Chem. Senses 27: 299-306,
2002
© Oxford University Press 2002
SYMPOSIUM: Proceedings of a Symposium on Functional Genomics in Neural Systems |
DNA Microarray Analysis of the Aging Brain
Departments of Genetics and Medical Genetics, University of Wisconsin, 445 Henry Mall, Madison, WI 53706, USA
Correspondence to be sent to: Tomas A. Prolla, University of Wisconsin-Madison, Departments of Genetics and Medical Genetics, 445 Henry Mall, Madison, WI 53706, USA. e-mail: taprolla{at}facstaff.wisc.edu
Abstract
To examine molecular events associated with brain aging and its retardation by caloric restriction (CR), we have employed high-density oligonucleotide arrays providing data on 6347 genes to define transcriptional patterns in two brain regions (cerebellum and neocortex). Male C57BL/6 mice were either fed normally or subjected to CR. To investigate aging, 5 month (young adult) and 30 month-old normally fed mice were compared. To study CR, 30 month-old control and CR mice were compared. In both brain regions, aging resulted in a gene expression profile suggestive of a marked inflammatory response, oxidative stress and reduced neuronal plasticity and neurotrophic support. In the brain, CR selectively attenuated the age-associated induction of genes encoding inflammatory and stress responses. In addition to providing an improved understanding of the aging process, the use of DNA microarrays generates panels of hundreds of transcriptional biomarkers of molecular aging, providing a new tool to measure biological age on a tissue-specific basis. These studies suggest that genomic approaches may be useful in understanding the molecular basis of the aging process in experimental animals.
Introduction
Several studies have identified age-associated alterations in olfaction in
humans, including increased odor detection threshold
(Deems and Doty, 1987
;
Stevens and Spencer, 1994),
decreased odor identification (Murphy
et al., 1997) and decreased odor discrimination
(Hulshof Pol et al.,
2000). Chemosensory decrements in the elderly may contribute to
poor nutrition, food poisoning and overexposure to environmentally hazardous
chemicals (Schiffman, 1997).
The molecular basis of age-related chemosensory decrements and how they relate
to cognitive and motor impairments remains unclear. In order to gain a better
understanding of general aspects of brain aging, we are currently using DNA
microarray analysis to study caloric restriction (CR) as a model system of
aging retardation in mammals (Lee et al.,
1999,
2000;
Kayo et al., 2001).
These studies have shown the feasibility and utility of gene expression
profiling to identify basic aspects of brain aging. The technology is readily
applicable to specific areas of aging investigation, including studies of the
molecular basis of age-related chemosensory alterations in experimental
animals.
CR retards the aging process
CR retards the aging process in laboratory rodents as characterized by a
delayed occurrence or complete prevention of a broad spectrum of
age-associated pathophysiological changes and a 30-50% increase in maximum
lifespan (Weindruch and Walford,
1998). The maximum lifespan of fish, rotifers, spiders and other
non-mammals is also extended by CR
(Weindruch and Walford, 1998).
An active area of research in biological gerontology concerns the mechanisms
by which CR retards the aging process. There are five classes of interrelated
and non-exclusive explanations for CR's mechanism: (i) decreases in oxidative
stress (Sohal and Weindruch,
1996); (ii) decreases in glycation or glycoxidation
(Kristal and Yu, 1992); (iii)
decreases in body temperature and circulating thyroid hormone levels
associated with a hypometabolic state
(Walford and Spindler, 1997);
(iv) alterations in gene expression and protein degradation
(Van Remmen et al.,
1995); and (v) neuroendocrine changes
(Nelson et al.,
1995).
Oligonucleotide microarrays as a tool to evaluate aging interventions
DNA microarrays are likely to revolutionize biomedical research through the
simultaneous analysis of gene expression patterns of whole genomes
(Lander, 1999). Large-scale
analysis of transcriptional responses has proven to be useful in the analysis
of complex biological phenomena such as metabolism in Saccharomyces
cerevisiae (DeRisi et al.,
1997) and tumorigenesis in humans
(Perou et al., 1999).
