Reduced mRNA and protein expression of CHCHD2 in erythrocytes of PD patients
A NanoString multiplex gene expression method was used to screen for potential mRNA biomarkers for PD within erythrocytes. A panel of 21 genes associated with PD or atypical Parkinsonism (see “Materials and methods” section for a complete list) chosen based on previous reports [6, 18, 20, 34, 43], was examined in a cohort consisting of 48 participants separated according to four diagnostic classifications: control, early-stage PD (Early PD), middle-stage PD (Mid PD), or late-stage PD (Late PD) patients according to Unified Parkinson’s Disease Rating Scale (UPDRS) score (Additional file 1: Table S1). mRNA expression frequency of each gene in erythrocytes of healthy control is shown in Additional file 1: Fig. S1, with SNCA, FBXO7, CHCHD2, PSEN1, LRRK2, VPS35, GATA1, MAPT, APOE, and PINK1 among the ten most highly expressed genes in erythrocytes. Of these, only CHCHD2 mRNA consistently demonstrated significantly reduced expression in all PD groups compared to the control group [Fig. 1a, F(3, 4) = 52.0, one-way ANOVA, n = 3; p < 0.01 for Early or Mid PD vs. control, p < 0.001 for Late PD vs. control]. This result was next confirmed by digital droplet PCR (ddPCR), which also detected significantly reduced CHCHD2 mRNA in erythrocytes across all PD groups compared to controls [Fig. 1b, F(3, 8) = 20.80, one-way ANOVA, n = 3; p < 0.01 or 0.001 for each PD group vs. the controls]. There were no differences, however, between PD groups with different disease severity. Having validated the mRNA result independently, we next examined the protein expression of CHCHD2 in erythrocytes of PD patients. As shown in Fig. 1c, the relative CHCHD2 protein levels were decreased significantly from 1.01 ± 0.07 in pooled erythrocytes of controls to 0.43 ± 0.12 in those of PD patients (p < 0.05, Mann–Whitney U test, n = 3).
Reduced expression of CHCHD2 in the post-mortem substantia nigra slices of PD patients
To probe whether expression of CHCHD2 is also reduced in the central nervous system (CNS) of PD patients, its expression was examined in the post-mortem substantia nigra slices obtained at autopsy from the brains of PD patients, along with age-matched controls, as well as in the animal model. Firstly, we examined the expression pattern of CHCHD2, both in the substantia nigra of human and wild type C57BL mice. Co-labeling of CHCHD2 with a pan-neuronal marker (NeuN) and a dopaminergic neuronal marker (Tyrosine hydroxylase, TH) revealed that CHCHD2 is expressed in nearly all neurons (> 90%, Fig. 2a for human, Fig. S2A for mice). CHCHD2 is also expressed in > 90% astrocytes (Fig. 2b for human, Fig. S2B for mice), but only in 10–30% microglia cells (Fig. 2c for human, Fig. S2C for mice). Subcellularly, CHCHD2 is located in mitochondria, as indicated by co-localization with Translocase of Outer Mitochondrial Membrane 20 (TOMM20) (Fig. S2D). Consistent with previous reports [19, 27], loss of dopaminergic neurons, indicated by TH staining, was detected (Fig. 2e–g, 13.2 ± 1.3 in Ctl vs. 4.6 ± 0.6 in PD per field, p < 0.001, multiple t-test). The number of CHCHD2 positive cells was also significantly lower, with 5.46 ± 0.6 per field in PD patients versus 15.2 ± 1.6 neurons in control (Fig. 2e–g, p < 0.001, multiple t-test). Fluorescence intensity of CHCHD2 protein in the surviving neurons of substantia nigra of PD patients also decreased markedly (Fig. 2e–f and h p < 0.001, Mann–Whitney U test).
We also detected the expression of CHCHD2 in post-mortem slices across different brain regions by immunohistochemistry. Consistent with the results obtained by immunofluorescence in Fig. 2e–h, the number of CHCHD2 positive cells (Fig. 3a, p < 0.0001, Mann–Whitney U test) and density of CHCHD2 in the surviving cells (Fig. 3b, p < 0.01, Mann–Whitney U test) of substantia nigra of PD patients were reduced significantly, compared with control. In contrast, the expression of CHCHD2 in frontal cortex (Fig. 3c, p > 0.05, Mann–Whitney U test) and cerebellum (Fig. 3d, p > 0.05, Mann–Whitney U test) were not changed. These results suggested CHCHD2 was preferentially reduced in the substantia nigra of PD patients.
