Postmortem Interval Influences α-Synuclein Expression in Parkinson Disease Brain

Duplications and triplications of the α-synuclein (SNCA) gene increase risk for PD, suggesting increased expression levels of the gene to be associated with increased PD risk. However, past SNCA expression studies in brain tissue report inconsistent results. We examined expression of the full-length SNCA transcript (140 amino acid protein isoform), as well as total SNCA mRNA levels in 165 frontal cortex samples (101 PD, 64 control) using quantitative real-time polymerase chain reaction. Additionally, we evaluated the relationship of eight SNPs in both 5′ and 3′ regions of SNCA with the gene expression levels. The association between postmortem interval (PMI) and SNCA expression was different for PD and control samples: SNCA expression decreased with increasing PMI in cases, while staying relatively constant in controls. For short PMI, SNCA expression was increased in PD relative to control samples, whereas for long PMI, SNCA expression in PD was decreased relative to control samples.


Introduction
One of the genes ubiquitously involved in Parkinson disease (PD, OMIM no. 168600) is the α-synuclein (SNCA) or PARK1/4 gene. The SNCA gene encodes two major transcripts: the full-length NM 000345.2 transcript and the NM 007308.1 transcript (corresponding to the NACP140, and the NACP112 protein isoforms, resp.). Missense mutations in SNCA [1][2][3], as well as duplications and triplications of the SNCA locus [4][5][6], have been shown to lead to familial PD in an autosomal dominant manner, suggesting that increased levels of SNCA are associated with PD risk. Several studies have compared sporadic PD and control SNCA mRNA levels, as well as α-synuclein protein levels in various tissues [7][8][9][10][11][12][13][14]. SNCA expression in human brain has been shown to be significantly different between sporadic PD cases and controls, although the direction of results varies among different studies (Table 1).
None of the past SNCA expression studies contrasting PD cases and controls used more than 32 total brain samples per brain region (Table 1). Small sample sizes reduce power to detect consistent effects and may have contributed to the conflicting results. We present the largest study to date contrasting SNCA expression between PD and control brain samples, with analysis performed for both full-length SNCA transcript (140 residue protein isoform, hereafter referred to as SNCA-FL), and total SNCA mRNA in the frontal cortex. Additionally, we analyze the relation of eight SNPs in the 5 and 3 regions of SNCA to SNCA expression levels.

pH Measurements.
The pH of all samples was measured following a previously established protocol [26]. A minimum of two pH measurements were taken for each brain sample and the average value of all readings was used.

Neuropathological Information.
Neuropathology reports were available for case and control samples. These reports were used to verify the PD diagnosis in the cases, and to evaluate the presence of Alzheimer disease (AD) characteristics in all brain samples. The AD variable for each brain was categorized as 0, 1, or 2 and was determined by a grading of plaques and the Braak score [27,28]. A value of 0 corresponds to brains that had no indication of Alzheimer pathology, a value of 1 corresponds to brains that had suggestive Alzheimer pathology, and a value of 2 corresponds to brains with unequivocal Alzheimer pathology (Supplementary Table 1

Endogenous Control Gene and the Analysis Method.
We considered both the Relative Standard Curve and the ΔΔCt methods for the real-time PCR quantification of SNCA expression. QARS (encoding for Glutaminyl-tRNA synthetase) was selected as a control gene, given its successful use in previous cortex expression studies [29,30]. Predesigned TaqMan primers for QARS (Hs00909458 g1), SNCA transcript NM 000345.2 (Hs00240907 m1, e.g., SNCA-FL), and all SNCA transcripts (Hs01103383 m1, total SNCA) were obtained from Applied Biosystems (Foster City, CA). Each sample was run in triplicate for each assay on an ABI PRISM 7900HT Sequence Detection system (Applied Biosystems, Foster City, CA). The control gene did not have the same PCR amplification efficiency as the SNCA target assays, which is a requirement for valid ΔΔCt calculations (Validation Experiment, ABI qRT-PCR manual). Therefore, the Relative Standard Curve Method was used to assess the expression data. The standard curves for the three assays were created from pooled cDNA from all available samples and were used to transform the Ct values into quantity units. For each sample that passed the QC filtering criteria, the quantity units for the SNCA-FL and total SNCA assays were standardized by division to the QARS control assay quantity value.

DNA Extraction.
DNA from the brain samples was extracted using QIAGEN's Puregene Core Kit A (QIAGEN, Valencia, CA) according to the manufacturer's protocol.

Genotyping.
Eight SNPs around the SNCA gene on chromosome 4 (Table 3) were genotyped in the available brain samples using the TaqMan technology implemented on an ABI PRISM 7900HT Sequence Detection system (Applied Biosystems, Foster City, CA). Pairs of SNPs rs356219-rs356229 (r 2 = 0.51) and rs4106153-rs1504489 (r 2 = 0.27) were in modest LD, as calculated by using all available brain samples.

Quality Control
2.5.1. RNA and DNA Extraction. Samples with RNA or DNA extraction yields below 5 μg after several attempts were removed from the study.

RT-PCR.
Samples were removed from the study if the variation in expression across the triplicate Ct values for any of the four gene expression assays was larger than 2 (Section 2.4).

