|Home | About | Journals | Submit | Contact Us | Français|
Alpha-synuclein has been implicated in Parkinson’s disease, yet the mechanism by which alpha-synuclein causes cell injury is not understood. Using a transgenic mouse model, we evaluated the effect of alpha-synuclein overexpression on gene expression in the substantia nigra. Nigral mRNA from wildtype and alpha-synuclein transgenic mice was analyzed using Affymetrix gene arrays. At three months, before pathological changes are apparent, we observed modest alterations in gene expression. However, nearly 200 genes were altered in expression at nine months, when degenerative changes are more apparent. Functional genomic analysis revealed that the genes altered at nine months were predominantly involved in gene transcription. As in human Parkinson’s disease, gene expression changes in the transgenic model were also modulated by gender. These data demonstrate that alterations of gene expression are widespread in this animal model, and suggest that transcriptional dysregulation may be a disease mechanism that can be targeted therapeutically.
Parkinson’s disease (PD) is a debilitating neurological disorder associated with dopaminergic cell loss in the substantia nigra (SN). Familial PD cases have been linked to several genes, including α-synuclein (α-syn), a 140-amino-acid protein that may function in neurotransmitter release (Abeliovich et al., 2000; Murphy et al., 2000). Families with mutated α-syn exhibit autosomal dominant PD (Athanassiadou et al., 1999; Kruger et al., 1998; Polymeropoulos et al., 1997; Zarranz et al., 2004), and gene multiplication leading to increased wildtype α-syn levels also causes disease (Singleton et al., 2003). Certain α-syn promoter polymorphisms are PD risk factors, suggesting that even modest increases in expression can predispose to disease (Farrer et al., 2001; Kruger et al., 1999; Pals et al., 2004; Tan et al., 2000; Wang et al., 2006). In patients without genetic mutations, α-syn is found in Lewy bodies and degenerating neurites (Irizarry et al., 1998; Spillantini et al., 1997).
Alpha-syn overexpression has been used to generate cellular and animal models of PD. Mutant α-syn or overexpression of wildtype α-syn induces cell death in dopaminergic cell lines and in primary dopaminergic cultures (Oluwatosin-Chigbu et al., 2003; Xu et al., 2002; Zhou et al., 2000; Zhou et al., 2002). Transgenic mice expressing mutant or wildtype α-syn show motor deficits, alterations in dopamine levels, and α-syn-positive inclusions (Hashimoto et al., 2003; Maries et al., 2003). Alpha-syn is also involved in the action of neurotoxins, such as MPTP and rotenone, used to model PD (Dauer et al., 2002; Betarbet et al., 2000).
The mechanism by which α-syn injures dopaminergic neurons remains unclear. A potential mechanism not closely examined is transcriptional dysregulation, in which interference with normal gene expression patterns leads to cellular dysfunction. Transcriptional dysregulation has been implicated in Huntington’s and Alzheimer’s diseases (Cha, 2000; Robakis, 2003). Using microarray methods, we have observed that PD causes gene expression alterations in human dopaminergic neurons (Cantuti-Castelvetri et al., 2007). Alpha-syn can also inhibit histone acetylation, a potent mechanism for altering gene expression (Kontopoulos et al., 2006).
Here we examined gene expression patterns in SN cells isolated from transgenic mice in which human wildtype α-syn is overexpressed under the platelet-derived growth factor β promoter. These mice develop neuronal α-syn inclusions in the SN and other brain regions (Masliah et al., 2000). At twelve months, they show motor impairments and decreased striatal dopaminergic terminals and tyrosine hydroxylase activity (Masliah et al., 2000). We used laser capture microdissection to isolate nigral cell RNA from transgenic and control mice, and microarray analysis to evaluate for alterations in gene expression induced by α-syn overexpression at two time points: 1) three months, when pathological changes are few, and 2) nine months, when α-syn inclusions are more widespread. Our data reveal the early effects of α-syn overexpression are modest. At the later time point, changes in gene expression are more prominent in these transgenic mice, and many of the altered genes have functions related to transcription.
