|Home | About | Journals | Submit | Contact Us | Français|
Young and middle-aged female mice were ovariectomized and given cyclic injections of either estradiol or vehicle treatments. During the fifth week after surgery the Morris water maze was used to assess cognitive function. Age and treatment effects emerged over the course of spatial training such that middle-aged vehicle treated mice exhibited deficits in acquiring a spatial search strategy compared to younger vehicle treated mice and middle-age estradiol treated mice. Following behavioral characterization, mice were maintained on their injection schedule until week seven and hippocampi were collected 24 hr after the last injection. Hippocampal RNA was extracted and genes responsive to age and estrogen were identified using cDNA microarrays. Estradiol treatment in middle-aged mice altered the expression of genes related to transcriptional regulation, biosynthesis, growth, neuroprotection, and elements of cell signaling pathways. Expression profiles for representative genes were confirmed in a separate set of animals using oligonucleotide arrays and RT-PCR. Our results indicate that estrogen treatment in middle-aged animals may promote hippocampal health during the aging process.
The hypothesis that estrogen therapies slow cognitive decline in normal aging and Alzheimer's disease continues to draw debate. Research in humans (MacLennan et al., 2006; Sherwin, 2006), primates (Lacreuse et al., 2002; Rapp et al., 2003), and rodents (Gibbs, 2000; Frick et al., 2002; Markham et al., 2002; Markowska and Savonenko, 2002; Foster et al., 2003) indicate that estrogen replacement may prevent or delay memory impairments, if initiated during middle-age, while reduced benefits of estrogen replacement are observed with increasing age (Foster et al., 2003). This suggests that middle-age represents a time limited window for estrogen's protective influence.
Research indicates that estrogen can influence cell growth and synaptic connectivity of the hippocampus in a manner opposite that observed during aging (McEwen et al., 1997; Gibbs and Aggarwal, 1998; Sandstrom and Williams, 2001; Adams and Morrison, 2003; Foster, 2005). Interestingly, estrogen improves memory function examined several days after treatment, suggesting the involvement of long-term and possibly genomic mechanisms (Sandstrom and Williams, 2001; Markowska and Savonenko, 2002; Rapp et al., 2003). However, the examination of genomic regulation is complicated by the fact that estrogen can influence transcription through nuclear receptors (e.g. ERα and ERβ) and rapid signal transduction cascades which influence the activity of several transcription factors. Moreover, age dependent changes in the expression of estrogen receptors and signaling cascade activity further complicate the study of estrogen's genomic mechanisms (Foster, 2005). The overwhelming complexity of estrogen signaling emphasizes the limits of studies that focus on a single gene or a single age.
Gene array technology provides a powerful tool for examining multifaceted transcription processes through the ability to monitor parallel expression of thousands of genes. However, the power of this technique is limited by the increased chance for Type I error associated with multiple comparisons. The current study uses microarray technology to test the hypothesis that estrogen interacts with age to influence transcription in the hippocampus. We employed a system of filtering in order to limit Type I error and verified age and estrogen sensitivity for a subset of genes.
Procedures involving animal subjects have been reviewed and approved by the Institutional Animal Care and Use Committee and were in accordance with guidelines established by the U.S. Public Health Service Policy on Humane Care and Use of Laboratory Animals. A total of sixty seven female C57/B6 mice (young: n = 28, 3-5 months; middle-aged: n = 39, 11-13 months) were obtained from National Institute of Aging. Animals were housed 3-5 per cage and maintained on 12:12 light: dark cycle. Following at least one week of habituation, mice were anesthetized (2 mg ketamine and 0.2 mg xylazine per 20 grams of body weight) and ovaries were removed through a small midline incision on the abdomen. All mice received ad lib access to food (Purina mouse chow, St Louis, MO) and water, until the surgery when they were placed on Casein based chow (Cincinnati Lab Supply, Cincinnati, OH), which is low in phytoestrogens found in soy based chow.