Postmortem Information.
Samples were excluded if their PMI information was missing.

Neuropathological Information.
Samples were removed when controls showed signs of Lewy bodies, or when PD was not confirmed neuropathologically (e.g., absence of Lewy bodies).
2.5.5. Age at Death. The age at death was available for cases and controls, but the range of values differed by disease status. All controls outside of ±5 years of the PD range (Table 2) were excluded from the case-control contrasts.
2.5.6. Genotyping. All samples with missing genotypes for more than 4 SNPs (less than 50% call rate) were removed from the genotype-expression part of the study. All genotyped SNPs had call rates higher than 92.12%.

Statistical Analyses.
The statistical analyses were performed using SAS 9.1 for Windows. The base 10 logarithm of the standardized SNCA-FL and total SNCA expression values was used for all analyses, to ensure the normal distribution of data required by the statistical tests performed. The distributions of the log 10 SNCA-FL and log 10 total SNCA expression values were examined within site, sex, and disease status subgroups. Samples were removed from analysis if they had SNCA-FL and total SNCA expression values lower than the 1st quartile minus 1.5 * interquartile range, or greater than the 3rd quartile plus 1.5 * interquartile range. In the final sample data-set, the SNCA-FL and total SNCA log 10 transformed expression values did not deviate significantly from the normal distribution (Shapiro-Wilk test), except for two of the total SNCA subgroups (SHRI/female/Control and HBTRC/male/PD).

Association of SNCA Expression with Disease Status.
We considered several covariates when looking at the association between SNCA expression and disease status, including sex, PMI, source of the specimen, pH, AD, and age at death. Since the PMI was highly correlated to the brain bank source (Table 2), the use of either PMI or site as a covariate yielded comparable results in our models. We retained PMI in these models. PMI and pH were selected as covariates because they were found to be associated with SNCA expression. Sex was retained due to a prior report, showing differences in SNCA expression between men and women [31]. Age at death was retained in the final model because it was significantly associated with SNCA expression in controls and was a modest confounder of the relationship between disease status and SNCA expression. AD was not associated and was not a confounder of SNCA expression; therefore, it was not included in the regression analyses. The interaction between PMI and disease status was included to adjust for the observed variation in SNCA expression between PD cases and controls at different PMIs.

Genotyping
Analysis. Eight SNPs were tested for association with disease status, as well as for association with SNCA-FL and total SNCA expression. Each SNP was evaluated in additive and, whenever the rare homozygote was present, recessive models. Disease association models included adjustment for sex and age at death. Expression models were analyzed for the set of all brains, as well as within only PD cases and only controls. The entire sample analysis included adjustment for disease status, sex, pH, age at death, as well as for the interaction between PMI and

Results
Unless otherwise stated, a significance level of α = 0.05 and log 10 expression values (see Section 2.5.7) were used for all tests.

Samples Excluded from the Final Analyses.
One control brain and one PD case from HBTRC were excluded from the study due to low DNA extraction yields. One control sample from HBSFRC and seven control samples from SHRI were excluded from further analyses due to low RNA yields. Eleven PD cases (5 from HBTRC, 5 from SHRI, 1 from HBSFRC) and eleven controls (5 from HBTRC, 5 from SHRI, 1 from HBSFRC) were excluded due to inconsistencies among the Ct values in replicates. One control and one PD case from HBSFRC were discarded due to missing PMI information. One HBTRC control showing Lewy bodies at the neuropathological exam, and one HBTRC PD case with very long duration of disease but no Lewy body pathology were removed from analysis. Four brain samples (1 control from HBTRC, 1 PD from HBSFRC, and 2 PD from SHRI) were outliers for both SNCA-FL and total SNCA expression assays and were, therefore, removed. Four PD cases from HBTRC with missing genotypes for at least 6 SNPs were removed from the genotype-expression analysis only. Eighteen controls from HBTRC that were outside the ± 5 years age at death range for the PD group were removed for the expression analyses. The description of the final set of samples used in the study is presented in Table 2.  We observed a significant association between age at death and SNCA expression values in controls, after adjustment for pH, PMI, and sex ( Figure 1  In a PMI, age at death, and sex-adjusted model, pH was significantly, positively associated with expression of total SNCA in PD cases (P < 0.0001) and controls (P = 0.002) and with SNCA-FL in PD cases (P = 0.0024), but not in controls (P = 0.14). We also observed a significant negative association between PMI and both SNCA-FL and total SNCA expression values in PD samples after adjustment for sex, age at death, and pH (P < 0.0001 for both SNCA-FL and total SNCA), but not in controls (total SNCA: P = 0.96; SNCA-FL: P = 0.54).

Correlations and
The correlation between duration of disease in PD samples and SNCA expression was not significant (total SNCA: P = 0.97; SNCA-FL: P = 0.64). Additionally, no significant association between the duration of disease in PD samples and SNCA expression was observed after adjustment for sex, age at death, and pH.

