Monday, February 26, 2024
BestWooCommerceThemeBuilttoBoostSales-728x90

Large-scale proteomics analysis of five brain regions from Parkinson’s disease patients with a GBA1 mutation – npj Parkinson’s Disease


Study design

The goal of this study was to systematically examine pathophysiological pathways in IPD and PD-GBA brain by performing non-targeted, mass spectrometry-based quantitative discovery proteomics followed by targeted proteomics of selected proteins for validation (Fig. 1). This unique tissue sample set included 21 IPD patients without a GBA1 mutation, 21 PD-GBA patients and 21 controls from five different brain regions (OCC, MTG, CG, STR and SN). The CG region contained only 52 samples (15 controls, 18 IPD and 19 PD-GBA), the STR region contained 61 samples (19 controls, 21 IPD and 21 PD-GBA) and the SN region contained 60 samples (20 controls, 20 IPD and 20 PD-GBA), for a total of 299 samples, a comparatively large number for human brain samples. Of these brain areas, 2 are associated with the nigrostriatal pathway (the SN and STR), while the other 3 regions all display some pathology in PD16,17 which may be responsible for non-motor symptoms (NMSs)18.

Fig. 1: Study design.

Non-targeted, mass spectrometry based quantitative discovery proteomics was performed on five human brain regions, namely the OCC, MTG, CG, STR and SN of 21 control, 21 IPD and 21 PD-GBA samples. Two of these regions are part of the nigrostriatal pathway and three are not. Subsequent to data collection, analysis was performed as indicated (green boxes). 20 proteins were chosen for targeted proteomics (yellow) in three brain regions (the CG, STR and SN), using the various criteria indicated in the text. Proteins in red were altered in at least one brain region in the non-targeted analysis; proteins in blue were selected as they directly impinge on the pathways discovered in the non-targeted study. The proteins are sorted according to the predominant pathway in which they are involved (black).

The groups were age- and sex-matched with no significant differences in the cause of death (control patients died from unrelated causes, whereas IPD and PD-GBA patients died from one or other issue related to PD). The PD patients were classified as IPD and most displayed characteristic Lewy body (LB) pathology. For PD-GBA brains13, GBA1 mutations included, among others, patients with the N370S and the L444P mutations which are found at high levels in association with PD3. The same samples were used for a recent non-targeted lipidomics study, in which elevation of levels of gangliosides was detected in four brain regions13, although no significant changes were observed in levels of GlcCer.

Subsequent to the non-targeted, mass spectrometry-based quantitative discovery proteomics, and based on the data described below, 20 proteins were chosen for validation by targeted proteomics using parallel reaction monitoring (PRM) (Fig. 1), in which samples were spiked with heavy labeled synthetic peptides (three per protein) to facilitate their identification. The proteins were chosen based on either changes in their levels in the non-targeted analysis (17 of the proteins; see Supplementary Fig. 1 for a comparison of the two methods) or on their possible involvement in one or other of the putative pathological pathways (3 of the proteins); the latter included transcription factor EB (TFEB) [a master regulator of transcription of lysosomal proteins19], matrix metalloproteinase 14 (MMP14) [a protein that is directly regulated by one of the proteins detected in the non-targeted analysis, namely 1,2-dihydroxy-3-keto-5-methylthiopentene dioxygenase (ADI120)] and transmembrane glycoprotein NMB (GPNMB) [a marker of the severity of symptoms in nGD21]. Targeted proteomics was conducted on three brain regions, the SN, STR and CG.

