Integrated Label Transfer for Oligodendrocyte Subpopulation

Profiling in Parkinson’s Disease and Multiple System Atrophy

Erin Teeple

, Pooja Joshi

, Rahul Pande

, Yinyin Huang

, Akshat Karambe

Martine Latta-Mahieu

, S. Pablo Sardi

, Angel Cedazo-Minguez

, Katherine W. Klinger

Amilcar Flores-Morales

, Stephen L. Madden

, Deepak K. Rajpal

and Dinesh Kumar

Translational Sciences, Sanofi, Framingham, MA, U.S.A.

Neurological and Rare Diseases Therapeutic Area, Sanofi, Chilly-Mazarin, France

erin.teeple,@sanofi.com

Keywords: CCA, Data Integration, Seurat, Label Transfer, Synucleinopathy, Parkinson’s Disease, MSA.

Abstract: Transfer of cell type labels as part of the comprehensive integration of multiple single nucleus RNA

sequencing (snRNAseq) datasets offers a powerful tool for comparing cell populations and their activation

states in normal versus disease conditions. Another potential use for these methods is annotation alignments

between samples from different anatomic areas. This study describes and evaluates an integration analysis

applied for profiling of oligodendrocyte lineage nuclei sequenced from human brain putamen region tissue

samples for healthy Control (n = 3), Parkinson’s Disease (PD; n = 3) and Multiple System Atrophy (MSA; n

= 3) subjects with label transfer to substantia nigra region tissue samples for healthy Control (n = 5) subjects.

PD and MSA are both synucleinopathies, progressive neurodegenerative disorders characterized by nervous

system aggregates of α-synuclein, a protein encoded by the SNCA gene. Histologic findings and genetic

evidence suggest links between oligodendrocyte biology and synucleinopathy pathogenesis. In this work, we

first identify disease-associated changes among transcriptionally distinct oligodendrocyte subpopulations in

putamen. We then apply label transfer methods to generalize our findings from putamen to substantia nigra,

a brain region characteristically impacted in PD and variably affected in MSA. Interestingly, our analysis

predicts oligodendrocytes in substantia nigra include a significantly greater proportion of an oligodendrocyte

subpopulation identified in putamen as most highly overexpressing SNCA in PD. Our results provide new

insights into oligodendrocyte biology in PD and MSA and our workflow provides an example of label transfer

methods applied for cross-dataset exploratory purpose.

1 INTRODUCTION

Synucleinopathies are a group of progressive

neurodegenerative disorders characterized by

nervous system aggregates of α-synuclein protein

(Coon, Cutsforth-Gregory & Benarroch, 2018). PD is

the most common synucleinopathy and the second

most common chronic neurodegenerative disorder,

affecting 1% of the population over age 60 (Tysnes &

Storstein, 2017). MSA occurs at a much lower

frequency than PD and has an estimated incidence

rate of 0.6 per 100,000 people (Vanacore, Bonifati,

Fabbrini, et al., 2001). Intracellular inclusions of α-

synuclein are observed on post-mortem microscopic

exam of central nervous system tissue (CNS) in both

PD and MSA, but the cellular location of α-synuclein

and patterns of CNS involvement differ between the

disorders. In PD, α-synuclein aggregates are observed

mainly as neuronal intracellular collections (Lewy

bodies) (Spillantini, Schmidt, Lee, et al., 1997). In

MSA, in contrast, α-synuclein aggregates occur most

frequently as oligodendroglial cytoplasmic inclusions

(Inoue, Yagishita, Ryo, et al.,1997; Hague, Lento,

Morgello, et al. 1997). Death of neurons in the

substantia nigra pars compacta is particularly

characteristic of PD, with lesion involvement

progressing from the brainstem and midbrain to the

neocortex observed over the course of the disease

(Del Tredici & Braak, 2016). While striatonigral

degeneration occurs to varying degrees in MSA,

concurrent and more variable involvement of the

cerebellum and autonomic nervous system are further

clinical features in MSA (Inoue et al., 1997; Hague,

et al., 1997).

SNCA is the gene encoding α-synuclein, a 140

amino acid protein known to participate in vesicle

Teeple, E., Joshi, P., Pande, R., Huang, Y., Karambe, A., Latta-Mahieu, M., Sardi, S., Cedazo-Minguez, A., Klinger, K., Flores-Morales, A., Madden, S., Rajpal, D. and Kumar, D.

