1.1 Human Samples

Fresh‐frozen substantia nigra and visual cortex samples were provided by the Queen Square Brain Bank (QSBB) at University College London. The brain donation program and protocols have received ethical approval for donation and research by the NRES Committee London – Central, and tissue is stored for research under a license issued by the Human Tissue Authority (No. 12198). Samples from 18 MSA patients (clinical diagnosis before postmortem evaluation: 8 MSA‐P, 10 MSA‐C; postmortem pathological evaluation: 3 MSA‐SND, 10 MSA‐OPCA, 5 MSA‐mixed), 12 PD patients, and 10 nonneurological controls (C) were included in this study. Demographical and clinical information was kindly provided by the QSBB (Table 1; the case list including all the clinical and pathological details is available in Supplementary Table 1). No personal information could be extracted from these data. The utilization of postmortem human samples was approved by the corresponding biobank ethics committees and by the Ethics Committee of the Medical University of Innsbruck (AN2016‐0012 358/4.2 359/4.1).

TABLE 1

Demographic and clinical data of patients with MSA, patients with PD, and C

					Post hoc analysis
	C (n = 10)	PD (n = 12)	MSA (n = 18)	P values, all	MSA vs C	MSA vs PD	PD vs C
Age (years), mean ± SD a	83.1 ± 4.63	77.42 ± 5.16	65.11 ± 6.18	< 0.0001	<0.0001	0.002	0.3128
Female, % (n) b	60 (6)	41.7 (5)	61.1 (11)	0.5394
Disease duration (years), mean ± SD c	NA	17 ± 7.39	7.39 ± 2.62	< 0.0001
HIF2A mRNA levels in substantia nigra, a.u., mean ± SD d	0.021 ± 0.011	0.018 ± 0.005	0.022 ± 0.01	0.3529
HIF2α protein levels in substantia nigra, a.u, mean ± SDd	0.011 ± 0.004	0.014 ± 0.007	0.023 ± 0.008	0.0002	0.0003	0.0084	>0.6688
HIF2A mRNA levels in visual cortex, a.u, mean ± SDd	0.010 ± 0.008	0.009 ± 0.005	0.014 ± 0.01	0.2197
HIF2α protein levels in visual cortex, a.u, mean ± SDd	0.012 ± 0.012	0.017 ± 0.012	0.021 ± 0.008	0.09	0.0892	0.9413	0.7207

NA, not applicable; SD, standard deviation; a.u., arbitrary units.

A more detailed table including all cases and the clinical and pathological information can be found in Supplementary table 1.

^a Kruskal‐Wallis with post hoc Dunn's test.

^b Chi‐square test.

^c Unpaired t test, 2‐tailed.

^d One‐way ANOVA with post hoc Bonferroni's test.

1.2 Molecular Analyses

Total RNA and proteins were obtained from frozen samples using Trizol Reagent (Thermofisher) according to the manufacturer's instructions. No differences were observed in RNA quality between MSA, PD, and C in the different brain regions, with A260/280 ratios within the expected range (substantia nigra: C, 1.93 ± 0.06; PD, 1.91 ± 0.04; MSA, 1.92 ± 0.05; visual cortex: C, 1.98 ± 0.01; PD, 1.97 ± 0.01; MSA, 1.96 ± 0.03). qRT‐PCR was performed using standard procedures. GAPDH, 18S, ACTB mRNA were used as housekeeping genes. All samples were run in duplicate, and results were analyzed using the 2−ΔCt method ⁹ and presented as relative gene expression normalized to the average cycle threshold for the 3 housekeeping genes. Protein quantification and Western blot were performed using standard procedures. Primary antibodies included anti‐HIF2α (Cell Signaling; 7096), anti‐HIF1α (BD; 610959), anti‐GAPDH (Sigma; G9545), anti‐α‐Tubulin (Abcam; ab7750), and anti‐ß‐tubulin (Sigma; T4026) as loading controls. Signal detection was performed using horseradish peroxidase–conjugated antirabbit (Cell Signaling; 7074) or antimouse (GE Healthcare; NA931) antibodies. Images were acquired using the Fusion FX system for Western blot and gel imaging. Relative protein levels were measured by densitometry using FUSION CAPT V16.09b software (Vilber Lourmat) and normalized to the average densiometric value for the 3 housekeeping proteins. ¹⁰ A reference sample was loaded in all gels for gel‐to‐gel normalization. All gels were run, transferred, incubated, and developed in parallel.