Data quality and sample validation

Prior to detailed data analysis, quality control was performed on all of the samples. First, principal component analysis (PCA) on all or most proteins detected in each region (OCC, 4411 proteins; MTG, 5002 proteins; STR, 4714 proteins; CG, 5082 proteins; SN, 4534 proteins), revealed that samples did not cluster according to age, gender or GBA1 mutation (in the case of PD-GBA samples) (Fig. 2a and Supplementary Figs 2 and 3), demonstrating that none of these factors influence the data. Each sample was also analyzed for the number of missing values (Supplementary Fig. 4), and in rare cases when a sample displayed a large number of missing values, it was removed from the analysis (where a missing value could either be due to a technical error or was close to or below the limit of detection), such as in the case of sample PG11 from the MTG and samples C13 and C15 from the CG. No clustering between the different samples was detected (Fig. 2b, Supplementary Fig. 5) which is not surprising since 42 of the 63 samples come from PD patients (with and without GBA1 mutations); the fact that control samples did not cluster can likely be explained by the high variability in human brain samples between individual patients and the relatively small number of changes in protein levels (see below). Next, the number of overlapping proteins between each brain region was analyzed (Fig. 2c), with most of the identified proteins detected in all regions, permitting their evaluation and comparison in different brain regions.

Fig. 2: Evaluation of the quality of the non-targeted, mass spectrometry-based quantitative discovery proteomics.
figure 2

a PCA plots showing sample clustering according to gender, age and GBA1 mutation for the SN; similar results were obtained for the other four brain regions (Supplementary Fig. 2). Controls, circles; IPD, triangles; PD-GBA, squares. The y and x axes represent sample variance. The color code is indicated for each panel (age, gender and GBA1 mutation). No clustering into groups was observed for any of these factors. b Sample clustering for the SN; similar results were obtained for the other regions. No clear clustering into sample groups (i.e. control; C, IPD; PD or PD-GBA; PG) was observed. c Venn diagram showing the number of detected proteins in each brain region and the overlap between different regions. Thus, 3938 proteins were detected in total in the five brain regions and, for instance, 111 proteins were found exclusively in the SN and 191 in the MTG. d Levels of β-sheet oligomers of human α-synuclein measured by ELISA using monoclonal antibody 5G4 (which specifically identifies aggregated α-synuclein)42,49 from all available samples from the OCC, MTG and CG and n = 12 for SN in each group. e Levels of MAP2 in five brain regions from the non-targeted proteomics. Boxes represent lower quartile, median and upper quartile (black). The whiskers represent the minimum and maximum values, up to 1.5-times the interquartile range from the bottom or the top of the box to the furthest data point within that distance, thus excluding outliers. The mean is in red. For d, the y axis is in μg/μg protein. For e, the y axis is in arbitrary units. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001, calculated using either ANOVA, followed by post-hoc pairwise comparisons, for the ELISA analysis or empirical Bayes moderation for the proteomics analysis.

An important factor in PD is the spread of α-synuclein, which is used to determine disease severity according to Braak staging16. Levels of β-sheet oligomers of α-synuclein, an aggregated form of α-synuclein, were analyzed by ELISA in all available control, IPD and PD-GBA samples for three brain regions and in 12 samples for the SN. According to Braak staging, the deposition of aggregated α-synuclein is observed in the SN in stage 3, in the MTG in stage 4, and in the OCC and CG in stage 5/616. In the PD samples used for the proteomics analysis, β-sheet oligomers of α-synuclein accumulated in the SN and in the MTG, but not in the CG and OCC (Fig. 2d), suggesting that the samples were mainly taken from stage 4 patients. Levels of oligomers were higher in the SN of PD-GBA patients than in IPD and lower in the MTG than in IPD (Fig. 2d).

We also analyzed levels of MAP2, a protein found in the dendrites of neurons22. MAP2 levels decreased significantly in the SN, where neuronal loss is extensive in PD2, in both the IPD and PD-GBA groups (Fig. 2e), and to a somewhat greater extent in the STR in PD-GBA compared to IPD. No reduction in MAP2 levels were observed in the other 3 brain regions indicating little or no neuronal loss in these regions.

