Cellular iron deposition patterns predict clinical subtypes of multiple system atrophy

Background: Multiple system atrophy (MSA) is a primary oligodendroglial synucleinopathy, characterized by elevated iron burden in early-affected subcortical nuclei. Although neurotoxic effects of brain iron deposition and its relationship with α -synuclein pathology have been demonstrated, the exact role of iron dysregulation in MSA pathogenesis is unknown. Therefore, advancing the understanding of iron dysregulation at the cellular level is critical, especially in relation to α -synuclein cytopathology. Methods: Iron burden in subcortical and brainstem regions were histologically mapped in human post-mortem brains of 4 MSA-parkinsonian (MSA-P), 4 MSA-cerebellar (MSA-C)


Introduction
Multiple system atrophy (MSA) is an atypical Parkinsonism characterized by α-synuclein (α-syn) pathology primarily in oligodendrocytes as glial cytoplasmic inclusions (GCI) and less in neurons as neuronal cytoplasmic and nuclear inclusions (NCI and NNI) (Kovacs, 2019).Neuronal degeneration accompanied by the accumulation of oligodendroglial α-syn pathology in the gray and in related white matter tracts severely affects the striatonigral and olivopontocerebellar systems of MSA patients.Mostly based on the predominant clinical phenotype, two subtypes are distinguished: 1) parkinsonian variant (MSA-P) with predominant striatonigral degeneration (SND), and 2) cerebellar variant (MSA-C) with olivopontocerebellar atrophy (OPCA) (Gilman et al., 1999;Graham and Oppenheimer, 1969;Lantos, 1998), while a third type associated with frontotemporal dementia and α-syn pathology in the frontal and temporal lobes (called atypical MSA or frontotemporal lobar degeneration-synuclein) has also been described (Aoki et al., 2015;Rohan et al., 2015).MSA α-syn pathology is seen to start in either the basal ganglia or the brain stem, but eventually affects both regions early in the disease (Brettschneider et al., 2018;Guo et al., 2022).Another characteristic feature of MSA pathology is a pronounced accumulation of iron in the same brain regions, in particular in the putamen (PUT), globus pallidus (GP), and the substantia nigra (SN) (Jellinger, 2003;Lee and Lee, 2019;Wenning et al., 2022).
Brain iron is vital for neuronal function, however, small fluctuations in its level becomes detrimental to neuronal health.Neurotoxic effects of elevated iron levels have been highlighted in Parkinson's disease (PD) research, another synucleinopathy, by an iron-dependent cell death mechanism called ferroptosis, which is characterized by mitochondrial damage and lipid peroxidation from increased production of reactive oxidative species (ROS) (Morris et al., 2018) Importantly, extensive evidence suggests a molecular relationship between iron accumulation and α-syn pathology (Lee and Kovacs, 2024).Iron is shown to 1) induce α-syn expression via the iron responsive element (IRE) motif of the α-syn messenger ribonucleic acid (mRNA) (Febbraro et al., 2012;Friedlich et al., 2007;Li et al., 2011;Rogers et al., 2011;Zhou and Tan, 2017)which up-regulated levels has previously been implicated for the development of α-syn pathology in genetic forms of PD and dementia with Lewy bodies (DLB) (Bradbury, 2003;Ibáñez et al., 2024.;Uchihara and Giasson, 2016), 2) induce α-syn aggregation in vitro and ex vivo by directly binding to α-syn at its C-terminal region (Bharathi, 2007;Davies et al., 2011;Kostka et al., 2008;Lu et al., 2011;Ostrerova-Golts et al., 2000;Uversky et al., 2001), and 3) enhance cell-to-cell transmission of α-syn aggregates by down-regulation of the autophagosome-lysosome fusion pathway (Xiao et al., 2018).Conversely, overexpression of, or aggregated forms of α-syn has been shown to induce iron dysregulation and iron-induced oxidative stress both in vivo and in vitro (Deas et al., 2016;Guo et al., 2021;Mi et al., 2021;Ortega et al., 2016;Shukla et al., 2021).
Due to such complex and interchangeable pathological mechanisms of iron and α-syn, the exact role of iron in the pathogenesis of MSA is yet undeciphered.Particularly, whether it serves a primary or secondary role to neurodegeneration is still unknown although its deposition is present early in the disease-course (Han et al., 2013;Kaindlstorfer et al., 2018;Lee and Baik, 2011).To elucidate its mechanism, careful mapping of the pathological iron burden at the cellular level and in relation to α-syn cytopathology in the early vulnerable regions of MSA brains is a critical step.Importantly, cellular iron homeostasis is distinct across cell populations, highlighting the possibility of a selective cellular vulnerability pattern to iron deposition in the diseased brains.We recently developed a method to evaluate cell-type specific iron accumulation and reported unique features in the tauopathy, progressive supranuclear palsy (PSP) (Lee et al., 2023).Here, applying our established methodologies to MSA, we deconstruct disease-associated iron deposition distinguished by neuropathological disease subtypes and explore possible underlying mechanisms by examination of select iron-related gene expressions.