Following surgery, mice were separated into four groups: young receiving β-estradiol 3-benzoate (Sinagra et al.) (n = 10), young receiving oil (n = 18), middle-aged receiving EB (n = 18), middle-aged receiving oil (n = 21). EB (Sigma) was dissolved in light mineral oil (Fisher Scientific, Pittsburgh, PA) to concentration of 0.5 mg/ml. Oil and EB (5μg) were injected subcutaneously at the nape of the neck in volumes of 0.05 ml. Injections were given on two continuous days of a five day cycle. The schedule for a series of eight cycles of injections and behavioral training is illustrated in Fig 1. Briefly, injections were initiated one week after surgeries and continued for a total of eight cycles during which animals were behaviorally tested on the water maze, starting 48 hr after completion of cycle five injections. A subset of animals employed for confirmation of cDNA microarray data received eight cycles of treatment in the absence of behavioral training. At 24 hr following the final injection of oil or EB, all animals were anesthetized with CO2 and decapitated. The brain was quickly removed and placed in ice cold artificial cerebral spinal fluid. Both hippocampi were removed, frozen in liquid nitrogen, and stored at -80°C. In some cases, uteri were also excised, excess tissue was removed and wet weight of uteri was immediately measured.
A circular black plastic pool (120 cm diameter) was filled with water (29 ±2°C, colored white with nontoxic white paint) to a level of 8 cm below the rim of the tank. The water maze task was located in a well-lit testing room. During testing on the cued discrimination task the pool was surrounded by a black curtain. The curtain was pulled back during spatial discrimination training in order to expose spatial cues in the room. A camera mounted above the center of the pool tracked the animal's movement.
Behavioral training began when animals were approximately 5 months and 12 months of age and was preformed during the light phase of the cycle. The cued discrimination training began during week four of hormone replacement, 48 hr after the fifth set of injections. A white flag was attached to a circular escape platform (10 cm diameter). Initially, mice were given a habituation swim in which they were placed in the pool and after 30 sec they were guided to the platform and permitted to climb up on the platform where they remained for 15 sec. Following habituation, mice received cued discrimination training consisting of four blocks with three trials per block. Each animal was released from one of four equally spaced starting locations (N, S, E, and W). The location of the escape platform and the release point were randomly assigned for each trial. The position of the escape platform and release point for each trial was kept consistent between animals. The mouse was allowed to swim until it located the escape platform. If the mouse failed to escape to the platform in 60 sec, it was then guided to the platform. After every trial, the animal remained on the platform for 15 sec. Between blocks, the animals were towel dried, placed back in the home cage, and warm air was blown over the cages. The inter-block interval was 15-20 min.
For spatial discrimination, the platform was localized just below the surface of the water and maintained in one quadrant (the goal quadrant) for the duration of testing. Training was conducted over three consecutive days and began in week five of the injection schedule, 48 hr after the sixth set of injections. Mice received four blocks of training per day, each block consisting of three trials. The release point for each trial was randomly assigned. The animal had 60 sec to escape during the trial. If the mouse did not escape within the allotted time they were guided to the platform and allowed to rest for 15 sec. A probe trial was performed during the penultimate trial for each day to determine the extent of learning. In addition, a probe trial to examine retention was conducted for the first trial on days two and three of spatial training. For probe trials, the platform was removed and the animal was released from the quadrant opposite the goal quadrant and allowed to swim freely for 60 sec. Following completion of each probe trial, the platform was returned to the pool and the animal placed on the platform for 15 sec to rest.
Behavioral data for cue and spatial discrimination tasks were recorded and analyzed by Water Maze 4.31 (Columbus Instruments, Columbus, OH). Measures included latency and path length to escape from the pool and time spent near the pool wall (thigmotaxis) during each trial. For probe trials, the time spent searching the goal quadrant and number of goal platform location crossings was determined.
Microarray analyses were performed on hippocampal tissues from each of the same behaviorally characterized animals (one chip per animal). Hippocampal RNA was isolated using Qiagen RNeasy Lipid Tissue Mini Kit (Qiagen, Germantown, MD) and RNA quality was analyzed on a subset of representative samples using Agilent Bioanalyzer (Santa Clara, CA). m17K mouse cDNA arrays were obtained from the JHU/NIA Microarray Facility (Bethesda, MD). Microarray procedure protocols used in the study are described in the National Institute on Aging Gene Expression and Genomic Unit website (http://www.daf.jhmi.edu/microarray/protocols.htm). Briefly, 5 μg total RNA from a single animal was reverse transcribed in the presence of 33P-dCTP, labeled cDNA was purified using QIAquick Nucleotide Removal Kit (Qiagen), diluted in hybridization buffer and hybridized to the m17k array for 16–18 h at 55 °C with rotation. Hybridized arrays were washed with 2 × SSC and 0.1% SDS one to two times for 15 min each at 65°C followed by one to two washes of 1 × SSC and 0.1% SDS at 65°C for 15 min each. The arrays were exposed to phosphoimager screens for 48 hr and scanned in a Molecular Dynamics Storm Phosphor Imager (Molecular Dynamics, Sunnyvale, CA) at 50 μm resolution. Array-Pro Analyzer (Media Cybernetics, Silver Spring, MD) was used to extract the log of the net signal (raw data minus background) from the scanned images. The data were exported into Microsoft Excel spreadsheets and converted to standard scores (z-score = (probe signal − mean signal for all probes across the array)/standard deviation of all probes across the array).
