Molecular and histological correlates of cognitive decline across age in male C57BL/6J mice

Abstract Introduction Increasing age is the number one risk factor for developing cognitive decline and neurodegenerative disease. Aged humans and mice exhibit numerous molecular changes that contribute to a decline in cognitive function and increased risk of developing age‐associated diseases. Here, we characterize multiple age‐associated changes in male C57BL/6J mice to understand the translational utility of mouse aging. Methods Male C57BL/6J mice from various ages between 2 and 24 months of age were used to assess behavioral, as well as, histological and molecular changes across three modalities: neuronal, microgliosis/neuroinflammation, and the neurovascular unit (NVU). Additionally, a cohort of 4‐ and 22‐month‐old mice was used to assess blood‐brain barrier (BBB) breakdown. Mice in this cohort were treated with a high, acute dose of lipopolysaccharide (LPS, 10 mg/kg) or saline control 6 h prior to sacrifice followed by tail vein injection of 0.4 kDa sodium fluorescein (100 mg/kg) 2 h later. Results Aged mice showed a decline in cognitive and motor abilities alongside decreased neurogenesis, proliferation, and synapse density. Further, neuroinflammation and circulating proinflammatory cytokines were increased in aged mice. Additionally, we found changes at the BBB, including increased T cell infiltration in multiple brain regions and an exacerbation in BBB leakiness following chemical insult with age. There were also a number of readouts that were unchanged with age and have limited utility as markers of aging in male C57BL/6J mice. Conclusions Here we propose that these changes may be used as molecular and histological readouts that correspond to aging‐related behavioral decline. These comprehensive findings, in the context of the published literature, are an important resource toward deepening our understanding of normal aging and provide an important tool for studying aging in mice.


INTRODUCTION
Aging is associated with a progressive decline in numerous functions and an increased incidence of frailty and disease Hou et al., 2019;Kane et al., 2019;Lopez-Otinet al., 2013).
Specifically in the aged brain, there is a loss of synaptic connections (Morrison & Baxter, 2012), increased neurodegeneration (Wyss-Coray, 2016), heightened neuroinflammatory responses (Spencer et al., 2017) from both microglia (Niraulaet al., 2017) and astrocytes (Boisvert et al., 2018), a greater number of infiltrating macrophages from the periphery (Scheiblich et al., 2020), vascular dysfunction , loss of blood-brain barrier (BBB) integrity (Benveniste et al., 2018;Kress et al., 2014), and degeneration of the auditory system (Kobrina et al., 2020), which can each contribute to a decline in cognitive function (Bettio et al., 2017;Weber et al., 2015). Research into agingrelated mechanisms has expanded rapidly over the past few years leading to many potential therapeutics to treat aging-related diseases in humans (Bakula et al., 2019;Hodgson et al., 2020). Studying behavior in aged mice as a model for human cognitive decline is necessary but remains challenging. Behavioral protocols need to be optimized for each age, strain, and animal source (Ryman & Lamb, 2006;Scearce-Levie, 2011;Sukoff Rizzo et al., 2018;Sukoff Rizzo & Silverman, 2016), which is time-consuming and often requires specialized equipment.
Additionally, aged animals are sensitive to environmental changes, and behavioral readouts can be variable within and between different cohorts and experimenters. Furthermore, interpreting cognitive decline in aged mice is complicated by the fact that aged animals also have motor impairments, so the readouts for many cognitive tasks are influenced by both cognition and ambulation. Here we aim to form a comprehensive profile of the molecular and histological changes that are robustly modulated with aging in male C57BL/6J mice, which is the most common inbred mouse strain used in the neuroscience field.
These endpoints are typically more straightforward to implement and do not suffer from the same variability issues as behavior. We propose that histological and molecular changes therefore may provide more granularity and be more consistent biomarkers of aging. While we will not opine on which is more functionally relevant than the other, we focus on three modalities: neuronal, microgliosis/neuroinflammation, and the neurovascular unit (NVU).

Animals
All animal handling and use was in accordance with Institutional Animal Care and Use Committee approved standard guidelines, protocol ALK-005. Male C57BL/6J mice were ordered from Jackson Laboratory (Sacramento, CA) and shipped to Alkahest prior to the start of each study. All animals were acclimated in house for at least 2 weeks prior to the start of the experiments. Upon arrival, all mice were housed with a unique identification number at standard temperature (22 ± 1 • C) and in a light-controlled environment (lights on from 7 am to 7 pm) with ad libitum access to food and water. To minimize the number of animals used per experiment, brains from cohorts 5-8 were sub-dissected and collected for 3 separate Animals of both ages were run together, and the experimenter was blinded to the age of the animals while performing and analyzing the experiment.