α-synuclein mice were originally generated by Masliah et al. (2000). A breeding colony from the Masliah D line was set up at Charles River Laboratories to generate transgenic and wildtype littermates. The use of mice was supervised by the Massachusetts General Hospital Animal Resources Program in accordance with the PHS policy on Humane Care and Use of Laboratory Animals. Prior to sacrifice, animals were euthanized by CO2 inhalation. Six gender-matched wildtype and six transgenic littermates were sacrificed at three months of age, and another six control and six transgenic mice were sacrificed at nine months of age. Brains were immediately frozen in isopentane and stored at -80°C. Brains were sectioned at eight μm by cryostat, mounted on uncoated glass slides, and stored at -80°C.
All of the following procedures were done under strict RNase-free conditions. Eight μm sections through the midbrain were first thawed and fixed with acetone for 40 seconds. To identify nigral neurons, these sections were immunostained using a primary mouse monoclonal antibody against tyrosine hydroxylase (TH; 5 minutes at 1:1500; Sigma) and a cy3-conjugated goat anti-mouse secondary antibody (5 minutes at 1:200; Jackson ImmunoResearch Labs, West Grove, PA) and then dehydrated (50, 70, 95, 100% ethanol, 5 seconds at each concentration, followed by xylene for several minutes). TH-stained substantia nigra pars compacta (SNpc) were laser captured from each section with an Arcturus PixCell II instrument (Mountain View, CA). Because the SNpc in mice is very densely packed with cells, we were unable to accurately count the number of captured TH-positive neurons and likely also captured some other cell types. To minimize differences in the amount of captured RNA, we attempted to capture cells from about eight nigral sections of comparable size for each animal. Global normalization of microarray data and normalization of quantitative PCR data with “housekeeping” genes were performed to correct for differences in the number of captured cells among animals.
RNA from laser-captured nigral cells was extracted using the Picopure™ RNA isolation kit (Arcturus) according to manufacturer’s specifications. Purified RNA was then amplified two rounds using the RiboAmp™ RNA amplification kit (Arcturus). This kit is based on a T7-RNA polymerase method that can linearly amplify total RNA quantities as small as 1 ng. After amplification, mRNA quality was assessed by an Agilent Bioanalyzer (Agilent, Palo Alto, CA). Selected samples had similar product sizes averaging approximately 500-1000 bp. Those samples with smaller average product sizes were discarded. Appropriate amplified RNA (aRNA) samples were sent to the Harvard Partners Center for Genetics and Genomics (HPCGG) facility to generate biotinylated aRNA that was then hybridized to Mouse Expression 430 2.0 arrays (Affymetrix, Santa Clara, CA). Each biotinylated aRNA sample from an individual animal was hybridized to one array, for a total of 24 arrays.
Validation PCRs were performed on a separate set of nigral neurons laser captured from wildtype and transgenic mice using the Arcturus Veritas system. RNA from these cells was extracted using the Picopure™ RNA isolation kit. The extracted RNA was reverse transcribed into first-strand cDNA using the SuperScript™ II reverse transcriptase kit (Invitrogen, Carlsbad, CA). cDNA was then precipitated using sodium acetate and ethanol with 1μg of glycogen as a carrier.