Image quantification and statistical analysis
Stained sections were scanned using the Huron TissueScope LE120 (Huron).Regional iron load of the GP, PUT, SN, pons base (PB), and the cerebellum (CB) white matter as well as α-syn inclusion load of the GP, PUT, PB, and CB were quantified by Perl's and 5G4 positivity using the HALO software (Indica Labs).Respective regions were identified on whole-slide scanned images and entirely annotated under the guidance of a neuropathologist (GGK) using the custom annotation tool.The PB was annotated by an oval region of interest capturing areas just above the tegmentum with crossing pontocerebellar fibers.Comparison between groups were analyzed by Mann-Whitney test with significance threshold of P = 0.05.For examination of cellular iron burden, images at 10.5× magnification were taken from each region in DAB-enhanced iron and immunohistochemistry double stained sections: GP (3 images), PUT (5 images), and SN (3 images).Number of GFAP+ astrocytes, MAP2+ neurons, HLA-DR+ microglia, and TPPP/P25+ oligodendrocytes both negative and positive for iron deposition were quantified using the object colocalization module, capitalising on the unique morphologies of our cell types of interest.To minimize bias in the quantification of the cellular iron load, we evaluated only cells that were sectioned and examined at the level of the nucleus across the cell types examined.The nuclei, however, were not inclusively evaluated for iron deposition.Parallelly, MSA-6 was excluded from the evaluation of oligodendrocytic S. Lee et al. iron accumulation in the GP and the PUT, as the oligodendroglial nuclei were particularly of dark-blue/black colour closely resembling DAB-iron positivity, which made it difficult to effectively exclude the nucleus from the quantitative analysis (Fig. S1).Moreover, SN was not assessed for neuronal iron burden to eliminate uncertainties in correctly differentiating neuromelanin from pathological iron load.For evaluation of cellular iron accumulation in α-syn-positive cell types (oligodendrocytes and neurons) in the strategic regions, the total number of cells and total number of cells positive for iron deposition were manually counted in seven snapshots at 21× magnification from each region of interest.The quantification values were pooled across MSA-C and MSA-P subgroups as well as across the entire MSA cohort by region and by cell types, and the percentage of iron positive cells for each cell type were calculated by assigning "1" to each iron-positive cell and "0" to each iron-negative cell and computing the mean.Comparisons between groups were analyzed using a Mann-Whitney test using GraphPad Prism (v.9).
Extent of gliosis and microglial activation in the GP, PUT, and SN was measured in each case by quantification of % tissue area of GFAP and HLA-DR immunoreactivity (μm 2 ) in the above-mentioned snapshots of DAB-enhanced iron and immunohistochemistry double stained sections using the HALO software.Iron deposition in the same area was quantified by % tissue positivity as well.Comparison between the means of MSA-P and MSA-C cases were analyzed using Student's t-test in GraphPad Prism with a significance threshold of P = 0.05.

Correlation and hierarchal cluster analysis
Correlations between regional iron and α-syn inclusion load in the disease-related anatomical regions as quantified above, as well as with the duration of disease (DoD) of each case were examined by computing the Spearman's r correlations using GraphPad Prism (v.9), with a significance threshold of P = 0.05.Unsupervised hierarchal cluster analyses of the same MSA cases were performed using SPSS Statistics (v.23), based on the 1) regional α-syn pathology load (% tissue area) of the GP, PUT, PB, and the CB, 2) regional iron load (% tissue area) in the GP, PUT, SN, PB, and the CB, 3) percentage of iron-positive GCIs in the GP, PUT, and SN, and finally, 4) the relative pattern of iron deposition in neurons, astrocytes, oligodendrocytes, and microglia in the same regions, as calculated by dividing the percentage of iron-positive cell type of interest by the summation of percentages of iron-positive cells in all four cell types, were used for the analyses.

Gene expression analysis
Gene expression analyses were performed in 7 MSA (MSA-cases 6, 7, 10-14) and 6 age-matched control cases from which frozen sample material was available (Table 1).Experiment was performed as in Lee et al. (Lee et al., 2023) using the Nanostring nCounter assay with a Custom CodeSet design of 26 iron-and oxygen-homeostatic genes (NanoString Technologies Inc.) (Table 2).Briefly, brain tissue of the frontal cortex, GP, PUT, and SN were micro-dissected and extracted for total RNA.100 ng of RNA from each sample were analyzed, which assay was performed at the Princess Margaret Genomics Centre (Toronto, Canada).Normalized data were analyzed and visualized using the ROSALIND® (https://rosalind.bio/)Platform for nSolver analysis.Pvalues were adjusted for multiple comparisons using the Benjamini-Hochberg method, however, the adjusted values were not used for thresholding differentially expressed genes (York et al., 2021).Complete data set of this experiment has been submitted and can be viewed on the Gene Expression Omnibus (GEO; # 24244197) (https://www.ncbi.nlm.nih.gov/geo/).