Custom designed oligo GEArray® Microarrays were obtained form SuperArray Bioscience Corporation (Frederick, MD). The custom arrays were designed to contain two or three probes for each gene of interest. Custom array procedures were followed according to SuperArray protocol (User Manual part #1018A version 3.0, 2005). Briefly total of 3 μg of total RNA from a single animals was used to make cRNA. cRNA purification was performed using ArrayGrade cRNA Cleanup Kit (SuperArray). UV spectrophotometry was used to quantify and assess the quality of cRNA. A total of 2 μg of biotin-labeled cRNA was hybridized rotating in 0.75 ml of pre-warmed GEAhyb Hybridization Solution (SuperArray) over night at 60°C. After overnight hybridization, membranes were washed with 5 ml of 2× SSC and 1% SDS for 15 min at 60°C and with 0.1× SSC and 0.5% SDS for 15 min at 60°C and incubated with streptavidin-AP conjugate (SuperArray,) (1:8,000). Array images were developed using CDP-star chemiluminescent substrate and imaged using a CCD camera (Chemic Doc XRS, Bio-Rad Laboratories, Hercules, CA). Images were analyzed using Array-Pro and data were exported to Microsoft Excel spreadsheets. The background signal was subtracted from the probe signal to provide net expression. The net expression for each probe of a specific gene was averaged across probes for that gene and the averages were normalized to the averaged signal derived from an internal cyclophilin A standard on the same membrane. Thus, the expression was calculated using the following formula: mRNA expression = [(average gene signal - background signal)/ (average cyclophilin A signal - background signal)].
In some cases, RNA from animals employed in the oligonucleotide array study was used for RT-PCR. RNA was converted to cDNA using the High Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA). Briefly 3 μg of total RNA from a single animal was incubated with appropriate reagents at 25°C for 10 min and then heated to 37°C for 120 min using 7300 Fast Real Time PCR System (Applied Biosystems). For relative quantification of RNA, 2.5 μl of cDNA was added to 12.5 μl of TaqMan® Universal PCR Master Mix (2×), 1.25 μl of 20× Gene Expression Assay Mix, and 8.75 μl of nuclease-free water for a total volume of 25 μl. The TaqMan® probes used for RT-PCR were selected from the Applied Biosystems Library and included; Histone deacetylase 2 (Hdac2) (TCAGTTGCTGGGGCTGTGAAATTAA), assay identification number Mm00515108_m1), Longevity assurance homolog 2 (Lass2) (GCACCGGACGCCGAGATGCTCCAGA), assay identification number Mm00504086_m1), POU domain, class 3, transcription factor 1 (Pou3f1) (GCAGCGGTGCCTCCGGCGCGCAGTT), assay identification number Mm00843534_s1), and the probe for Glyceraldehyde-3-phosphate dehydrogenase (Gapdh) was used as an internal control (TGAACGGATTTGGCCGTAATTGGGCG), assay identification number Mm99999915_g1). Thermal cycle conditions were set at 2 min at 50°C, 10 min at 95°C and cycled between 15 sec at 95°C and 1 min at 60°C for 40 cycles. Relative quantification was determined with 7300 Fast Real-Time PCR System and SDS Software 1.3.1 analysis software (Applied Biosystems). Each sample was examined in triplicate and the relative quantities (RQ) were normalized by the level of Gapdh. The three RQ values were averaged for each animal and the means for animals in the treated groups were normalized by the mean for the control samples (young oil treated) in order to derive the fold change.
In general, repeated measures analyses of variance (ANOVAs) were used to establish main effects and interactions on behavioral measures. Follow-up ANOVAs either collapsing the data across days of training or examining behavior within each day were employed to localize specific differences.