Expression Results.
Interestingly, the PMI was determined to modify the relationship between expression and disease status for both SNCA-FL and total SNCA. For PMI of 5.5 hours or less (23 controls, 45 cases), PD cases had higher total SNCA expression (β = 0.1501, P = 0.0319) and higher SNCA -FL expression (β = 0.1195, P = 0.0051) than controls. For PMI of 10 hours or more (23 controls, 48 cases), PD cases had lower total SNCA expression (β = −0.2716, P = 0.0005) and lower SNCA-FL expression (β = −0.1708, P = 0.0093) than controls. The presented results were adjusted for sex, age at death, pH, and PMI. The predicted regression lines for total SNCA and SNCA-FL for PD cases and controls after adjustment for age at death, pH, PMI, sex, and disease status-PMI interaction are shown in Figure 2 and Supplementary Figure 2, respectively.

eSNP Results.
None of the eight SNPs had a significant nominal P-value for association with disease status in our data. Additionally, none of the SNPs had a P-value associated with expression that was significant after Bonferroni adjustment for multiple testing. Nevertheless, significant nominal P-values (Table 3) were obtained for the following SNPs: rs924033 for SNCA-FL expression using the additive model in controls only, and rs1560488 for total SNCA expression using the recessive model in controls only.

Discussion
The presence of the α-synuclein protein (α-syn) in Lewy bodies [34], together with the findings of SNCA gene mutations [1][2][3], SNCA gene duplications, and triplications in familial PD [4][5][6], and SNCA SNP associations in PD genome-wide association studies [32] make this gene a focal point of PD research. The association of increased gene dosage with PD risk strongly suggests that increased levels of α-syn increase risk for PD. Yet, SNCA gene expression studies have yielded inconsistent results with several reporting reduced SNCA mRNA levels in PD versus control brains. In this study of 101 PD and 64 control brains, we found significant differences for the effect of post mortem interval on SNCA levels in PD. These results may shed light on the previous contradictory expression findings and support the hypothesis that PD is associated with increased SNCA levels.
We detected increased expression of the full-length SNCA transcript, as well as overall SNCA gene expression, in PD compared to control brains at PMI up to 5.5 hours. Additionally, we observed significantly decreased levels of full-length and total SNCA in PD compared to control brains at PMI longer than 10 hours. The result obtained for short PMI suggests the presence of biologically increased levels of SNCA expression in PD compared to normal brains, while the apparently conflicting findings between short and long PMI groups could be attributable to a more rapid degradation rate of SNCA, and possibly other transcripts, in PD brains. The presence of increased RNA degradation activity in PD compared to normal brains is conceivable, given the large differences in expression profiles between normal and affected brains [29]. Additionally, although little information is available on the differential mRNA degradation levels between neurologically healthy and diseased brains, there exists a prior report of correlation between PMI and pH (and indirectly between PMI and certain mRNAs level [26]) in Alzheimer disease, but not in control brains [35].
Previous SNCA expression studies have consistently used relatively small numbers of brain samples with mixed PMI values (commonly above 10 hours), precluding an accurate assessment of the effect of PMI [36]. Therefore, the previous conflicting results (Table 1) might be an artifact of both small sample sizes and heterogeneous PMI values. It is important to note, however, that the different sources of tissue in our study were also related to different PMI values (Table 2). Nearly all of the short PMI samples were from the SHRI brain bank. Therefore, we cannot exclude the possibility that the source of the tissue also influences SNCA expression levels. Nevertheless, our statistical analyses included all major variables that are commonly considered in expression studies, and we do not know of any other differences that may exist among the different brain tissue sources and would influence SNCA expression. Additionally, given the previous knowledge of frontal cortex homogeneity in terms of expression [11], it is unlikely that variation within the Brodmann area 9 from different brain banks is a factor in the observed findings.
We acknowledge as a possible limitation for our SNCA RT-PCR expression study the use of a single control gene. To address this potential problem, we tried to evaluate the obtained RT-PCR expression data by using expression results from a recent microarray study [37]. The microarray experiment was performed on the One-Color Agilent 60mer Whole Human Genome Microarray, which contains a single 3 UTR probe for the SNCA gene repeated 10 times on the chip. The microarray experiment included a subset of 26 PD and 23 control samples from the RT-PCR study. The range of PMI values for the microarray samples did not allow the interaction study presented in this paper to be tested. Nevertheless, we could evaluate the correlation between the two RT-PCR SNCA probes and the median expression value of the microarray SNCA probe, which measures total expression of the gene. As expected, the correlation between the total SNCA and the microarray SNCA probe (r = 0.68, P = 6.6E−8) was strong and better than the correlation between the SNCA-FL and the microarray SNCA probe (r = 0.43, P = 0.001). These correlation results imply the validity of the RT-PCR data.
Our study suggests that sporadic PD is associated with increased SNCA mRNA levels in samples with short PMI. The observation of higher SNCA expression in controls among samples with longer PMI suggests that SNCA transcripts may degrade more rapidly in PD than in normal brain; this result points to the importance of brain samples with short PMI for an accurate evaluation of RNA levels in PD. Therefore, brain banks such as the Sun Health Research Institute, which can provide samples with very low PMI to the research community, are valuable for future neurodegenerative research.