Abstract
Impairments in synaptic dynamics and stability are observed both in neurodegenerative disorders and in the healthy aging cortex, which exhibits elevated dendritic spine turnover and decreased long-term stability of excitatory connections at baseline, as well as an altered response to plasticity induction. In addition to the discrete gain and loss of synapses, spines also change in size and strength both during learning and in the absence of neural activity, and synaptic volume has been associated with stability and incorporation into memory traces. Furthermore, intrinsic dynamics, an apparently stochastic component of spine volume changes, may serve as a homeostatic mechanism to prevent stabilization of superfluous connections. However, the effects of age on modulation of synaptic weights remain unknown. Using two-photon excitation (2PE) microscopy of spines during chemical plasticity induction in vitro and analyzing longitudinal in vivo 2PE images after a plasticity-inducing manipulation, we characterize the effects of age on volumetric changes of spines of the apical tuft of layer 5 pyramidal neurons of mouse primary somatosensory cortex. Aged mice exhibit decreased volumetric volatility and delayed rearrangement of synaptic weights of persistent connections, as well as greater susceptibility to spine shrinkage in response to chemical long-term depression. These results suggest a deficit in the aging brain’s ability to fine-tune synaptic weights to properly incorporate and retain novel memories. This research provides the first evidence of alterations in spine volumetric dynamics in healthy aging and may support a model of impaired processing and learning in the aged somatosensory system.
Significance Statement
Aging is known to impact cognitive functions and sensory processing, yet the underlying mechanisms at the synaptic level remain unclear. This study investigates dendritic spine dynamics of layer 5 pyramidal neurons in the aging somatosensory cortex. By employing two-photon excitation microscopy and analyzing spine volume changes during plasticity induction, we reveal that aged mice exhibit decreased volumetric volatility and delayed synaptic weight rearrangement. These alterations may impair the brain’s ability to fine-tune synaptic weights, which is crucial for incorporating and retaining new memories. This research provides the first evidence of altered spine volumetric dynamics in healthy aging, suggesting a potential mechanism for impaired processing and learning in the aged cortex. Understanding these changes could aid in mitigation of age-related cognitive decline.
Introduction
Increasing age is associated with decreases in cognitive function even without overt pathology (Kelly et al., 2006). Aged individuals show deficits in processing in major sensory modalities including hearing, vision, and touch (Thornbury and Mistretta, 1981; Kenshalo, 1986; Ebaid and Crewther, 2019; Felipe, 2019). In particular, aging is associated with decreased somatosensory thresholds (Thornbury and Mistretta, 1981) and deficits in spatial (Stevens and Patterson, 1995), textural (Skedung et al., 2018), and vibratory discrimination (Stuart et al., 2003). However, the mechanisms leading to these changes are not yet known.
Dendritic spines—small, membranous protrusions from the dendritic shaft that receive the majority of excitatory inputs onto neurons in the brain (Harris and Stevens, 1989)—are a potential site of dysfunction in aging. Spines are gained and lost both at baseline and in response to learning (Wilbrecht et al., 2010), and these dynamics are altered in the aged cortex (Mostany et al., 2013; Davidson et al., 2019; Voglewede et al., 2019) and in age-related neurodegenerative disease (Ortiz-Sanz et al., 2020; Mijalkov et al., 2021). The volume of the dendritic spine head correlates closely with the size of the postsynaptic density, number of AMPA receptors, and number of docked vesicles at the presynaptic site (Harris and Stevens, 1989; Matsuzaki et al., 2001; Arellano et al., 2007), allowing the use of volume measurements as a proxy for synaptic strength. Synapses modulate their strengths through Hebbian (Bliss and Collingridge, 1993; Coultrap et al., 2014; Noguchi et al., 2019; Huang et al., 2020; Runge et al., 2020; Tazerart et al., 2020) and homeostatic (Turrigiano et al., 1998; Hengen et al., 2013; Ma et al., 2019; Wu et al., 2021) plasticity mechanisms, incorporating novel information and tuning their input–output curves to maintain activity within biologically constrained ranges. In addition to these rule-based, activity-dependent mechanisms, spine volumes and strengths also undergo stochastic intrinsic dynamics, exhibiting size changes proportional to their volumes and tending toward the distribution median (Kasai et al., 2003; Yasumatsu et al., 2008; Ishii et al., 2018; Humble et al., 2019). Thus, the distribution structure of spine volumes remains constant despite flux of the spines composing it. The combination of discrete and volumetric dynamics allows both for changes in network connectivity and the permutation of strength rankings of stable synapses (Minerbi et al., 2009). As a spine’s morphology and prior stability may predict its future persistence (Loewenstein et al., 2015), changes in these parameters in aging may partly explain differences in discrete dynamics.
In this study, we utilized a dataset previously acquired in our laboratory (Voglewede et al., 2019) using in vivo two-photon laser scanning microscopy to measure discrete dynamics of dendritic spines in anesthetized mice at baseline and in response to plasticity induction. Images of dendritic fragments were reanalyzed to track the relative volumes of spines at 24 h intervals for baseline and poststimulation time points to determine how aging affects modulation of the strength of existing synapses. We saw a decrease in the dynamics of both overall distributions and individual spines in aged mice, which implied a deficit in the response of the aged cortex to sensory information. Aged mice exhibit a flattening of multiplicative dynamics following the final whisker stimulation session, preluding the delayed decrease in spine turnover observed in Voglewede et al. (2019). We also used 2PE microscopy to track the volumes of spines in acute cortical slices at 15 min intervals at baseline and following chemical long-term potentiation (cLTP) and depression (cLTD). We found that spines of young mice respond to cLTP with a gradual increase in spine volumes that is absent in aged mice, while spines of aged mice exhibit a transient decrease in volumes following cLTD that is absent in young mice. Together, these data suggest that the aged somatosensory cortex may be impaired in its ability to rapidly assimilate novel inputs into existing networks through fine-tuning of synaptic weights, potentially destabilizing network connectivity through altering spine dynamics following sensory experience.
Materials and Methods
Animals
For in vivo imaging experiments, Thy1-eGFP-M mice [The Jackson Laboratory, 007788, Tg(Thy1-EGFP)MJrs/J; RRID: IMSR_JAX:007788; Feng et al., 2000] at ages 3–5 months for young adults and 18–21 months for aged adults were used. Only male mice were used, as it has been reported (Alexander et al., 2018) that estrous cycle stage may affect whisker stimulation-induced plasticity. Food and water were provided ad libitum, and mice were group housed under a 12 h light/dark cycle. For slice experiments, strain and age ranges remained as above, but both males and females were used. All procedures were approved by the Tulane University Institutional Care and Use Committee and were performed per the NIH Office of Laboratory Animal Welfare’s Public Health Service Policy on Humane Care and Use of Laboratory Animals and Guide for the Care and Use of Laboratory Animals.