Heterogeneity in regional iron deposition by MSA subtype
Recent MRI studies have begun to report differences in the regional iron load of MSA-C and MSA-P disease subtypes (Ito et al., 2017;Sugiyama et al., 2019).However, to date no post-mortem studies have compared regional iron deposition distinctively in disease subtypes.Using classical Perl's iron staining, we examined the regional iron load in the GP, PUT, SN, CB, and PB of MSA-C and MSA-P cases, as well as in an MSA case which showed both parkinsonian and cerebellar symptoms (MSA-C/P) (Fig. 1).First, we confirmed the clinicopathological subtype of the cases by quantification of the α-syn inclusion load in the basal OPCA-SND subtype refers to MSA cases with comparable neuropathological involvement of both the olivopontocerebellar and striatonigral systems.MSA case 5 show movement disorder but also unequivocal gait impairment.MSA case 12 show both parkinsonian and cerebellar symptoms.Roman numerals in brackets represents stages.Cases with FFPE samples were used for histological analysis, and frozen samples for gene expression analysis.Abbreviations: -indicates absence of pathology or duration of illness not applicable, cerebral amyloid angiopathy (CAA), duration of disease (DoD), formalin-fixed paraffin-embedded (FFPE), multiple system atrophyparkinsonian type (MSA-P), multiple system atrophy-cerebellar type (MSA-C), striatonigral degeneration (SND), olivopontocerebellar atrophy (OPCA), primary agerelated tauopathy (PART), argyrophilic grain disease (AGD), age-related tau astrogliopathy (ARTAG).
S. Lee et al. ganglia (GP and PUT) and the brainstem (PB and CB) as detected by 5G4 immunohistochemistry (Fig. 1A).All MSA-P cases showed higher inclusion burden in the basal ganglia compared to the brainstem, consistent with the neuropathological MSA-SND subtype and MSA-C cases showed higher inclusion burden in the brainstem, confirming MSA-OPCA subtype.Accordingly, the MSA-C/P case showed comparable 5G4 positivity in the two systems (Fig. 1A).Consistent with findings by Jellinger et al. (Jellinger et al., 2005), the MSA-P cohort presented with minimal involvement of α-syn pathology in the olivopontocerebellar systems, however, all MSA-C cases showed pathological involvement of the striatonigral system.The anatomical deposition of iron, as examined by Perl's staining, varied by the disease subtype in our MSA cohort.The iron load (% tissue positivity) in the subcortical regions (GP, PUT, and SN) were higher in MSA-P compared to MSA-C, which was particularly distinguishable by a significantly higher putaminal deposition of iron in the MSA-P group (p = 0.029; Fig. 1B).The MSA-C group showed similar levels of iron deposition across the three subcortical regions examined, being lowest in the PUT.In contrast, iron load was highest in the PUT compared to the GP and SN in MSA-P (Fig. 1B) Parallelly, the diseased case with equivalent involvement of both systems (MSA-C/P) showed the most homogenous deposition of iron across the three subcortical regions (data not shown).Interestingly, MSA-P cases showed a trend of greater gliosis and microglial activation compared to MSA-C cases in the subcortical regions -which relationship with the greater iron levels in MSA-P cases may be validated in a larger cohort (Fig. S2).Iron deposition in the CB and PB was negligible compared to the subcortical regions (Fig. 1A).Particularly, the PB was deemed absent of iron deposition.However, case MSA-9 (MSA-C) was an exception, showing comparable deposition of iron in the CB white matter and a particularly notable burden in the dentate nucleus (Fig. S3).
We then explored the relationship among the regional α-syn and iron deposition in the regions examined by computing Spearman's r correlations (Fig. 1C).Generally, the regional deposition of iron was positively associated within the subcortical system and the brainstem, and the α-syn load of GP and PUT, and CB and PB was positively correlated with each other (r = 0.88, p = 0.003; r = 0.77, p = 0.021 respectively).However, no particular correlation was found between the α-syn and iron load in the regions examined.Only the α-syn load in the GP and the PUT significantly correlated with lower iron load in the CB (r = − 0.83, p = 0.008; r = − 0.93, p = 0.001 respectively) (Fig. 1C).The extent of gliosis and microglial activation were not able to explain the variance in pathological iron deposition as well (Data not shown).Importantly however, a uniform trend of negative correlation was found between the duration of disease and iron deposition in all regions evaluated, whereas a uniform trend of positive correlation was found between the duration of disease and α-syn load in all regions.(Fig. 1C) When examining these correlations among the regional iron and α-syn deposition separately in disease subtypes, different patterns were observed, supporting distinct iron-related mechanisms in disease subtypes (Fig. S4).Interestingly, the MSA-C cohort showed a uniform pattern of negative association between iron deposition and α-syn load in all regions examined (Fig. 1D).