For analysis of cDNA microarrays, the data underwent a two phase filtering process. The first phase was designed to exclude probe sets in which a substantial number of chips did not exhibit reliable detection/hybridization and remove probes with uncertain function. This was done by first calculating the mean and standard deviation of the blank probe sets for each microarray. A cut off was chosen as three standard deviations above the mean for the blanks. Scores below this value were considered as lacking specific hybridization. The probe was considered absent for a treatment group if the majority, greater than 60%, of the arrays in this group exhibited hybridization below this cut-off. Finally, the probe set was removed from further consideration if the probe set was judged absent for at least three of the four treatment groups. The remaining probe sets were further filtered to remove expressed sequence tags and probes for hypothetical proteins and pseudo genes that did not have an indication of biological or molecular function as expressed through gene ontology (GO). Finally, when a gene was represented by multiple probes, the probes for the gene were averaged for each animal and the averages were used as measures of expression.
To generate a list of likely age or estrogen sensitive genes, a second filter was employed using statistical sorting. The alpha level was set at 0.025 in accordance with previous studies (Blalock et al., 2003) and t-tests were used to detect differences associated with age or treatment. Once a subgroup of genes was identified as probable age sensitive, the stringency of the alpha value was reduced (e.g. p < 0.01) to examine the effects of the other variable (e.g. EB treatment) for this subgroup of genes. False discovery rate (FDR) was calculated as the number of expected false positives/number of genes observed to change.
Uterine weight was measured in a subset of animals (middle-aged EB treated (n = 14) 140 ± 6 mg; middle-aged oil treated (n = 14) 40 ± 6 mg; young EB treated (n = 5) 106 ± 7 mg; young oil treated (n = 13) 39 ± 4 mg). An ANOVA indicated a significant effect of age [F(1,42) = 7.6, p < 0.01] and treatment [F(1,42) = 174.3, p < 0.0001].
Forty seven mice received the treatment schedule outlined in Figure 1 (young EB = 10, young oil = 12, middle-aged EB = 11, middle-aged oil = 14) and were tested for cue and spatial discrimination. Figure 2 illustrates the performance on the cue task. Two-way repeated measures ANOVAs across the training blocks indicated a significant effect of training on escape latency [F (3,129) = 32.6, p < 0.0001] in the absence of age or treatment effects. A significant effect of training [F (3,129) = 20.5, p < 0.0001] on escape path length was observed in the absence of age or treatment effects. No age, treatment, or training effects were observed for swim speed (data not shown).
The behavioral measures indicated improved performance across days of training for all groups. In addition, age × treatment interactions were observed due to poorer performance by middle-aged oil treated mice. An ANOVA on escape latency (Fig 3A) indicated a significant effect of training [F(2, 86) = 41.2, p < 0.0001] (Fig 3A) and an interaction of day × age × treatment [F(2, 86) = 4.5, p < 0.05]. Examination within each day indicated a tendency for an age × treatment interaction on Day 3 (p = 0.06), due to longer escape latencies for middle-aged oil treated mice.
Additional evidence for an age × treatment interaction was provided by analysis of escape path length (Fig 3B). An ANOVA revealed significant effects of training across days [F(2, 86) = 35.0, p < 0.0001] and a day × age × treatment interaction [F(2, 86) = 6.7, p < 0.005]. Examination within each day indicated an age × treatment interaction on Day 3 [F(1, 43) = 5.0, p < 0.05], again with the poorest poor performance observed in the middle-aged oil treated group. Finally examination of the percent of the escape latency time the animals spent along the wall (i.e. thigmotaxis) decreased over days of training F(2, 86) = 27.0, p < 0.0001] in the absence of age or treatment effects (Fig 3C).
Probe trial measures confirmed that middle-aged oil treated mice were impaired in acquiring the spatial discrimination. A repeated measures ANOVA for percent time in the goal quadrant for the three acquisition probe trials demonstrated a significant effect of training [F(2,86) = 25.1, p < 0.0001] and an interaction of day × age × treatment [F(2,86) = 3.3, p < 0.05]. ANOVAs within each day revealed an age × treatment interaction on Day 2 [F(1,43) = 7.2, p < 0.01] and ANOVAs within each age and treatment condition indicated an age difference for oil treated mice [F(1,24) = 9.5, p < 0.01] with poorer performance for middle-aged mice (Fig 4A).
