Parkinson’s disease (PD) is the second most common neurodegenerative disease (Dorsey et al., 2007; Alzheimer’s Association, 2014), characterized by motor syndromes, such as resting tremor, bradykinesia, rigidity, and postural instability, or non-motor symptoms such as dementia, depression, and sleep disorders (Kalia and Lang, 2015). The prevalence of PD is approximately 1% in the population over the age of 60 years, but reaches up to 5% in individuals older than age 85 years, highlighting the impact of advancing age on the risk of this disease (Ascherio and Schwarzschild, 2016; Lim et al., 2019). The economic burden of PD is up to $23 billion USD per year (Findley, 2007). As life expectancy and aging populations increase worldwide, the number of people with PD is expected to increase by more than 1.5-fold by 2030, making this disease an urgent global health issue (Dorsey et al., 2007).

Although both genetic and environmental factors likely contribute to the development of PD, it shows a high heritability (Blauwendraat et al., 2019a; Li et al., 2019). With the development of high throughput technologies in the last two decades (Jansen et al., 2017; Sandor et al., 2017; Siitonen et al., 2017), thousands of studies have focused on the genetic architectures of PD to understand its pathogenesis (Kalia and Lang, 2015; Labbe et al., 2016; Blauwendraat et al., 2019b). Up to now, genome-wide association studies (GWAS) on the contribution of genetics in a common form of PD (sporadic PD) identified more than 100 loci that were associated with disease risk at a genome-wide significance level, which could sever as potential biomarkers and risk factors for PD (International Parkinson Disease Genomics Consortium, Nalls et al., 2011; Nalls et al., 2014; Chang et al., 2017; Blauwendraat et al., 2019a; Nalls et al., 2019). Additionally, over 20 genes were regarded as PD-causing genes (Blauwendraat et al., 2019b). Meanwhile, we have replicated PRKN (Guo et al., 2008), PINK1 (Tang et al., 2006; Guo et al., 2008), PARK7 (Tang et al., 2006; Guo et al., 2008), ATP13A2 (Guo et al., 2008), PLA2G6 (Shi et al., 2011), CHCHD2 (Liu et al., 2015), RAB39B (Kang et al., 2016), TMEM230 (Yan et al., 2017), GCH1 (Xu et al., 2017), and other genes (Tian et al., 2012; Yang et al., 2019) in patients with PD in China. Furthermore, we performed multiplex ligation-dependent probe amplification assays and whole-exome sequencing for 1,676 unrelated patients with PD in a mainland Chinese population. Our data indicate that PD patients with an age at onset (AAO) of 40 years may benefit from genetic counseling, especially those from families with a recessive inheritance pattern (Zhao et al., 2020).

Gene expression studies have been a ubiquitous methodology in biomedical research to deeper understand pathological mechanisms driving diseases (Amar et al., 2015). Meanwhile, gene expression in the patient-personalized medicine may predict the fate of disease, treatment effectiveness, as well as may improve diagnostics and epidemiological disease at a population level (Gire et al., 2014; Fischer et al., 2015; Roychowdhury and Chinnaiyan, 2016; Judge et al., 2020). It is to be noted that gene expression in human brain plays an important role in the development of PD. However, whether PD-associated genes (PAGs) present with a specific expression pattern is unclear. To investigate the specific genetic components of PAGs expression, we systematically characterized the expression pattern of PAGs. This study is expected to provide a comprehensive genetic landscape of PAGs for improving the understanding of the pathogenesis of PD.

The PAGs were integrated with PD-causing genes from a systematic review and PD-risk genes (Blauwendraat et al., 2019b) from iPDGC Locus Browser¹ (Grenn et al., 2020), which is collecting three GWAS researches (Iwaki et al., 2019; Nalls et al., 2019; Foo et al., 2020). This gene set was further used to investigate the tissue specificity, excitatory-inhibitory neuron expression profile, as well as the spatiotemporal expression pattern.