Pathological iron deposition shows selective vulnerability patterns at the cellular level in MSA brains
To better understand the pathological accumulation of iron in MSAaffected brains, we evaluated its cellular distribution in astroglia, neurons, microglia, and oligodendrocytes in the vulnerable nuclei (GP, PUT, SN) (Fig. 2A).The cellular deposition of iron differed by cell types but exhibited a consistent vulnerability pattern across the regions examined (Fig. 2B).In all regions, iron burden was highest in the microglia (p < 0.001 for all comparisons), which was most prominent in the PUT (36.990% iron-positive microglia) compared to the GP (21.600%) and a Functions of genes described (Lee et al., 2023).Abbreviations: Neurodegeneration with Brain Iron Accumulation (NBIA), Iron responsive element (IRE), red blood cell (RBC), divalent metal transporter 1 (DMT1), six-transmembrane epithelial antigen of the prostate (STEAP).
S. Lee et al. the SN (21.860%).Neuronal iron burden was consistently lowest across the cell types examined, showing minimal deposition (GP: 1.918%, PUT: 0.654%, SN not evaluated for neuronal iron deposition; p < 0.001 for all).Astrocytic and oligodendrocytic iron deposition were relatively comparable in the vulnerable nuclei, but astrocytic iron deposition was comparably higher in the PUT (p = 0.036) and the SN (p < 0.001), and lower in the GP (p < 0.001) than that of the oligodendrocytes (Fig. 2B).
Next, we examined for a possible difference in the pattern of cellular iron deposition by clinical subtypes.Interestingly, we found differences in the cellular vulnerability patterns (Fig. 2B).Microglial deposition of iron was significantly lower in MSA-C compared to MSA-P in all regions (p < 0.001 for all comparisons), and the difference was most prominent in the PUT (MSA-C: 8.901%, MSA-P: 30.470%).In MSA-P, microglial deposition was highest compared to the other cell types throughout the examined regions, followed by either astrocytes or oligodendrocytes, and the least in neurons.In contrast, iron deposition was highest in astrocytes in the GP of MSA-C brains, which was significantly higher than that in MSA-P brains (p < 0.001).Neuronal iron burden was consistently higher in MSA-C cases compared to those of MSA-P (GP: p = 0.019, PUT: p < 0.001).Generally, the level of iron deposition was more homogenous across the cell types in MSA-C.Importantly, heavy iron load was also observed outside the cellular bodies in these regions of MSA brains, particularly in the PUT (Fig. 2C).

Cellular iron deposition predominantly associates with oligodendrocytic α-syn cytopathology in MSA
We next examined the relationship between cellular iron deposition and α-syn cytopathology in the vulnerable nuclei of MSA brains, by evaluating the proportion of 5G4-α-syn positive oligodendrocytes and neurons accumulating iron as detected by DAB-enhanced Perl's iron staining.Interestingly, in all regions examined, α-syn-positive neurons were completely negative for iron deposition (Fig. 2D).On the other hand, α-syn-positive oligodendrocytes showed variable iron burden across the vulnerable nuclei, with the highest deposition observed in the SN (24.310%; p < 0.001 for all comparisons) (Fig. 2E).Iron deposition in α-syn-positive oligodendrocytes were significantly lower in the GP (10.120%) and the PUT (8.409%).Again, we further examined the Fig. 1.Iron deposition in early-affected regions of MSA brains.A. Heatmap representing regional 5G4-α-syn inclusion load (% tissue area) in the basal ganglia (BG; GP + PUT) and the brainstem (BS; PB + CB), as well as the regional iron load (% tissue area) in the same systems and the SN of MSA cases.B. Box-violin plot representing iron deposition in each region distinctly in MSA-C and MSA-P groups.Iron deposition is generally higher in the basal ganglia in MSA-P cases compared to MSA-C, and higher in the brainstem of MSA-C, although iron deposition in the PB is neglectable in both groups.The greatest difference between the two diseased groups is in the putamen, where iron deposition is highest among the subcortical regions in MSA-P, and lowest in MSA-C.Significance (*p < 0.05) of the comparison between groups by Mann-Whitney U test are indicated above the error bars C. Matrix representing Spearman's r correlations among the evaluated regional iron and α-syn pathology load, as well as with the duration of disease (DoD) in all MSA cases examined.Abbreviations: multiple system atrophy-parkinsonian (MSA-P), multiple system atrophy-cerebellar (MSA-C), multiple system atrophy with both cerebellar and parkinsonian symptoms (C/P), globus pallidus (GP), putamen (PUT), substantia nigra (SN), pons base (PB), cerebellum (CB).
cellular association at the level of clinical subtypes.The percentage of iron-positive GCIs were consistently lower in the MSA-C subtype across the regions examined, which was significant compared to the MSA-P subtype in the GP and the PUT (GP: 4.545% in MSA-C, p < 0.001; PUT: 3.644%, p < 0.001) (Fig. 2E).In the MSA-C (MSA-9) case which showed abnormal accumulation of iron in the cerebellar regions, we examined their cellular localization and the proportion of 5G4-α-syn positive cell types accumulating iron in the cerebellum (Fig. S5).Parallel to observations in subcortical regions, α-syn positive neurons were negative for iron deposition.However, 29.09% of α-syn positive oligodendrocytes showed positivity for iron deposition, which was higher than that observed in subcortical regions of MSA-C cases.Abbreviations: glial fibrillary acidic protein (GFAP), microtubule-associated protein 2 (MAP2), major histocompatibility complex II cell surface receptor (HLA-DR), tubulin polymerization-promoting protein (TPPP/P25), 3,3′-Diaminobenzidine (DAB), multiple system atrophy-cerebellar (MSA-C), multiple system atrophyparkinsonian (MSA-P), glial cytoplasmic inclusion (GCI), neuronal cytoplasmic inclusion (NCI), globus pallidus (GP), putamen (PUT), substantia nigra (SN), and standard error of the mean (SEM).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)S. Lee et al.

