Age-related changes in cerebellar and hypothalamic function accompany non-microglial immune gene expression, altered synapse organization, and excitatory amino acid neurotransmission deficits

Aging mice have functional deficits

We began by examining how aging affected mouse functional status: food intake, water intake, and physical activity. These behaviors were evaluated for 13-18 days in a custom-designed home cage behavioral monitoring (HCM) system following 5 days of acclimation [7]. We tested cohorts of young (2-3 mo), middle-aged (12-13 mo), and aged (21-24 mo) BALB and C57BL/6 mice. These two strains have distinct breeding lineages [8], and thus provide an indication of how genetic diversity might influence age-related CNS phenotypes. In C57BL/6 mice, locomotion (Figure 1. left panel) markedly decreased with age (p<0.011, F_2,30=5.2, one-way ANOVA), with a 32% difference noted between young and aged mice by post hoc testing (p<0.013). By contrast, BALB mice demonstrated no age-related locomotor decreases. Furthermore, false discovery rate analysis (Supplemental Table 1A) showed that aged C57BL/6 mice were more likely to have mobility impairments when compared to either young C57BL/6 or aged BALB mice. For example, aged C57BL/6 mice had more impairments in locomotor-associated behaviors when compared to young C57BL/6 mice (p<0.042, χ² test) or when compared to aged BALB mice (p<0.026). Similarly, aged C57BL/6 mice had more impairments of nonlocomotor movement-associated behaviors when compared to young C57BL/6 mice (p<0.046) or when compared to aged BALB mice (p<0.05).

Further dissection of this mobility deficit revealed a significant decrease in the overall number of locomotor bouts occurring during the circadian dark cycle as well as decreased locomotor speed (Figure 2A). This decrease in locomotor bout frequency was incompletely compensated by an increase in locomotor bout duration (Figure 2B). Similar changes are observed across aging human cohorts [9]. Furthermore, locomotor paths of young C57BL/6 mice (Figure 2C) were significantly straighter (p<<0.001; across-mouse variability p<<0.001, within-mouse variability no effect) compared to locomotor paths of aged C57BL/6 mice, suggesting an age-related impact on cerebellar function (and related to tandem gait impairment commonly seen in older adults). By contrast, we observed no age-related difference in overall locomotion, locomotor bouts, or locomotor path straightness in BALB mice (Supplemental Figure 1A, B respectively, green traces).

By contrast, aging BALB mice demonstrated dysregulated overall energy balance. Normal aging is characterized by body weight maintenance, and decreased basal metabolic rate, activity, and food consumption. Both BALB and C57BL/6 mice show significant weight loss between the middle-aged and aged cohorts (Figure 1, center panel, p<0.05, p<0.0002 respectively, one sided t-test). However, we found no significant differences between middle-aged and aged BALB mice in peak oxygen consumption ($\dot{\text{V}}$O₂), carbon dioxide production ($\dot{\text{V}}$CO₂), or adiposity (Supplemental Figure 2A). Aged BALB mice also maintain activity levels similar to those of young BALB mice, while aged C57BL/6 mice significantly decrease their activity. Furthermore, aged BALB mice tend to consume less chow (normalized to body weight) compared to middle-aged BALB mice, while aged C57BL/6 mice tend to consume more chow compared to middle-aged C57BL/6 mice (Figure 1, right panel). False discovery rate analysis further demonstrated that aged BALB mice showed impairments in more feeding-associated behaviors compared to aged C57BL/6 mice (p<0.036; Supplemental Table 1B). Further analysis of energy balance dysregulation in aged BALB mice revealed a significant decrease in the overall number of circadian dark cycle feeding bouts and decreased bout food intake (Figure 3A). Again, this decrease in feeding bouts was incompletely compensated by an increase in feeding bout duration, leading to a 10-15% decrease in overall feeding in aged BALB mice (compared to young cohort, Figure 3B). In C57BL/6 mice, no age-related difference in overall food intake or feeding bouts was appreciated (Supplemental Figure 2B). Age-related changes in feeding patterns similar to aged BALB mice have also been reported in older human populations [10].

Altered CNS immune gene expression in aged mice

The molecular events underlying the above-described functional impairments remain poorly understood. To obtain an unbiased view of mRNA transcriptional changes associated with these behavioral deficits, we used whole mouse genome microarrays. Whole tissue cerebellar and hypothalamic mRNA expression was determined and compared between cohorts of young, middle-aged, and aged BALB and C57BL/6 mice (per Methods). Transcripts obtained originate from neuronal soma and dendrites, with minimal contribution from axonal projections [11]. Signal strength characteristics across all chips were similar, as were cross-sample correlations for each of the individual groups, suggesting good technical microarray performance (Supplemental Figure 3A-D for BALB hypothalamus, BALB cerebellum, C57BL/6 hypothalamus, and C57BL/6 cerebellum, respectively).