Integrated Label Transfer for Oligodendrocyte Subpopulation Proﬁling in Parkinson’s Disease and Multiple System Atrophy.

DOI: 10.5220/0010915400003123

In Proceedings of the 15th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2022) - Volume 5: HEALTHINF, pages 219-227

ISBN: 978-989-758-552-4; ISSN: 2184-4305

219

exocytosis, endocytosis, and neurotransmitter vesicle

cycling (Maroteaux, Campanelli, & Scheller, 1988).

Studies have also localized this protein to the cell

nucleus, where direct interaction with DNA and

histone proteins (Pinho, Paiva, Jercic, et al., 2019)

and modulation of DNA damage responses (Goers,

Manning-Bog, McCormack, et al. 2003) have been

reported. The complete spectrum of α-synuclein

activities and how this protein may contribute to the

development and progression of each of the

synucleinopathies are not fully understood.

Genome-wide association studies (GWAS) and

transcriptomic profiling of human brain tissue have

highlighted potential roles for glial cells in the

synucleinopathies, with particularly strong evidence

implicating oligodendrocyte biology in PD (Bryois,

Skene, Hansen, et al., 2020; Smajic, Prada-Medina,

Landoulis, et al., 2020; Nalls, Blauwendraat,

Vallerga, et al., 2019; Reynolds, Botia, Nalls, et al.,

2019). Inadequate metabolic support for neurons,

overactive stress and inflammatory response

signaling, and dysfunctional autophagy have been

suggested as mechanisms by which oligodendrocytes

might contribute to PD development (Teeple, Jindal,

Kiragasi, Annaldasula, et al., 2020; Bryois, et al.,

2020; Reynolds, et al., 2019). Oligodendrocyte-

specific differentially expressed genes have also been

linked with variants significantly associated with PD

risk by GWAS in analyses of snRNAseq data from

healthy human donor substantia nigra tissues

(Agarwal, Sando, Volpato, et al.,2020) and mouse

nervous system single-cell data (Bryois, et al., 2020).

The present analysis was undertaken in order to

further profile the relationship between midbrain

oligodendrocyte population heterogeneity and α-

synuclein biology as part of a comprehensive analysis

of PD and MSA snRNAseq data with label transfer

methodologies found in this analysis to reveal new

insights (Teeple, Joshi, Pande, Huang, et al., 2021).

1.1 Related Work

The development and ongoing refinement of single

nucleus RNA sequencing (snRNAseq) techniques

have greatly advanced our ability to understand the

heterogeneity and functional activities of cell

populations in the brain and nervous system. Cells are

the basic unit of the multicellular organism, but

although cells in the brain share DNA, each differs in

its transcriptional activities, epigenetic modifications,

and functions in and responses to its

microenvironment (Duran, Wei, & Wu, 2017).

Neuronal cells in the brain form densely

interconnected, diversified networks where structure

and cell functional activation states support and

coordinate dynamic and complex processes, for

example memory encoding, vision, and motor

coordination. Non-neuronal cell populations

intermixed in these cellular networks support

neuronal metabolism, facilitate signal transmission,

and modulate vascular flow and immune responses,

among many other activities (Duran et al., 2017).

Sequencing of nuclei in a tissue sample yields a

unique molecular identified (UMI) count matrix. This

matrix includes integer counts of the number of RNA

molecules for each feature (gene) identified in each

nucleus (one nucleus per cell). In the analysis of

snRNAseq data, variations in gene counts between

nuclei are used both to cluster cell types (by similar

patterns of gene expression in nuclei) as well as for

differential expression analysis where different cell

groups are compared with respect to their mean

expression of different genes. Pre-processing of

snRNAseq data includes initial filtering of UMI data

tables to remove low quality rows (ie those nuclei

with few counts or very many, which likely represent

data for empty droplets or multiplets, respectively)

and cells with very high percentages of mitochondrial

genes (Hao, Hao, Andersen-Nissen, et al., 2021).

These filtering steps are undertaken to ensure high

quality data are used for downstream analyses.

Variations in sequencing depth may result in

different numbers of molecules being detected in

different cells. Normalization of UMI count matrices

is therefore performed to address this technical

variability as a preprocessing step. Options for

normalization include log normalization of gene

expression measurements for each cell followed by

scale factor multiplication (Hao et al. 2021) as well as

an alternative method, sctransform, which takes

sequencing depth as a covariate in a generalized

linear model and yields the residuals of a regularized

negative binomial regression for use as effectively

normalized data (Hafemeister & Satija, 2019). The

sctransform modelling framework has been proposed

as a method by which to remove technical

characteristics from data while preserving cell-to-cell

biological heterogeneity.