Cellular pattern of iron deposition predicts MSA subtypes
To further examine the implications of the distinct regional and cellular patterns of iron deposition in MSA pathophysiology, we performed unsupervised hierarchal clustering of the MSA cases using the regional and cellular deposition patterns of α-syn pathology and pathological iron (Fig. 3).First, the regional α-syn pathology load in the GP, PUT, PB, and the CB (% tissue positivity) of these cases grouped the MSA-P cases together and grouped the MSA-C cases with the MSA-C/P case and identified one MSA-C case as an outlier to the two groups (Fig. 3A).Using the regional iron load in the GP, PUT, SN, PB, and the CB, the two disease subtypes could not be distinguished (Fig. 3B).Frequency of iron-accumulation in α-syn-affected GCIs in these regions alone also did not correctly cluster cases by the disease subtype (Fig. 3C).The pattern of iron deposition across the four cell types, or in other words, the relative deposition of iron in astrocytes, neurons, oligodendrocytes, and microglia in the subcortical regions, clustered the MSA-P cases together with the MSA-C/P case, and also grouped the MSA-C cases together, with one MSA-C case which showed abnormal iron deposition in the cerebellar and dentate nucleus, as an outlier to the two groups (Fig. 3D).

Differential expression of iron and oxygen homeostasis-related genes in vulnerable nuclei of MSA brains
To probe the molecular pathways involved in the observed pathological accumulation of iron in MSA vulnerable regions, we examined mRNA expression changes of select iron and the closely related oxygen homeostatic genes, as well as ten major Neurodegeneration with Brain Iron Accumulation (NBIA)-associated genes in the GP, PUT, SN, and additionally, in the frontal cortex of MSA and control brains.
In the basal ganglia, iron homeostatic genes were generally increased in expression in the diseased group (Fig. 4).Interestingly, exceptions were those that are involved in ferrous iron homeostasis, SLC11A2 both -IRE and non-IRE forms (GP: fold change = − 1.196 and − 1.481, p = 0.465 and 0.187 respectively; PUT: fold change = − 1.316 and − 2.070, p = 0.336 and 0.090) which function in cellular ferrous iron uptake, and FTH gene in the PUT (fold change = − 1.321, p = 0.276) which function in cellular iron storage by reduction of cellular ferrous iron into the proper ferric form.However, none of the examined iron-homeostatic genes were found to be significantly dysregulated in these regions (Table 3).In the PUT, an NBIA-associated gene that functions in biosynthesis of coenzyme A, COASY, was significantly upregulated (fold change = 1.553, p = 0.036), which may be indicative of mitochondrial dysregulation.The frontal cortex generally showed a similar pattern of gene expression as the GP and the PUT (Fig. 4).The SN showed more variable changes in the expression of iron homeostatic genes with an overall down-regulated trend, among which the SLC11A2 (non-IRE) gene functioning in non-regulated form of cellular non-transferrinbound iron (ferrous) uptake was found to be significantly downregulated (p = 0.020) (Table 3).Interestingly, the expression of oxygen homeostatic genes was significantly dysregulated in the diseased SN by an up-regulation of the HBA gene (p = 0.009) and a down-regulation of the NGB gene (p = 0.037).