Y-maze ambulation
To measure distance and velocity, the same Y-maze protocol was used as described in section 2.3.1. However, all movements were recorded and tracked for analysis using ANY-maze software (Stoelting Co., Wood Dale, IL), which allows for measurement of the total distance and velocity for the duration of the test. Animals of both ages were run together, and the experimenter was blinded to the age of the animals while performing and analyzing the experiment.

Barnes maze
The Barnes maze is a circular maze with a diameter of 118 cm approximately 95 cm off the ground, consisting of 40 holes with a diameter of 5 cm aligned in three concentric circles. Each day, a hole was designated as the escape hole, where a small black box was placed beneath the hole and provided a space below the maze that the mouse could climb into.
To create an aversive environment and motivation to find the escape hole, the maze was illuminated with two large flood lights and a fan blew over the maze, creating palpable wind and a constant background noise of approximately 60 Hz. Two walls and two curtains surrounded the maze, each of which displayed distinct visual cues. Mice were habituated to the room for at least 20-30 min prior to the start of testing.
The testing ran for four consecutive days, with five trials each day. Mice were given 90 s to find and enter the escape hole after being placed in the center of the maze. If mice failed to identify the escape hole in that time, they were guided to the hole and encouraged to stay inside for 30 s. The inter-trial latency was 10 min. For the first 2 days of training (trials 1-10), the escape hole remained unchanged. For the second 2 days of testing (trials 11-20), the escape hole location was changed at the start of each day but was kept consistent for the trials occurring on that day (11)(12)(13)(14)(15)(16)(17)(18)(19)(20). Analysis began as soon as the mouse was placed in the center of the maze and concluded either once the mouse was inside the escape hole for >3 s or at a duration of 90 s. After each trial, the maze and escape hole were wiped down thoroughly with 70% ethanol. All movements were recorded and tracked for analysis using CleverSys Software. Animals of both ages were run together, and the experimenter was blinded to the age of the animals while performing and analyzing the experiment. The Barnes maze assay was performed in the same cohort of mice (cohort 2) as the Y-maze experiment, and these behavioral tests were run approximately 1 week apart.

Grip strength
Mice were habituated to the room for at least 20 min prior to testing.
After habituation, each mouse was gently lifted by the base of the tail to the height of the grip bar and allowed to grab the bar with an overhand grip. The mouse was gently pulled to ensure a tight grip and then continuously pulled at a slow, constant horizontal speed until the grip was broken. Steps were repeated for a total of four trials per mouse and peak tension (grams of force) was recorded for each mouse using a grip strength meter (Columbus Instruments, Columbus, OH). At the end of the testing, the body weight of each mouse was recorded. The average pull for each mouse was calculated and normalized to body weight.

Histology
Mice were anesthetized with 2,2,2-tribromoethanol (Avertin, T48402-25G, Sigma Aldrich) and subsequently perfused with 0.9% saline transcardially. The brains were dissected and cut sagittally in two even halves. One half was snap frozen in dry ice for protein and RNA anal-  CD68/Iba1 images were acquired using the Hamamatsu Nanozoomer 2.0HT at 20×. Quantification was done using percent thresholded area of the entire hippocampus region using ImagePro software by a single experimenter blinded to age.
GFAP antibody (ab53554, Abcam) was used at a concentration of 1:1000. First, images were acquired using a Zeiss LSM800 confocal microscope. The 6 z-stack (1 μm step size) images in the CA1 region of the hippocampus were acquired at 40×. Maximum intensity projections of each z-stack were quantified using ImageJ (National Institutes of Health, Bethesda, MD) for percent GFAP thresholded area and total GFAP cell count. Next, images were acquired using the Axio Scan.Z1

Plasma protein quantifications
Blood was collected by cardiac puncture in syringes containing 250 mM EDTA (BM-711, Boston BioProducts, Ashland, MA). Plasma was isolated by centrifugation at 1000 x g for 15 min at 4 • C and immediately frozen on dry ice. Mouse plasma was diluted 1:1 in PBS and then shipped on dry ice to Eve Technologies in Calgary, Canada. Single sam-ples were analyzed using a multi-plex Luminex technology assay for cytokines and chemokines or cell adhesion molecules. Quantitative data was sent in an Excel sheet after completion of the data acquisition and analysis.

qPCR
RNA was isolated from hippocampal brain tissue using the RNeasy Mini Kit ( Chemidoc and quantified using Image Lab 6.0 (Bio-Rad) software. Samples were randomized across gels and run blinded in single replicates.
A bridging sample was run to normalize across multiple blots, and band intensities of AQP4 and CD13 were additionally normalized to Actin loading control.