Since such arrays are synthesized based on sequence information, they provide
a direct link between differential gene expression patterns and information
readily available in public genomic databases. We have used Affymetrix
oligonucleotide arrays containing tens of thousands of gene-specific
oligonucleotides synthesized in situ using light-directed,
solid-phase combinatorial chemistry
(Lipshutz et al.,
1999). The Affymetrix arrays used initially represented 6347 genes
and contained 40 specific oligonucleotide probes, which are 25 bp long, for
each gene represented in the array. Twenty of these oligonucleotides represent
perfect matches (PM probes) to the gene of interest along its length. Another
20 oligonucleotides contain a single mismatch (MM probes) as compared with the
PM probes. In order to calculate an average signal intensity, hybridization
intensity from each MM probe is subtracted from that of its PM counterpart and
the values are averaged for the 20 pairs.
In a typical experiment, mRNA populations from individual animals are used to generate cRNA pools (see Materials and methods), which are then hybridized to individual oligonucleotide arrays. Because samples are applied to different microarrays and the data from such microarrays are compared, normalization between arrays is required. Such normalization can be performed by using the hybridization intensity of specific genes as internal controls, in the same manner that one uses the hybridization of housekeeping genes such as GAPDH or ß-actin in Northern blots. Alternatively, the total hybridization intensity of DNA microarrays can be used as normalization factors. Figure 1 demonstrates the image of two such arrays and the fluorescence patterns of two genes, derived from the comparison of heart samples from young (5 month) and old (30 month) C57BL6/J mice (unpublished results).
|
We are currently characterizing the gene expression profile associated with
the aging process in several organs of mice, including skeletal muscle,
cerebellum, hippocampus, neocortex, kidney, liver, heart and lymphocytes.
Although this review emphasizes results that we have obtained in the brain
using DNA microarrays that contain 
6500 genes, we are currently using
third generation Affymetrix murine genome U74 DNA chips that can obtain gene
expression information on >30 000 cDNAs and ESTs.
Materials and methods
Animals and diets
Details on the methods employed to house and feed male C57BL/6 mice, a
commonly used model in aging research with an average lifespan of 30
months, were recently described (Pugh
et al., 1999). Briefly, mice were purchased from Charles
River Laboratories (Wilmington, MA) at 1.5 months of age. After receipt in
Madison, the mice were housed singly in the specific pathogen-free Shared
Aging Rodent Facility at the Madison Veterans Administration Geriatric
Research, Education and Clinical Center, and provided a nonpurified diet (PLI
5001; Purina Labs, St Louis, MO) and acidified water ad libitum for 1
week. The mice were then allocated into two groups and fed one of two nearly
isocaloric (4.1 kcal/g), semipurified diets. Each mouse in the control
group was fed 84 kcal/week of the control diet (TD91349; Teklad, Madison, WI)
which is 5-20% less than the range of individual ad libitum
intakes. This dietary intake was used so that the control mice were not obese
and retained motor activity up to the age of sacrifice. Each mouse subjected
to CR was fed 62 kcal/week of the restricted diet (TD91351; Teklad), resulting
in a 26% reduction of caloric intake. The latter diet was enriched in protein,
vitamins and minerals such that CR and control mice were fed nearly identical
amounts of these components. The fat component, corn oil, was at the same
level (13.5%) in both diets, leading to a 26% reduction in fat intake for the
CR mice. The adult body weights of the mice averaged 32 g for controls
and 23 g for those on CR. Mice were killed by rapid cervical dislocation,
autopsied to exclude animals showing overt disease and several tissues
collected and immediately flash-frozen in liquid nitrogen and stored at
-80°C. All aspects of animal care were approved by the appropriate
committees and conformed with institutional guidelines.