Discussion
Early MRI observations of iron accumulation in the subcortical nuclei of MSA patients has well-established iron dysregulation in the pathophysiology of MSA (Han et al., 2013;Lee and Baik, 2011).In the wide spectrum of neurodegenerative diseases that are associated with elevated brain iron levels, MSA is often simply considered as a disorder with a high burden of iron particularly in the PUT (Kaindlstorfer et al., 2018).Our histological evaluation of the subcortical iron load, however, revealed heterogeneous distribution of iron that showed distinct patterns of regional deposition in MSA subtypes, particularly highlighted by a contrast in the putaminal iron burden (Fig. 1).Although examination in a larger cohort is required to confirm the stratification, consistent findings in recent iron-sensitive MRI studies of MSA-C and MSA-P patients support our results (Ito et al., 2017;Sugiyama et al., 2019).Distinct patterns of iron accumulation in the subcortical regions, despite involvement with α-syn pathology in both disease subgroups, suggest that iron-associated pathomechanisms may differ in the two subtypes, further stratifying MSA-P/SND and MSA-C/OPCA subtypes as distinct pathogenic entities.The difference in the correlational relationship of the regional deposition levels of iron and α-syn pathology between the two disease subtypes (Fig. S4) further support this idea.
To explore these findings in a more comprehensive manner, we evaluated the distribution of subcortical iron detected by DAB-enhanced Perl's staining in different cell types.Consistent with observations in Alzheimer's Disease (AD) (Kenkhuis et al., 2021), PD (Guo et al., 2021), and in normal aging (Ashraf et al., 2018), iron accumulation in MSA was prominently observed in the microglia across all regions examinedmore prominently in the MSA-P subtype.As activated subcortical Fig. 3. Hierarchal clustering analysis of MSA cases.Representative dendrograms using Average Linkage (Between Groups). A. Unsupervised cluster analysis using regional α-syn pathology load (% tissue positivity) of the GP, putamen, CB, and the PB correctly groups MSA-P cases (MSA 1-4).Case showing both cerebellar and parkinsonian symptoms (MSA 5) was grouped with MSA-C cases (MSA 9,6,8), and one MSA-C case (MSA 7) was identified as an outlier.B. Cluster analysis using GP, putamen, SN, CB, and PB iron load (% tissue positivity) did not effectively group the diseased cases.C. Iron positivity in GCI alone also do not distinguish MSA subtypes.D. The relative deposition of iron across neurons, astrocytes, oligodendrocytes, microglia in the GP, putamen, and SN groups MSA-P cases together including the case with both cerebellar and parkinsonian symptoms, and groups the MSA-C cases together, with one MSA-C case (MSA 9) showing abnormally high level of iron in the cerebellum as an outlier.Abbreviations: Alpha-synuclein (α-syn), multiple system atrophy (MSA), glial cytoplasmic inclusions (GCI), MSAparkinsonian (MSA-P), MSA-cerebellar (MSA-C), globus pallidus (GP), substantia nigra (SN), pons base (PB), and cerebellum (CB).microglia are an early feature of MSA disease progression, phagocytic processing of cellular debris in the extracellular space most likely contributes to the observed microglial deposition of iron in the diseased brains.Accordingly, semi-quantitative analysis in our cohort revealed higher microglial activation in the MSA-P cohort compared to MSA-C cohort in all regions examined (Fig. S2).Such iron-bearing microglia may act as an important player in disease pathogenesis by a ferroptosisinduced neurodegeneration.A recent study using human iPSC-derived tri-culture system of microglia, astrocytes, and neurons to examine the contribution of iron accumulation to neurodegeneration has demonstrated microglia to be highly sensitive for iron-related changes and susceptible to ferroptosis, consistent with findings in single cell culture vulnerability assays (Jiao et al., 2022;Ryan et al., 2023).Importantly, the removal of microglia from the same tri-culture system prevented neuronal cell death, further supporting an important role of microglial iron loading in the disease pathogenesis (Ryan et al., 2023).In contrast, the MSA-C/OPCA cases showed a heterogeneous pattern of cellular iron deposition across these regions in which astrocytes demonstrated relatively either similar or higher burden than that of the microglia.Astrocytes are found to be relatively resistant to iron-toxicity as well as serving a protective role against ferroptosis, further supporting distinct contributions of iron to disease progression in MSA-P and -C (Jiao et al., 2022;Pelizzoni et al., 2013;Ryan et al., 2023;Wang et al., 2022).Importantly, our unsupervised hierarchical cluster analysis revealed such different patterns of cellular iron accumulation alone to group the MSA cases by the clinical subtypes.Neuropathological studies demonstrate that clear distinction of SND and OPCA type of MSA is difficult as the pathology shows significant overlap (Ozawa et al., 2004), therefore parkinsonian and cerebellar classification of MSA is made based on early predominating clinical symptoms.Our post-mortem evaluation of different iron deposition patterns at the regional and cellular levels in the two disease subtypes even at end-stage disease, suggest that iron dysregulation may not be a simple uniform consequence of local α-syn pathology, but a dynamic agent that could affect the course of disease.
Examining iron accumulation distinctly in α-syn-affected cells, 10.1% and 8.4% showed GCI-positivity for iron deposition in the GP and the putamen respectively, and 24.3% in the SN.This is in sharp contrast with our previous observations in PSP (Lee et al., 2023), where 74.5% and 65.3% of tau-affected astrocytes were positive for iron deposition in the very early-affected GP and the SN respectively (Lee et al., 2023).These differences suggest that iron-associated pathomechanisms differ across neurodegenerative proteinopathies distinctively in relation to the build-up of protein pathologies.Furthermore, NCIs were negative for iron deposition in the examined nuclei of MSA brains.Although the loss of early-affected neuronal cells could account for the lack of association, evidence for neuronal mechanisms to minimize iron uptake in physiological aging and pathological conditions of elevated iron levels suggest otherwise (Griffiths and Crossman, 1996;Quintana et al., 2006;Ryan et al., 2023).Parallelly, the level of cellular iron deposition is also generally lower in MSA compared to that in PSP (Lee et al., 2023).Accordingly, iron was notably observed in spaces outside the cellular bodies in the examined nuclei of MSA brains.These iron species may function in pathological processes such as oligomer seeding, but also in neurodegeneration by the reduction of ferric iron to ferrous iron by extracellular ferrireductases, and the production of ROS by participation  of ferrous iron in the Fenton reaction (Kostka et al., 2008;Singh et al., 2014).MRI studies demonstrating correlation between iron deposition and local atrophy in MSA patients (Barbagallo et al., 2016;Lee et al., 2015;Matsusue et al., 2008), as well as our finding of a uniform pattern of negative correlation between the disease duration and iron levels in the key pathology-related regions support this concept (Fig. 1C).GCIs, although not as abundant compared to iron bearing tau-positive astrocytes in PSP, show greater association with cellular iron deposition than α-syn-unaffected oligodendrocytes in the diseased brains, and may also contribute to such neuronal toxicity.Altogether with a general upregulation of iron homeostatic genes in the same regions of MSA brains as revealed by Nanostring gene expression analysis, our findings provide insight into cell-type specific contribution of iron dysregulation in MSA-affected brains.(Fig. 5).Through phagocytosis, heavy microglial iron deposition promotes disease toxicity via ferroptosis, and comparable astrocytic iron deposition in MSA-C subcortical regions may be either be protective or destructive.Intracellular iron in oligodendrocytes contribute to α-syn pathology, which in turn contribute to cellular iron deposition and accelerated iron cycling into the extracellular spaces to induce local toxicity.Protected against excess iron uptake, such mechanism is not initiated in neurons.Examination of a more comprehensive selection of iron homeostasis-related genes in a larger diseased cohort may identify key dysregulated genes that we were not able to identify in this studypossibly distinct at subtype-specific levels.Moreover, iron-  (Jiao et al., 2022;Ryan et al., 2023).Increased microglial iron uptake also induce up-regulation of CXCL8 expression and secretion into the extracellular environment (Ryan et al., 2023).Astrocytes, under aging and pathological conditions, are seen to increase iron uptake by up-regulation of DMT1 (Lu et al., 2017;Xia et al., 2021).Moreover, astrocytes in neurodegenerative brains, including synucleinopathies, are seen to engulf extracellular debris by pinocytosis and phagocytosis (Morales et al., 2017;Morizawa et al., 2017;Yang et al., 2022).Astrocytic iron deposition in MSA-C subcortical regions may be either be protective or destructive.In oligodendroglia, increased iron uptake induce expression of α-syn through IRE-IRP mechanisms (Febbraro et al., 2012).Increase in α-syn levels potentially induce pathological aggregation of the protein (Shukla et al., 2021) which aggregates then sequester free iron by direct binding (Bharathi et al., 2007;Davies et al., 2011;Kostka et al., 2008;Lu et al., 2011;Ostrerova-Golts et al., 2000;Uversky et al., 2001).Consequent reduction of the labile iron pool maintains the expression of iron uptake proteins to allow consistent cellular uptake of iron (Zhou and Tan, 2017).Cellular iron export via ferroportin is maintained as well, suggesting an accelerated iron cycling in cells which iron builds up in the extracellular space.Protected against excess iron uptake in physiological aging and pathological conditions by downregulation of TfR and up-regulation of FPN (Griffiths and Crossman, 1996;Jiao et al., 2022;Quintana et al., 2006;Ryan et al., 2023), such mechanism may not be initiated in neurons.Exported ferric iron is reduced to the toxic ferrous form by extracellular ferrireductase including α-syn, which produce reactive oxygen species to induce neurodegeneration.Abbreviations: Alpha-synuclein (α-syn), Zrt-and Irt-like protein 14 (Zip14), divalent metal transporter 1 (DMT1), transferrin-transferrin receptor complex (Tf-TfR1), lactoferrin-lactoferrin receptor (LfR-Lf), T-cell immunoglobin and mucine domain (Tim-2), Low-density lipoprotein receptor-related protein 1 (LRP1), iron responsive element-iron regulatory protein (IRE-IRP), Six-transmembrane epithelial antigen of prostate 3 (STEAP3), labile iron pool (LIP), ferroportin (FPN), hephaestin (HEPH), ceruloplasmin (Cp), reactive oxygen species (ROS), and ferritin light chain (FTL).
S. Lee et al. related cellular mechanisms may differ in pontocerebellar regions of MSA-C cases.
In the SN, significant dysregulation in the gene expression of hemoglobin-alpha and neuroglobin provide a further novel insight into hypoxia-related disease mechanisms in MSA.Chronic hypoxia has been demonstrated to be involved in the pathophysiology of MSA by a significant increase in the protein levels of HIF2α marker in the SN compared to PD and age-matched controls (Heras-Garvin et al., 2020).Meanwhile, neuroglobin is thought to function in the scavenging of ROS and has been linked to neuroprotection in different neurodegenerative diseases by an overexpression in stroke, hypoxia, and ischemia (Baez et al., 2016;Burmester et al., 2000;Sun et al., 2001;Van Leuven et al., 2013;Yu et al., 2012).The contradicting down-regulation of neuroglobin in MSA suggests a faulty compensation for oxidative stress which may exacerbate neurodegeneration and disease progression (Ozawa et al., 2004).