Reduced mRNA and protein expression of CHCHD2 in erythrocytes and brains of A53T α-synuclein mice
Having observed both CNS and peripheral reduction of CHCHD2 expression in PD patients, we next sought to explore the potential underlying mechanisms in a PD mouse model. For this purpose, we used the A53T α-synuclein transgenic mouse model (M83 line), an extensively utilized model where progressive pathology and behavior are driven by expression of mutated α-synuclein [8]. A53T+/+ mice were used at 10 months old, when neurologic defects are readily detectable. As shown in Fig. 4, the relative mRNA and protein expression of CHCHD2 in erythrocytes was significantly lower in A53T+/+ mice (0.27 ± 0.03 for mRNA, p < 0.01, Mann–Whitney U test, n = 6, Fig. 4a; and 0.29 ± 0.04 for protein, p < 0.01, Mann–Whitney U test, n = 3, Fig. 4b) compared to wild type control mice (1.11 ± 0.20 for mRNA; 1.02 ± 0.14 for protein). In the substantia nigra, real time PCR and Western blot results clearly showed that the mRNA (Fig. 4c, p < 0.01, Mann–Whitney U test, n = 6) and protein expression (Fig. 4d, p < 0.05, Mann–Whitney U test, n = 3) of CHCHD2 were significantly decreased.
We also detected a reduced expression of CHCHD2 by immunofluorescence across several brain regions, including substantia nigra (Fig. 5a, b), the rest of midbrain (Fig. 5c, d), frontal cortex (Fig. 5e, f) and cerebellum (Fig. 5g, h) in A53T+/+ mice compared to non-transgenic wild type mice (all p <0.05, Mann–Whitney U test, n = 3). Consistent with previous reports [8, 10, 35], aggregated α-synuclein was increased in these regions (all p < 0.05, Mann–Whitney U test, n = 3). A correlation analysis indicated a negative correlation between the expression of CHCHD2 and the level of aggregated α-synuclein (r = − 0.978, p < 0.001 in substantia nigra, Fig. S3A; r = − 0.943, p < 0.01 in the rest of midbrain, Fig. S3B; r = − 0.952, p < 0.01 in frontal cortex, Fig. S3C; r = − 0.978, p < 0.001 in cerebellum, Fig. S3D).
Reduced mRNA and protein expression of CHCHD2 in a cellular model of PD
To further explore the potential mechanisms leading to decreased CHCHD2 expression in PD patients, wild type and A53T α-synuclein plasmids were transfected in MN9D cells, a line that is derived from dopaminergic cells and previously utilized in multiple in vitro PD models [5, 15, 37]. As expected, transfection of the cells with α-synuclein vector resulted in increased mRNA [Fig. S4A, p < 0.01, F(2, 6) = 20.95, one-way ANOVA, n = 3] and protein [Fig. S4B, F(2, 9) = 14.37, one-way ANOVA, n = 4; p < 0.01 for A53T vs. Vector] expression of α-synuclein. The overexpression of α-synuclein also dramatically reduced the number of CHCHD2 mRNA transcripts, regardless of whether the vector was wild type or A53T α-synuclein [Fig. 6a F(2, 6) = 35.31, one-way ANOVA, n = 3; p < 0.001 for Wt α-syn vs. Vector, p < 0.01 for A53T α-syn vs. Vector]. Consistent with the human data, the protein expression of CHCHD2 was also reduced markedly[Fig. 6b, F(2, 6) = 20.75, one-way ANOVA, n = 3; p < 0.05 for Wt or A53T α-syn vs. Vector].
α-synuclein negatively regulates the expression of CHCHD2, possibly via modulation of P300
The results shown above suggest that reduced CHCHD2 protein is largely attributable to a decreased level of mRNA, likely secondary to overexpression of α-synuclein. To further investigate potential links between α-synuclein overexpression and CHCHD2 level, we asked whether there is a direct interaction of α-synuclein with the promoter of CHCHD2. Initial chromatin immunoprecipitation (ChIP) experiments failed to demonstrate any direct interaction between α-synuclein and CHCHD2 promoter (Additional file 1: Fig. S5A). Therefore, we next investigated whether α-synuclein could negatively regulate the expression of CHCHD2 indirectly. Because it has previously been shown that α-synuclein could negatively regulate protein kinase C expression by reducing the expression and activity of p300 histone acetyltransferase [13], we tested whether there is a direct interaction between p300 and the promoter of CHCHD2. As shown in Fig. 6c, ChIP results distinctly showed that p300 can directly bind the promoter of CHCHD2. Additionally, overexpression of wild type or mutant α-synuclein in MN9D cells reduced the interaction of p300 with the promoter of CHCHD2 (p < 0.001, F (2, 6) = 73.44, one-way ANOVA, n = 3).