1.3 Statistical Analyses

Most statistical analyses and all graphs were performed in GraphPad Prism version 8.0 (GraphPad Inc.). Data are expressed as mean ± SD; P ≤ 0.05 was considered statistically significant. All individual measurements constituted biological replicates. Samples were evaluated for normal distribution using D'Agostino and Pearson's omnibus normality test. Comparisons were done with 1‐way analysis of variance (ANOVA) with Bonferroni's test for HIF2α protein, HIF2A gene expression, and postmortem interval (PMI). A 2‐tailed t test was used to analyze disease duration differences between PD and all MSA cases and ANOVA with Bonferroni's test for PD and MSA variants. The Kruskal‐Wallis test with Dunn's test and the chi‐square test were used to evaluate age and sex differences, respectively, between groups. Univariate analysis of variance for HIF2α with covariate age and fix factors diagnosis and sex was performed with SPSS version 24 (IBM) to evaluate the possible confounding effect of age. Correlations were studied using linear regression analysis.

2.1 MSA Brains Show Signs of Chronic Hypoxia

The hypoxia‐inducible factors (HIFs) are key in the cellular adaptation to low oxygen levels. ¹¹ These transcription factors are DNA‐binding heterodimers consisting of alpha and beta subunits. ¹¹ Genes coding for HIFα subunits are constitutively transcribed and translated in normal oxygen conditions, but the resultant proteins are degraded by the proteasome through their consecutive hydroxylation by prolyl‐hydroxylases (PHDs) and ubiquitination by von Hippel‐Lindau ligase in an oxygen‐dependent manner ¹¹ (Fig. 1A). However, when oxygen availability is scarce, PHDs reduce their activity, and HIFα proteins are stabilized and transported to the nucleus. There HIFα proteins heterodimerize with HIFβ subunits and activate a plethora of different transcriptional programs that are cell specific ¹² (Fig. 1A). HIF1α and HIF2α constitute the 2 main factors driving the cellular adaptation to hypoxia, in which HIF1α is primarily active during the acute phase of hypoxic adaptation, and HIF2α dominates during later, more chronic phases of hypoxia11, 13, 14, 15 (Fig. 1A).

FIG. 1.

Signs of chronic hypoxia in MSA. (a) Schematic representation of the regulation of HIFα in normoxia and acute or chronic hypoxia. Genes coding for HIFα subunits are constitutively transcribed and translated. However, under normal oxygen conditions (normoxia), the resultant proteins are rapidly hydroxylated, ubiquitinated, and degraded via proteasome in an oxygen‐dependent manner. Under acute hypoxic conditions, the decrease in oxygen levels impairs the hydroxylation, which results in HIF1α and HIF2α protein stabilization. Under chronic hypoxic conditions, HIF1α protein is degraded, and HIF2α drives the cellular adaptation. (b) Representative Western blot images showing HIF2α and HIF1α protein levels in extracts from human substantia nigra. Samples from control (c) and PD and MSA subjects were utilized. GAPDH, α‐tubulin, and β‐tubulin were used as loading controls. A protein extract from HEK cells exposed to DMOG, an agent that mimics hypoxic conditions by inhibiting PHDs, was used as a positive control (right lane). (c) Violin plot illustrating HIF2α protein levels in control and PD and MSA cases based on the quantification of Western blots. Controls, white circles; PD, black squares; MSA, green triangles; black triangles, PD cases that were clinically misdiagnosed as MSA. ANOVA, analysis of variance. ***P < 0.001, **P < 0.01 (Bonferroni's test). (d) Correlation analysis of HIF2α levels and disease duration in MSA and PD cases. MSA only: r ² = 0.1919, P = 0.069; PD only: r ² = 0.1211, P = 0.2677; MSA + PD: r ² = 0.1385, P = 0.0428. PD, black squares; MSA, green triangles; black triangles, PD cases that were clinically misdiagnosed as MSA. [Color figure can be viewed at wileyonlinelibrary.com]