Since neuronal loss occurs concomitantly with the loss of components of the dopaminergic pathway in PD, we examined levels of 4 proteins involved in this pathway, tyrosine hydroxylase (TH), dopa decarboxylase (DDC), solute carrier family 18 member A2 (SLC18A2) and solute carrier family 6 member 3 (SLC6A3) (Fig. 3a). A significant reduction in levels of all 4 proteins was observed in both the SN and the STR (the nigrostriatal regions which are most affected in PD2); note that these 4 proteins were not detected in the other 3 brain regions. The extent of reduction of these proteins was larger in the STR of PD-GBA samples compared to IPD (Fig. 3a), consistent with greater axonal damage in PD-GBA compared to IPD, and also consistent with suggestions that PD-GBA is a more severe form of IPD23,24. Levels of one of the proteins, namely TH, was analyzed by targeted proteomics and compared to the non-targeted data, which gave a good correlation in the STR and in the SN (Supplementary Fig. 6).

Fig. 3: Analysis of proteins associated with dopaminergic pathways shows a more significant loss of these components in PD-GBA compared to IPD in the STR.
figure 3

a Boxplots of levels of four key proteins related to the dopaminergic pathway in the SN (upper panel) and STR (lower panel) from the non-targeted proteomics. b GPNMB levels detected by targeted proteomics in the STR and SN. Boxes represent lower quartile, median and upper quartile (black). The whiskers represent the minimum and maximum values, up to 1.5-times the interquartile range from the bottom or the top of the box to the furthest data point within that distance, thus excluding outliers. The mean is in red. The y axis is in arbitrary units. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001, calculated using empirical Bayes moderation.

Next, levels of GPNMB, which can be used as a marker for determining the severity of nGD disease21, were increased in the SN in PD-GBA (GPNMB was analyzed by targeted proteomics since this protein was not detected in the non-targeted analysis) (Fig. 3b). In conclusion, proteomics analysis is consistent with the classification of the patient samples in the clinical data, and consistent with the notion that PD-GBA is a more severe form of PD than IPD.

Brain region-specific similarities and changes in protein levels between IPD and PD-GBA

PD is characterized by loss of neurons in the SN. The SN projects to the STR and a significant loss of these projected axons is observed in the STR, which causes a lack of dopamine in this region, leading to pathology. However, other brain regions are affected in PD pathogenesis, such as the three other tissues chosen for this study, the OCC, MTG and CG. These brain regions exhibit altered functional connectivity17, display Lewy body pathology at different disease stages16 and are responsible for some of the NMSs observed in PD25. Analysis of changes in protein levels in all five regions indicate that the SN and CG display the largest number of changes in protein levels (Fig. 4). As might be expected based on PD pathology, the highest number of changes in protein levels is in the SN, with more proteins downregulated in IPD versus control (280 downregulated and 140 upregulated) but a similar number is found for up- or downregulated proteins in PD-GBA versus control (249 downregulated and 262 upregulated). Perhaps more unexpectedly, the next most affected region was the CG, in which differentially-expressed proteins are more upregulated in both PD groups (IPD versus control, 156 downregulated proteins and 213 upregulated; PD-GBA versus control, 154 downregulated and 351 upregulated). The relatively large extent of changes in the CG is consistent with the overactivation of this brain region in PD, as shown using functional magnetic resonance imaging26. The STR is not one of the most altered regions, though it is one of the most affected regions in PD. In all regions except for the STR, more proteins were upregulated then downregulated in PD-GBA versus IPD.

Fig. 4: Differentially-expressed proteins in all five brain regions.
figure 4

Volcano plots using non-targeted proteomics data. The brain region is indicated (created using BioRender.com). Each plot shows a comparison between the three groups (i.e. IPD versus control, PD-GBA versus control or PD-GBA versus IPD). The dashed vertical lines (red) indicate fold-change ≥ 1.5 and the horizontal red lines indicate p ≤ 0.05, calculated using empirical Bayes moderation. Individual proteins are shown, with red indicating proteins whose levels were elevated and blue proteins whose levels were reduced, with the number of such proteins also indicated.

Levels of only two proteins were altered in all five brain regions in one or other of the groups (Fig. 5a), namely GCase and ADI1, which were differentially expressed between IPD and PD-GBA in all brain regions (with statistical significance in four regions) (Fig. 5b, c). ADI1 has not been previously reported to play a role in PD pathogenesis, but its levels were reduced in IPD, although unexpectedly, not in PD-GBA. This data could not be validated by targeted proteomics since only one out of three peptides was detected, and this peptide was not one of the 8 detected peptides identified in the non-targeted analysis.

