Our PCA established that each region of the mouse CNS possesses its own unique transcriptome signature that distinguishes it from other regions regardless of the age, gender or diet of the animal. Although previous studies have shown that different CNS regions have unique patterns of gene expression [32
], our study demonstrates that the region-specific patterns of gene expression are maintained regardless of the age of the animal, its dietary energy intake and its gender. These signature transcriptomes presumably represent the molecular basis of the phenotypic differences in the cells that comprise the different CNS regions and, by extension, regional functionality. Changes in many different common and cell type-specific genes in the CNS are known to occur in response to environmental factors, including age, diet, exercise, activity in neuronal circuits, and injury or disease [11
]. However, our findings suggest that such epigenetic responses to the environment do not alter the fundamental transcriptome 'fingerprints' that distinguish different regions of the CNS.
The false discovery rate (FDR), an approach widely applied to microarray data analysis, allows the researcher to balance the size of the candidate gene list against its quality in order to enhance confidence in the validity of the data, and is particularly well-suited to large datasets such as ours. However, we found that the variability in gene expression was considerably larger in samples from old compared to young animals (Figure S4 in Additional data file 1), a result consistent with a recent study [40
]. The increased variation with advancing age results in higher p
values, and q-values as well since the q-value is computed from the p
value. Inasmuch as aging is considered a stochastic process, it should be expected that the effect of natural aging on the gene expression is more variable when compared with gene expression effects of more well-defined and dramatic experimental manipulations or disease states. Our PCA analysis (Figure ) also provided evidence that the factor of age has only minor effects on the CNS regional transcriptomes. FDR is typically applied to large (robust) effects of factors on gene expression [41
]. Because many of the genes that were significantly affected by age in our study exhibited relatively small changes, the FDR was not, therefore, applied to this dataset. However, it should be noted that the increased variance in p
values in the old cohort may confound findings concerning the numbers of genes identified as changed in the old group compared to the younger group.
A change in the expression of a gene during aging might contribute to a decline in function and degeneration of neural cells or, instead, might be an adaptive response to aging. The greater number of genes upregulated by aging in the spinal cord may represent a superior ability of cells in the spinal cord to adapt to aging, perhaps because it is the most primitive part of the CNS and the most essential for survival. On the other hand the relatively greater proportion of white matter (olidodendrocytes) in the spinal cord may also contribute to the greater effect of aging on the spinal cord transcriptome compared to the four brain regions examined. Interestingly, many genes downregulated during aging in brain regions were upregulated in the spinal cord, including genes of the UPS, for example (Table S3 in Additional data file 1). In contrast to the spinal cord, very few genes in the striatum were affected by aging and CR and most of those that did respond, including those in the UPS, were downregulated with age and upregulated by CR. These findings suggest the striatum may be prone to age-related diseases, such as Huntington's and Parkinson's, because its transcriptome does not respond adaptively during aging.
Prior gene expression studies of brain aging included only young and old animals of one gender, typically used microarrays with relatively few genes and often analyzed pooled RNA samples resulting in negligible statistical power [11
]. We therefore analyzed RNA isolated from five different CNS regions from male and female mice of three different ages and two different diets (three to five mice analyzed for each age, gender and diet) using a large mouse gene array. Inclusion of the middle-aged group revealed a caveat with previous studies of 'aging' in which comparisons are made between young and old individuals only. We found that many genes that would have been considered sensitive to aging in a young versus old comparison are, in fact, changed only between young and middle ages with no further change between middle and old age. Indeed, in the cerebellum, 64% of the age-responsive genes followed the latter pattern (Figure ). In the hippocampus, 27% of the genes that were significantly affected by age changed between young and middle age, and then returned to the young level in old age. On the other hand, we found that it was extremely rare for the expression level of a gene to change in one direction between young and middle age, and in the opposite direction between middle and old age.