In vivo plasticity
Cranial window surgery
Cranial window surgery was carried out as described previously (Mostany and Portera-Cailliau, 2008; Holtmaat et al., 2009; Mostany et al., 2013; Voglewede et al., 2019). Anesthesia was induced with 5.0% isoflurane and subsequently maintained at 1.5–1.7%. Mice were then given subcutaneous injections of carprofen (5.0 mg/kg; Zoetis) and dexamethasone (0.2 mg/kg; MWI) to prevent inflammation and brain swelling. Mice were secured in a stereotaxic frame. A 4 mm craniotomy was performed using a pneumatic dental drill over the barrel field of the primary somatosensory cortex, centered 3 mm lateral to midline and 1.95 mm caudal to the bregma. A 5 mm glass coverslip (#1; Electron Microscopy Sciences) was then secured over the craniotomy using cyanoacrylate glue and dental acrylic (Lang Dental). A custom-made titanium bar (9.5 × 3.2 × 1.1 mm) was affixed using dental acrylic to the frontal bone of the skull to allow for the mouse to be fixed to the microscope stage using a small screw. Two to three weeks were allowed for recovery before imaging. This length of time was chosen as it has previously been demonstrated to be sufficient for recovery from inflammation as measured by neuronal morphology and glial morphology and has been employed successfully in our laboratory in multiple experiments (Holtmaat et al., 2009; Mostany et al., 2013; Alexander et al., 2018; Voglewede et al., 2019; Davidson et al., 2019).
Intrinsic optical imaging
Mice were anesthetized with 5.0% isoflurane and affixed to a custom-built intrinsic optical signal (IOS) microscope. Imaging was performed through the cranial window, as previously described (Grinvald et al., 1986; Alexander et al., 2018; Voglewede et al., 2019). A reference image of the vasculature was taken under a green (535 nm) LED array. The focus was then placed 250–350 μm below the dura. IOSs from the barrel cortex were taken under two arrays of red LEDs using a camera (Pantera 1M60; Dalsa), frame grabber (64 Xcelera-CL PX4; Dalsa), and custom-written MATLAB (MathWorks; RRID: SCR_001622) scripts. The whiskers were bundled and attached using dental wax to a glass capillary tube affixed to a piezo bender actuator (Physik Instrumente). IOS acquisition consisted of 30 trials of 1.5 s of rostrocaudal whisker stimulation at 10 Hz, separated by 20 s breaks. Frames 0.9 s before stimulation onset (baseline) and 1.5 s after stimulation (response) were collected, and the response signal for each trial was divided by the baseline signal. These were then summed to obtain the signal map, and an outline was placed over the vasculature reference image to identify activated cortex. During 2PE imaging, cells were chosen from within this region.
In vivo 2PE imaging
Mice were anesthetized with isoflurane (5.0% for induction, 1.0–1.5% for imaging) and secured to a custom-built 2PE microscope using the head cap bar installed during the cranial window surgery. Imaging of dendritic spines (Mostany et al., 2013; Alexander et al., 2018; Voglewede et al., 2019) was performed using a Ti:sapphire laser (Chameleon Ultra II; Coherent) at 910 nm, a 40 × 0.8 NA water-immersion objective (Olympus) and ScanImage software written in MATLAB (RRIG: SCR_014307; Pologruto et al., 2003). A reticle within the microscope ocular was used to identify a point of origin within the vasculature map, allowing coordinates of individual dendritic fragments to be recorded and reached via a micromanipulator controller (Sutter Instrument). Classification of cells as layer 5 (L5) pyramidal neurons was accomplished by verification of GFP expression and a soma depth 400–700 µm below the dura. In the Thy1-GFP line M mice used (Feng et al., 2000), neuronal expression of GFP is contained primarily to layer 5 pyramidal neurons with occasional expression in layers 2 and 3 and none in layer 4. Distance below the dura was thus used to ensure pyramidal neurons were not located supragranularly. For each cell, 5–8 dendritic fragments of dendritic tuft within layer I (15–115 μm below the pia) were imaged chronically every 24 h for 5 d. High-magnification stacks (40× optical with 5× digital magnification; 0.14 × 0.15 μm/pixel; 1.5 μm apart) were collected for analysis of dendritic spines.
Whisker stimulation
Mice were anesthetized with isoflurane (5.0% for induction, <1.0% for stimulation) and affixed to a platform using the titanium headbar. Whiskers were bundled and attached with dental wax to a piezo bender actuator oscillating rostrocaudally at 8 Hz (Carvell and Simons, 1990) for 10 min (Megevand et al., 2009; Gambino et al., 2014; Alexander et al., 2018; Voglewede et al., 2019; Williams and Holtmaat, 2019). Stimulation was performed within a 2 h window every 24 h for 3 consecutive days, beginning immediately after the second baseline imaging session.
Chemically induced structural plasticity
Brain slicing
Mice were anesthetized by isoflurane inhalation before being decapitated. The brain was rapidly removed and submerged in ice-cold sucrose solution containing the following (in mM): 234 sucrose, 25 NaHCO3, 7.0 MgCl2, 7.0 glucose, 2.5 KCl, 1.25 NaH2PO4, 0.5 CaCl2 with a pH of 7.3–7.4 and oxygenated by bubbling with 95% O2, 5% CO2. The brain was mounted to a detachable stage using cyanoacrylate-based glue and then sliced in 350 μm increments in the coronal plane on a vibratome (VT 1000 S; Leica Biosystems) while bathed in the iced sucrose solution. Slices were incubated for 15 to 30 min at 33–34°C in artificial cerebrospinal fluid (aCSF) containing the followinh (in mM): 125 NaCl, 25.0 NaHCO3, 25.0 glucose, 2.5 KCl, 2.0 CaCl2, 1.25 NaH2PO4, 1.0 MgCl2, with a pH of 7.3–7.4 and bubbled with 95% O2, 5% CO2. Slices were then maintained in room temperature aCSF for at least 1 h for recovery before being moved to the imaging chamber.