Statistical analysis
All data were analyzed using GraphPad Prism 8 (GraphPad Software, San Diego, CA). Sample sizes were similar to those employed in the field and all experimental n values reflect biological replicates of individual mice unless otherwise stated. For n > 10 with normally distributed data, parametric tests were used, and for n < 10 and data with a non-normal distribution, non-parametric tests were used. If technical replicates were used, it is stated explicitly within the methods section.
Technical replicates reflect samples replicates from the same mouse, such as ROI. Statistical significance was defined as p <0.05.
When two groups were compared in the motor and cognitive tests, data were analyzed using a Mann-Whitney U test. Average maximum grip strength across 4 trials was normalized to individual mouse body weight and then analyzed using a mixed-effects analysis with repeated measures with main effects of age and trial, followed by Mann-Whitney test. For Y-maze performance, two separate cohorts of mice were run and data were pooled across two experiments. Data were analyzed using a three-way repeated measures ANOVA for interaction between arm x age x experiment followed by Wilcoxon matched-paired signed rank tests. For Barnes maze performance, data were tested first for a normalized distribution and then analyzed using a mixed-effects analysis with repeated measure with main effects of age and trial.
The total number of DCX-positive cells per DG was estimated by counting the number of positive cells from 6 tissue sections and multiplying the sum of the number counted per section by 12, as an estimate for the total hippocampal volume. Mice with less than 6 quantifiable sections were excluded from the analysis. The thresholded percent area of CD68 and Iba1 were measured from 5-6 hippocampi per mouse using Image-Pro 9.2 software (Media Cybernetics). Mice with less than 5 quantifiable sections were excluded from the analysis. Ki67-and BrdU-positive cells were counted from 5 dentate gyri per mouse and CD3-CD45-double positive cells were counted from the hippocampus and SVZ of 5 hemibrain sections per mouse, and then the counts were summed. Mice with less than 5 quantifiable sections were excluded from the analysis. BrdU and Ki67 data were analyzed using nested t-tests. DCX, CD68, Iba1, SVZ CD3/CD45, and hippocampus parenchyma CD3/CD45 data were analyzed using nested one-way

Impaired cognitive and motor function with age
In humans, aging leads to a progressive decline in cognitive function (Klimova et al., 2017) and, in mice, has been shown to cause impairments in cognitive tasks including the Morris and radial arm water mazes and contextual fear conditioning (Murphy et al., 2006;Villeda et al., 2014;Weber et al., 2015). We found that 20-to 22-monthold aged mice had impairments in the hippocampal-dependent spatial learning and memory tasks, Y-maze ( Figure 1A) and Barnes maze ( Figure 1B), compared to young 2 to 3-month-old mice. However, aging also leads to declines in gait, motor function, and strength in both humans  and C57BL/6J mouse strains (Murphy et al., 2006;Villeda et al., 2014). We tested locomotor function in young and aged mice and showed that aged mice traveled shorter distances ( Figure 1C) and had a 62% reduced velocity ( Figure 1D) relative to young mice while exploring the Y-maze. Next, we assessed forearm grip strength between young and aged mice and identified that aged mice generated significantly less pulling force ( Figure 1E). For this task, we used 6.5-month-old young mice to ensure there was no difference in body weight between groups ( Figure 1F). The impairments in motor function and strength with age confound the interpretation of cognition in both the Y-maze and Barnes maze and highlight one of the challenges with behavior in aged animals. Therefore, we sought to outline molecular and histological changes that occur at the same time as the impairments in cognition and motor function.  Raber et al., 2004;Saxe et al., 2006). In the DG, these newborn neurons functionally integrate into neuronal networks and contribute to cognitive processing (Kozareva et al., 2019;Toni & Schinder, 2015). To measure neurogenesis, we examined the newborn neuron marker DCX in the DG using histology and show a dramatic decrease by 6 months of age with little neurogenesis occurring by 18-24 months of age (Figure 2A,B). However, using bulk hippocampal qPCR, Dcx gene expression was only modestly reduced ( Supplementary Fig. 1A), indicating that histology is a more robust readout for age-related neurogenesis changes. Additionally, the Age-related reductions in synaptic density and expression of genes related to synaptic function occur in both humans and rodents, and these changes correlate with cognitive deficits (Bishop et al., 2010;Blalock et al., 2003;Lee et al., 2000;Xu et al., 2018;Yankner et al., 2008). We found that excitatory synaptic density decreased between 12 and 18 months of age in the Schaffer collateral synapses of the CA1 hippocampal region, which is essential for activity-dependent synaptic plasticity (Bishop et al., 2010), as measured by juxtaposed pre-synaptic Synapsin and post-synaptic PSD-95 ( Figure 2E,F). However, the gene expression of Syn1 and Dlg4, the genes encoding Synapsin-1 and PSD-95, respectively, were unchanged by qPCR from bulk hippocampal tissue with age ( Supplementary Fig. 1D-E), while gene expression of neuron-specific Tuj1 had a small stepwise reduction with age, which is only significant at 24 months of age ( Supplementary Fig. 1F). Taken together, these data suggest that histology may be a better readout for the small synaptic changes that occur with healthy aging in mice, while bulk qPCR may be better suited for detecting larger changes to neuronal morphology or number.