Fig. 5: Two proteins are differentially expressed in all five brain regions.
figure 5

a Venn diagram of differentially expressed proteins between IPD and PD-GBA. Proteins which were differentially expressed between IPD and PD-GBA and between one of the PD groups versus the control in all five brain regions are indicated. Number of peptides ≥ 2, p ≤ 0.05. b Boxplots displaying non-targeted proteomics results of ADI1 levels in all five brain regions. c Boxplots (upper panel) displaying nontargeted proteomics of GCase levels in all brain regions measured in the analysis. For both b and c, the box represents lower quartile, median and upper quartile (black). The whiskers represent the minimum and maximum values, up to 1.5 times the interquartile range from the bottom or the top of the box to the furthest data point within that distance, thus excluding outliers. The mean is shown in red. For b and c (upper panel), the y axis is in arbitrary units. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001, calculated using empirical Bayes moderation. GCase activity (lower panel) (n = 12 for control, IPD and PD-GBA, chosen blindly); *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001, calculated using ANOVA, followed by post-hoc pairwise comparisons. For c (lower panel), the y axis is in pmol/mg/min. Error bars indicate +/- standard deviation. d, The non-targeted proteomics data from c is plotted according to GBA1 mutation classified as severe [G232E, R131C, L444P, R463C, RecA456P, RecNciI (red)], mild [N370S (orange)] and PD risk factors [E326K, T369M (blue)]27,50,51. Each point indicates GCase levels in an individual patient (arbitrary units).

In contrast, GCase levels decreased in PD-GBA in all brain regions irrespective of the GBA1 mutation (Fig. 5c). This is somewhat unexpected since most of the mutations in the PD-GBA group are point mutations which are not necessarily expected to affect protein expression or stability, particularly mutations which are risk factors for PD but are not known to cause GD, such as E326K or T369M27. The non-targeted analysis was validated by the targeted analysis (Supplementary Fig. 7). The reduction in GCase levels was also confirmed by GCase enzymatic assays inasmuch as PD-GBA samples had lower levels of GCase activity than either controls or IPD samples (n = 12 for each group; Fig. 5c). There was no observed association between protein levels detected by non-targeted proteomics and the GBA1 mutation (Fig. 5d), and the same mutation sometimes gave different levels of GCase expression (see for instance GCase levels for the E326K mutation in the CG). Finally, a recent suggestion implied that GCase activity decreases with age28, but we did not detect an age-dependent decrease in GCase levels from control patients by non-targeted proteomics (Supplementary Fig. 8).

Changes in cellular and metabolic pathways in IPD and PD-GBA

We next analyzed changes in cellular and metabolic pathways using ingenuity pathway analysis (IPA) (https://www.qiagenbioinformatics.com/products/ingenuitypathway-analysis). We first examined changes in each brain region independently, followed by a compared comparison of the altered pathways in all 5 regions, and expressed data as either fold-change (Fig. 6a) or p value, calculated using the right-tailed Fisher’s Exact Test (Fig. 6b). These two factors measure different parameters with fold-change (activation z-score) taking into account levels of activators and inhibitors of classified pathways and accordingly predict whether a pathway is activated or inactivated, whereas p value indicates how many differentially expressed proteins were altered; changes in both are highly suggestive that the pathway in question has been moderated under the specific experimental conditions.