Chemical stimulation and imaging
Following the room temperature recovery, slices were moved to the imaging chamber of an ex vivo two-photon fluorescence microscopy imaging rig (SliceScope Pro 6000; Scientifica). Slices were continuously perfused with aCSF using a custom-built gravity pump at a rate of 2 ml/minute. Apical dendritic branches of layer 5 pyramidal neurons within barrel cortex were located with reference to a mouse brain atlas (Franklin and Paxinos, 2008) and imaged every 15 min for 45 min. For controls, standard aCSF was perfused in the bath throughout imaging. For cLTP experiments, following the second baseline imaging time point, aCSF was switched to magnesium-free aCSF to remove NMDA receptor magnesium blockade. This solution was identical to aCSF except for the removal of MgCl2. After 10 min of magnesium-free aCSF perfusion and the third baseline images, the bath was switched to aCSF with the addition of 100 µM glycine (Tocris) to induce neuronal activity and plasticity (Abe et al., 1990; Musleh et al., 1997; Zhang et al., 2014). Following 10 min of stimulation, the bath was returned to the original aCSF. Dendritic branches continued to be imaged every 15 min for another 45 min for three baseline and three poststimulation time points. For cLTD experiments, the bath was switched to aCSF containing 50 µm (S)-3,4-dihydroxyphenylglycine (DHPG, Tocris), a potent agonist of group I metabotropic glutamate receptors mGluR1 and mGluR5 demonstrated to cause rapid long-term depression of synapses (Palmer et al., 1997; Fitzjohn et al., 1999; Lodge et al., 2013). For each cell, 2–4 dendritic fragments of dendritic tuft within layer I (15–115 μm below the pia) were imaged. High-magnification stacks (40× optical with 5× digital magnification; 0.089 × 0.0.089 μm/pixel; 1.5 μm apart) were collected for analysis of dendritic spines.
Analysis
Spine volume measurements
Dendritic spine volumes were measured from 2PE laser scanning microscopy stacks from Voglewede et al. (2019, used with permission) and in vitro slices using routines within ScanImage software that were modified to enable acquisition of background-subtracted, normalized integrated fluorescence intensities (NII). Visible spines projecting 1/3 or more beyond the cross section of the adjacent dendritic shaft were manually annotated by demarcation of the spine beginning at the point of contact with the dendritic shaft and extending to the end of the spine head. For volume measurement, we assumed that spines are roughly spherical in shape and thus took measurements from the individual slice with the largest cross section for each spine. Regions of interest (ROIs) including the fluorescence of the spine, nearby background, and a portion of the adjacent dendritic shaft were then manually outlined for each dendritic spine. The average adjacent background was subtracted from the value of each pixel within the spine ROI, and the sum of the resulting values was divided by the average pixel value of the shaft. Thus, NII of a spine was given by the following:
Spine volume distributions and dynamics
Normality and log-normality of spine volume distributions were tested using the Shapiro–Wilk test before and after log-transformation, respectively. As spine volumes were better fit by a log-normal distribution, data was log-transformed in order to use parametric tests. Volume gain, loss, turnover, and log volume differences within age groups were calculated with repeated-measures one-way ANOVA with Bonferroni’s post hoc tests to correct for multiple comparisons. We define volume turnover as the sum of the magnitude of volume gain and loss per unit length in micrometer since the prior time point and report it per 100 μm of dendrite. As volumetric difference distributions deviated significantly from normality despite little skew, comparisons were made using nonparametric tests. When considering the population of spines combined from all neurons, one-way repeated-measures ANOVA was performed followed by bootstrapping of F statistics using custom-written Python scripts in order to determine robustness and reliability of results. The relationship between volume, volume changes, age, and time point were examined using multiple linear regression, with categorical variables of age for between-group tests, and time point, for within-group tests.
Synaptic weight rearrangement
Rearrangement of synapse rankings by volume was quantified using the rank-biased overlap measure (RBO; Webber et al., 2010). This measure compares rankings of items between two lists
U and
V, giving greater weight to those items at the top of the reference list
U. The rows of spine volumes (one row containing the volumes of an individual spine over time) are sorted in descending order according to their volume at the reference time point. The volumes are then converted to ordinal rankings. The RBO algorithm iterates through the list of spine rankings at increasing depths from 1 to N, with N being the number of all spines. The number of spines present in both the reference and comparison list at a given depth
d is divided by the depth to give the agreement
Statistical analysis
Statistical analyses were carried out using GraphPad Prism (GraphPad Software; RRID: SCR_002798) or custom Python scripts written in Python 3.10. Significance was set at p < 0.05. In all the figures, *p < 0.05, **p < 0.01, ***p < 0.001. Error bars depict the standard error of the mean (SEM) unless otherwise specified.
Results
Spine volume turnover in aged mice is higher at baseline due to increased discrete spine dynamics
For the examination of structural plasticity in vivo, we analyzed two time points prior to and three time points following introduction of whisker stimulation (Fig. 1a) to obtain measurements of dendritic spine relative volumes at baseline and in response to ongoing plasticity induction. Images were taken at an interval of 24 h in order to determine the effects of aging on properties of volumetric dynamics in the interval leading up to Day 4 in Voglewede et al. (2019), with each stimulation imaging session occurring 24 h following the corresponding stimulation. The days Baseline 1 through Stimulation 3 in our study correspond with −d1 through +d3 in Voglewede et al. (2019). Our previous study found that whisker stimulation resulted in increased discrete spine dynamics in young mice but had no effect in aged mice. At Day 4, following multiple sessions of whisker stimulation, aged mice began to exhibit a delayed decrease in discrete spine dynamics. We thus sought to determine whether this change was preceded by more subtle differences in volumetric dynamics. Representative images of a dendritic fragment on 2 consecutive days showing dendritic volume changes as well as an example of the spine, background, and shaft regions used to calculate spine volume are in Figure 1b.
Experimental timeline and spine volume measurement procedure. a, Following craniotomy and installation of a cranial window, mice were allowed 2–3 weeks to recover, and location of the barrel cortex was determined via intrinsic optical imaging (IOS). Two-photon imaging of dendritic spines was carried out longitudinally every 24 h for multiple baseline time points. Beginning after Baseline 2 imaging, whisker stimulation was carried out after each imaging session for multiple days (purple box). For our analysis, we examine two baseline days and three peristimulation days. b, Composite image of a dendritic fragment from a young mouse imaged at Baseline 2 and Stimulation 1. For these representative images, best projection images were created by using ImageJ to stitch together overlapping portions of the dendritic branch using the clearest segment in each slice. Relative volume measurement was carried out by summing all pixel intensity values within a spine region of interest (1) in the slice in which the spine profile was largest. A nearby background region (2) was then traced and its mean pixel intensity value multiplied by the area in pixels of (1) was subtracted from the previous sum. Finally, a region of adjacent dendritic shaft was traced (3) and the background-subtracted sum was divided by the average pixel intensity value of this region to produce the spine NII. Examples of spines decreasing (blue arrows) or increasing (orange arrow) in volume over the course of 24 h.