Heightened microgliosis and elevated proinflammatory cytokines with age
Neuroinflammation is a major hallmark of aging and disease (Jansen et al., 2019; and numerous changes in microglia, which are the resident macrophages of the central nervous system, are impacted by animal age, including proliferation (Long et al., 1998), reactivity (Hefendehl et al., 2014), motility (Damani et al., 2011;Hefendehl et al., 2014), gene expression (Harry, 2013;Hart et al., 2012), and secretion of inflammatory cytokines (Ye & Johnson, 1999;Yu et al., 2002). Using CD68 and Iba1 to mark microglia in the hippocampus, we found a stepwise increase in microgliosis with age ( Figure 3A,C). Furthermore, there was increased gene expression of the proinflammatory genes Tnfa, Cd11b, and Il1a analyzed by qPCR from bulk hippocampal tissue ( Figure 3D,F). Interestingly, while these genes are predominantly expressed by microglia (Bohlen et al., 2017), they did not show the same stepwise progression as histological evaluation, but rather a sharp increase at 12 or 24 months of age. We also identified a subset of inflammatory genes that are unchanged with age, including Nfkb and Il4 (Supplementary Fig. 2A-B), suggesting that bulk gene expression may not be a robust readout of age-related microgliosis.
Circulating factors in the blood can have significant impacts on brain health, including neurogenesis, proliferation, myelination, synaptic plasticity, vascular remodeling, and cognition (Katsimpardi et al., 2014;Ruckh et al., 2012;Villeda et al., 2011Villeda et al., , 2014. Additionally, the contributions of inflammaging-the small yet persistently increased levels of proinflammatory signaling with age-are becoming increasingly more appreciated (Goronzy & Weyand, 2019;Lopez-Otin et al., 2013;Salminen et al., 2012). We examined the plasma levels of two circulating cytokines that are known to mediate microglia activation: IP-10/CXCL10 (Clarner et al., 2015) and MIG/CXCL9 (Ellis et al., 2010), and we found that levels of IP-10 and MIG increased with age ( Supplementary Fig. 2C,D). Taken together, these results suggest that increased microgliosis and heightened expression of a subset of hippocampal and circulating proinflammatory cytokines occur at the same time as age-related cognitive and motor decline in mice and could be used as molecular or histological readouts.
Indeed, a line graph representation of GFAP along the vessels suggests an increase with age ( Figure 4D,G). This age-related increase in GFAP seemed to be largely in the vascular region ( Figure 4E), but there was a trending increase in the surrounding perivascular region as well ( Figure 4F). There is also an increase in the astrocytic endfoot protein AQP4 measured from total cortical lysates by western blot

Increased T cell infiltration into the brain with age
One consequence of inflammaging and BBB dysfunction is the increase in infiltrating T cells into the brain in both humans (Dulken et al., 2019;Gemechu & Bentivoglio, 2012;Loeffler et al., 2011;Moreno-Jimenez et al., 2019;Moreno-Valladares et al., 2020) and mice (Dulken et al., 2019;Gemechu & Bentivoglio, 2012;Mrdjen et al., 2018;Ritzel et al., 2016;Stichel & Luebbert, 2007). Susceptibility to T cell infiltration is partially related to the BBB leakiness of the brain region (Loeffler et al., 2011), and infiltration of T cells is greatly enhanced in human patients with AD (Itagaki et al., 1988;Rogers et al., 1988;Togo et al., 2002), in mouse models of AD (Ferretti et al., 2016;Mrdjen et al., 2018), and following injury (Muzio et al., 2010;J. Wang et al., 2015). Infiltration into the hippocampus and SVZ are of particular interest due to their functions as neurogenic niches. T cells have been identified in the SVZ of aged mouse brains with single cell RNA sequencing (Dulken et al., 2019;Ogrodnik et al., 2021) and an increase in cytotoxic CD8+ T cells have been found in various regions of the aged mouse brain by histology (Propson et al., 2021). We used histological markers to quantify T cells in the hippocampus and SVZ across age. There was a stepwise increase in CD3+CD45+ T cells within the hippocampal parenchyma ( Figure 5A,D) and within blood vessels ( Figure 5B,D) with increasing age. Additionally, there was a large increase in T cells at the SVZ with age ( Figure 5C), suggestive of BBB impairment or recruitment of peripheral immune cells to the brain during aging.