For each age group, two genes whose levels were altered in the microarray analysis were selected to validate microarray data. The two genes selected for validation were altered at both time points in the microarray analysis and were expressed at high intensities. 20mer, synthetic PCR primers were designed using Primer3 (http://frodo.wi.mit.edu). Primers against cyclin D (NM_007631) were 5’ tgaacccaaggaggaatcag 3’(forward) and 5’ gaagcccaaattcaccaaac 3’ (reverse; product size 256 bp). Primers against TCDD-inducible poly (ADP-ribose) polymerase (NM_178892) were 5’ ctggaaccctgagatccttg 3’ (forward) and 5’ gacacgatgggttgatttcc 3’ (reverse; product size 153 bp). For real-time quantitative PCR (QPCR), first strand cDNA created from extracted RNA was incubated with appropriate forward and reverse primers and SYBR® Green PCR Master Mix (Applied Biosystems, Foster City, CA) in a 96-well plate. QPCR was performed using an iQ5Cycler (BioRad, Hercules, CA) set to the following protocol: 1 cycle of denaturation at 95°C for 10 minutes; 50 cycles of denaturation at 95°C for 30 seconds, annealing at 57°C for 30 seconds, and polymerization at 72°C for 45 seconds; and finally, 80 0.5°C increases in temperature (starting at 55°C) to collect melting curve data. For each primer set, we used a standard curve with known concentrations of cDNA to calculate primer efficiency and to quantitate PCR products. The quantity of each PCR sample was calculated using the ΔΔCt method (Fink et al., 1998; Livak and Schmittgen, 2001). We used both pyruvate decarboxylase (forward primer 5’ gctggagaaggacctgattg 3’; reverse primer 5’ ccagctcaacctcaaactcc 3’; product size 191 bp) and hypoxanthine phosphoribosyltransferase (forward primer 5’ tgttgttggatatgcccttg 3’; reverse primer 5’ tgcgctcatcttaggctttg 3’; product size 107 bp) levels to normalize PCR results, as both genes were unchanged between transgenic and wildtype mice.
Microarray analysis was performed as previously described (Cantuti-Castelvetri et al., 2007). Chips were developed, scanned, and normalized using global scaling. All quality control parameters calculated by Affymetrix GCOS Software were monitored. The images of the chips were analyzed to find spotted or damaged array regions, and the graphical analysis of all chips through this step provided evidence that the data was of good quality (Quackenbush, 2002). All data were normalized using the GeneChip Robust MultiChips Analysis (GCRMA) algorithm (Cope et al., 2004) performed with ArrayAssistLite (Stratagene, La Jolla, CA). Normalized data was then statistically analyzed with multiclass analysis followed by two-tailed unpaired t-tests using the Significance Analysis of Microarrays (SAM) algorithm with SAM 2.0 plug-in for Excel (Efron and Tibshirani, 2002; Larsson et al., 2005; Tusher et al., 2001).
Any probes associated with transcripts that were no longer listed in the EntrezGene or Unigene databases were excluded, as were probes which were present for less than half the animals in a given experimental group (transgenic, wildtype, female transgenic, female nontransgenic, etc). Probes with intensities lower than background in all samples were also filtered out (log2 intensity <3). Average expression for each probe/gene was calculated for each of the following groups of animals: 1) all three-month-old transgenic animals, 2) all three-month-old wildtype animals, 3) all nine-month-old transgenic animals, 4) all nine-month-old wildtype animals, 5) three-month-old female transgenic mice, 6) three-month-old female wildtype mice, 7) three-month-old male transgenic mice, 8) three-month-old male wildtype mice, 9) nine-month-old female transgenic mice, 10) nine-month-old female wildtype mice, 11) nine-month-old male transgenic mice, and 12) nine-month-old male wildtype mice. Average expression for a given probe was calculated for all transgenic mice and for all wildtype mice, and the ratio of average transgenic mouse expression to average wildtype mouse expression was calculated for each probe. This ratio was expressed as a log2 ratio in our computer analyses programs. Probes were considered to be different between two groups if the log2 ratio between the two groups was greater than 0.5 (equivalent to 1.4-fold difference) or less than -0.5 (equivalent to 0.7-fold difference) and if the p value was less than 0.05 for the unpaired t-test. To reduce the possibility of a type I statistical error and to control for the variance within each probe set considered in the analysis, we also estimated the false discovery rates (represented by q values) using SAM and limited the final lists of differentially expressed genes to those probes with q values ≤ 20%.
Hierarchical clustering on median normalized samples using cosine correlation with complete linkage was performed for each selected probe sets lists on all three-month samples and on all nine-month samples using SpotFire DecisionSite for Functional Genomics 8.0 (Fig. 1; Eisen et al., 1998).