We first compared discrete dynamics of dendritic spines between age groups at baseline to align our results with those in Voglewede et al. (2019). Aged mice had greater spine gain (young, 2.14 ± 0.26; aged, 3.5 ± 0.34; t(26) = 3.020; p = 0.0056; unpaired t test; Fig. 2a), a trend toward greater spine loss (young, 2.9 ± 0.30; aged, 3.655 ± 0.2663; t(26) = 1.793; p = 0.0846; Fig. 2b), and greater spine turnover (young, 5.1 ± 0.52; aged, 7.127 ± 0.55; t(26) = 2.690; p = 0.0123; Fig. 2c) compared with young mice, consistent with the original findings. Though we observed lower overall magnitudes for each metric, this is likely due to exclusion of spines projecting in the z-direction, which were included in quantification of discrete spine dynamics but which precluded accurate volume measurements in the present study. Our densities, which would be expected to exclude roughly 30–50% of spines, are approximately half of what was observed in prior studies (Mostany et al., 2013; Voglewede et al., 2019), thus putting the true densities in line with these experiments.
Increased spine volume loss and turnover in aged mice primarily result from elevated spine discrete dynamics. Discrete spine dynamics, Young adult (orange) versus aged (blue) baseline spine gain (a), loss (b), and turnover (c). Volumetric analysis of all spines, Spines were divided into those that decreased in volume or were lost, and those that increased in volume or were gained. Volume gain (d), loss (e), and turnover (f). Volumetric analysis of persistent spines, Spines were divided as before but contributions from spine gain or loss were excluded. Volume gain (g), loss (h), and turnover (i). Unpaired t tests. For all datasets: n = 13 cells/6 mice (young adult), 15 cells/6 mice (aged); bar graphs are presented as mean ± SEM; individual cells marked as black circles; *p < 0.05.
We next examined volume loss, gain, and overall volume turnover (defined as volume gained + volume eliminated per 100 µm of dendritic shaft) of dendritic spines between age groups at baseline. No difference was observed between age groups for volume gain (young, 128.3 ± 39.87 a.u.; aged, 129.3 ± 38.03 a.u.; t(27) = 0.3779; p = 0.7085; unpaired t test; Fig. 2d). Greater volume loss was observed in aged versus young adult mice (young, 114.2 ± 32.20 a.u.; aged, 145.1 ± 29.49 a.u.; t(27) = 2.695; p = 0.0120; unpaired t test; Fig. 2e). Combining volume loss and gain, aged mice exhibited greater volume turnover (274 ± 47.89 a.u.) versus young adult mice (238 ± 40.55 a.u.; t(27) = 2.197; p = 0.0386; unpaired t test; Fig. 2f). These results were in agreement with those observed for spine turnover and were primarily the result of elevated discrete dynamics in aged mice, as when analysis was restricted only to spines present at both baseline time points no differences were observed for volume gain (young, 72.13 ± 7.30 a.u.; aged, 63.08 ± 6.10 a.u.; t(27) = 0.9565; p = 0.3473; Fig. 2g), loss (young, 56.91 ± 7.60 a.u.; aged, 67.97 ± 6.31 a.u.; t(27) = 1.126; p = 0.2700; Fig. 2h), or turnover (young, 129.0 ± 7.56 a.u.; aged, 131.0 ± 9.26 a.u.; t(27) = 0.1666; p = 0.8689; Fig. 2i). This suggests that at baseline, age-related differences in changes in the level of excitatory input onto apical dendrites of L5 pyramidal neurons in S1 are primarily due to altered rates of formation and elimination of spines.
Spine volume distributions exhibit decreased volatility in aged mice
We next examined the distributions of dendritic spine volumes for individual neurons and for spines pooled across all neurons within each age group, focusing on spines which persisted across all five time points. Spine sizes formed a skewed, heavy-tailed distribution (Fig. 3a). This is in line with the log-normal distributions previously reported in apical dendrites of layer 5 pyramidal auditory cortex in mice (Loewenstein et al., 2011), though other shapes such as gamma distributions have been reported (Benavides-Piccione et al., 2013). Normality tests of individual and pooled spine distributions before log-transformation indicated that the raw distributions were significantly non-normal at all time points for both age groups (p < 0.0001; Shapiro–Wilk test). Following log-transformation, distributions were visually less-skewed (Fig. 3b). However, results of the Shapiro–Wilk test were mixed, with the transformed distributions of some neurons still significantly deviating from normality. A model comparison test indicated that the majority of neurons exhibited a likelihood ratio >95% of coming from a log-normal distribution with pooled spines exhibiting a 100% likelihood ratio at all time points. Thus, for the purposes of the following analyses, we assumed spine sizes were sampled from a log-normal distribution and performed parametric tests on log-transformed volumes.
Pooled spine volume distributions exhibit decreased volatility in aged mice. a, Spine volume distributions are skewed and heavy tailed and can be approximated by a log-normal distribution. b, Following log-transformation, spine volume distributions are more symmetric. c, d, Mean log volume of individual neurons over time in young mice and aged mice, respectively. e, f, Mean log volume of pooled spine distributions over time in young and aged mice, respectively. One-way repeated-measures ANOVA with Bonferroni’s post hoc tests to correct for multiple comparisons. Orange and blue background highlights time points following whisker stimulation sessions. F statistics were bootstrapped 5,000 times for spine volume distributions of (g) young and (h) aged mice in order to assess the reliability of results in Figure 3. Histograms depict frequency of F values. Solid curves depict the smoothed frequency distribution. Dotted lines show the F value from the original pooled spine data. Error bars indicate SEM for individual neurons and 95% confidence intervals for pooled spines. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
To determine the response of neuronal spine volumes in response to whisker stimulation, a repeated-measures ANOVA was performed to compare the effects of stimulation on mean log volume. When considering individual neurons, there was no statistically significant difference in mean log volume between time points (young: F(4,12) = 2.147, p = 0.117, Fig. 3c; aged: F(4,10) = 1.631, p = 0.2080, Fig. 3d). When considering individual spine volumes as samples taken from a more representative population of dendritic spines, a repeated-measures ANOVA showed a significant effect of time for young (F(4,952) = 10.10; p < 0.0001; Fig. 3e) but not aged (F(4,1013) = 0.2611; p = 0.8939) spines. Young mice exhibited a general upward trend in population mean log volume over time, with a significant increase in volume from Baseline 1 (M = 3.072; SD = 0.609) to Baseline 2 (M = 3.115; SD = 0.584), and from both Baseline 1 and 2 to Stimulus 2 (M = 3.157; SD = 0.5649), while the spine population in aged mice remained flat. As such a test would be expected to pick up even minute differences as statistically significant, this suggests that the volume distribution of pooled persistent spines in aged mice is quite stable relative to that in young mice. To assess the robustness and reliability of the results for pooled spine volumes, we bootstrapped F statistics by resampling from the original spine volume trajectories with replacement for 5,000 replications. A bootstrapped p value was then calculated as the number of F statistics higher than our original F divided by the total number of replications. For young mice, we observed a bootstrapped p of 0.033 (Fig. 3g). For aged mice, the observed bootstrapped p was 0.744 (Fig. 3h). These indicate that were the null hypothesis that time/stimulation has no effect on mean log volume of our distribution true, we would observe an effect as extreme as that observed in our original data 3.3 or 74.4% of the time, respectively. The results of the bootstrap thus indicate that our results are robust and not simply due to chance variations within the data.