In addition to technical effects, joint analysis of

multiple samples presents further challenges, as this

requires matching cell subpopulations across

datasets. Stuart et al. 2019 have proposed and

implemented a comprehensive strategy for

integration of single cell datasets (Stuart, Butler,

Hoffman, et al., 2019). Applying concepts from

statistical learning, their approach combines single

cell datasets through the application of canonical

correlation analysis (CCA) and mutual nearest

HEALTHINF 2022 - 15th International Conference on Health Informatics

220

neighbors profiling for the task of identifying

‘anchors’, pairwise correspondences of cell states

between datasets. These anchor correspondences,

once identified, are then used to transform multiple

UMI tables into a shared space for integrated

comparisons -- also the transformation derived from

creating an integrated reference can then be applied

to other query datasets for aligned comparisons.

Integrated reference datasets may be used to

transfer predicted cell type labels to a query sample,

efficiently labelling newly processed data. Another

possible use for these methods, however, is

exploration of research datasets to examine whether

cell transcriptomic states identified in one dataset

may resemble particular subpopulations in a query

data set. Here, we present such an analysis, using an

integrated reference constructed from Control, MSA,

and PD putamen data with Control substantia nigra

data as our query.

1.2 Problem Formulation and Aims

Clustering of integrated snRNAseq data is performed

after filtering, normalization, and integration. The

features (genes) in the integrated data with the highest

cell to cell variation are used for clustering. PCA is

performed on the subset of these highly variable

genes for dimension reduction, followed by

unsupervised clustering using optimization of a

modularity function for different parameter settings

to generate cluster solutions (Hao, et al., 2021).

Seurat version 4, the software package used in this

analysis, uses a graph-based clustering approach that

includes selection of a resolution parameter based on

stratification of cell-specific feature (gene

expression) markers among identified clusters. When

working with a single integrated dataset, cluster

identities can be assigned by feature differential

expression comparisons, with subsets of genes whose

expression is particular to certain clusters used for

annotation of cell types.

When clustering a single or integrated dataset,

feature expression patterns selected by the above

workflow for clustering (and the final cluster

solutions themselves) will depend on sample

composition. For example, analyzing an snRNAseq

dataset with many neurons and few other glial cell

types will result in the selection of highly variable

genes which differ among neurons as the most highly

variable features and clustering will likely separate

more neuron subclusters than other types. Using such

a neuron-predominant reference in a query dataset

without many neurons would have limitations, in that

the features selected for use in the transformation may

Figure 1: Data analysis workflow schematic.

not perform well for identifying non-neuronal cell

subpopulation clusters. Thus, an issue to consider

when using snRNAseq label transfer for exploratory

research as presented here is whether cell type

compositions may be similar. For profiling

oligodendrocytes in substantia nigra using a putamen

reference, then, a first step to be taken before

examining the results of label transfer is to assess the

generalizability of the reference to the query dataset.

Therefore, we include in this workflow an annotation

of nuclei population types for both putamen and

substantia nigra to first confirm that oligodendrocyte

nuclei are being broadly correctly identified by label

transfer predictions. We then explore the results of

label transfer among oligodendrocyte subpopulations

as the next stage of the analysis.

2 METHODS AND PROCEDURES

Post-mortem fresh-frozen unfixed human putamen

samples were each obtained through partnerships

with licensed organizations with completed pre-

mortem consent for donation and ethical committee

approval for sample acquisition and use (Teeple et al,

2021). Samples used for single-nucleus RNA

sequencing (snRNA-seq) were putamen tissue

sections from nine human donors (n =3 per group,

Integrated Label Transfer for Oligodendrocyte Subpopulation Proﬁling in Parkinson’s Disease and Multiple System Atrophy

221

Figure 2: Putamen sample nuclei integrated and clustered.

Plots are UMAP of principle components coloured by

cluster identity. Expression levels for type-specific markers

are shown in violin plots by cluster.

mean age in years ± SD: Control, 78.7 ± 9.5; PD, 79.7

± 5.5; MSA, 65.0 ± 10.6).