Sample preparations and hybridizations
Total RNA was extracted from frozen tissue using TRIZOL reagent (Life
Technologies) and a power homogenizer (Fisher Scientific) with the addition of
chloroform for the phase separation before isopropyl alcohol precipitation of
total RNA. Poly(A)+ RNA was purified from the total RNA with
oligo-dT-linked Oligotex resin (Qiagen). One microgram of poly(A)+
RNA was converted into double-stranded cDNA (ds-cDNA) using SuperScript Choice
System (Life Technologies) with an oligo dT primer containing a T7 RNA
polymerase promoter region (Genset). After second strand synthesis, the
reaction mixture was extracted with phenol/chloroform/isoamyl alcohol. Phase
Lock Gel (5 Prime 
3 Prime Inc.) was used to increase ds-cDNA recovery.
The ds-cDNA was collected by ethanol precipitation. The pellet was resuspended
in 3 µl of DEPC-treated water. In vitro transcription was
performed using a T7 Megascript Kit (Ambion) with 1.5 µl of ds-cDNA
template in the presence of a mixture of unlabeled ATP, CTP, GTP and UTP, and
biotin-labeled CTP and UTP [bio-1 1-CTP and bio-16-UTP (Enzo)]. Biotin-labeled
cRNA was purified using a RNeasy affinity column (Qiagen). The amount of
biotin-labeled cRNA was determined by measuring absorbance at 260 nm.
Biotin-labeled cRNA was fragmented randomly to sizes ranging from 35 to 200
bases by incubating at 94°C for 35 min in 40 mM Tris-acetate, pH 8.1, 100
mM potassium acetate and 30 mM magnesium acetate. The hybridization solutions
contained 100 mM MES, 1 M [Na+], 20 mM EDTA and 0.01% Tween 20. In
addition, the hybridization solutions contained 50 pM oligonucleotide B2 (a
biotin-labeled control oligonucleotide used for making grid alignments), 0.1
mg/ml herring sperm DNA and 0.5 mg/ml acetylated BSA. The final concentration
of fragmented cRNA was 0.05 g/l in the hybridization solutions. Hybridization
solutions were heated to 99°C for 5 min followed by 45°C for 5 min
prior to placing in the gene chips. cRNA (10 µg) was placed in the gene
chip. Hybridizations were carried out at 45°C for 16 h with mixing on a
rotisserie at 60 r.p.m. Following hybridization, the hybridization solutions
were removed and the gene chips were installed in a fluidics system for washes
and staining. The fluidics system (Affymetrix GeneChip Fluidics Station 400)
performed two post-hybridization washes (a non-stringent wash and a stringent
wash), staining with streptavidin—phycoerythrin and one post-stain wash.
The gene chips were read at a resolution of 6 µm using a Hewlett Packard
GeneArray Scannar. Data collected from two scanned images were used for the
analysis.
Data collection
All calculations were performed by Affymetrix software. To determine the
quantitative RNA abundance, the average of the differences representing PM
minus MM for each genespecific probe family was calculated, after discarding
the maximum, the minimum and any outliers beyond three standard deviations.
For example, to calculate fold changes (FC) between data sets
obtained from young (y) versus old (o) mice, the following formula was used:
 |
Alternatively, if the Qfactor, a measure of the
non-specific fluorescence intensity background, was larger than the smallest
of either SIy or SIo, the FC
was calculated as:
 |
The Qfactor is automatically calculated for different regions of the microarray and therefore minimizes the calculation of spurious fold changes by replacing very low average signal intensities in the FC equation. Average of pairwise comparisons were made between study groups, using Excel software. As an example, in our studies each 5-month-old mouse was compared with each 30-month-old mouse generating a total of nine pairwise comparisons. Correlation coefficients between animals of the same experimental group were >0.98 in all cases, indicating very small variability between animals.
Statistical Analysis
A key challenge to the development of statistical methods for microarray
data is that sample size (i.e. the number of animals, tissues or independent
replicas in tissue culture experiments) is often small, but the number of
measurements is very large. Additionally, the expression levels for individual
genes may not be independent. Fold changes, which are typically used to report
microarray data, are measures of magnitude and not direct measures of strength
of evidence for an effect. When comparing large numbers of genes between two
specific age groups, it is likely that a number of genes will differ in
expression by 2-fold or more (the usual cutoff in DNA microarray experiments)
between the experimental groups by chance (type 1 errors). In other words,
such changes would not be observed if a different set of individuals from the
two age groups was compared. The observed number of false positives depends on
the individual variance of genes, because genes that have large variations in
expression levels are more likely to result in type 1 errors. It has been
calculated (Miller et al.,
2001) that in a study design containing three young versus three
old mice (such as we employed) and an average gene coefficient of variance
(CV) of 30, 58 false positives would be generated by chance if the
expression of 10 000 genes is analyzed. We have estimated that if the number
of genes to be screened is 6347 (as contained in the MU6500 DNA chip) the per
gene level required is 0.0000081 to keep the overall experiment-wide type
1 error to be 0.05, if one uses conventional statistic methodology. In order
to have 99% power to detect a difference of 0.2 standard deviations, 2400
samples (mice in our case) would have to be sampled for microarray testing.