A repeated measures ANOVA for platform crossings indicated a main effect of training [F(2,86) = 22.6, p < 0.0001] and age [F(1,86) = 4.6, p < 0.05] and a training × age × treatment interaction [F(2,86) = 3.3, p < 0.05]. ANOVAs within each day revealed an age × treatment interaction on Day 1 [F(1,43) = 4.2, p < 0.05] and Day 2 [F(1,43) = 6.0, p < 0.05] and ANOVAs within each age and treatment condition indicated an age difference for EB treated mice on Day 1 [F(1,19) = 8.5, p < 0.01] and confirmed poorer performance of middle-aged oil treated animals on Day 2 [F(1,24) = 9.5, p < 0.005] (Fig 4B).
An ANOVA examining the percent time in the goal quadrant during the retention probe trials delivered as the first trial on Days 2 and 3 indicated a significant effect of training [F(1,43) = 5.3, p < 0.05] and a significant age × treatment interaction [F(1,43) = 5.1, p < 0.05]. Analyses within each day revealed differences on Day 2 due to poorer performance by middle-aged oil treated mice such that treatment effects were observed for middle-aged animals [F(1,23) = 5.1, p < 0.05] and age difference were observed for oil treated mice [F(1,24) = 8.9, p < 0.01] (Fig 4C).
Examination of platform crossings during the retention probe trials confirmed age × treatment effects (Fig 4D). An ANOVA indicated a main effect of training [F(1,42) = 6.7, p < 0.05] and a tendency for an age × treatment interaction (p = 0.07). ANOVAs for each day localized an age × treatment interaction to the Day 2 probe trial [F(1,42) = 8.8, p < 0.005]. Analysis of treatment effects during the Day 2 retention probe trial indicated more crossings in middle-aged EB treated mice relative to aged matched oil treated mice [F(1,22) = 4.7, p < 0.05]. Examination of age differences indicated an age effect for oil treated mice [F(1,24) = 9.9, p < 0.005].
Microarray analyses were performed on hippocampal RNA from each of the behaviorally characterized animals (one chip per animal). cDNA microarray data were not obtained for 10 animals due to poor hybridization or poor quality of RNA. The remaining 37 microarrays (young EB = 10, young oil = 8, middle-aged EB = 9, middle-aged oil = 10) were submitted to the first phase filtering process to eliminate probes with low hybridization and uncertain biological function. This initial filtering process resulted in 4217 probes that were eligible for the second phase of filtering to identify likely genes that are sensitive to age and treatment effects. Our previous work and that of others suggests that a substantial number of genes are altered between young and middle-aged animals. In order to examine the relative influence of age and treatment, t-tests with alpha set at 0.025 were used to estimate the number of genes influenced by age regardless of treatment. An age difference was observed for 567 probes. In contrast, with alpha set at 0.025, only 187 probes exhibited a treatment effect regardless of age. An examination of treatment effects within each age group suggested that treatment effects were more common in middle-aged animals. Only 58 probes exhibited differences between oil and EB treatment in young animals, while 244 probes were influenced by EB treatment in middle-aged animals.
Our interest is for genes which exhibit EB effects in a manner opposite that observed during aging. If EB has effects opposite that of age, it is likely that the probes of interest are missed by collapsing across treatments or ages. Accordingly, in order to identify candidate aging genes for further analysis, we used statistical sorting according to expression differences between middle-aged oil treated and young oil treated mice, before examining EB effects only in middle-aged mice. Using t-tests with alpha set at 0.025, the number of probes expected to reach significance by chance alone is 105. A comparison indicated that 570 probes (FDR = 0.18) were differentially expressed in oil treated middle-aged mice relative to oil treated young mice. These 570 probes were classified as potential age-related genes, and this set of genes was used to test for EB effects. Because we were only interested in effects opposite that of aging, we employed one tailed t-tests (alpha set at 0.025) with the direction specified as opposite that observed for aging and compared middle-aged oil treated and middle-aged EB treated mice. For the 570 potential age-related genes, 132 probes (FDR = 0.1) were observed to exhibit transcription changes in a manner opposite that observed for aging. Gene expression profiles of the 132 age-EB sensitive genes were examined using the Unigene EST Profile Viewer from the National Center for Biotechnology Information and a cut off for expression for the gene of interest was set at ≥ 10-4 out of every one million transcripts normally found in neural tissue, brain or dorsal root ganglion. This procedure resulted in 119 age-EB sensitive genes that are moderately to highly expressed in the brain (Table 1&2). Using gene ontology, Swiss-Prot protein knowledgebase, and literature searches, the age-EB sensitive genes were characterized according to likely biological processes. In most cases, genes could be categorized as involved in more than one biological function; however, the majority could be classified as involved in transcription, apoptosis/cell health, receptor/cell signaling pathways, cell growth/structural organization, cholesterol/lipid metabolism, and protein metabolism.
