.M. and R.C. performed all immunostaining, qPCR, and Western blotting assays; collected and analyzed data for each assay; and gen-