To analyze the biological roles of the differentially expressed genes, we used the Database for Annotation, Visualization and Integrated Discovery (DAVID) 2007 (Dennis et al., 2003; Hosack et al., 2003). DAVID 2007 is a freely-available, NIH-sponsored data-mining tool that uses several databases containing gene-specific functional data, such as Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), the Protein Information Resources (PIR), and others. DAVID calculates the probability that particular functional categories are overrepresented in a given data set using a one-tailed Fisher exact probability for overrepresentation (or EASE score) using a Gaussian hypergeometric probability distribution. This distribution describes sampling without replacement from a finite population made of two types of elements: genes that belong to a particular functional category and genes that do not belong to that particular category.
For the functional analysis, we used only those differentially expressed probe sets that had a q value ≤ 20%. First, the DAVID functional annotation tool was used to generate those categories defined by GO database annotations that were overrepresented. All GO categories that had a p value ≤ 0.05 and which contained at least 3% of all regulated genes were considered significant, regardless of how broad or specific the categories were. The functional annotation clustering tool in DAVID was then used as a second level of functional analysis. This tool evaluated each annotation term by the genes that it was comprised of from a given data set and then clustered that term with similar annotations from other ontology databases based on the amount of genes that were co-associated. Only those functional clusters that gave a median p value ≤ 0.05 were considered significant.
We used laser capture microdissection to isolate nigral cell RNA from both α-syn transgenic and wildtype mice. After amplification, RNA quality was assessed by an Agilent Bioanalyzer. Samples selected for microarray hybridization had similar product sizes averaging 500-1000 base pairs. Those samples with smaller-than-expected average product sizes were discarded, and more nigral cell RNA was captured and amplified. For each age group, we isolated and amplified RNA from gender-matched groups of six transgenic and six wildtype mice. The resultant aRNA samples were labeled and hybridized to Affymetrix Mouse Expression microarrays. Each aRNA sample from an individual animal was hybridized to one array, for a total of 24 arrays. Background intensity levels for all arrays were similar with a mean 58.9 ± 6.8 (SEM). Average signal for present probes across all arrays was 388.6 ± 22.2 (SEM). As a group, the 24 arrays had a mean 42.0 ± 0.8% (SEM) of probes classified as “present.”
We initially examined nigral neurons from mice at three months of age, a point at which there is no apparent histological abnormality in the SN. Data normalized by the GCRMA algorithm (Cope et al., 2004) was statistically analyzed with multiclass analysis followed by two-tailed unpaired t-tests using the SAM algorithm (Efron and Tibshirani, 2002; Larsson et al., 2005; Tusher et al., 2001). We considered probes to be different between wildtype and transgenic mice if the log2 ratio between the groups was greater than 0.5 (equivalent to 1.4-fold difference compared to wildtype) or less than -0.5 (equivalent to 0.7-fold difference compared to wildtype) and if the p value was less than 0.05 for the unpaired t-test. 237 genes were differentially expressed between transgenic and wildtype littermates (Supp. Table 1). 180 of these genes have been identified in the Entrez Gene and Unigene databases with names instead of expressed sequence or RIKEN numbers. Hierarchical clustering performed using the data from these differentially expressed genes appropriately classified all of the samples from three-month-old mice, although it was clear that there was some variation within the groups in the strength of the phenotype defined by this set of genes (Fig. 1a).
Among this list of genes, we validated two genes by quantitative PCR. The two genes chosen were cyclin D and TCDD-inducible poly (ADP-ribose) polymerase (Tiparp), which were both expressed at high intensities at three months. We collected SNpc by laser capture microdissection from six transgenic and six wildtype mice. Expression levels were determined in cDNA samples produced from the isolated neurons without any intervening amplification steps. The expression levels of these two genes as determined by QPCR were similar to the microarray results (Table 1).