In order to provide some validation of the findings, custom oligonucleotide arrays containing probes for 10 representative age-EB sensitive genes and a cyclophillin A (Ppia) control were constructed. The genes were selected in order to have a balance of those that exhibited an increase (5 genes) and decrease (5 genes) with age. In addition, genes were selected to represent the various biological processes including transcription (Ldb2, Hdac2, Pou3f1), apoptosis/cell health (Foxo3a), receptor/cell signaling (Kctd3, Cbln1), cell growth/structural organization (Ppfia1), cholesterol/lipid metabolism (Lass2), and protein metabolism (Fbxw8, Ube2r2). RNA was obtained from an independent set of young oil (n = 6), middle-aged oil (n = 7), and middle-aged EB treated (n = 7) mice which received the same ovariectomy-injection schedule (see Fig 1) in the absence of behavioral training and cRNA was hybridized to the oligonucleotide arrays.
For two probes (Ppfia1, Ube2r2) hybridization of at least 8 of the 20 chips exhibited an absence of signal and data were considered unreliable for parametric statistical analysis. The signal for the remaining genes was normalized by the level of Ppia on the array. These normalized scores were then divided by the mean for young animals in order to determine directional changes and fold changes. Predicted direction of change due to age or EB treatment was based on results of the cDNA arrays. For the remaining 8 genes, the predicted direction of change associated with age, increasing or decreasing in middle-aged oil treated mice relative to young oil treated mice, was confirmed for all genes except Pou3f1, in which the mean response for middle-aged oil treated mice was elevated relative to young oil treated mice (Fig 5). Furthermore, the predicted direction of change for EB treatment in middle-aged EB treated mice relative to middle-aged oil treated animals was observed in 8 of the 8 genes. Thus, for the set of eight genes, the relative direction of change due to age and EB treatment could be predicted correctly ~94 % of the time.
Genes were separated according to the expected directional influence of age and EB and repeated measures ANOVAs were employed to examine group effects. A repeated measures ANOVA across the 5 genes expected to increase with age and decrease with EB treatment (Ldb2, Foxo3a, Hdac2, Lass2, Kctd3) indicated a difference across groups [F(2,68) = 4.10, p < 0.05]. Subsequent ANOVAs comparing young oil treated relative to middle-aged oil treated indicated a significant [F(1,44) = 6.91, p < 0.05] increase in expression with age. Although, the mean for each gene was reduced in the middle-aged EB treated group relative to middle-aged oil treated animals, an ANOVA examining EB treatment effects in middle-aged animals indicated no difference. For three genes (Fbxw8, Pou3f1, Cbln1) expected to decrease with age and increase with EB treatment, a repeated measures ANOVA indicated a difference across the genes [F(2,34) = 5.24, p < 0.05] and subsequent ANOVAs indicated a treatment effect in aged animals due to an increased expression associated with EB treatment [F(1,24) = 10.67, p < 0.01].
An examination of the oligonucleotide arrays revealed several genes that were expected to exhibit at least a 2 fold increase in middle-aged oil or EB treated animals compared to young oil treated mice (e.g. Hdac2, Lass2). Thus, RNA isolated for the oligonucleotide array study was also employed for RT-PCR studies (3-4 animals per group) to further validate the gene profiling results (Table 3). The results confirmed that histone deacetylase 2 and LAG1 longevity assurance homolog 2 increased by at least 2 fold in middle-aged oil animals. Furthermore, RT-PCR of octamer-binding transcription factor 6 (Pou3f1) confirmed an increased expression in middle-aged animals treated with EB relative to young oil treated mice.
While middle-aged oil treated mice exhibited learning over the course of training, age and treatment interactions were observed for spatial discrimination escape latency and path length, as well as probe trial measures due in part to poorer performance by middle-aged oil treated mice. In contrast, the performance of middle-aged EB treated animals was similar to young mice and treatment effects were not observed for younger animals. The poor performance of middle-aged oil treated mice was not due to sensory-motor deficits since there was no difference in latency or swim speed on the cue discrimination task. Furthermore, no age or treatment effects were observed for thigmotaxis, measured as the percent time swimming along the pool wall, suggesting that the disparity in performance was not due to differences in anxiety.