Aging evokes modest changes in overall CNS gene expression (Figure 4, Venn diagram). Lists of genes differentially expressed between young and aged mice are provided in Supplemental Table 1 (BALB hypothalamus, BALB cerebellum, C57BL/6 hypothalamus, C57BL/6 cerebellum respectively). These changes occurred in both a strain- and site-specific manner (Figure 4 comparing BALB hypothalamus and C57BL/6 cerebellum; Supplemental Figure 4 comparing BALB cerebellum and C57BL/6 hypothalamus). In whole cerebellar and hypothalamic tissue of C57BL/6 and BALB mice, respectively, there was marked overexpression of transcripts characterized by ontology as having immune/defense function (p<<0.001 for both regions; Figure 4, also see Supplemental Table 2). Ontology analysis suggested that these transcripts belonged to specific functional categories, including classical major histocompatibility complex I (e.g. H2-D1, H2-K1), atypical major histocompatibility complex I (e.g. H2-Q1, H2-Q7), complement (e.g. C3, C1q, C4b), cytokine/chemokine (e.g. Ccl6, Ccl12), pattern recognition receptors (e.g. Tlr2, Clec7a, Lgals3, Trem2, F_cRs, Lilrb3), and cell adhesion molecules (e.g. Itgax, Lyz, Timp1). Predicted protein-protein interactions (STRING 9.1; string-db.org) further suggest that many genes overexpressed in the C57BL/6 cerebellum and BALB hypothalamus form membrane-bound signaling complexes (e.g. Fcgr3-Fcer1g-Tyrobp-Lilrb3-Ms4a6d-C1q; Clec7a-Tyrobp-Emr1-Ctss-Ly86-Igsf6-Mpeg1; Lgals3-Cd68-Capg-S100a4-Anxa4-Lgals3bp) that converge to activate NFκB. RT-qPCR confirmed these age-related changes in gene expression for selected loci in whole hypothalamic tissue (Supplemental Figure 5). 16% of the cerebellar genes differentially expressed between young and old C57BL/6 mice are known NFκB targets [12]; this value rises to 33% by including NFκB target predictions by TRANSFAC (ver 7.0 Public 2005) [13]. Similarly, 10% of the hypothalamic genes differentially expressed between young and old BALB mice are known NFκB targets, rising to 23% by including TRANSFAC predictions. Intriguingly, in human cerebellum we find age-related increases in expression of mRNAs homologous to those identified in the C57BL/6 mouse cerebellum (Figure 5). These changes occurred across all of the above categories, including MHC I molecules (HLA-B), complement (C3, C4b), PRRs (Lilrb3, Lgals3) and CAMs (Itgax, Spp1).

Immune genes do not localize to microglia

Age-related increases in the CNS expression of immune/defense related transcripts have been attributed to neuroinflammation [14–16]. If this were the case, then a majority of these changes should localize to microglia. Microglia are resident CNS mononuclear phagocytes, and account for the vast majority of CNS immunocytes. Furthermore, microglial activation is a key and necessary step supporting neuroinflammation. We used a multi-step magnetic bead based separation process combined with molecular and functional validations [17] to obtain highly enriched populations (65-90% CD11b(+) total cells within suspension) of both parenchymal and perivascular microglia from the cerebellum and hypothalamus of young, middle-aged, and aged BALB and C57BL/6 mice. RT-qPCR gene expression assays found that these microglia clearly did not contribute to most of the observed age-related changes in whole tissue gene expression (Figure 6). Microglia only contributed to the age-associated C3 increase in the C57BL/6 cerebellum (Figure 6A). In the BALB hypothalamus (Figure 6B), expression of C1q and Lilrb3 were decreased in microglia derived from aged mice (p<0.01). These results strongly suggest that most of the age-related changes in cerebellar and hypothalamic immune/defense gene expression do not localize to CNS microglia.