Clearly, novel statistical methods for the analysis of DNA microarray data are
the most urgent need in the field.
Our own study design of the gene expression alterations in the aging brain
(Lee et al., 2000)
involved a first step of data filtering that removed most genes with high
CVs, since we only reported genes that had a fold change >1.3 +
the SEM. After removal of genes that did not fit these criteria, the average
CV for all remaining genes (a few hundred) was only 20.9. According
to Miller's simulation of 10 000 genes with this particular average
CV (Miller et al.,
2001), one would expect to observe only six `false positive genes'
displaying a >2x alteration in expression and no such genes
displaying a >3-fold alteration in gene expression. We observed >12
genes displaying a 3-fold or greater alteration in gene expression in the
cerebellum of mice, indicating that the majority of age-related alterations in
gene expression are unlikely to be type 1 errors.
A `two-step approach' to reduce type 1 errors has also proposed
(Miller et al.,
2001). This involves an initial, exploratory selection of genes
that satisfy a given statistical criteria (using an initial level), followed
by `confirmation testing' of these genes with an independent group of animals
(using a second level). However, it is unclear if this method indeed helps
investigators to make fewer type 1 (false positive) and type 2 (false
negative) errors. We have argued that the two-stage procedure offered by
Miller et al is unnecessary to hold the overall experiment-wise type
1 error rate to some desired level (Prolla
et al., 2001). Based on examination of many microarray
data sets, we feel that it is likely that (i) many genes have dependent
(correlated) expression levels (i.e. are co-expressed) and (ii) with profound
independent variables such as aging, the null hypothesis of no group
differences in expression is false for many genes, but the relative difference
in expression across the two groups may be modest (e.g. 30-100%). Therefore,
we are developing a sequence of procedures that involve finite mixture
modeling and bootstrap inference in order to address significance testing for
microarray data (Allison et al.,
2001). This set of procedures is based on the idea that when many
statistical tests are conducted, one obtains a distribution of test statistics
and corresponding P-values and that there is information available in
this distribution that can be exploited. In particular, the method employs the
observation that under the alternative, more likely hypothesis (i. e. that
there is in fact a difference between groups) the distribution of
P-values for individual genes will not be uniform.
Our initial analysis involving finite mixture modeling and bootstrap
inference of our data collected in aging skeletal muscle in mice (three
mice/age group, 6347 genes screened) indicates that even among genes for which
there is a real difference in expression, we only expect P-values to
be 0.17 on average if three animals/group are used in the analysis
(Allison et al, 2001).
We have also determined that as long as the P-value for an individual
gene is <0.35, there is a >50% chance that the gene identified is a gene
for which there is a real difference in expression. These results underlie the
fact that, given the small sample size in most microarray studies, power is
low and that conventional significance testing with an level of 0.05 or
smaller would lead to many false negatives (i.e. misses). Therefore, false
negatives, as opposed to false positives, are likely to represent the major
statistical issue in DNA microarray analysis.
Results and discussion
The gene expression profile of aging in the mouse brain
Of the 6347 genes surveyed, only 67 (1%) displayed an increase >1.7-fold
in expression levels with aging in the neocortex, whereas 63 (1%) displayed an
increase of >2.1-fold expression with aging in the cerebellum (see
www.wisc.edu/genetics/CATG/prolla/data/aging/index.html
for a full listing of genes). A subset of these genes was used for a
comparison between oligonucleotide microarray data and TaqMan real-time
quantitative PCR assays, which demonstrated excellent overall agreement
between the methods (Lee et al.,
2000). Of the upregulated genes, 20% (14/67) and 27% (17/63) could
be assigned to an immune or inflammatory response in the neocortex and
cerebellum, respectively (Table
1). Transcriptional alterations of several genes in this category
were shared by the two brain regions, although fold changes tended to be
higher in the cerebellum. Interestingly, we saw a concerted induction of the
complement cascade components C4, C1qA, C1qB and C1qC. Complement proteins are
found in the senile plaques characteristic of Alzheimer's disease
(Akiyama et al.,
2000).