The results are consistent with previous studies in humans (Foster, 2006) and rodents (Verbitsky et al., 2004; Ziegler and Gallagher, 2005), which indicate mild cognitive impairments emerge in mid-life and cognitive weakening continues with advancing age. Further, age-related impairments may be enhanced by hormone deprivation, and estrogen treatment in humans (Sherwin, 2005; MacLennan et al., 2006) and animals (Frick et al., 2002; Markham et al., 2002; Markowska and Savonenko, 2002; Foster et al., 2003; Daniel et al., 2006) may delay the decline of certain cognitive processes. Interestingly, beneficial effects of estrogen treatment may not be evident if the treatment is initiated after long-term hormone deprivation or the behavioral demands of the task do not reveal an age difference (Markowska and Savonenko, 2002; Ziegler and Gallagher, 2005; Daniel et al., 2006). The interaction of aging and hormonal status on cognition suggests that middle-age may provide an important window for examining hormonal influences on markers of brain aging.
For cDNA arrays examined across treatment groups, 567 probes exhibited age differences; confirming that a considerable number of genes alter their expression during middle-age. In contrast, only 187 genes exhibited altered expression in response to EB treatment, independent of age, suggesting that a relatively low number of hippocampal genes are sensitive to our EB treatment. However, the effects of age are relatively chronic, and the number of EB responsive genes, and magnitude of EB effects, is likely to be a function of the time between treatment and sample collection. Furthermore, when EB effects were examined within each age group, the number of probes increased four fold between middle-aged animals relative to young animals (244 versus 58 probes) indicating that older animals are more sensitive to hormonal status.
While the set of EB responsive genes in aged animals is likely to be important for determining treatment effects on hippocampal function, the current study focused on a subset that were age and EB responsive. Age-EB responsive genes were distinct from those normally seen with learning, which often involve synapse specific molecules (Irwin, 2001; Luo et al., 2001; Cavallaro et al., 2002; D'Agata and Cavallaro, 2003; Leil et al., 2003). The difference might be due to the fact that hippocampi were harvested several days after training when training effects may have dissipated. Thus, the set of age-EB genes may contribute to overall hippocampal function rather than being induced by behavioral training. Estrogen has enduring effects on synaptic plasticity, memory, the growth, and vitality of neurons, processes which depend on transcriptional regulation. Microarray results indicate hormonal status of middle-aged animals regulates transcriptional mechanisms; including an EB associated reduction in transcription repressors (Hdac2, Zik1, Sap18) and an increase in products which enable transcription/translation (Gtf3c2, Bop1, Sfrs7, Tcf12) including Mms19l, a possible estrogen receptor coactivator of transcription (Wu et al., 2001). The findings suggest EB influenced the production of components of the transcription apparatus.
Estrogen modulates structural plasticity of dendritic spines (Gould et al., 1990; Li et al., 2004), which maybe age specific (Miranda et al., 1999). EB treatment was associated with genes for biosynthesis (Gmppa, Axot, Rpl23), protein folding (Tcp1, Cct6a, Erp29), and vesicle/protein transport (Copg, Vps28, Vps33a, Rab11fip2, Kifc2). An increase was observed for genes associated with growth (Nrp1, Emp1, Sema3e) (Wulf and Suter, 1999; Pozas et al., 2001), structural organization (Bsg, Lrtm1, Mical2, Flnb) (Naruhashi et al., 1997; Fan et al., 1998; Zhang et al., 1998; Sheen et al., 2002; Terman et al., 2002; Lauren et al., 2003), regulation of neural connections and glutamate receptor trafficking (Pcdhgc3, Lmtk2, Ppfia1) (Serra-Pages et al., 1998; Wu and Maniatis, 1999; Ko et al., 2003; Kawa et al., 2004; Dunah et al., 2005). Finally, neurogenesis is altered by aging and estrogen (Kempermann et al., 2004; Galea et al., 2006; Saravia et al., 2007) and we observed a shift in expression of genes supporting neurogenesis, dendritic formation, and synaptogenesis (Pou3f1, Vldlr, Adam19, Tcf12, Emp1) (Frantz et al., 1994; Wulf and Suter, 1999; Kurisaki et al., 2002; Uittenbogaard and Chiaramello, 2002; Niu et al., 2004; Sinagra et al., 2005). Together, these results suggest estrogen's actions on hippocampal function in middle-age include trophic influences.
