A novel super-resolution microscopy platform for cutaneous alpha-synuclein detection in Parkinson's disease

Abstract

Alpha-synuclein (aSyn) aggregates in the central nervous system are the main pathological hallmark of Parkinson's disease (PD). ASyn aggregates have also been detected in many peripheral tissues, including the skin, thus providing a novel and accessible target tissue for the detection of PD pathology. Still, a well-established validated quantitative biomarker for early diagnosis of PD that also allows for tracking of disease progression remains lacking. The main goal of this research was to characterize aSyn aggregates in skin biopsies as a comparative and quantitative measure for PD pathology. Using direct stochastic optical reconstruction microscopy (dSTORM) and computational tools, we imaged total and phosphorylated-aSyn at the single molecule level in sweat glands and nerve bundles of skin biopsies from healthy controls (HCs) and PD patients. We developed a user-friendly analysis platform that offers a comprehensive toolkit for researchers that combines analysis algorithms and applies a series of cluster analysis algorithms (i.e., DBSCAN and FOCAL) onto dSTORM images. Using this platform, we found a significant decrease in the ratio of the numbers of neuronal marker molecules to phosphorylated-aSyn molecules, suggesting the existence of damaged nerve cells in fibers highly enriched with phosphorylated-aSyn molecules. Furthermore, our analysis found a higher number of aSyn aggregates in PD subjects than in HC subjects, with differences in aggregate size, density, and number of molecules per aggregate. On average, aSyn aggregate radii ranged between 40 and 200 nm and presented an average density of 0.001-0.1 molecules/nm². Our dSTORM analysis thus highlights the potential of our platform for identifying quantitative characteristics of aSyn distribution in skin biopsies not previously described for PD patients while offering valuable insight into PD pathology by elucidating patient aSyn aggregation status.

Keywords: Parkinson’s disease; alpha-synuclein aggregates; biomarker; density-based spatial clustering of applications with noise (DBSCAN); direct stochastic optical reconstruction microscopy (dSTORM); early diagnosis; fast optimized cluster algorithm for localizations (FOCAL); super-resolution microscopy.

Figure 1

Steps and methodology of skin biopsy processing for dSTORM. (A) Skin biopsy taken from the upper back (C7). (B) The skin contains various innervated structures, such as erector pili muscles, sweat glands, autonomic nerves, and cutaneous nerves. (C,D) Confocal image of a sweat gland from PD patient: p-aSyn (red), neuronal marker PGP9.5 (green) and areas of co-localization (yellow). (E) dSTORM reconstructed image of part of the sweat gland innervation (insert in D) shows p-aSyn (red) and PGP9.5 (green) and their co-localization (yellow). Insert-a close-up view showing single-molecule distribution of both p-aSyn (red) and PGP9.5 (green).

Figure 2

Nerve cells enriched with p-aSyn are less well-preserved. (A,B) Reconstructed dSTORM images showing p-aSyn localization (red) and that of a neuronal marker (green) in sweat gland innervation in a PD patient (A) and an HC subject (B). (A) and (B) insert - a close-up view showing single molecules distribution of both t-aSyn (red) and PGP9.5 (green). (C) The ratio of PGP9.5 ROI volume to t-aSyn ROI volume within each dSTORM image is not significantly different between HC subjects (blue) [Median: 0.89±0.125, n = 40] and PD patients (red) [Median: 0.97±0.295, n = 40], with a p-value of 0.165. (D) The ratio of the number of PGP9.5 localizations to t-aSyn localizations is significantly different between HC subjects (blue) [Mean (median) – 0.485±0.03 (0.47)] and PD patients (red) [Mean (median) – 0.59±0.04 (0.56)], with a p-value of 0.026. (E) The ratio of PGP9.5 ROI volume to p-aSyn ROI volume within each dSTORM image is not significantly different between HC subjects [Median: −18.5e+09±10.7e+09, n = 60] and PD patients [Median: −21.6e+09±5e+09, n = 60], with a p-value of 0.39. (F) The ratio of the number of PGP9.5 localizations to p-aSyn localizations is significantly higher in HC subjects (blue) [Mean (median) – 0.74±0.2 (0.29)] compared to PD patients (red) [Mean (median) – 0.51±0.05 (0.41)] (each point represents a ratio in a single dSTORM image), with p-value of 0.03. *p < 0.05.

Figure 3

Aggregates’ radii decrease and aggregates’ densities increase when increasing maPC from 0 to 2,500 to 5,000 using FOCAL¹, FOCAL^PC1, and FOCAL^PC2. (A–C) Clustering visualization and parameters comparison. Aggregate identification in the same area is shown using FOCAL¹ (A), FOCAL^PC1 (B), and FOCAL^PC2 (C). Inserts show the same 2 aggregates under the different FOCAL analyses. (D) Aggregate properties, such as the number of localizations in the aggregates, its radius, and density are listed for the 2 aggregates in the inserts for the 3 FOCAL parameters shown above. (E) Distribution of aggregates’ radii identified by each algorithm. FOCAL¹ (purple), FOCAL^PC1 (gray), and FOCAL^PC2 (red). (F) Distribution of aggregates’ densities identified by each algorithm. FOCAL¹ (purple), FOCAL^PC1 (gray), and FOCAL^PC2 (red). (G) Scatter plot showing the correlation between aggregates’ radius and density as detected in the different FOCAL analyses. The mean numbers of aggregates per image in FOCAL¹ [85.79], in FOCAL^PC1 [78.403], and in FOCAL^PC2 [55.919]. The mean percentage of localizations identified as aggregates in FOCAL¹ = 49.1%, in FOCAL^PC1 = 49.5%, and in FOCAL^PC2 = 49.6%.

Figure 4

PD patients display a higher number of aggregates compared to HC subjects. (A,B) Visualization of aggregates identified by FOCAL^PC2 in an HC subject (A) and a PD patient (B). (C) Distribution of aggregates’ radii shows that PD patient-derived images contain a larger number of aggregates with medium radii than do HC subject-derived images. (D) Distribution of aggregates’ densities shows that PD patient-derived images display a larger number of aggregates with low-medium densities. (E) Number of aggregates is significantly higher in PD patients than in HC subjects [means – HC: 42.45, PD: 55.92]. (F) Radii of aggregates are significantly larger in PD patients than in HC subjects [means – HC: 69.43, PD: 74.98]. (G) Densities of aggregates are significantly higher in HC subjects than in PD patients [means – HC: 0.1785, PD: 0.1263]. (H) Number of localizations per aggregates is significantly higher in HC subjects than in PD patients [means – HC: 217.66, PD: 203.68]. The mean percentage of localizations identified as aggregates in HC subjects is 20.3% and in PD patients is 21.6%. Means are marked with a + sign, and the 25^th percentile, median, and 75th percentile are marked by the bottom, middle, and top black lines outlining the whisker box, respectively. Each outlier data point is marked with a diamond. *p < 0.05, **p < 0.01, ****p < 0.0001.

Figure 5

t-aSyn aggregates are larger and less dense compared to p-aSyn aggregates in both PD patients and HC subjects. Comparison of t-aSyn and p-aSyn clustering in PD patients and HC subjects using FOCAL^PC2. (A,B) Visualizations of clusters of t-aSyn (A) and p-aSyn in a PD patient (B). (C–E) t-aSyn aggregates are larger and less dense compared to p-aSyn aggregates for both PD patients and HC subjects. ****p < 0.0001.