Fig. 6: Mitochondrial pathways and specific respiratory electron transport complexes are reduced to a larger extent in PD-GBA than in IPD.
figure 6

Pathway analysis of the results of the five brain regions compared to each other, according to fold-change (a) and to p value (b). This analysis was conducted using IPA software, which also considers known inhibitors or activators of the pathway and scores the pathway according to the level of change in these factors, resulting in a predicted activation score of the pathway; a lower score means the pathway is inhibited and a higher score means the pathway is activated. IPA calculates a p value using the right-tailed Fisher’s Exact Test. For a, blue indicates reduced activation of the pathway and orange indicates elevated activation of the pathway. For b, a darker shade of purple indicates higher statistical significance. Number of peptides ≥ 2, p ≤ 0.05. c, Analysis by PERMANOVA of mitochondrial complexes. This statistical test takes into account changes in all the subunits of each complex. p values are colored according to the scale shown, where values below 0.05 are in red and values above 0.05 are in blue. Grey indicates cases in which the number of subunits is greater than the number of tested samples (due to limitations of the PERMANOVA test). The SN was removed from the analysis as it included changes in all complexes in both groups, which is probably due to neuronal death in this region.

Only two pathways were altered according to both parameters, namely oxidative phosphorylation (Fig. 6a, b), and sirtuin signaling, although changes in the latter were relatively small and not consistent. A number of pathways only changed in one of the analyses, including mitochondrial dysfunction and ceramide signaling (p value) along with tRNA charging (activation z-score). In the case of oxidative phosphorylation, the activation z-score, which takes into account expression levels of known inhibitors and activators of the pathway, suggests that the pathway was inactivated to a larger extent in IPD compared to controls and to an even larger extent in PD-GBA, with the exception of the SN where the pathway was affected more in IPD than in PD-GBA. Likewise, changes in oxidative phosphorylation were larger according to p value for PD-GBA than for IPD in all regions. Further analysis of the pathway of oxidative phosphorylation by PERMANOVA revealed that specific complexes were affected to a larger extent in PD-GBA than in IPD (specifically complexes III and IV) (Fig. 6c). Finally, targeted proteomics validated the non-targeted analysis for cytochrome C oxidase subunit 6 C (COX6C), a subunit of mitochondrial complex IV (Supplementary Fig. 9).

One of the pathways, which was significantly affected was ceramide synthesis, at least in the CG (Table 1, Supplementary Table 1). Two subunits of serine palmitoyl transferase were down-regulated in IPD but either found at control levels (SPTLC1) or elevated in PD-GBA versus IPD (SPTLC1 and SPTLC2) (Table 1 and Fig. 7a); changes in SPTLC2 were validated by targeted proteomics (Supplementary Fig. 10A). Levels of three members of the ceramide synthase (CerS) family were elevated in either or both IPD or PD-GBA (CerS6, Table 1; CerS4 and CerS2, Table 1 and Fig. 7b, c), with CerS4 validated by targeted proteomics (Supplementary Fig. 10B) (CerS2 was not analyzed by targeted proteomics). Further validation was obtained by comparing C20- (generated by CerS429) and C24-ceramide (generated by CerS230) levels in the same samples (data taken from ref. 13) which, at least in the case of CerS4, revealed a significant elevation of C20-ceramide levels in PD-GBA (Fig. 7b). Finally, CerS4 and CerS2 activity measurements were consistent with the proteomics and lipidomics data (Fig. 7b, c) and there was a reasonable correlation between levels of ceramides in individual patients and CerS activity (Fig. 7d), Together, this data indicates that the ceramide synthesis pathway is altered in the CG of PD-GBA samples.

Table 1 Elevation of ceramide synthesis-related proteins in the CG of PD-GBA samples.
Fig. 7: Elevation of the ceramide synthesis pathway in the CG of PD-GBA.
figure 7

a Boxplots indicating levels of SPTLC2 by non-targeted proteomics. b Boxplot of non-targeted proteomics for CerS4 levels, along with levels of C20-ceramide (data from ref. 13); purple indicates which samples were used for the CerS activity assay in the plot below (n = 6 for each group). c Boxplot of non-targeted proteomics of CerS2 levels in the CG with levels of C20:0-ceramide (taken from ref. 13) in the same samples, with purple indicating which samples were used for the CerS activity assay (n = 9 for each group). d Pearson correlation matrices of CerS4 activity and CerS2 activity compared with lipidomics data for PD-GBA and for IPD. Correlation coefficients are indicated. Axes represent CerS activity data (pmol/mg protein/min) versus the lipidomics data (pmol/mg protein). For all boxplots, the box represents lower quartile, median and upper quartile (black). The whiskers represent the minimum and maximum values, up to 1.5 times the interquartile range from the bottom or the top of the box to the furthest data point within that distance, thus excluding outliers. The mean is shown in red. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, calculated using either empirical Bayes moderation for the proteomics analysis or ANOVA, followed by post-hoc pairwise comparisons, for the activity assays.