To explore the tissue specificity of PAGs in different tissues, we sourced the expression table of 54 tissues across 948 donors from the Genotype-Tissue Expression (GTEx)² database (GTEx Consortium et al., 2017). The average expression level of each gene in each tissue was then calculated. For each gene, tissues with the top 50% expression levels in the 53 tissues were defined as preferential expression tissue. In addition, a Reads Per Kilobase per Million mapped reads (RPKM) value of 1 is generally used for defining the threshold of expressed genes. Fisher’s exact test was used to calculate the P-value for each tissue.

We analyzed the inhibitory-excitatory neuron expression profile of PAGs based on single-nucleus RNA sequencing data which was downloaded from the Allen Brian Map database³. This data set contains 75 transcriptionally distinct cell types from 15,928 nuclei, including 45 GABAergic (inhibitory) neuron types, 24 glutamatergic (excitatory) neuron types, and six non-neuronal types. For each gene in each nucleus, the RPKM value, i.e., the mRNA expression level, was calculated based on the counts per million. Then, the average expression level of each neuronal type for each gene was counted. For the sake of comparison and analysis, we normalized the mean expression values by applying a conversion formula, as follows:

We then used normalized mean expression values to assess the transcriptional profiles in inhibitory and excitatory neurons, and differences in transcription levels were statistically evaluated by the Wilcoxon rank test.

Developmental human brain RNA sequencing data were extracted from the BrainSpan database⁴, which contained 524 tissue-samples of expression data across different developmental stages (from the fetal period to the adult period) and 16 brain regions. To characterize the spatio-temporal expression pattern of PAGs, we performed a weighted gene co-expression network analysis (WGCNA) crossing all tissue samples from BrainSpan to cluster the spatio-temporal expression patterns of PAGs using the standard method with a power of seven. The expression values were normalized by a conversion formula, as follows:

In each co-expression module, the temporal-pattern was divided into 12 time-nodes (six before birth time-nodes and six after birth time-nodes); whereas the spatio-pattern was divided into 16 different brain districts, including the hippocampus, amygdala, striatum, mediodorsal nucleus of thalamus, cerebellar cortex, and 11 cerebral cortical districts.

In total, we sourced 21 known PD-casuing genes (Blauwendraat et al., 2019b) and 89 PD risk genes (Table 1) from iPDGC Locus Browser⁷ (Grenn et al., 2020), which is collecting three GWAS researches (Iwaki et al., 2019; Nalls et al., 2019; Foo et al., 2020). We noted that three PD-causing genes, such as LRRK2, SNCA, and VPS13C, were also regarded as PD-risk genes, the other genes in which have suggestive genetic evidence that confer risk to PD from GWAS. Totally, 107 PAGs were collected in this study.

With the development of biotechnology, numerous candidate genes associated with PD have been reported from thousands of published studies. The ability to comprehensively investigate these PAGs from expression patterns in human, would be helpful for studying the pathogenesis of PD (Blauwendraat et al., 2019b).