To reduce the possibility of a type I statistical error and to control for the variance within each probe set considered in the analysis, we also estimated the false discovery rate for each gene (represented by the q value) altered at three months and eliminated those genes that had a q value > 20%. By this more conservative statistical approach, we reduced the initial list of 237 genes to 29 (Table 2). Six of these genes were downregulated in the transgenic mice, while 23 were upregulated. 22 of the 29 genes were associated with gene names other than expressed sequence or RIKEN numbers. None of the altered genes in this list include molecules previously linked to PD. All upregulated genes from the initial 237 gene list were found in this more conservative list, while only six of the initial 214 downregulated genes had q values ≤ 20%. Because of the small number of genes modulated at this early time point, further classification by ontological analysis was not informative.
We performed similar laser capture microdissection and array experiments with nine-month-old mice, which have widespread α-syn inclusions in the brain, to evaluate gene expression changes at a later point in disease pathogenesis. We found that the magnitude of the transcriptional dysregulation, as measured by the number of altered genes, was much larger at this late stage of the disease model, and the identities of the genes modified were different than those observed to be altered at the earlier time point.
As with the three month data, we performed an initial analysis using criteria of p value ≤ 0.05 and a log2 ratio of at least ±0.5. Using these filters, we found that 278 genes were altered between nine-month-old α-syn and wildtype mice (Supp. Table 2). 211 of these genes have been identified in the Entrez Gene and Unigene databases with names other than expressed sequence or RIKEN numbers. Hierarchical clustering using this gene set revealed that the data readily classified the transgenic and control animals (Fig. 1b). Indeed, the distinctions between the groups were much more striking than at three months, suggesting that at a later stage in the model the effect of α-syn overexpression is more uniform across the population of animals studied. Among the nine-month list of genes, we validated cyclin D and Tiparp by QPCR. The expression levels of these two genes as determined by QPCR were similar to the microarray results (Table 1).
As with the three-month microarray data, we also estimated the false discovery rate for each gene altered at nine months and eliminated those genes that had a q value > 20%. This conservative statistical approach reduced the initial list of 278 genes to 179 (Table 3). In contrast to the three-month time point, the predominant effect at nine months was a reduction of gene expression, with more genes downregulated than upregulated in the transgenic mice. The mRNA levels of 165 genes were decreased, while the mRNA levels of 14 genes were elevated in the nine-month-old α-syn mice. Some of the genes that were altered in the nine-month-old group included an E3 ubiquitin protein ligase (Cbl), a proteosomal subunit (Psmd14), and other molecules linked to PD, including the neurotrophin receptor TrkB (von Bohlen und Halbach et al., 2005), and 14-3-3θ(Berg et al., 2003; Kawamoto et al., 2002). Interestingly, among the downregulated genes was ceruloplasmin, which is decreased in Wilson’s disease (Scheinberg and Gitlin, 1952), a genetic disorder with Parkinsonian symptoms.
We next compared the identity of the genes altered in the three-month-old and nine-month-old groups. Surprisingly, we found no genes that overlapped between the three-month and nine-month groups when analyzing those lists limited by q value ≤ 20%. When looking at the less conservative gene lists (in which genes were not eliminated by q value criterion), nine genes showed alterations in expression level at both time points. These genes included calmodulin 1, cyclin D2, disheveled associated activator of morphogenesis 1, fibronectin type III domain containing 4 (Fndc4), hematological and neurological expressed sequence 1 (Hn1), nucleoporin 62, PCTAIRE-motif protein kinase 2 (RIKEN 6430598J10), Tiparp, and expressed sequence AI55357. Seven of these genes showed concordant downregulation at both three and nine months, while Fndc4 and Hn1 showed alterations in opposing directions at the two time points.