Nuclei Isolation: Samples were stored at -80°C. For

tissue lysis and washing of nuclei, tissue sections

were added to 1 mL lysis buffer (Nuclei PURE lysis

buffer, Sigma) and thawed on ice. Samples were then

Dounce homogenized with PestleAx20 and

PestleBx20 before transfer to a new tube, with the

addition of additional lysis buffer. Following

incubation on ice for 15 minutes, samples were then

filtered using a 30 mM MACS strainer (MACS

strainer, Fisher Scientific), centrifuged at 500xg for 5

minutes at 4°C using a swinging bucket rotor (Sorvall

Legend RT, Thermo Fisher), and then pellets were

washed with an additional 1 mL cold lysis buffer and

incubated on ice for an additional 5 minutes. Samples

were then centrifugated at 500g for 5 minutes at 4°C

and then were resuspended in 1mL Nuclei PURE

Storage Buffer (Nuclei PURE storage buffer, Sigma).

Sample washing was performed until the supernatant

cleared. A final resuspension was then prepared in

0.6mL wash buffer before NeuN/Dapi staining and

FACS sorting was performed. For NeuN/Dapi and

FACS sorting, from 0.6 mL nuclei sample, 540 mL,

30 mL, and 30 mL were aliquoted into tubes for

sample and controls and then 10X Dapi/NeuN buffer

was added to tubes for a final 1X concentration.

Tubes were then incubated on ice for 30 minutes, with

inversion every 10 min. Following incubation,

samples were spun at 500xg for 5 min, supernatant

removed, and samples were resuspended in 600 ul

Wash buffer for samples (300 ul for control tubes).

Nuclei then underwent filtering and sorting using BD

Bioscience InFlux Cell Sorter.

Library Preparation and NovaSeq Sequencing:

Libraries were prepared according to 10xGenomics

protocol for Chromium Single Cell 3’ Gene

Expression V3 kit. NovaSeq sequencing was

performed according to illumine NovaSeq 6000

protocol. UMI count matrices generated by

Cellranger V3.0.2.

2.1 Sample Integration and Annotation

A workflow schematic for data integration and

analysis steps for putamen and substantia nigra

samples in shown in Fig. 1.

Putamen: Summary information for final UMI count

matrices for nuclei by individual samples together

with nucleus barcodes and gene labels were loaded

with R version 4.0.0/RStudio for sample integration

and unsupervised clustering using Seurat Package

version 4.0.1. For Quality Control (QC), nuclei were

filtered following standard protocols based on

examination of violin plots. Cutoffs 200 <

nFeature_RNA < 9000 and percent.mt < 5 were used.

Filtered matrices were individually normalized by

sample according to Seurat workflows for

SCTransfom. After quality filtering, 87,086 total

nuclei were included in the final dataset. Sample

integration was performed using the R package Seurat

using the FindIntegrationAnchors and IntegrateData

functions for 3000 variable features. Clustering

resolution 0.5 and 30 dimensions were used for the

final clustering. Broad type annotations were

assigned based on expression of canonical markers:

oligodendrocyte precursor cell (OPC; VCAN),

oligodendrocyte (OLIGO; MOG, MBP), neuron

(NEUR; RBFOX3, SNAP25, GAD1, GAD2,

NRGN), astrocyte (ASTRO; GFAP, AQP4, GJA1),

microglia (MICRG; CSF1R), and vascular

leptomeningeal cells (VLMC; SLC6A13). UMAP

HEALTHINF 2022 - 15th International Conference on Health Informatics

222

Figure 3: Substantia nigra sample nuclei integrated and

clustered. Plots are UMAP of principle components

coloured by cluster identity. Expression levels for type-

specific markers are shown in violin plots by cluster.

plots for cluster assignments together with violin

plots for expression of broad types markers, and types

annotations are shown in Fig. 2.

Substantia Nigra: Data was downloaded from

supplementary files from Agarwal et al. 2020 for

substantia nigra samples (5943 nuclei) obtained from

five human donors and sequenced using the 10x

Genomics Chromium Platform (Agarwal, et al.,

2020). These files are accessible online through the

NCBI interface: https://www.ncbi.nlm.nih.gov/geo/

query/acc.cgi?acc=GSE140231. Integration and

broad cell type annotations were performed for

Substantia Nigra data separately in R using the Seurat

package, version 4.0.1. A similar workflow as for

putamen samples was followed. Pre-processing cut-

offs were selected based on initial QC plots: 200<

nFeatureRNA < 7500 and percent.mt<5. Data were

normalized at the individual sample level using

SCTransform and then integrated using

FindIntegrationAnchors and IntegrateData as

described in the Seurat data integration workflow

with 3000 variable features and cluster resolution 0.5.