|
Similar to previous data obtained from the gastrocnemius muscle
(Lee et al., 1999),
genes involved in a stress response (consistent with a state of oxidative
stress and accumulation of damaged proteins) accounted for 24% (16/67) and 13%
(8/63) of genes highly induced in the neocortex and cerebellum, respectively,
of old animals (Table 2). These
included the four cathepsins (lysosomal proteases) and the several heat-shock
factors (e.g. Hsp40, Hsp27, Hsp59). Also induced with age was the gene
encoding ubiquitin, a major stress-inducible transcript that marks altered
proteins for degradation by the ubiquitin/ATP-dependent proteosome pathway.
The accumulation of intracellular inclusions containing misfolded,
ubiquitinated proteins is a hallmark of several neurodegenerative disorders
(Koshy et al., 1997;
Arnold et al., 1998).
Possibly, the molecular basis for the induction of chaperones and lysosomal
proteases in the aged brain is protein oxidation subsequent to inflammation
and mitochondrial ROS production.
|
We also observed an induction of the early response genes junB and c-fos, which are co-induced in response to neocortical injury or hypoxic stress and GFAP, a marker of reactive astrocytosis induced in the aging brain (Table 1). Taken as a whole, our observations support the concept that the aging process in the brain is associated with a state of heightened immune reactivity and oxidative stress accompanied by the accumulation of altered or misfolded proteins. Therefore, a chronic state of oxidative stress appears to be a common transcriptional feature of aging in both brain and skeletal muscle.
Several transcripts which declined in expression encode nuclear genes
involved in mitochondrial functions. These include NADP transhydrogenase,
cytochrome-C-oxidase subunit VIII and the ATP synthetase-gamma and delta
chains which are all either components of the mitochondrial electron transport
system or support its function. This transcriptional profile suggests that
mitochondrial function may be compromised in the aged brain, parallel to
observations in skeletal muscle (Lee
et al., 1999).
CR is associated with a transcriptional reprogramming in the brain
Of the largest age-associated changes in the neocortex (1.7-fold or more),
30% (34/114) were prevented either completely or partially by CR. Remarkably,
the effect of CR on age-associated alterations in gene expression was highly
dependent on transcript class: CR influenced only 20% (2/10) of the observed
decreases in expression of genes involved in neuronal growth and plasticity,
whereas it prevented 50% (8/16) of the induction of stress response and 65%
(11/17) of immune-response-related genes. A similar pattern was observed in
the cerebellum. The effects of CR on immune- and stress-related transcripts
agree with studies suggesting that reductions in both autoimmunity
(Weindruch and Walford, 1998)
and oxidative damage (Sohal et
al., 1994) occur in the brain of CR mice.
As was observed in muscle, CR shifted the activity of several genes that were unaltered in expression with age (see www.wisc.edu/genetics/CATG/prolla/data/aging/index.html for a full listing of genes affected by CR in the neocortex). As compared with age-matched controls, CR induced the expression of 120 genes by 1.7-fold or more in the neocortex, which represent 1.9% of the genes surveyed (Table 3). One of the largest classes of transcripts induced by CR was comprised of growth and neurotrophic factors (9%). This class included the developmentally regulated homeobox genes Hox2.5, Hoxb3 and Hoxa6, all of which may be involved in neural development, and the gene encoding neuroserpin, a factor that promotes neuronal plasticity.
|
Of the four genes upregulated by CR in the neocortex which were assigned to
the stress response class, two encode I-
B 
chain and I-traf,
which act to attenuate stress responses by inhibiting the NF-kB signaling
pathway involved in the response to oxidative stress and inflammation in the
nervous system (Kaltschmidt et
al., 1999). CR also upregulated genes which modulate the
inflammatory response such as Ia2 and Ia5 (encoding
interferon isoforms 2 and 5) and Relb, a gene encoding a transcription
factor which may suppress inflammation through NF-B-mediated signaling
(Ryseck et al.,
1996).