Functional genomic analysis of the data from the nine-month-old mice was performed to evaluate whether the nine-month data set was enriched in any particular functional categories. We used the DAVID 2007 functional annotation tool (http://david.abcc.ncifcrf.gov/home.jsp; Dennis et al., 2003; Hosack et al., 2003), which classifies each gene into one or more functional categories using annotations from over 40 ontological databases, including the GO database, and then determines which functional categories are represented in the data set more frequently than expected by chance alone. We used the more conservative nine-month list of 179 genes with upregulated and downregulated genes combined to perform this analysis. We considered significantly overrepresented those functional categories in which the number of genes altered exceeded the number expected with a p value ≤ 0.05. We limited the functional annotation chart generated by DAVID to those categories defined by GO database annotations (http://www.geneontology.org/). Table 4 reveals 39 GO ontological categories that contain at least 3% of the regulated genes. We did not exclude broad categories from this ontology table and instead included all parent and children categories that revealed a p value ≤ 0.05 and contained at least 3% of all regulated genes. Categories that were enriched in our data set included GTPase regulator activity, cytoskeletal protein binding, ubiquitin cycle, protein amino acid phosphorylation, transcription, and signal transduction.
We next performed a higher order functional analysis using the functional annotation clustering tool in DAVID, which evaluates each annotation term by the genes that it is comprised of from a given data set and then clusters that term with similar annotations from other ontology databases based on the amount of genes that are co-associated. By this clustering method, we found six functional clustering groups that had a median p value ≤ 0.05 (Table 5). Five of these six functional clusters were associated with DNA transcription and related categories, and one was associated with signaling. Among these co-clustering functional categories, 14.89% of genes altered at nine months were involved in transcription, 13.62% bind DNA, 12.34% bind zinc ions, 4.26% comprise transcription factor complexes, and 8.09% have transcription factor activity. Interestingly, 2.13% show steroid hormone receptor activity. Based on this functional annotation clustering, transcription appeared to be the predominant biological function that was altered in the nine-month group.
Because gender has recently been found to play an important role in transcriptional changes in human PD (Cantuti-Castelvetri et al., 2007), we evaluated whether gender modified the gene expression changes in the α-syn transgenic mice. The primary analysis described above was conducted in a group of animals where the number of mice of each gender was balanced. To analyze the influence of gender on gene expression patterns between wildtype and transgenic mice, we compared the average gene expression of female α-syn mice to female wildtype littermates and did a similar comparison of male transgenic to male wildtype mice. The group sizes in the gender analysis were necessarily smaller than in the primary analysis, and therefore, may lack statistical power to detect gene expression alterations. For this analysis, we limited the gender-specific data sets by the following criteria: 1) p value ≤0.05, and 2) log2 ratio of at least ±0.5. At both time points, females showed a greater dysregulation in gene expression than males did. 226 genes were altered in three-month-old transgenic females compared to wildtype females, and 332 genes were altered in nine-month-old transgenic females by at least a log2 ratio of ±0.5. The majority of the genes were downregulated in the transgenic female mice compared to gender-matched controls. Expression of 191 genes was decreased, while 35 genes showed increased expression at three months. At nine months, 193 genes showed decreased expression, and 85 had elevated levels in the transgenic females.
In contrast, fewer genes showed alterations in expression in the transgenic males compared to male controls. In three-month-old transgenic males, alterations in gene expression were apparent in 173 genes, while in nine-month-old transgenic males, 188 genes were altered by at least a log2 ratio of ±0.5. Unlike the females, transgenic males had more genes that increased in expression. At three months, 73 genes had decreased expression levels, and 100 had increased levels in transgenic males. At nine months, a decrease in expression was found in 77 genes, and an increase was seen in 111 genes in transgenic males.
A comparison of the identities of the genes regulated in the two genders revealed that the difference between the males and females was not simply the result of a difference in magnitude of regulation. Few of the same genes were altered in both genders at either age. Six genes were altered in both male and female transgenic mice at three months, and 20 were altered in both genders at nine months. Because our group sizes in the gender analysis were small, interpretation of our gender-based findings is limited. While we cannot conclude which genes are necessarily crucial to gender-related differences in α-syn toxicity, this gender analysis does point to the general conclusion that gender impacts the gene expression changes observed in α-syn mice.