Conclusions
The first cellular mapping of iron accumulation in the vulnerable regions of human post-mortem MSA brains reveals selective cellular vulnerability patterns in pathologic iron accumulation that are distinct in MSA-C/OPCA and MSA-P/SND, suggesting subtype-specific ironassociated disease mechanisms in MSA that has not been found for hypoxia-related events (Heras-Garvin et al., 2020).We reveal cellular vulnerability pattern to pathological iron deposition as a novel neuropathological feature that effectively distinguish MSA parkinsonian and cerebellar subtypes, distinctly from α-syn pathology, reinforcing the involvement of iron dysregulation in neurodegeneration, and disease pathogenesis.MSA is a rare synucleinopathy and accordingly, the limitation of the current study is the relatively low number of MSA cases examined, especially in the context of patient heterogeneity highlighted by the disease duration range of 3-10 years in our MSA cohort.Therefore, examination in other cohorts as well as examination using other quantitative measures of cellular iron concentration in different cell types (such as inductively coupled plasma mass spectrometry and synchrotron X-ray spectromicroscopy) would complement our histologybased observations and interpretation of disease subtype-specific cellular vulnerability patterns to pathological iron deposition.
Our findings demonstrating different cellular localization of pathological iron in MSA and PSP suggest distinct contributions of iron in the pathogenesis of iron-associated neurodegenerative proteinopathies.Our discovery provides insight for novel therapeutic strategies for disease modification, particularly that of iron chelation at the disease and cellular-specific levels.To date, iron chelators actively examined in synucleinopathies, such as deferiprone in PD (Devos et al., 2022), target the sequestration of global iron content, which is challenged by interference of the vital physiological functions of essential iron in different cellular systems.Specifically targeting the major iron accumulating key population, or the associated pathways by a combinatory approach which is thought to be the next generation of iron chelation strategies, may confer better clinical outcomes.