Volume distributions of lost and gained spines are also skewed and heavy tailed
Spine volume and postsynaptic density covary over time and are positively correlated with spine stability (Cane et al., 2014), resulting in smaller spines being more likely to be lost than their larger counterparts. As aged mice have been demonstrated to gain and lose more spines per unit time than young mice, we hypothesized that the distributions of volumes of these transient spines may be altered in aged mice. However, to our knowledge, the volume distributions of gained and lost spines have not been reported. Plotting the distributions of volumes on the previous day of all spines lost (Fig. 4a) and gained (Fig. 4b) in each age group across all time points together with the distributions of all spines at baseline shows that lost and gained spines also have skewed, heavy-tailed distributions. These thus resemble the parent populations but exhibit a smaller mean log volume. This implies that the probability of spine loss within a given time interval is proportional to its volume. Over longer intervals, the mean log of the population of gained or lost spines might be expected to asymptotically approach the mean log of the steady-state population of all spines; over shorter intervals, it would tend to zero. This potentially allows for use of the mean log volume of lost spines across two different time intervals in the construction of a time constant for spine survival within the steady-state population.
Volume distributions of gained and lost spines show differences with age. Distributions of lost (a) and gained (b) spine volumes plotted against volumes of all spines at baseline show a similar structure with a smaller mean log volume. Black bars denote the median. Two-sample Kolmogorov–Smirnov test. c, Comparison between age groups of mean log volume of lost spines by neuron. d, Comparison between age groups of mean log volume of gained spines by neuron. For unpaired t tests, n = 14 neurons/6 mice (young) and 15 neurons/6 mice (aged). Black bars denote the mean, error bars denote the SEM. *p < 0.05, ***p < 0.001.
There was a significant difference between the distributions of gained (Dg = 0.1154; p = 0.0007) but not lost (Dl = 0.0657; p = 0.0880) spines when compared between age groups by the two-sample Kolmogorov–Smirnov test (Fig. 4b). When compared within neurons by unpaired t test, aged mice exhibited a significantly lower mean log volume for gained (young, 2.705 ± 0.02714; aged, 2.580 ± 0.04321; t(27) = 2.403; p = 0.0234; Fig. 4d) but not lost (young, 2.684 ± 0.0338; aged, 2.591 ± 0.05213; t(27) = 1.466; p = 0.1541; Fig. 4c) spines.
Spine volume change distributions exhibit decreased volatility in aged mice
The greater movement of the spine mean log volume in young mice over time compared with aged mice (Fig. 3) raises the question of whether the distributions of volumetric changes of spines within the populations are also less volatile in aged mice. The volume distributions can be interpreted as the steady-state structure resulting from the dynamical properties of individual spines but say little about those movement outside of the properties of set point and multiplicative dynamics governing their movements. We examined the distributions of volume changes of individual spines and how these are affected by stimulation. Plotting volume change distributions for individual neurons and pooled persistent spines for young mice showed that, at baseline, volume changes form a leptokurtic distribution with a mean slightly larger than zero (Fig. 5a). Repeated-measures one-way ANOVA was performed to compare the effects of stimulation on volume changes. There was again an effect of stimulation for spines of young (F(3,952) = 8.326; p < 0.0001; Fig. 5b) but not aged (F(3,952) = 2.321; p = 0.0772; Fig. 5c) mice. As with the pooled spine volumes, we performed bootstrapping of the F statistic for 5,000 replications to assess the reliability of these results. The bootstrapped p values for young and aged mice were 0.078 (Fig. 5d) and 0.261 (Fig. 5e), respectively. This indicates that though our spine volume change results are potentially more susceptible to sampling variability than our results for spine volumes, they are still reasonably robust. Together, these suggest that both the volume and volume change distributions of spines are less responsive to plasticity induction in aged pyramidal neurons of somatosensory cortex.
Spine volume change distributions are less responsive to plasticity induction in aged mice. a, Distributions of spine volume changes for individual neurons (light gray lines) and pooled spines (dark gray line). b, Mean volume change over time for spines of young mice. c, Mean volume change over time for spines of aged mice. One-way repeated-measures ANOVA with Bonferroni’s post hoc tests to correct for multiple comparisons. Orange and blue background highlights time points following stimulation sessions. F statistics were bootstrapped 5,000 times for spine volume difference distributions of (d) young and (e) aged mice in order to assess the reliability of results in Figure 5. Histograms depict frequency of F values. Solid curves depict the smoothed frequency distribution. Dotted lines show the F value from the original pooled spine data. Error bars indicate 95% confidence intervals for pooled spines. *p < 0.05, **p < 0.01.
Large spines show delayed relative reordering in aged mice
Given the correlation between spine size and synapse strength, differences in the volatility of spine volumes may manifest in a change in the rate at which spines reorder themselves in terms of their relative potential contribution to depolarization of their parent cell (Harris and Stevens, 1989; Iñiguez et al., 2022). Previous studies have examined how the ranking of spines by volume or strength evolves over time (Minerbi et al., 2009) but have used metrics such as Kendall tau distance that do not take into account the strength of the synapses that have reordered. That is, if the largest spine on a neuron shrinks significantly so that it decreases in rank by 50, this may cause a greater impact on the postsynaptic neuron than if the smallest spine on the neuron increases in rank by 50. We thus sought to find a metric that would take into account the initial importance of each synapse to the cell while also allowing for new elements to appear or disappear from the rankings as spines are gained or lost by the neuron.
We thus employed a permutation measure known as rank-biased ordering (Webber et al., 2010). This measure compares two lists—the reference list and the comparison list—at increasing depths. Here, the reference list is a column containing spine volume rankings at a given time point after ordering by volume—which is simply an ordered list containing the numbers 1 through the number of spines. The comparison list contains the volume rankings for each spine following volumetric changes observed at a later time point. Thus, a given index within the two lists refers to the same spine, while the value at that index refers to its ranked volume. At each depth the overlap of items in the lists is calculated and weighted according to an exponentially decreasing weighting function whose decay is governed by the parameter p. The contribution of overlap at each depth is summed with that of prior depths until the end of the list is reached, resulting in a number between 0 and 1 that provides a measure of the similarity between the two lists. A value of 1 indicates that the lists are identical and thus that the spines have remained in the same order by volume, while a 0 indicates that they share no elements and thus that there has been complete turnover. Intermediate numbers indicate intermediate levels of reordering and connectivity change. When applied solely to persistent spines, the measure quantifies volumetric reordering, while allowing for loss and gain of spines provides a broader picture covering both fine-tuning and connectivity changes. This thus allows us to compare the rate at which two spine populations diverge from their initial ordering while giving more weight to larger spines. The weighting formula and example calculations are provided in Figure 6a,b.