In this study, we examined the effect of α-syn overexpression in mice on gene expression in the SNpc. Our data showed that α-syn overexpression caused alterations in a small number of genes early in the pathological process, before any cellular or behavioral changes were apparent in the transgenic mice. Alterations in gene expression were more pronounced at a later stage of disease when pathological changes were apparent, and these late changes affected different genes than those modified early. These late gene expression changes were focused around transcription-related processes.Our analysis also suggested that gender modified the alterations of gene expression in this model.
We found striking differences between early changes in gene expression and those observed after pathological changes had clearly developed in the α-syn model. Prior to the onset of pathological changes, gene expression changes in the α-syn transgenic mice were modest. By conservative statistical criteria, we found only 29 genes to be altered. In contrast, 179 genes were altered in older transgenic mice. The larger gene list from the nine-month-old mice analysis reflects in part a reduction in variability among the animals, which presumably is a result of a more stable state of α-syn-related pathology at this later stage. When limiting our data sets by false discovery rates, we found no genes altered at both early and late time periods. Only nine genes were found altered in transgenic mice at both time periods when we evaluated our less stringent gene lists not limited by false discovery rate. Two of these nine genes are implicated with cell cycle regulation – calmodulin 1 (Rasmussen and Means, 1989) and cyclin D2 (Jena et al., 2002). The other seven genes do not seem to share any particular biological processes, with their functions still largely unknown.
The most striking result of our study is that many of the genes altered later in the disease model were directly related to transcriptional regulation, nuclear function, and DNA binding. These data support the view that transcriptional dysregulation is an important consequence of α-syn overexpression and, therefore, may be critical to the pathogenesis of synucleinopathies. Alpha-syn has been identified within cell nuclei and has been shown to associate with histones in vitro (Goers et al., 2003; Maroteaux et al., 1988). Studies by Kontopoulos et al. (2006) have provided direct evidence that α-syn can inhibit histone acetylation in both mammalian cell culture models and transgenic Drosophila. The effect of α-syn on histone acetylation appears critical to α-syn-induced toxicity, as treatment with histone deacetylase inhibitors rescues against α-syn-induced toxicity (Kontopoulos et al., 2006). The present study confirms the transcriptional effect of α-syn in a mouse model for PD.
It is important to recognize that microarray studies do not directly assess transcription. The parameter measured is the abundance of specific mRNAs, which can be affected not only by transcription but also by alterations in RNA half-lives. While we cannot exclude an effect of α-syn on the stability of some transcripts, the large number of alterations as well as the observed changes in mRNAs for many components of the transcriptional process suggests that transcriptional dysregulation is the predominant effect underlying the changes.
Another significant finding is the confirmation in this mouse model that gender influences gene expression changes induced by α-syn overexpression. This analysis was prompted by the recent observation that gender has a marked effect on gene expression in human dopaminergic SN neurons, influencing the patterns of gene expression in both normal brain and the response to PD (Cantuti-Castelvetri et al., 2007). Other microarray studies have also revealed gender differences in gene expression in normal human brain (Vawter et al., 2004), but there is little additional data available on the interaction of gender and disease state on neuronal gene expression in either humans or animal models. The importance of gender is emphasized by the fact that the incidence of PD is nearly twice that in men as in women (Baldereschi et al., 2000; Van Den Eeden et al., 2003). Since our study was originally designed with gender-matched animal groups, we were able to examine the effect of gender directly, although the number of animals in the subgroup analysis is necessarily smaller and lacks statistical power compared to the primary analysis. Nevertheless, we found substantial gender-based differences in the response to α-syn overexpression. We observed that α-syn caused more gene changes among female mice, and most altered genes were different between genders. At present it is not known whether mice exhibit gender-based differences in vulnerability to α-syn toxicity, yet in other animal PD models, males have exhibited increased susceptibility to MPTP (Dluzen and McDermott, 2000; Freyaldenhoven et al., 1996; Miller et al., 1998) and 6-hydroxydopamine (Murray et al., 2003; Tamas et al., 2005). Our findings suggest that issues of gender should be carefully considered in the design of experimental studies using animal models of PD.