A comparison of our data with those of previous gene array analyses performed on RNA samples from the cerebral cortex of mice [11
] and humans [36
] revealed only five genes that were significantly affected by age in all three studies (Table S7a in Additional data file 1). Two of the genes (vimentin
) encode astrocyte cytoskeletal proteins previously shown to be upregulated in aging and neurodegenerative disorders [42
]. The other three genes encode a cell adhesion molecule (ICAM2), a protein that interacts with SIRT1 and p53 in cellular stress response signaling (NDRG1) [43
] and a putative energy and nutrient sensor (FRAP1) [44
]. An additional nine genes were common to the cerebral cortex datasets of the two mouse studies and included those encoding proteins involved in cell senescence, mitochondrial translation initiation and protein phosphorylation (Table S7b in Additional data file 1). A comparison of our mouse cerebellum dataset with that of Lee et al
] identified eight genes significantly affected by aging (Table S7c in Additional data file 1), including those encoding proteins involved in proteolysis (ubiquitin-specific peptidase 46 and cathepsin Z), transforming growth factor-β signaling (TGFβ receptor 3) and nitric oxide signaling (endothelial nitric oxide synthase). A comparison of our mouse cortical dataset with the human frontal cortex data [36
] identified nine genes significantly affected by aging (Table S7d in Additional data file 1), including those involved in calcium signaling (voltage-dependent calcium channel beta-2 subunit and calcium/calmodulin-dependent kinase 3), neurotransmitter signaling (gamma-aminobutyric acid (GABA-A) receptor subunit beta 3) and oxidative stress responses (thioredoxin interacting protein). A comparison of hippocampal gene expression profiles in young (2 month old) and middle-age (15 month old) C57BL/6 mice identified 35 genes as being significantly upregulated; they included genes related to synaptic plasticity, inflammation, oxidative stress and protein processing [45
]. Blalock et al
] performed microarray analyses of approximately 2,000 genes in hippocampi from young, middle-aged and old rats, and correlated changes in gene expression with performance of the rats on learning and memory tasks. Many of the AAGs in the latter two studies were in the same functional categories as AAGs in our study, including calcium signaling, oxidative stress and proteolysis.
In addition to functional classes of genes documented in previous studies of brain aging and neurodegenerative disorders, our findings identified Werner/telomere-interacting proteins and the Wnt signaling pathway as being highly responsive to both aging and dietary energy restriction throughout the CNS. Werner is a DNA helicase that plays a pivotal role in DNA repair and telomere function [31
]. Loss-of-function mutations in Werner cause a premature aging syndrome that includes neurological abnormalities. A brain imaging study of two siblings with Werner's syndrome provided evidence for reduced cerebral energy metabolism compared to age-matched control subjects [46
]. The latter findings, evidence for reduced energy metabolism in the brain during normal aging [47
], and our finding of significantly reduced expression of Werner in the brain during normal aging suggest a possible role for Werner in age-related compromise of brain function. Werner interacts with several telomere-associated proteins also implicated in cellular senescence (Figure ). Increasing evidence suggests that telomerase and other telomere-associated proteins play roles in neuronal plasticity and survival [48
]. The responsiveness of several telomere-associated proteins to aging and CR suggests roles for telomere modifications and DNA damage and repair processes in CNS aging. In addition to playing major roles in CNS development [29
], the Wnt signaling pathway has been implicated in adult neural plasticity and the pathogenesis of neurodegenerative disorders [50
]. Several genes in the Wnt signaling pathway were prominently affected by aging and CR in several CNS regions in our study, including those encoding nemo-like kinase, α-catenin and calcium/calmodulin-dependent kinase 2. Each of the latter proteins is involved in mechanisms of signaling associated with the pathogenesis of neurodegenerative disorders [51
], suggesting a role for age-related perturbation in Wnt signaling in the disease processes.
Chromosome mapping of genes that were differentially expressed in mice of different ages and/or in response to CR revealed a wide distribution of genes with some physical clustering of responsive genes within the genome. The latter findings are consistent with the concept that aging is a complex process and that evolutionary adaptations to aging, if they exist, may or may not involve geographic clustering of functionally related genes.
Despite the existence of many phenotypic differences between males and females, the vast majority of gene expression analyses have been performed on males, and direct comparisons of CNS transcriptome responses of males and females to aging and environmental factors are lacking. However, analysis of the expression of 4,000 genes in trained and untrained muscles of young and old men and women revealed that more genes were affected by gender than by age or training [54
]. In our study, numerous genes, spanning functional categories, were differentially expressed in the CNS of males and females (Tables S5a-d in Additional data file 1). In general, genes involved in protein degradation, oxidative stress resistance and cell survival were expressed at higher levels in females compared to males, suggesting a superior ability of brain cells in females to resist age-related oxidative and metabolic stress. Interestingly, there was considerable variability in the numbers of genes affected by gender among CNS regions, with the hippocampal transcriptome being the most sensitive to gender and the striatum and cerebellum the least sensitive. The proteins encoded by genes differentially expressed in the CNS of males and females may determine gender-specific differences in behaviors, responses to dietary energy intake and susceptibility to age-related dysfunction and disease.