CR was associated with a reduction of 1.7-fold or greater in the expression of 84 genes in the neocortex. The largest functional class (11%) consisted of genes involved in protein synthesis. Inflammatory related transcripts downregulated by CR include the proinflammatory cytokine Mcp5. A state of lowered endogenous oxidative stress in CR mice is also suggested by reduced expression of the chaperone-encoding genes Hsp27 and Cypa, and also the genes encoding the p65 and p100 subunits of NF-kB. Taken as a whole, these results suggest that metabolic alterations induced by CR may profoundly affect brain aging.
The gene expression profile of aging and CR in rhesus monkey vastus lateralis muscle
We recently reported the effects of aging and adult-onset CR on the gene
expression profile of 7070 genes in the vastus lateralis muscle from rhesus
monkeys (Kayo et al.,
2001). The average lifespan of rhesus monkeys in our colony is
27 years and the maximum lifespan is 40 years. Gene expression
analysis of old rhesus monkeys (mean age 26 years) was compared with young
animals (mean age 8 years). Aging resulted in a selective upregulation of
transcripts involved in inflammation (a class not observed to be upregulated
in the mouse gastrocnemius muscle) and oxidative stress, and a downregulation
of genes involved in mitochondrial electron transport and oxidative
phosphorylation. Middle-aged monkeys (mean age 20 years) subjected to CR since
early adulthood (mean age 11 years) were studied to determine the gene
expression profile induced by CR. CR resulted in an upregulation of
cytoskeletal protein encoding genes and also a decrease in the expression of
genes involved in mitochondrial bioenergetics. Surprisingly, we did not
observe any evidence for an inhibitory effect of adult-onset CR on age-related
changes in gene expression. These results indicate that the induction of an
oxidativestress-induced transcriptional response may be a common feature of
aging in skeletal muscle of rodents and primates, but the extent to which CR
modifies these responses may be species-specific. However, critical
differences in the experimental design of our mouse and monkey studies may
explain the lack of efficacy of CR in retarding age-associated transcriptional
changes in monkeys. Most notable are the early life onset of CR in the mouse
study and the study of truly old CR mice (versus middle age CR monkeys) in the
mouse study.
Implications for aging research
These data underscore the utility of high-density oligonucleotide microarray gene expression analysis in the study of complex biological phenomena such as the aging process. Further, gene chips are useful tools for evaluating the impact of nutritional interventions by providing a global, unbiased view of gene expression alterations. The promise of this approach derives from the ability to provide (eventually) a genome-wide screen within a single experimental setup. Proteomics is expected to offer similar opportunities.
However, certain limitations are associated with data generated by microarrays. First, the observed collection of gene expression alterations in aging whole tissues such as muscle and brain is complex, reflecting the presence of diverse cell types. Second, the ability of oligonucleotide microarrays accurately to measure mRNA levels for thousands of genes has predated the development of statistical tools to minimize the occurrence of false positives and false negatives. Additionally, changes in mRNA levels may not always result in a parallel alteration in protein levels.
None the less, the complete or partial prevention of the majority of the observed aging alterations by CR in mouse muscle and the targeted prevention of stress response and inflammatory transcripts in the brain provide biological validation for this approach. Further, these data suggest that gene expression patterns can be used to assess the biological age of a tissue. Extension of our study to other organs should result in the identification of hundreds of tissue-specific biomarkers of aging, facilitating the elucidation of aging mechanisms and the development of interventions.
Acknowledgments
Supported by NIH grant R01 CA78723 (T.A.P.). T.A.P. is a recipient of the Shaw Scientist (Milwaukee Foundation), Burroughs Wellcome Young Investigator and Basil O'Connor (March of Dimes) awards.