Fig. 2 .
Fig. 2. Cellular mapping of pathological iron deposition in the vulnerable regions of MSA brains.A. Cellular iron deposits in MSA astrocytes (GFAP), neurons (MAP2), microglia (HLA-DR), and oligodendrocytes (TPPP/P25).Iron is visualized inside cellular bodies (magenta) by dark-brown DAB deposits.B. Percentage of astrocytes, neurons, microglia, and oligodendrocytes positive for iron deposits in the GP, putamen, and SN of MSA brains.Iron accumulation is greatest in microglia across all regions, particularly in the putamen.Neurons consistently show lowest iron deposition.The pattern of cellular iron accumulation differs by MSA-C and P subtypes.Data are presented as mean ± SEM * 100%.Significance (p < 0.05) was determined by a Mann-Whitney test, indicated above the error bars.*p < 0.05 and ****p < 0.0001 vs. MSA-C, # p < 0.05 and #### p < 0.0001 vs. astrocytes-pooled, ΔΔΔΔ p < 0.0001 vs. neurons-pooled, ×××× p < 0.0001 vs. microglia-pooled.Black dots indicate % iron-positive cell value for each cases examined C. Iron deposits outside cellular bodies in the putamen.Inset shows higher magnification image.D. iron deposition in α-syn-affected oligodendrocytes and neurons.The different cell types are distinguishable by 5G4-α-syn morphology.α-Syn-affected neurons are negative for iron deposition.E. Iron accumulation in α-syn-affected oligodendrocytes (GCI) in the GP, putamen, and the SN.Iron burden in the affected oligodendrocytes is highest in the SN.The association of cellular iron accumulation in the affected oligodendrocytes are comparably lower in the GP and the putamen.Data are presented as mean ± SEM * 100%.Significance (p < 0.05) was determined by a Mann-Whitney test, indicated above the error bars.****p < 0.0001 vs. MSA-C, ◆◆ p < 0.01 and ◆◆◆◆ p < 0.0001 vs. GP-pooled, ×××× p < 0.0001 vs. PUT-pooled.Black dots indicate % iron-positive inclusion values for each cases examined.Abbreviations: glial fibrillary acidic protein (GFAP), microtubule-associated protein 2 (MAP2), major histocompatibility complex II cell surface receptor (HLA-DR), tubulin polymerization-promoting protein (TPPP/P25), 3,3′-Diaminobenzidine (DAB), multiple system atrophy-cerebellar (MSA-C), multiple system atrophyparkinsonian (MSA-P), glial cytoplasmic inclusion (GCI), neuronal cytoplasmic inclusion (NCI), globus pallidus (GP), putamen (PUT), substantia nigra (SN), and standard error of the mean (SEM).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4 .
Fig. 4. Gene expression changes of iron and oxygen homeostatic genes in vulnerable regions of MSA brains.Heatmap representing differential expression of iron and oxygen-homeostasis related genes in MSA compared to age-matched controls.Oxygen homeostasis-related genes are colour-coded in orange, iron homeostasis genes in green, and Neurodegeneration with Brain Iron Accumulation (NBIA)-associated genes in magenta.Red box represents up-regulation and blue represents down-regulation in the MSA cohort.Asterix indicates significance (p < 0.05) in fold change.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5 .
Fig. 5. Iron dyshomeostasis in MSA.Cellular mapping of pathological iron deposition in the early-affected subcortical regions of MSA brains provide insight into cell-type specific contribution of iron dysregulation in MSA disease pathology.Through phagocytosis of extracellular debris, heavy iron deposition in ferroptosisvulnerable microglia leads to up-regulation of SEC24B expression, which together with increased cellular labile iron pool (LIP) promote ferroptosis and neuronal toxicity(Jiao et al., 2022;Ryan et al., 2023).Increased microglial iron uptake also induce up-regulation of CXCL8 expression and secretion into the extracellular environment(Ryan et al., 2023).Astrocytes, under aging and pathological conditions, are seen to increase iron uptake by up-regulation of DMT1(Lu et al., 2017;Xia et al., 2021).Moreover, astrocytes in neurodegenerative brains, including synucleinopathies, are seen to engulf extracellular debris by pinocytosis and phagocytosis(Morales et al., 2017;Morizawa et al., 2017;Yang et al., 2022).Astrocytic iron deposition in MSA-C subcortical regions may be either be protective or destructive.In

Table 1
Clinical, demographic and neuropathologic data of the cases included in the study.

Table 3
Differentially expressed genes in MSA brain regions.