Analysis of the spatiotemporal gene expression pattern is a widely accepted method for evaluating cellular function at the transcription level in a specific tissue (Li et al., 2017). In this study, we comprehensively analyzed the expression pattern of PAGs with a novel landscape, and many meaningful results were achieved. First, we demonstrated that PAGs were more likely to have higher expression levels in brain tissues than in other tissues. Second, we found that most PAGs had broad expression profiles in excitatory neurons rather than in inhibitory neurons, suggesting that excitatory neurons are more likely associated with PD (Zhang et al., 2016). Furthermore, an interesting study, focused on analyzing the specific gene expression pattern in different cell types associated with PD risk and described a single-cell atlas in human substantia nigra and cortex (Agarwal et al., 2020). According to this atlas, we investigated whether the PAGs were enriched in any cell types. For cortex, it showed that the PAGs were enriched in excitatory neurons (P = 1.18e-05) and other four cell types, including astrocyte (P = 7.64-e03), microglia (P = 0.048), oligodendrocytes (P = 0.01), oligo-precursor cells (P = 1.78e-03). We noted that the PAGs were also enriched in inhibitory neurons (P = 0.07), which is close to be significant (Supplementary Table 2). These results were consisted with our conclusion in this study. For substantia nigra, we also find that the PAGs enriched in Astrocytes (P = 4.62e-03), Dopaminergic neurons (P = 2.60e-05), GABAergic neurons (P = 1.92e-03), Oligodendrocytes (P = 2.39e-03), Oligo-precursor cells (P = 0.02), but not Endothelial (P = 0.15) in substantia nigra (Supplementary Table 2). Notably, the PAGs were most enriched in dopaminergic neurons with the most significant p-value. Third, we identified two spatiotemporal expression patterns (M1 and M2) associated with PD. More PAGs belonged to M1 than to M2, highlighting that the high expression level in the adult period is associated with neurodegeneration. However, M2 genes including two PD-causing genes (VPS35, GIGYF2), were highly expressed during the embryonic and early to-middle fetal period, suggesting that they may associated with neurodevelopment. For example, GIGYF2 was identified to be associated with macrocephaly (Guo H. et al., 2018). This finding underlines the importance of further research on the mechanisms contributing to the correlation between PD and neurodevelopmental disorders (Hu et al., 2017). Notably, NUS1, which belongs to M2, is a novel PD risk gene, which was found to be associated with the early onset of PD in a Han Chinese population in our previous study (Guo J.F. et al., 2018). However, the innovative results in this study were based on the analysis of bioinformatics; thus, the pathogenic mechanisms will need to be verified with cell or animal experiments. Moreover, 47 PAGs including 10 known PAGs in M1, were steadily highly expressed in the adult period (i.e., M1), which is consisted with the feature of adult onset of PD. It can be inferred that other 37 PAGs (Supplementary Table 1), co-expressed with these 10 known PAGs, acting on the same signaling pathway. Our study adds to the evidence that these 37 PAGs contribute to PD. Moreover, another spatiotemporal expression patterns (i.e., M2) which highly expressed during the embryonic and early to-middle fetal period, suggest that not all the PAGs are highly expressed in adult periods, and the pathogenesis of these genes is worth further investigation, which may be helpful to the molecular typing of PD. These hint that the genes, enriched in the postmitotic cells and progenitors or developmental cell states, might involve into different unknown signal ways, respectively. We highly encourage further study to perform more comprehensive analysis with more data.

PD is an age-dependent neurodegenerative condition. The onset age of PD has been widely reported to be related to the variation in phenotypes and genotypes (Wickremaratchi et al., 2011; Kilarski et al., 2012; Nalls et al., 2015). PD-causing genes, such as ATP13A2 (Ramirez et al., 2006), PLA2G6 (Paisan-Ruiz et al., 2009), FBXO7 (Di Fonzo et al., 2009), PRKN (Lucking et al., 2000; Djarmati et al., 2004), PINK1 (Bonifati et al., 2005; Kumazawa et al., 2008), and PARK7 (Djarmati et al., 2004) have been suggested be strongly associated with AAO. Based on two integrated studies on analyzing the AAO data and genetic data from MDSgene database and Gene4PD database, respectively, we confirmed the association between commonly reported PD-causing genes and AAO. Interestingly, we found that most of the PD-causing genes were belonged to M1, suggesting that the PD-causing genes may be highly expressed after birth. Data from both databases showed very similar trends of each gene, which further examining the relationship between PD-causing genes, AAO and expression pattern.

In conclusion, we prioritized 107 functionally related PAGs and characterized the genetic landscape of the PAGs by analyzing the expression patterns, associations between AAO and expression pattern. This provided novel insight among the AAO, PAGs, and spatiotemporal expression patterns, as well as provided novel clues for understanding the pathology of PD.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

BL, BT, and JL: study design. BL, GZ, KL, ZW, ZF, XW, TL, YZ, YW, QC, YH, and JL: literature search and data collection. BL, GZ, KL, ZW, BT, and JL: figures. BL, GZ, ZF, XW, TL, YZ, YW, QC, LD, and JL: data analysis. BL, GZ, YW, QC, JG, BT, and JL: data interpretation. BL, GZ, KL, ZW, ZF, XW, TL, YZ, YW, QC, YH, LD, JG, BT, and JL: writing of the manuscript draft and all revision stages. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.