High-dose LPS induces BBB impairment
While BBB impairment in aged humans is well known (Montagne et al., 2015), changes to the BBB in aged mice are less well characterized and the impairment in BBB leakiness is reported to be less robust (Sumbria et al., 2018). To measure BBB leakiness, we administered sodium fluorescein (NaF, 0.4 kDa) by IV tail vein injection and examined fluorescence in brain tissue 4 h later. Indeed, we found that aged mice (22 month) do not have overt BBB leakiness compared to younger (4 month) animals ( Figure 6A). To determine if aged mice may be more susceptible to BBB damage, we used a high, acute dose of lipopolysaccharide (LPS, 10 mg/kg), which has previously been reported to increase barrier leakiness 6 h following administration (Bien-Ly et al., 2015). High-dose LPS induced leakiness in both young and aged mice, and this leakiness was exacerbated with age ( Figure 6A), indicating impaired maintenance of the BBB in aged mice following chemical insult.
LPS has been well studied across multiple labs due to its potent effects and relative ease of use in animal models. LPS administration causes hundreds of genes to be differentially expressed (Chen et al., 2020). Furthermore, LPS increases soluble plasma levels of cell adhesion molecules (CAMs), which are released from endothelial cells in response to damage (Gotsch et al., 1994;Kisucka et al., 2009;Ley et al., 2007;Petri et al., 2008;Rossi et al., 2011). For example, P-selectin is increased following acute neuroinflammation and blocking it prevents neutrophil recruitment into the brain parenchyma (Bernardes-Silva et al., 2001) and leads to improved BBB integrity (F. Wu et al., 2015).
We identified that high-dose LPS leads to significant increases in soluble E-Selectin, ICAM-1, and P-Selectin in the plasma of both young and aged mice ( Figure 6B,D), suggesting widespread endothelial damage in response to LPS.

DISCUSSION
We identified changes in neurogenesis, proliferation, synaptic density, microgliosis, neuroinflammation, astrocytes, and pericytes at the NVU, and T cell infiltration into the brain during healthy aging in male C57BL/6J mice and propose the specific techniques that can be used to quantify these changes. Due to the many challenges with cognitive and behavioral testing in mice, we propose these molecular and histological changes may be used as readouts associated with aging-related cognitive and motor decline. The challenges of measuring behavior in aged mice include optimization of protocols, specialized equipment, and variability within and between aged cohorts. Furthermore, interpreting cognitive decline in aged mice is complicated by the fact that aged animals also have motor impairments. The readouts for many cognitive tasks are influenced by both cognition and ambulation. Finally, blinding of behavioral experiments is confounded by the obvious differences in size and appearance between young and aged animals. Here we aim to form a comprehensive profile of the molecular and histological changes that are robustly modulated with aging in male C57BL/6J mice and more straightforward to implement across labs.
Additionally, we identified a number of readouts that were unchanged TA B L E 1 Age-specific changes in male C57BL/6J mice

Modality Change with aging Current study Citations
Behavior

Neurogenesis and proliferation
Reduced Figure  Proliferation Increased Figure 3A-C (Long et al., 1998;Weber et al., 2015) Dystrophy Activated shape, increased size
with age and have limited utility as robust markers of aging in male C57BL/6J mice.
There are a few limitations to the results presented here. We are only reporting results from male mice in the one strain C57BL/6J. As a result, these conclusions can only be generalized within this population of animals. Others have published the differences in female mice or across different aged strains, and we point the reader to these published studies for additional references (Adelof et al., 2019;Kohama et al., 1995;Tran et al., 2021;Weber et al., 2015;Xiong et al., 2018).
Broadly, the results reported here across the three modalities of neurons, microglia, and NVU cell types are recapitulated in other strains of mice and across sex. However, the specific timelines and magnitudes are distinct between background strain and sex. While many of these endpoints have been previously reported, the additional data here, and the bringing together of multiple biological mechanisms, is significant as aging is a multimodal process and must be considered holistically.
These results, along with reports from the literature, summarized in Table 1, are essential tools for understanding aging processes and development of therapeutics for gerontological disease.

ACKNOWLEDGMENTS
We

CONFLICT OF INTEREST
All authors were full-time employees of Alkahest, Inc. at the time they contributed to the experiments in this manuscript.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.