The mouse model used here was selected because it demonstrates progressive dopaminergic dysfunction with a motor phenotype (Masliah et al., 2000), but the nature of the model imposes some limitations on our study. Because the mouse SNpc is much more densely packed with cells than in the human SNpc, we dissected the entire SNpc region for this study, and not individual neurons. Thus, the transcriptional changes may reflect alterations in glia as well as neurons. It is also important to note that this transgenic α-syn mouse does not fully recapitulate all pathological aspects of PD. This mouse does show evidence of dopaminergic dysfunction and mild motor impairment, but it does not show loss of dopaminergic nigral neurons (Masliah et al., 2000). Another pathological difference is that α-syn inclusions are much more widespread in Masliah mouse brains than are Lewy bodies found in PD brains. Therefore, many of the expression changes that we saw may be more broadly applicable to other synucleinopathies, such as dementia with Lewy Bodies. All other published α-syn transgenic mice also fail to show dopaminergic neuron loss, yet these models vary in the appearance of α-syn-positive inclusions and severity of motor disability (Hashimoto et al., 2003; Maries et al., 2003). Transgenic mice expressing the A53T α-syn mutant show more severe motor impairments (Giasson et al., 2002; Lee et al., 2002), and some of the other transgenic mice have α-syn inclusions more typical of α-syn aggregates found in PD (Kahle et al., 2000; Lee et al., 2002).
While there are accepted principles for microarray analyses, different approaches to normalization and selection may lead to different outcomes regarding which genes are defined as differentially expressed (Hoffmann et al., 2002). Significance Analysis of Microarrays (SAM) is currently the method most commonly used that corrects for the large number of comparisons inherent in microarray experiments by estimating false discovery rate (Efron and Tibshirani, 2002; Larsson et al., 2005; Tusher et al., 2001). We limited false positives by excluding those genes whose q values were > 20%. However, by using this conservative statistical method, we likely eliminated some genes that are truly altered between wildtype and transgenic animals. Indeed, by QPCR we did confirm alterations in two genes identified by less stringent criteria, Tiparp and cyclin D2, yet both of these genes had high q values in the three-month animals and were thus eliminated from the more stringent three-month list. Attempts to validate directly individual genes introduce additional issues, especially with regard to the selection of genes for normalization. Although genes such as beta-actin and GAPDH have often been used for this purpose, expression of such “housekeeping” genes can vary, especially in neurodegenerative disease (Gutala and Reddy, 2004; Radonic et al., 2004; Vandesompele et al., 2002). Indeed, our array data does show that GAPDH levels are altered in the nine-month-old α-syn transgenic mice. We chose pyruvate decarboxylase and hypoxanthine phosphoribosyltransferase, both of which were unchanged in our microarray lists, to verify that similar results were obtained regardless of the gene chosen for normalization.
Given these considerations, we have decided to publish two sets of gene lists in this paper: 1) the three-month and nine-month tables (Tables 2 & 3) restricted by q values in addition to fold-change and p value criteria; and 2) the supplementary tables (Supp. Tables 1 & 2) restricted only by fold-change and p value criteria. In addition, the full gene array dataset from this study is publicly available on the Array Express site and permits further analysis and comparison with other studies.
In conclusion, we have explored the molecular changes induced by α-syn at a transcriptional level by comparing α-syn transgenic mice to wildtype mice by microarray analysis. Significant transcriptional changes occur in the SNpc of older transgenic mice, and these changes are influenced by gender. A large proportion of the altered genes are involved in transcriptional regulation. The next step is to evaluate whether some of the differentially expressed genes are protective or, alternatively, can magnify α-syn toxicity in cellular and animal models. This study reinforces the concept that targeting transcriptional dysregulation as a mechanism may be a useful therapeutic approach for PD.
Supported by the MGH/MIT Morris Udall Center of Excellence in PD Research (NIH NS38372) and the Parkinson’s Association of Alabama.
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.