Brain aging involves enhanced inflammation and stress responses (Prolla, 2002; Verbitsky et al., 2004) and EB treatment is neuroprotective (Singh et al., 2006). Middle-aged mice exhibited reduced expression of genes for preventing oxidative (Akr1c13) and ischemic damage (Egln2), as well as genes preventing damage from inflammation (C1qbp). EB increased transcription of these protective proteins, and reversed an elevation in the expression of genes associated with stress (Foxo3a, Rps9, Myd116) and inflammation (Socs7). EB also increased the expression of mRNA coding for enzymes involved in DNA repair (Mms19l, Giyd2, Polb), ubiquitin ligase activity, and protein degradation (Derl3, Fbxw8, Psmd3, Psmb5, Ube2r2). This effect is consistent with studies indicating that improved performance in aged animals is associated with transcriptional up-regulation of genes associated with the proteasome (Blalock et al., 2003; Burger et al., 2007). Furthermore, an increased expression of Fbxw8 has been associated with memory consolidation in young animals (Cavallaro et al., 2001). Thus, an EB dependent increase in neuroprotective genes may contribute to maintenance of hippocampal function with age.
Some aging-EB responsive genes have been linked to premature aging (Tbl2, Il13ra1), transcriptional changes in the hippocampus of senescence-accelerated mice (Dusp12) (Meng et al., 1998; Kyng et al., 2003; Cheng et al., 2007), and aging in different organisms (Lass2) (Obeid and Hannun, 2003), or organs (Peli1) (Cheng et al., 2004; Chelvarajan et al., 2006). Some genes may be associated with biological markers of aging including altered lipid/cholesterol metabolism (Abca2, Vldlr, Acsl5, Srebf1, Apoa1) (Foster, 2006), lipid composition (Ptdss2) (Giusto et al., 2002), mitochondrial function (Atp5g1, Aco2) (Poon et al., 2006), and structural changes (Ctsb) (Bednarski et al., 1997). Indeed, specific genes may be markers for an aging hippocampus. Similar changes were found in CA1 of aged rats, including decreased Ilkap (i.e. protein phosphatase 2C), Polb, Bsg, and Cct6a and an increase in Nrp1(Blalock et al., 2003). Importantly, this transcriptional profile was reversed by EB treatment in middle-aged females.
Many of the aging-EB responsive genes can be classified under one of the following cell signaling processes: cytokine signal transduction (Socs7, Il13ra1, Peli1, Irak4), G-protein coupled receptor signaling (Adcyap1r1, Gpr125, Cbln1, Tmem11, Trh), and phosphatase/kinase activity (Ilkap, Lmtk2, Adck1, Hs1bp3, Dusp12, Irak4, Nagk, Pace4, Ppfia1, Ppp1r2). Age-related changes in these signaling cascades may contribute to altered synaptic plasticity and memory impairments (Lynch, 1998; Foster, 1999). Thus, estrogen influences on these signaling pathways may preserve physiological processes involved in memory (Foster, 2005).
EB treatment of ovariectomized middle-age mice reversed transcriptional markers of brain aging which have been previously described. It may be important that altered transcription emerges by middle-age, prior to cognitive decline, possibly in response to oxidative stress, or altered neural activity (Blalock et al., 2003; Lu et al., 2004). Unchecked, these processes may initiate cascades for altered cell signaling and transcriptional regulation, resulting in a decrease in neuronal growth and enhanced inflammation. Thus, middle age may be a critical period for treatments that can delay aging processes that begin in middle-age and cumulate over time to weaken cognition. In this regard, it may be important the EB treatment of middle-aged animals counteracts changes in transcription for proteins associated with cell signaling cascades, and reverses a decline in markers for neuroprotection, biosynthesis, and neurite growth.
This work was supported by NIH MH 059891 and the Evelyn F. McKnight Brain Research Foundation.
Disclosure Statement: The authors of this manuscript have no actual or potential conflicts of interest. Procedures involving animal subjects have been reviewed and approved by the Institutional Animal Care and Use Committee and were in accordance with guidelines established by the U.S. Public Health Service Policy on Humane Care and Use of Laboratory Animals.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.