Synapse changes accompany altered PRR expression

In this light it is interesting to note that recent studies have found that pattern recognition molecules play an important role during both synaptogenesis and synaptic pruning phases of CNS development [18]. In the developing mouse thalamus, MHC Class I molecules demonstrate an activity-dependent expression pattern [19], and loss of these molecules leads to decreased synaptic pruning [20] and altered visual cortex plasticity [21]. Similar disruptions of synaptic pruning or plasticity are observed with loss of C1q (in the thalamus), PirB (in the primary visual cortex), and H2-K^b/H2-D^b (in the cerebellar Purkinje layer) [22–24]. MHC Class I molecules also regulate hippocampal synaptic activity [25]. We thus tested the hypothesis that increased expression of these PRR transcripts was associated with altered excitatory synapse organization. Given our earlier findings, we predicted that age-related synaptic changes should be more pronounced in the cerebellum of C57BL/6 mice compared with BALB mice, and in the hypothalamus of BALB mice compared with C57BL/6 mice. To measure synaptic counts in these regions at highest accuracy, we employed a modified array tomography protocol. Ultrathin (90 nm) sections were prepared from the cerebellum of BALB and C57BL/6 mice from young and aged cohorts; thin (10 μm) sections were prepared from the hypothalamus of BALB and C57BL/6 mice from young and aged cohorts (Supplemental Methods). Cell nuclei were visualized with DAPI, and excitatory synapses were visualized with antibodies directed against Vglut1. We chose C3 as a molecule increased in expression in both BALB hypothalamus and C57BL/6 cerebellum. Figure 7A shows representative photomicrographs in young and old C57BL/6 cerebellum, as well as quantified puncta counts (quantification workflow provided in Supplemental Figure 6); Figure 7B shows the equivalent figure for the young and old BALB hypothalamic arcuate nucleus.

In the cerebellar internal granule cell layer, we observe a marked age-associated increase in both Vglut1 and C3 expression in C57BL/6, but not BALB, mice (Figure 7A; 2-way ANOVA with mouse strain and age as primary factors; strain × age interaction p<0.04 for Vglut1; p<0.001 for C3). Specifically, in C57BL/6 mice, internal granule cell layer Vglut1 expression increases nearly 20% with age; in BALB mice, Vglut1 baseline expression is lower in young mice, and does not significantly change with age. We did not find any size difference in counted puncta. No significant differences were appreciated in either Vglut1 or C3 expression in aged BALB mice compared to young BALB counterparts (Supplemental Figure 7A). There were no age-related differences noted in overall cell count in either C57BL/6 or BALB mice, making neurodegeneration an unlikely explanation for this phenotype.

We further note that in the hypothalamic arcuate nucleus, we observe a marked age-associated increase in both Vglut1 and C3 expression in BALB (Figure 7B), but not C57BL/6 (Supplemental Figure 7B), mice (2-way ANOVA with mouse strain and age as primary factors; strain × age interaction p<0.001 for Vglut1; p<0.007 for C3). While young BALB and C57BL/6 mice both show the same degree of Vglut1 expression, with aging we find a nearly 20% increase in BALB arcuate Vglut1 expression that is not seen in the aged C57BL/6 mouse. Again, no size difference was noted in counted puncta, and there were no age-related differences in arcuate hypothalamus cell counts. These findings suggest that aging increases the number of excitatory synapse components in CNS regions where we observe increased expression of immune, complement, and pattern recognition genes with age, but does not change excitatory synapse organization in CNS regions where these age-related patterns of gene expression do not occur.

Functional impairment in aged synapses

To determine if the increased synapse counts observed in the aged C57BL/6 cerebellum and aged BALB hypothalamus were associated with a functional outcome, we assessed excitatory amino acid activity across the entire cerebellar internal granule cell layer and hypothalamus using manganese-enhanced MRI (MEMRI, Supplemental Methods). Neuronal Mn⁺⁺ transport occurs in an activity-dependent manner analogous to calcium; however, Mn⁺⁺ is paramagnetic, causing reduction of T₁ in proportion to concentration, which can be measured using T₁ weighted imaging. Thus, Mn⁺⁺ enhances MRI tissue signals in direct proportion to overall excitatory amino acid activity. Our results show no age-related increase in either C57BL/6 cerebellar internal granule cell layer (Figure 8A,C) or BALB hypothalamic arcuate nucleus (Figure 8B,D) excitatory amino acid activity. These findings do not rule out the possibility of increased excitatory amino acid activity below the MEMRI detection threshold in aged compared to young mice; however, changes of this magnitude are below expected biological signaling thresholds, and would require electrophysiology to verify. To rule out the possibility that age-related vascular disease contributed to our phenotypes, we further evaluated diffusion tensor images for all subjects. There were no significant age-related changes in water diffusion characteristics or vascular disease in either C57BL/6 or BALB mice (Supplemental Figure 8). These findings support the conclusion that the increased number of excitatory synapses present in the aged C57BL/6 cerebellar internal granule cell layer and aged BALB arcuate hypothalamic nucleus have a currently uncharacterized functional deficit.