The number of PCs used for clustering (n = 30) was

chosen to optimize separation between clusters.

Broad cell types were assigned for each cluster based

on marker expression levels as for putamen (Fig. 3).

2.2 Differential Gene Expression

Differentially expressed genes for PD versus Control

and MSA versus Control were identified within each

cluster using the Seurat FindMarkers() function and

the MAST package (Finak, McDaid, Yajima, et al.,

2015) for differential gene expression analysis

comparisons. Pathway enrichment analysis for

differentially expressed genes was performed using

Qiagen Ingenuity Pathway Analysis (IPA) software

(Kramer, Green, Pollard, & Tugendreich, 2014) using

adjusted p-value<0.05 and abs(log2 fold change)

cutoff 0.35. For identification of cluster marker

genes, FindMarkers was used with the MAST

package for the comparison of the selected cluster

versus all other nuclei. Functional enrichments for

markers were queried using the Enrichr platform

(Xie, Bailey, Kuleshov, et al., 2021).

Table 1: Broad Cell Types Proportions.

Cell Type

Tissue Source -

Condition

Mean

Proportion ±

Standard

Deviation

Oligodendrocyte Putamen - Control

Putamen – PD

Putamen - MSA

Subst. Nigra - Control

66.5±14.3

64.2±24.5

64.6±13.2

63.8±16.9

Neuron Putamen - Control

Putamen – PD

Putamen - MSA

Subst. Nigra - Control

14.5±6.6

13.9±14.8

18.7±12.1

5.5±5.5

Astrocyte Putamen - Control

Putamen – PD

Putamen - MSA

Subst. Nigra - Control

9.4±5.9

12.6±7.4

7.4±1.4

16.0±8.6

Microglia Putamen - Control

Putamen – PD

Putamen - MSA

Subst. Nigra - Control

4.1±0.8

6.3±2.1

4.8±1.2

5.4±3.7

OPC Putamen - Control

Putamen – PD

Putamen - MSA

Subst. Nigra - Control

5.1±1.6

2.4±0.3

3.8±1.5

8.4±4.5

VLMC Putamen - Control

Putamen – PD

Putamen - MSA

Subst. Nigra - Control

0.4±0.4

0.7±0.5

0.9±0.4

Integrated Label Transfer for Oligodendrocyte Subpopulation Proﬁling in Parkinson’s Disease and Multiple System Atrophy

223

2.3 Label Transfer

Label transfer was performed in Seurat using the

FindTransferAnchors and TransferData functions to

predict substantia nigra nuclei type for broad cell

types and subpopulation clusters as identified in

putamen reference samples. Accuracy of broad types

classifications was calculated using annotations made

for the substantia nigra dataset as ground truth. To

examine which cluster markers identify

oligodendrocytes in the 4-OLIGO subcluster, the

function FindMarkers, using MAST for differential

expression testing, was applied for cluster 4-OLIGO

in Control putamen samples and for nuclei predicted

to belong to 4-OLIGO in Control substantia nigra in

comparison to all other sample nuclei.

3 RESULTS

3.1 Broad Cell Types

After quality filtering, nuclei from human putamen

tissue samples included Control (n = 3 donors; 22,297

nuclei), PD (n = 3 donors; 32,301 nuclei), and MSA

(n = 3 donors; 32,488 nuclei). Data for nuclei from

substantia nigra were Control (n = 5 donors; 6,018

nuclei). For both putamen and substantia nigra

samples, oligodendrocytes were found to be the

dominant cell type. Table 1 presents a summary of

broad cell types proportions for each tissue type and

condition.

3.2 Oligodendrocytes in Putamen

Unsupervised clustering of integrated putamen

sample data identified eight oligodendrocyte clusters

from their transcriptomic features. Pathway

enrichment analysis for differentially expressed

genes in PD versus Control oligodendrocyte nuclei

and MSA versus Control oligodendrocyte nuclei

revealed differences in gene expression changes

between PD and MSA. In IPA comparison pathway

enrichment analysis, more prominent differences in

expression of genes linked with unfolded protein

responses and stress signalling were observed in PD

oligodendrocytes (Fig. 4). SNCA expression among

cell clusters was also compared, revealing

oligodendrocyte clusters 4-OLIGO and 5-OLIGO as

subpopulations with the most pronounced increases

in SNCA expression in PD while this expression

pattern was absent in MSA and in Control

oligodendrocytes (Fig. 5).