References
Akiyama, H., Barger, S., Barnum, S., Bradt, B., Bauer, J., Cole, G.M., Cooper, N.R., Eikelenboom, P., Emmerling, M., Fiebich, B.L., Finch, C.E., Frautschy, S., Griffin, W.S., Hampel, H., Hull, M., Landreth, G., Lue, L., Mrak, R., Mackenzie, I.R. and McGeer, P.L. (2000) Inflammation and Alzheimer's disease. Neurobiol. Aging, 21,383 -421.[Web of Science][Medline]
Allison, D.B., Gadbury, G., Heo, M., Fernandez. J., Lee, C.-K, Prolla, T.A. and Weindruch., R. (2001) Statistical methods for the analysis of microarray gene expression data.Comput. Stat. Data Anal. (in press).
Arnold, J., Dawson, S, Fergusson, J., Lowe. J., Landon, M. and Mayer, R.J. (1998) Ubiquitin and its role in neurodegeneration. Prog. Brain. Res.,117 , 23-34.[Web of Science][Medline]
Deems, D.A. and Doty, R.L. (1987) Age related changes in the phenyl ethyl alcohol odor detection thresholds. Trans. Penn. Acad. Ophthalmol. Otolaryngol.,39 , 646-640.
DeRisi, J.L., Iyer. V.R. and Brown, P.O.
(1997) Exploring the metabolic and genetic control of gene
expression on a genomic scale. Science,278
, 680-686.
Hulshof Pol, H.E., Hijman, R., Baare W.F.C., van Eekelen, S.
and van Ree J.M. (2000) Odor discrimination and task
duration in young and older adults. Chem. Senses,25
, 461-464.
Kaltschmidt, B., Sparna, T. and Kaltschmidt, C. (1999) Activation of NF-kappa-B by reactive oxygen intermediates in the nervous system. Antiox. Redox. Sig.,1 , 129-144.
Kayo. T., Allison, D.B., Weindruch, R. and Prolla,
T.A. (2001) Influences of aging and caloric restriction
on the transcriptional profile of skeletal muscle from rhesus monkeys.Proc. Natl Acad. Sci. USA
, 98,5093
-5098.
Koshy, B.T. and Zoghbi, H.Y. (1997) The CAG/polyglutamine tract diseases: gene products and molecular pathogenesis. Brain. Pathol., 7,927 -942.[Web of Science][Medline]
Kristal, B.A. and Yu, B.P. (1992) An emerging hypothesis: synergistic induction of aging by free radicals and Maillard reactions. J. Gerontol. Biol. Sci.,47 , B107-B114.
Lander, E.S. (1999) Array of hope.Nat. Genet. , 21(Suppl.),3 -4.[Web of Science][Medline]
Lipshutz, R.J., Fodor, S.P., Gingeras, T.R. and Lockhart, D.J. (1999) High density synthetic oligonucleotide arrays. Nat. Genet., 21,20 -24.[Web of Science][Medline]
Lee, C.K., Klopp, R.G., Weindruch, R. and Prolla T.
A. (1999) Gene expression profile of aging and its
retardation by caloric restriction. Science,285
, 1390-1393.
Lee, C.K., Weindruch, R. and Prolla, T.A. (2000) Gene expression profile of the ageing brain.Nat. Genet. , 25,294 -297.[Web of Science][Medline]
Miller, R.A., Galecki, A. and Shmookler-Reis, R.J.
(2001) Interpretation, design and analysis of gene array
expression experiments. J. Gerontol. A. Biol. Sci. Med.
Sci., 56A,B52
-B57.
Murphy, C., Nordin, S. and Acosta, L. (1997) Odor learning, recall and recognition memory in young and elderly adults. Neuropsychology,11 , 126-137.[Web of Science][Medline]
Nelson, J.F., Karelus, K., Bergman, M.D. and Felicio, L.S. (1995) Neuroendocrine involvement in aging—evidence from studies of reproductive aging and caloric restriction. Neurobiol. Aging., 16,837 -843.[Web of Science][Medline]
Perou, C.M., Jeffrey, S.S., van de Rijn, M., Rees, C.A., Eisen,
M.B., Ross, D.T., Pergamenschikov, A., Williams, C.F., Zhu, S.X., Lee, J.C.,
Lashkari, D., Shalon, D., Brown, P.O. and Botstein, D.