Cellular changes are not restricted to any one organelle and do not differ between IPD and PD-GBA

We next determined whether changes in levels of proteins associated with different subcellular locations differ between IPD and PD-GBA. Indeed, a recent study suggested an excessive burden of lysosomal storage disorder gene variants in PD6. Of the 54 proteins identified in this study6, levels of 23 were altered in one or other brain region in either IPD or PD-GBA by non-targeted proteomics (Supplementary Table 2), with most decreased, although many were elevated in PD-GBA in the CG (Fig. 8a). Levels of acid sphingomyelinase (ASM), which is also a risk factor for PD31, decreased in the IPD group in the OCC and in the PD-GBA group in the MTG, but was elevated in the CG in the PD-GBA group (confirmed by targeted proteomics; Supplementary Fig. 11). Similarly, levels of two other proteins, cathepsin A (PPGB) and cathepsin D (CATD) were reduced in the OCC, MTG and SN, whereas PPGB was elevated in the CG in both PD groups. While this data is not definitive about whether there are global changes in lysosomal proteins in IPD and PD-GBA, they are nevertheless consistent with the notion that the lysosome plays a critical role in PD pathology. However, upon differentiating between proteins, which are located in different intracellular compartments, the mitochondria, endoplasmic reticulum and plasma membrane all displayed a similar number of differentially expressed proteins as the lysosome (Table 2 and Supplementary Table 3).

Fig. 8: Alterations in levels of LSD-related proteins in IPD and PD-GBA.
figure 8

a Heatmap of LSD-related proteins which differ in at least one of the brain regions in any of the groups in the non-targeted proteomics. All five brain regions are displayed. Ratios of protein levels are shown for IPD versus control, PD-GBA versus control and PD-GBA versus IPD. Blue indicates a ratio of < 1 and red a ratio of > 1; grey indicates not detected. Reduced levels of these proteins are observed in most brain regions in both PD groups, with the exception of the CG, where they appear to be more elevated, especially in the PD-GBA group. b Boxplots displaying non-targeted proteomics data of ASM, PPGB and CATD levels in all five brain regions. The box represents lower quartile, median and upper quartile (black). The whiskers represent the minimum and maximum values, up to 1.5 times the interquartile range from the bottom or the top of the box to the furthest data point within that distance, thus excluding outliers. The mean is shown in red. The y axis is in arbitrary units. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001, calculated using empirical Bayes moderation.

Table 2 Several cell compartments are altered in both PD groups, but to a larger extent in PD-GBA in most brain regions.



Source link

Related Articles

Leave a Reply

[td_block_social_counter facebook="beingmedicos1" twitter="being_medicos" youtube="beingmedicosgroup" style="style8 td-social-boxed td-social-font-icons" tdc_css="eyJhbGwiOnsibWFyZ2luLWJvdHRvbSI6IjM4IiwiZGlzcGxheSI6IiJ9LCJwb3J0cmFpdCI6eyJtYXJnaW4tYm90dG9tIjoiMzAiLCJkaXNwbGF5IjoiIn0sInBvcnRyYWl0X21heF93aWR0aCI6MTAxOCwicG9ydHJhaXRfbWluX3dpZHRoIjo3Njh9" custom_title="Stay Connected" block_template_id="td_block_template_8" f_header_font_family="712" f_header_font_transform="uppercase" f_header_font_weight="500" f_header_font_size="17" border_color="#dd3333"]
- Advertisement -spot_img

Latest Articles