Figure 4: Pathway enrichments for oligodendrocyte nuclei

differentially expressed genes. (grey dot: p-adj>0.05).

3.3 Predicted Cell Types

Using profiled putamen nuclei as a reference, the

accuracy of oligodendrocyte nuclei classification by

label transfer for substantia nigra oligodendrocytes

was 98%. A summary of accuracy across all cell types

is shown in Fig. 6. Prediction of oligodendrocyte

subtypes was then performed, which was found to

Figure 5: Comparative proportions and average expression

of SNCA in oligodendrocyte lineage clusters.

HEALTHINF 2022 - 15th International Conference on Health Informatics

224

Figure 6: Prediction of broad cell types from putamen

reference and confusion matrix with class accuracies.

identify a greater than expected number of

oligodendrocytes as the 4-OLIGO type. Predicted

subpopulations are identified in label transfer based

on similarities in gene expression patterns, and these

gene expression patterns may be functionally

annotated by gene set enrichment analysis relative to

pathway and function annotation references. Marker

genes are genes differentially expressed within a

cluster relative to alll other nuclei. We compared

cluster markers for 4-OLIGO nuclei in putamen with

marker genes for the predicted 4-OLIGO subcluster

of substantia nigra. Remarkably, 302 genes were

identified as shared markers for both the 4-OLIGO

cluster in putamen and the predicted 4-OLIGO nuclei

in substantia nigra. Functional enrichments for the

putamen 4-OLIGO gene set and the common gene

markers for predicted cluster nuclei are shown in Fig.

7. Prominent among these enriched pathways and

functions are microtubule binding and folding.

Figure 7: Predictions shown in UMAP project and predicted

nuclei population proportions. Overlap of markers for 4-

OLIGO cluster and functional enrichments.

Integrated Label Transfer for Oligodendrocyte Subpopulation Proﬁling in Parkinson’s Disease and Multiple System Atrophy

225

4 CONCLUSIONS

Recent methods developed for single cell and single

nucleus sequencing have enabled more

comprehensive studies and profiling of

oligodendrocytes in different brain diseases (Agarwal

et al., 2020; Smajic, et al., 2020; Jakel, Agirre, Falcao,

et al., 2019) as well provided new avenues for

exploratory and comparative analyses. In this study,

we apply label transfer methods to generalize our

disease versus control putamen region comparisons to

another brain region to newly identify an expanded

subpopulation of transcriptionally similar nuclei in

substantia nigra. While we observe a greater

predicted proportion of an oligodendrocyte subtype in

substantia nigra which is identified as overexpressing

SNCA in PD in putamen, it remains to be further

understood how functional activities in

oligodendrocyte subpopulations relate to α-synuclein

biology and synucleinopathy disease processes.

Oligodendroglial cytoplasmic inclusions of α -

synuclein protein are described as the predominant

neuropathological finding in MSA; neuronal α -

synuclein aggregates are described as being more

prominent in PD, although varying degrees of

neuronal and oligodendroglial involvement are

reported in both disorders (Jellinger, 2018;

Henderson, Trojanowski, & Lee, 2019. Gillman,

Wenning, Low, et al., 2008). SNCA mutations,

duplications, and triplications have been causally

linked with familial PD in multiple studies (Ibanez,

Bonnet, Debarges, et al., 2004; Polymeropoulos,

Lavedan, Leroy, et al., 1997; Singleton, Farrer,

Johnson, et al., 2003). While genetic variants within

the SNCA locus have also been associated with MSA

in a few studies (Scholz, Houlden, Schulte, et al.,

2009; Kiely, Asi, Kara, et al., 2013), the connection

between SNCA overexpression in oligodendrocytes

and MSA is less clear. Cell-to-cell transmission of

highly pathogenic misfolded α-synuclein proteins

from neurons to oligodendrocytes has been

hypothesized as one potential explanation for the

prominent oligodendroglial inclusions observed in

MSA (Peng, Gathagan, Covell, et al., 2018). Our

observation of lower levels of oligodendrocyte SNCA

expression in MSA versus PD may lend some further

support to this theory. Yet it remains to be further

explored how increased SNCA expression in

oligodendrocytes may relate to PD pathogenesis and

how disease mechanisms may vary between PD,

MSA, and other synucleinopathies.