(1999) Distinctive gene expression patterns in human mammary
epithelial cells and breast cancers. Proc. Natl Acad. Sci.
USA, 96,9212
-9217.
Prolla, T.A., Allison, D.B. and Weindruch, R. (2001) Response to `Interpretation, design and analysis of gene array experiments'. J. Gerontol.,56A , B1-B4.
Pugh, T.D., Oberley, T.D. and Weindruch, R.
(1999). Dietary intervention at middle age: caloric
restriction but not dehydroepiandrosterone sulfate increases lifespan and
lifetime cancer incidence in mice. Cancer Res.,59
, 1642-1648.
Ryseck, R.P., Weih, F., Carrasco, D. and Bravo, R. (1996) RELB, a member of the REL/NF-KAPPA-B family of transcription factors. Braz. J. Med. Biol. Res.,29 , 895-903.[Web of Science][Medline]
Schiffman, S.S. (1997) Taste and smell
losses in normal aging and disease. J. Am. Med. Assoc.,278
, 1357-1362.
Sohal, R.S and Weindruch, R. (1996) Oxidative stress, caloric restriction, and aging.Science , 273,59 -63.[Abstract]
Sohal, R.S., Agarwal, S., Candas, M., Forster, M.J. and Lal, H. (1994) Effect of age and caloric restriction on DNA oxidative damage in different tissues of C56BL/6 mice.Mech. Ageing Dev. , 76,215 -224.[Web of Science][Medline]
Stevens, J.C. and Spencer, N.A. (1994)
Olfactory detection of a complex versus a simple substance in advanced
age. Chem. Senses, 19,365
-369.
Van Remmen, H., Ward, W.F., Sabia, R.V. and Richardson, A. (1995) Gene expression and protein degradation. In Masoro, E.J. (ed.), Handbook of Physiology. Oxford University Press, New York, pp. 171-234.
Walford, R.L. and Spindler, S.R. (1997) The response to calorie restriction in mammals shows features also common to hibernation—a cross-adaptation hypothesis. J. Gerontol. Biol. Sci., 52,B179 -B183.[Abstract]
Weindruch, R. and Walford, R.L. (1998)The Retardation of Aging and Disease by Dietary Restriction . C.C. Thomas, Springfield, IL.
Accepted December 7, 2001
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  Y. Yamamoto, T. Tanahashi, T. Kawai, S. Chikahisa, S. Katsuura, K. Nishida, S. Teshima-Kondo, H. Sei, and K. Rokutan Changes in behavior and gene expression induced by caloric restriction in C57BL/6 mice Physiol Genomics, November 6, 2009; 39(3): 227 - 235. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Judge and C. Leeuwenburgh Cardiac mitochondrial bioenergetics, oxidative stress, and aging Am J Physiol Cell Physiol, June 1, 2007; 292(6): C1983 - C1992. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. P. de Magalhaes and G. M. Church Genomes Optimize Reproduction: Aging as a Consequence of the Developmental Program Physiology, August 1, 2005; 20(4): 252 - 259. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E.-S. Han and M. Hickey Microarray Evaluation of Dietary Restriction J. Nutr., June 1, 2005; 135(6): 1343 - 1346. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. J. McLean and D. G. Le Couteur Aging Biology and Geriatric Clinical Pharmacology Pharmacol. Rev., June 1, 2004; 56(2): 163 - 184. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Kaput and R. L. Rodriguez Nutritional genomics: the next frontier in the postgenomic era Physiol Genomics, January 15, 2004; 16(2): 166 - 177. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. A. LYNCH Long-Term Potentiation and Memory Physiol Rev, January 1, 2004; 84(1): 87 - 136. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. Chowers, D. Liu, R. H. Farkas, T. L. Gunatilaka, A. S. Hackam, S. L. Bernstein, P. A. Campochiaro, G. Parmigiani, and D. J. Zack Gene expression variation in the adult human retina Hum. Mol. Genet., November 15, 2003; 12(22): 2881 - 2893. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. S. McClintock High-throughput Expression Profiling Techniques Chem Senses, March 1, 2002; 27(3): 289 - 291. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||