Cumulative and consecutive RBO of persistent spines show delayed reordering of large spines in aged mice. a, Depiction of relative remodeling of synapses. The largest two spines have swapped rankings due to a decrease in volume of the largest spines. The smallest spine has changed volume but retains the same ranking. b, Example calculations of two potential reorderings demonstrating the effect of swapping items higher or lower in the list. c, RBO of persistent spines in young (orange) and aged (blue) mice compared against rankings at Baseline 1. d, RBO of persistent spines when each time point is compared against the prior time point. For c and d, the parameter p was set to 0.992, giving 90% of the weight to the top 200 (∼20%) spines. e, f, As in a and b, but with parameter p set to 0.9995, giving 30% of the weight to the top 200 spines. Shaded areas depict subsampled 95% confidence intervals. Flattening of weights shows less divergence of reordering. g, RBO of all spines in young (orange) and aged (blue) mice compared against rankings at Baseline 1. h, RBO of all spines when each time point is compared against the prior time point. For g and h, the parameter p was set to 0.992. i, j, As in e and f, but with parameter p set to 0.9995 for decreased top-weighting. Shaded areas depict subsampled 95% confidence intervals. Flattening of weights shows less divergence of reordering.
The combined weight of elements 1 through d within the list of ranks is given by the following equation:
Larger spines of aged mice show decreased tendency to shrink following multiple rounds of stimulation
Langevin dynamics are a defining feature of the volume changes of individual spines in hippocampus (Yasumatsu et al., 2008) and cortex (Ishii et al., 2018). Spines continuously undergo apparently stochastic movements even after the silencing of network activity or blockade of calcium-dependent plasticity mechanisms (Yasumatsu et al., 2008). The change in a spine’s volume between time points tends to be proportional to the original size of the spine, a property known as multiplicative dynamics. Furthermore, small spines tend to grow larger, while large spines shrink, with overall movement toward a set point near the median of the population distribution. These properties together give rise to the characteristic log-normal distributions observed earlier (Fig. 3a). As small changes in these dynamics have the potential to cause large changes in population structure and spine stability (Humble et al., 2019), we sought to determine whether the relationship between spine volume and subsequent volume change were altered in aged mice. Multiple regressions were run at each time point to predict volume change from initial volume and age (Table 1).
Multiple linear regression of relationship of volume change with initial volume and age
The emergence of age as a significant predictor following multiple stimulations suggested a change in the relationship between spine volume and subsequent volume change. Plotting of linear regressions of initial volume versus volume change for each time point (Fig. 7, top row) illustrates this, with a large divergence in the slope of the regression lines following Stimulus 3. To better understand the nature of this change, we also plotted a rolling window average of 100 spine volumes against the average volume change of the window for each time point (Fig. 7, bottom row). This suggests that while volume changes are similar between age groups for smaller spines, large spines in aged mice exhibit less decrease in volume on average after repeated stimulation.
Linear regressions and rolling window average show less negative slope of volume–volume change relationship following multiple days of stimulation in aged mice. Linear regressions (top row) were plotted for each time point to show changes in volume–volume change relationship over time between age groups. A rolling window average volume of 100 spines was plotted against the average volume change within the window to better visualize population differences between young and aged mice (bottom row).
The change seen in multiplicative dynamics after Stimulus 3 has two potential interpretations: (1) aged large spines might exhibit less volume decrease on average because they exhibit less volume change overall or (2) because they now have a lower tendency to decrease in volume. In order to distinguish between these possibilities, we looked at the rolling average of 100 spine volumes against the average absolute volume change (Fig. 8a) and against the window standard deviation (Fig. 8b). However, neither the absolute value change nor the standard deviation appear decreased in the large spines of aged relative to young mice, implying that the large spines have not become less volatile, but rather exhibit decreased attraction to original set point.
Absolute volume change and standard deviation as a function of initial volume are similar between young and aged mice following the third stimulation despite a less negative slope of multiplicative dynamics in aged mice. For the imaging time point following the third stimulation—at which point spines of aged mice exhibited a sharp decrease in the slope of volume change versus volume—a rolling window average of 100 spine volumes was plotted against the window average absolute volume change (a) or the window standard deviation (b). Slopes of linear regression lines of these curves were similar between age groups, suggesting that large spines exhibit similar magnitudes of changes in volume.
Fast intrinsic dynamics of spines of young and aged barrel cortex are similar at baseline
While the above experiments tracked the fluctuation of spine volumes at 24 h intervals, changes in spine volume can be observed on the timescale of minutes to hours. We thus sought to determine whether spines of aged mice also exhibit differences in short timescale dynamics at baseline and in response to chemical plasticity induction. To this end, we used 2PE microscopy to image spines of apical dendrites of layer 5 pyramidal neurons of barrel cortex in acute coronal slices at 15 min intervals for 90 min under three protocols: (1) during application of standard aCSF (control; Fig. 9), (2) with induction of cLTP via glycine-containing aCSF after 45 min, and (3) with induction of cLTD via DHPG-containing aCSF after 45 min (Fig. 10).
Baseline volumetric spine dynamics are similar in young and aged mice in vitro. Composite images of dendritic fragments imaged at T-30, T0, and T30 for (a) young and (b) aged. All representative images are best projections. T0 corresponds to the time point at which chemical plasticity induction occurs in Figure 10. c, Mean log volume of spine distribution over time in young mice (orange). ΔV1 denotes the interval (delta) between the first two time points, ΔV2 between the second two time points, and so on. d, Mean log volume of spine distribution over time in aged mice (blue). e, Mean of volume change distribution over time for young mice. f, Mean of volume change distribution over time for aged mice. g, h, Linear regressions of initial volume versus subsequent volume change for spines aggregated over time.