Our observation of different oligodendrocyte

transcriptional changes suggests that PD and MSA,

while both synucleinopathies, may differ in their

pathological mechanisms. Future work, including the

analysis of greater numbers of patient samples is

needed to verify and generalize our observations. In

addition, further studies are needed to examine how

gene expression changes relate to protein levels by

orthogonal analytic methods. The oligodendrocyte

subpopulations profiled here exhibit distinctive

functional activities which may offer promising

therapeutic targets for these debilitating and often

lethal diseases.

ACKNOWLEDGEMENTS

We thank Dr. Srinivas Shankara for critical review

and insightful feedback on this paper.

This work was supported by Sanofi. E.T., P.J., R.P.,

Y.H., A.K., M.L.M., S.P.S., A.C.M., K.W.K.,

A.F.M., S.L.M., D.K.R., and D.K. are employees of

Sanofi and may hold shares and/or stock options in

the company.

REFERENCES

Coon, E.A., Cutsforth-Gregory, J.K. Benarroch, E.E.

(2018). Neuropathology of autonomic dysfunction in

the synucleinopathies. Movement Disorders, 33(3):

349-58.

Tysnes, O., Storstein, A. (2017). Epidemiology of

Parkinson’s disease. Journal of Neural Transmission,

124: 901-5.

Vanacore, N. Bonifati, V., Fabbrini, G., Colosimo, C., et.

al. (2001). Epidemiology of multiple system atrophy.

ESGAP Consortium. European Study Group on

Atypical Parkinsonisms. Neurol Sci., 22(1): 97-9.

Spillantini, MG, Schmidt, ML, Lee, VM, Trojanowski JQ,

Jakes, R, Goedert, M. (1997). Alpha-synuclein in Lewy

bodies. Nature, 388(6645):839-40.

Inoue, M., Yagishita, S., Ryo, M., Hasegawa, K., Amano,

N., Matsushita, M. (1997). The distribution and

dynamic density of oligodendroglial cytoplasmic

inclusions (GCIs) in multiple system atrophy: a

correlation between the density of GCIs and the degree

of involvement of striatonigral and

olivopontocerebellar systems. Acta Neuropathol,

93(6):585-91.

Hague, K., Lento, P., Morgello, S., Caro, S., Kaufmann, H.

(1997). The distribution of Lewy bodies in pure

autonomic failure: autopsy findings and review of the

literature. Acta Neuropathol, 94(2):192-6.

Del Tredici, K., Braak, H. (2016). Review: Sporadic

Parkinson's disease: development and distribution of α-

synuclein pathology. Neuropathol Appl Neurobiol.

42(1):33-50.

Maroteaux, L., Campanelli, J.T., Scheller, R.H. (1988).

Synuclein: a neuron-specific protein localized to the

HEALTHINF 2022 - 15th International Conference on Health Informatics

226

nucleus and presynaptic nerve terminal. J Neurosci,

8(8):2804-15.

Pinho, R., Paiva, I., Jercic, K.G., Fonseca-Ornelas, L., et al.

(2019). Nuclear localization and phosphorylation

modulate pathological effects of alpha-synuclein. Hum

Mol Genet, 28(1):31-50.

Goers, J., Manning-Bog, A.B., McCormack, A.L., Millett

I.S., et al. (2003). Nuclear localization of alpha-

synuclein and its interaction with histones.

Biochemistry. 22;42(28):8465-71.

Bryois, J., Skene, N.G., Hansen, T.F., Kogelman, L.J.A., et

al., Genetic identification of cell types underlying brain

complex traits yields insights into the etiology of

Parkinson's disease. Nat Genet, 2020. 52(5): p. 482-

493.

Smajić, S., Prada-Medina, C.A.., Landoulis, Z.., Dietrich,

C., et al., (2020). Single-cell sequencing of the human

midbrain reveals glial activation and a neuronal state

specific to Parkinson’s disease. medRxiv. (pre-review)

Nalls, M.A., Blauwendraat, C., Vallerga, C.L., Heilbron,

K., et al. (2019). Identification of novel risk loci, causal

insights, and heritable risk for Parkinson's disease: a

metaanalysis of genome-wide association studies.

Lancet Neurol., 18(12):1091-102.

Reynolds, R.H., Botia, J., Nalls, M.A., et al. (2019).

Moving beyond neurons: the role of cell type-specific

gene regulation in Parkinson's disease heritability. NPJ

Parkinsons Dis, 5:6.

Teeple, E., Jindal, K., Kiragasi, B., Annaldasula, S., et al.,

(2020). Network Analysis and Human Single Cell

Brain Transcriptomics Reveal Novel Aspects of Alpha-

Synuclein (SNCA) Biology. BioRxiv. (pre-review).