We next examined the potential mechanisms by which α-synuclein expression alters the function of p300, by measuring the effect of α-synuclein overexpression on p300. Significant reductions in the mRNA (Fig. 6d, p <0.001, F(2, 6) = 130.2, one-way ANOVA, n = 3) and protein expression (Fig. 6e, p <0.001, F(2, 6) = 41.95, one-way ANOVA, n = 3) of p300 were seen 48h after transfection of α-synuclein vector. Direct protein interaction between p300 and α-synuclein was also demonstrated by reciprocal co-immunoprecipitation experiments (Additional file 1: Fig. S5B). Immunofluorescence results showed that p300 was mainly expressed in the nucleus, while it was less expressed in the cytoplasm of vector transfected Mn9D cells. However, overexpression of wild type or A53T α-synuclein significantly reduced the nuclear expression of p300 (Fig. 6f, p < 0.0001, F(2, 154) = 12.8, one-way ANOVA). Western blot of the protein of isolated subcellular fractions showed an altered distribution, with a higher proportion of p300 in the cytoplasm, compared to a lower amount in the nucleus after α-synuclein (A53T) overexpression compared to vector-transfected cells (Fig. 6g, p < 0.05, Mann–Whitney U test, n = 3). All the above results suggested that α-synuclein may inhibit the expression of CHCHD2, possibly by reducing the expression and the nuclear distribution of p300.
Exploration of erythrocytic CHCHD2 as a PD biomarker in a large cohort
Having demonstrated that expression of CHCHD2 is significantly reduced in the substantia nigra as well as in erythrocytes in PD patients, we turned our attention back to its utility as a more convenient biomarker for PD diagnosis. To achieve this goal, we further examined CHCHD2 mRNA in erythrocytes of individual PD patients in a cohort of 340 subjects including patients with early- (n = 73), middle- (n = 98), and late-stage PD (n = 34) along with controls (n = 135). As shown in Fig. 7a, CHCHD2 mRNA was reduced significantly in all PD groups compared to controls, as detected by ddPCR [F (3, 332) = 26.11, one-way ANOVA; p < 0.001 for each PD group vs. control], consistent with our discovery cohort results.
Because PD is an age-related disorder, not evenly distributed by sex, we next investigated the impact of both sex and age dependence on CHCHD2 mRNA levels isolated from erythrocytes. Two-way ANOVA comparisons of male and female control and PD patient samples revealed no significant difference in CHCHD2 mRNA levels between sexes [Additional file 1: Fig. S6A, p = 0.65, F (1, 326) = 0.2077], while linear regression analysis revealed no correlation between CHCHD2 mRNA levels and age in either group (Additional file 1: Fig. S6B, control: p = 0.49, R = 0.06; PD: p = 0.35, R = − 0.07).
We also evaluated CHCHD2 mRNA as a diagnostic marker in erythrocytes. Analysis of Receiver Operating Characteristic (ROC) curve was performed to evaluate the diagnostic accuracy of CHCHD2 mRNA levels in the total PD cohort as well as PD patients at early-stage. The results were similar for both analyses, and when the specificity was anchored to ≥ 80%, sensitivity for detecting early PD versus healthy controls was 80.82%, yielding a final diagnostic value of 85.38% for PD vs. controls (Additional file 1: Table S2 and Fig. 7b).
Finally, we compared the CHCHD2 mRNA levels in erythrocytes with disease severity, including worsening of motor symptoms and mild cognitive impairment (MCI), and disease duration by comparing CHCHD2 mRNA to the UPDRS part III on-state motor scores, Montreal Cognitive Assessment (MoCA) score, and time following diagnosis in the validation cohorts of PD patients (n = 205 total; cases < 50 years were not excluded due to the lack of age dependence as determined above). Once again, as determined by linear regression analyses, no significant association could be determined between CHCHD2 mRNA levels and the disease severity (p = 0.49, R = − 0.05 for UPDRS motor), disease duration (p = 0.62, R = 0.04) or MOCA scores (p = 0.11, R = 0.11) (Additional file 1: Fig. S6C–E).