Aged mice show impaired structural cLTP and enhanced structural cLTD of spines compared with young mice. a, Composite images of dendritic fragments from young and aged mice imaged at the two baseline time points and 30 min after cLTP induction via glycine. All representative images are best projections. b, Mean log volume of dendritic spines from barrel cortex slices of young mice at baseline and following glycine-mediated induction of cLTP. c, As in b but for aged mice. d, Composite images of dendritic fragments from young and aged mice imaged at the two baseline time points and 30 min after cLTD induction via DHPG. All representative images are best projections. e, f, As in b and c, but with DHPG-mediated induction of cLTD. n = 5 mice for all groups. One-way repeated-measures ANOVA with Dunnet’s post hoc test for multiple comparisons against the final time point using standard ACSF prior to chemical plasticity induction (Baseline 2 for glycine experiments, Baseline 3 for DHPG experiments). *p < 0.05, **p < 0.01.
To establish baseline intrinsic dynamics and volatility, spines were pooled by age group (young, n = 166 spines/3 mice; aged, n = 228 spines/5 mice) and compared across time points by one-way repeated-measures ANOVA following log-transformation. Max projection representative images of dendritic spines across the imagine session are shown for young (Fig. 9a) and aged (Fig. 9b) mice. Spine distributions of neither young (F(5,165) = 0.117; p = 0.9506; Fig. 9c) nor aged (F(5,227) = 1.115; p = 0.3402; Fig. 9d) mice exhibited a significant change in mean log volume over time. Volume difference distributions between time points were also compared. Again, neither young (F(4,165) = 0.1324; p = 0.9665; Fig. 9e) nor aged (F(4,227) = 1.082; p = 0.3631; Fig. 9f) mice differed over time in spine volume change distributions. Spine volumes and changes were thus aggregated over time points and multiple linear regression was carried out to determine the relation of volume change with initial volume and age. The regression was significant (F(2,1657) = 34.52; p < 0.0001) with volume arising as a significant predictor of volume change (F(1,1657) = 68.31; t = 8.265; p < 0.0001). Age was not a significant predictor of volume change (F(1,1657) = 0.223; t = 0.473; p = 0.637). Individual linear regressions of aggregated spine volumes against volume change were plotted for young (Fig. 9g) and aged (Fig. 9h) spines. While regression line slopes were negative and significantly different from zero (young, slope = −0.1101, p < 0.0001; aged, slope = −0.08782, p < 0.0001), this appears slight relative to that observed in vivo over 24 h intervals, as expected, particularly when taking into account differences in the normalized integrated intensity units between our in vivo and ex vivo setups.
Aged mice exhibit altered susceptibility to chemically induced structural potentiation and depression
We next examined the effect of chemical induction of potentiation or depression on spine volumes through a 10 min application of either 100 µM glycine (for cLTP) or 50 µM DHPG (for cLTD) into the bath following the initial two imaging time points. For cLTP, magnesium-free aCSF was introduced following the second time point in order to remove magnesium blockade of NMDA receptors. Following 10 min of stimulation, standard aCSF was reintroduced to allow for washout, with imaging continuing for three more time points. Representative images of dendritic fragments from young and aged mice at baseline and following glycine treatment are shown in Figure 10a. Dendritic spine volumes were log-transformed and the means compared over time within each group by one-way repeated-measures ANOVA. Following cLTP induction, young (n = 5; F(5,4) = 19.54; p = 0.0012; Fig. 10b) but not aged (n = 5; F(5,4) = 0.6706; p = 0.5608; Fig. 10c) mice showed an effect of glycine treatment. Post hoc multiple comparisons against the baseline prior to cLTP induction showed a significant increase in mean log volume 30 min (Baseline 2, 4.09 ± 0.048 a.u.; Glycine 2, 4.253 ± 0. 046 a.u.; p = 0.0493; Dunnet’s post hoc test) and 45 min (Glycine 3, 4.280 ± 0.048 a.u.; p = 0.0075) after introduction of glycine. Following cLTD induction, an opposite effect was seen. Representative images of dendritic fragments from young and aged mice at baseline and following glycine treatment are shown in Figure 10d. Aged (n = 5; F(5,4) = 9.161; p = 0.0123; Fig. 10f) but not young (n = 5; F(5,4) = 2.430; p = 0.1391; Fig. 10e) showed an effect of DHPG treatment. Post hoc multiple comparisons against baseline prior to cLTD induction showed a significant decrease in mean log volume 30 min after introduction of DHPG (Baseline 3, 4.137 ± 0.072 a.u.; DHPG 2, 3.772 ± 0.046 a.u.; p = 0.0131) as well as a trend toward a decrease after 15 min (DHPG 1, 3.881 ± 0.0730 a.u.; p = 0.07303). Forty-five minutes after DHPG treatment, spine volumes had recovered (DHPG 3, 4.066 ± 0.086 a.u.; p = 0.8827). Taken together, these data suggest that spines of aged mice are less responsive to rapid structural potentiation, mirroring results seen in vivo, and more susceptible to structural depression, at least transiently.
Discussion
Decreases in sensory acuity during aging correlate with changes in organization and representation in the brain. Impaired tactile acuity in aging humans has been linked to enlarged representations of the hands in the somatosensory cortex (Kalisch et al., 2009) and increased excitability of neurons within this region (Lenz et al., 2012). Similar observations of increased excitability of the somatosensory cortex and a correlation of expansion of hindpaw representation with impaired gait in rodents suggest that this may be a general feature of the aging cortex (David-Jürgens et al., 2008; David-Jürgens and Dinse, 2010).
Aging appears to be associated with an increase in spine turnover on apical dendrites of pyramidal neurons at baseline within multiple primary cortical regions (Mostany et al., 2013; Davidson et al., 2019; Huang et al., 2020). In addition, aged individuals often show declines in memory and processing speed and require more repetitions to learn a task (Zanto et al., 2010; Tagliabue et al., 2020). Following multiple daily repetitions of a sensory stimulus, spine turnover in aged mice was observed to decrease to levels seen in young mice at baseline (Voglewede et al., 2019). In contrast to young mice, which exhibited an immediate increase in turnover, aged mice showed no changes in spine turnover for several days. However, it is unknown whether changes in volumetric volatility or intrinsic dynamics precede this delayed shift in plasticity.
Our in vivo data suggest that the dynamics of both population distribution structure and individual dendritic spines are decreased in aged mice (Voglewede et al., 2019). In both groups, spine volumes are constantly changing and appear to follow stochastic, multiplicative dynamics at baseline (Yasumatsu et al., 2008; Minerbi et al., 2009; Loewenstein et al., 2011). These intrinsic dynamics may seem counterproductive to long-term storage of memory traces, but biological and computational studies show that they improve function of neural networks and that slight deviations in their properties can result in learning deficits (Nagaoka et al., 2016; Ishii et al., 2018; Humble et al., 2019). During learning, patterned activity may result in formation of recurrently connected cell assemblies (Liu et al., 2012; Nabavi et al., 2014), which self-stabilize via Hebbian mechanisms (Matsuzaki et al., 2001, 2004; Smith et al., 2003; Asrican et al., 2007; Zito et al., 2009). Without intrinsic dynamics, such assemblies recruit new neurons through spurious firing correlations, degrading the original memory trace while simultaneously resulting in the runaway aggregation of large spines through positive feedback loops. Homeostatic plasticity provides one mechanism by which informational structure may be preserved through scaling of synaptic weights within neurons or firing rates within circuits (Turrigiano et al., 1998; Hengen et al., 2013; Ma et al., 2019; Wu et al., 2021). However, homeostatic plasticity preserves the synaptic volume rankings and thus cannot solve the issue of integration of unrelated connections within an assembly. Through intrinsic dynamics, accidental recruitment becomes less likely, as weights will tend to be pushed back toward the set point before incorporation into the cell assembly.