Agarwal, D., Sando, C., Volpato, V., Caffrey, T.M., et al.,

(2020). A single-cell atlas of the human substantia nigra

reveals cell-specific pathways associated with

neurological disorders. Nat Commun, 11(1): 4183.

Teeple, E., Joshi, P., Pande, R., Huang, Y. et al., (2021).

Single Nuclei Sequencing of Human Putamen

Oligodendrocytes Reveals Altered Heterogeneity and

Disease-Associated Changes in Parkinson’s Disease

and Multiple System Atrophy. BioRxiv (pre-review).

Duran, R.C.D., Wei, H., Wu, J.Q. (2017). Single cell RNA

sequencing of the brain. Clinical and Translational

Medicine, 6:20. DOI 10.1186/s40169-017-0150-9.

Hao, Y., Hao, S., Andersen-Nissen, E., Mauck, W.M., et al.

(2021). Integrated analysis of multimodal single-cell

data, Cell, 184(13): 3573-87.

Hafemeister, C., Satija, R. (2019). Normalization and

variance stabilization of single-cell RNA-seq data using

regularized negative binomial regression. Genome Biol,

20: 296.

Stuart, T., Butler, A., Hoffman, P., Hafemeister, C., et al.

(2019). Comprehensive Integration of Single-Cell Data.

Cell, 13:177(7):1888-1902.e21.

Finak, G., McDavid, A., Yajima, M., Deng, J., et al., (2015).

MAST: a flexible statistical framework for assessing

transcriptional changes and characterizing

heterogeneity in single-cell RNA sequencing data.

Genome Biol, 16:278.

Kramer, A., Green, J., Pollard, J., Tugendreich, S. (2014).

Causal analysis approaches in Ingenuity Pathway

Analysis. Bioinformatics, 30(4):523-30.

Xie, Z., Bailey, A., Kuleshov, M.V., Clarke, D.J.B., et al.

(2021). Gene set knowledge discovery with Enrichr.

Current Protocols, 1: e90.

Jakel, S., Agirre, E., Falcao, A.M., van Bruggan, D., et al.,

(2019). Altered human oligodendrocyte heterogeneity

in multiple sclerosis. Nature, 566(7745): 543-547.

Jellinger, K.A. (2018). Multiple System Atrophy: an

oligodendroglialsynucleinopathy. J Alzheimers Dis,

62(3):1141-1179.

Henderson, M.X., Trojanowski, J.Q., Lee, V.M. (2019).

alpha-Synuclein pathology in Parkinson's disease and

related alpha-synucleinopathies. Neurosci Lett, 709:

134316.

Gilman, S., Wenning, G.K., Low, P.A., Brooks, D.J., et al.

(2008). Second consensus statement on the diagnosis of

multiple system atrophy. Neurology, 71(9): 670-6.

Ibanez, P., Bonnet, A.M., Debarges, B., Lohmann, E., et al.

(2004). Causal relation between alpha-synuclein gene

duplication and familial Parkinson's disease. Lancet,

364(9440): 1169-71.

Polymeropoulos, M.H., Lavedan, C., Leroy, E., Ide, S.E., et

al. (1997). Mutation in the alpha-synuclein gene

identified in families with Parkinson's disease. Science,

276(5321): 2045-7.

Singleton, A.B., Farrer, M., Johnson, J., Singleton, A., et

al., (2003). alpha-Synuclein locus triplication causes

Parkinson's disease. Science, 302(5646): 841.

Scholz, S.W., Houlden, H., Schulte, C., Sharma, M., et al.

(2009). SNCA variants are associated with increased

risk for multiple system atrophy. Ann Neurol, 65(5):

610-4.

Kiely, A.P., Asi, Y.T., Kara, E. Limousin, P., et al. (2013).

alpha-Synucleinopathy associated with G51D SNCA

mutation: a link between Parkinson's disease and

multiple system atrophy? Acta Neuropathol, 125(5):

753-69.

Peng, C., Gathagan, R.J., Covell, D.J., Medellin, C., et al.

(2018). Cellular milieu imparts distinct pathological

alpha-synuclein strains in alpha-synucleinopathies.

Nature, 557(7706): 558-563.

Integrated Label Transfer for Oligodendrocyte Subpopulation Proﬁling in Parkinson’s Disease and Multiple System Atrophy

227