Therefore, to evaluate the presence of hypoxia in human brains, we analyzed the protein levels of HIF1α and HIF2α (Fig. 1B). In agreement with the main hypothesis of the present study, Western blot analyses indicated the presence of a hypoxic environment in the substantia nigra of MSA patients compared with PD patients and controls, characterized by the significant accumulation of the hypoxia marker HIF2α (MSA vs C, P = 0.0003; MSA vs PD, P = 0.0084; Fig. 1B,C and Table 1). The differences between groups remained significant after performing univariate analysis including age as a covariate (MSA vs C, P = 0.021; MSA vs PD, P = 0.05) and confirmed that changes in HIF2α levels were determined by diagnosis (P = 0.022) but not age (P = 0.496) or sex (P = 0.502), discarding a possible confounding effect. No differences in HIF2α protein levels were observed between MSA variants defined according to clinical symptoms (MSA‐P vs MSA‐C, P > 0.9999; Supplementary Fig. 1A and Supplementary Table 2) or postmortem pathological subtype (MSA‐SND vs MSA‐OPCA, P > 0.9999; MSA‐SND vs MSA‐mixed, P > 0.9999; MSA‐OPCA vs MSA‐mixed, P > 0.9999; Supplementary Fig. 1B and Supplementary Table 3). Interestingly, 2 of the PD patients with higher levels of HIF2α were originally misdiagnosed as MSA‐P, suggesting a possible association between the presence of hypoxia and the clinical presentation (Fig. 1C and Supplementary Fig. 1A,B). Furthermore, gene expression analysis showed no differences in HIF2A (also known as EPAS1) mRNA levels between groups (Table 1, Supplementary Fig. 1C–E, and Supplementary Tables 2–3). This finding indicated that the increase of HIF2α protein was not because of upregulation of the HIF2A gene (HIF2α vs HIF2A: r ² = 0.01613, P = 0.4473), but of protein stabilization under hypoxic conditions in MSA. In addition, we evaluated molecular signs of hypoxia in the visual cortex, a brain region less affected in α‐synucleinopathies, in which we observed a numerical increase of HIF2α protein in MSA cases compared with controls, although not significantly (MSA vs C: P = 0.0892); see Table 1 and Supplementary Figure 2A,B. As in the substantia nigra, higher HIF2α protein was not associated with upregulation of the HIF2A gene (HIF2α vs HIF2A: r ² = 0.05844, P = 0.1329; Supplementary Fig. 2C). We did not detect HIF1α protein in the tissue samples of patients or controls (Fig. 1B and Supplementary Fig. 2A).

That HIF2α levels were significantly higher in MSA cases compared with controls in areas more affected by the pathology may suggest a possible association between hypoxia and disease progression. Therefore, we hypothesized that hypoxia may also have an impact on disease duration in α‐synucleinopathies. However, correlation analysis did not show a significant relationship between disease duration and hypoxia level in MSA (r ² = 0.1919, P = 0.069; Fig. 1D) or in PD (r ² = 0.1211, P = 0.2677; Fig. 1D). Nevertheless, the small number of cases per group and the presence of a similar disease duration in most MSA cases may reduce the statistical power of the correlation analysis. A significant negative relationship between hypoxia and disease duration was only observed when cases from both synucleinopathies were pooled together (r ² = 0.1385, P = 0.0428; Fig. 1D), suggesting that the presence of hypoxia in MSA may at least partly contribute to the faster progression compared with PD. However, the latter may artificially result from the lack of significant change of HIF2α in PD cases, together with a much longer disease duration in PD than MSA.

Finally, correlation analyses of PMI and HIF2α protein levels in the substantia nigra or visual cortex demonstrated that the presence of hypoxia in MSA brains was independent of the time elapsed between death and the freezing of the sampled tissue (substantia nigra: r ² = 0.03268; P = 0.2643; visual cortex: r ² = 0.001221; P = 0.8305), therefore discarding PMI as a possible confounding factor.