Modeling the intrinsic dynamics of the fmr1KO mouse model of fragile X syndrome, which shows spine volume distributions similar to wild-type mice (Cruz-Martín et al., 2010; Ishii et al., 2018; Humble et al., 2019) found that stronger dynamics resulted in cell assemblies unable to properly self-stabilize and suggested this as a potential cause for some of the cognitive issues associated with the syndrome. The slower reordering of synaptic weights and poststimulation divergence of multiplicative dynamics we observe in aged mice may cause a similar phenomenon. In aged mice, the largest spines may exhibit excess stabilization in the short term, rearranging their rankings at a slower pace. Following repeated stimulations, rearrangement accelerates but larger spines of aged mice appear to break their tendency to decrease in volume. A possible consequence of intrinsic dynamics is the weakening of synapses strengthened by accidental correlations in activity and a breakdown in this mechanism could result in integration of unrelated traces within cell assemblies. Dysfunction could thus explain the loss of differentiation between representations observed in aged cortex (Bowman et al., 2019; Simmonite and Polk, 2022) as engrams begin to cross-link and subsume one another.
Previous studies observe that the rate of spine gain and loss can be modeled by the passage of small spines beyond a “reflecting boundary” at the lower end of the volume distribution (Yasumatsu et al., 2008), with small spines more likely to be lost due to their proximity to this boundary. This may explain the decrease in spine loss observed, with spines possessing volumes above the median volume increasing their probability of staying at the upper end of the lost spine volume distribution shown in Figure 4a. While this might preserve stronger connections important to circuit integrity, it may come at the expense of integration of new information by decreasing the ability of the circuit to adapt and requiring further patterned activity onto a subset of the smaller, more volatile spines to induce growth and subsequent stabilization. However, it has been shown that whisker stimulation results in a decrease in the density of persistent newly gained spines in aged mice (Voglewede et al., 2019). Given our observation that newly formed spines in aged mice are smaller compared with those in young mice, decreased persistence of these nascent spines may be due to their higher probability of movement back below the boundary.
Our study also found that spines of aged and young mice exhibit qualitatively different responses to induction of chemical plasticity. Spines of aged mice appeared less responsive to glycine-mediated cLTP (Kim et al., 2020; Cho et al., 2023) while appearing to be more susceptible to DHPG-mediated cLTD, in line with previous results showing that aged synapses are more vulnerable to low-frequency stimulation induction of LTD or reversal of LTP in rats (Norris et al., 1996). One possibility, raised in Voglewede et al. (2019), is that elevated spine turnover in aged animals may be due to an overdriven system at baseline, resulting in a failure to respond to further stimuli. Thus, the lack of response to glycine stimulation may simply be an analog of this ceiling effect. Similarly, despite high prevalence of significant decrease in at least one sensory modality with age (Völter et al., 2021), few studies have examined the effect of sensory deprivation on the aging cortex. Our observation that aged barrel cortex is more susceptible to cLTD provides some support for existing theories of cognitive decline in aging such as the sensory deprivation hypothesis, which suggests that a decrease in sensory input results in atrophy of cortical systems (Oster, 1976; Valentijn et al., 2005), and the information degradation hypothesis, which posits that age-related degradation of perceptual inputs affects downstream higher-order cognitive processes resulting in greater expenditure of resources to interpret the weakened signals (Bruffaerts et al., 2019).
Further study is necessary to determine the molecular mechanisms underlying observed changes in intrinsic dynamics and reordering in aging. As these fluctuations occur even with blockade of calcium- and activity-dependent plasticity mechanisms (Yasumatsu et al., 2008; Minerbi et al., 2009), it is believed that they may partly result from stochastic assembly and disassembly of actin filaments (Kasai, 2023). Age-related effects on proteins involved in actin dynamics are thus targets for future research. For instance, the serine/threonine protein kinase calcium–calmodulin-dependent kinase II (CaMKII) plays a vital role in the remodeling of the actin network of the postsynaptic site following synaptic stimulation (Khan et al., 2019). CaMKII is typically activated via Ca2+ influx through NMDAR receptors. However, Ca2+/CaM-independent activity of CaMKII may be altered via oxidation by reactive oxygen species, which are elevated with aging (Bodhinathan et al., 2010; Khan et al., 2019), thus affecting spontaneous actin remodeling. Since activity independence of intrinsic dynamics has only been demonstrated in young mice (Yasumatsu et al., 2008; Minerbi et al., 2009; Ishii et al., 2018), this property should also be confirmed in the aged brain.
Taken together, these results suggest that incorporation of new information in the aged brain may require repeated patterned activity to overcome a neural system that may prioritize stabilizing existing information. Failure of spines in aged mice to respond to stimulation is likely to result in a decrease in the number of novel cell assemblies that self-stabilize or in linkage of unrelated cell assemblies (Humble et al., 2019; Dorkenwald et al., 2022). On the other hand, both whisker stimulation and chemical plasticity induction result in broad, unpatterned activation of the cortex, and it is possible that targeted approaches may reveal that the aged cortex is more amenable to learning than it currently appears. Further study is necessary to determine whether changes in spine volumetric dynamics causally contribute to changes in turnover in aging or merely precede them.
In summary, we provide the first study examining volumetric fluctuations of dendritic spines in the aging somatosensory cortex and linking them to previous observations of altered discrete spine dynamics. Aged mice exhibit reduced synaptic weight reordering, a dampened response to stimulation, and heightened vulnerability to synaptic depression, supporting at a finer scale prior studies showing reduced plasticity and failure to incorporate novel sensory experiences.
Footnotes
This work was supported by a Louisiana Board of Regents Graduate Research Fellowship (LEQSF(2013-18)-GF-17) to R.L.V and by grants from the National Institute on Aging (R01AG047296, R01AG074489, and R56AG072676) to R.M.
The authors declare no competing financial interests.
- Correspondence should be addressed to Ricardo Mostany at rmostany{at}tulane.edu.
