Identification and Quantification of Oligodendrocyte Precursor Cells in Multiple System Atrophy, Progressive Supranuclear Palsy and Parkinson's Disease

Abstract

Multiple system atrophy is a neurodegenerative disorder characterized pathologically by abnormal accumulations of α‐synuclein in the cytoplasm of oligodendrocytes, which are termed glial cytoplasmic inclusions (GCIs). Oligodendrocytes are responsible for myelinating axons and providing neurotrophic support, but in MSA, myelin loss, axonal loss and gliosis are consistent features suggesting that GCIs play a central role in disease pathogenesis. Oligodendroglial, myelin and axonal degeneration are also features of multiple sclerosis (MS) in which recent studies have highlighted the robust remyelination capacity of the central nervous system (CNS). The cells responsible for remyelination are called oligodendroglial precursor cells (OPCs). In this study, we investigated the role of OPCs in the pathogenesis of MSA and progressive supranuclear palsy (PSP), a neurodegenerative disease in which neuropathological changes include oligodendroglial inclusions composed of microtubule‐associated protein tau. Despite the lability of OPC‐specific antigens, we successfully identified OPCs and demonstrated that tau and α‐synuclein do not accumulate in OPCs. We also showed that the density of OPCs was increased in a white matter region of the MSA brain, which is also severely affected by GCIs and myelin degeneration. These findings raise the possibility that OPCs could be available to repair disease‐associated damage in MSA, consistent with their biological function.

Introduction

Multiple system atrophy (MSA) is a sporadic, neurodegenerative disorder characterized by a combination of Parkinsonism, cerebellar ataxia and autonomic dysfunction 26. Histologically, the MSA brain shows atrophy, neuronal loss and gliosis, changes that are particularly severe in the striatonigral (StrN) and/or the olivopontocerebellar (OPC) structures 8, leading to the neuropathological stratification of MSA into StrN predominant, OPC predominant and mixed StrN/OPC subtypes 16. In contrast, the cerebral cortex is only mildly affected 19. Neurodegeneration is associated with the accumulation of α‐synuclein in the cytoplasm of oligodendrocytes 27, termed glial cytoplasmic inclusions (GCIs) or Papp‐Lantos bodies, the hallmark lesion of MSA 18. This accumulation of α‐synuclein, which is also abnormally phosphorylated and biochemically insoluble 8, classifies MSA as an α‐synucleinopathy. Although α‐synuclein inclusions also occur within neurons in MSA, GCIs are thought to play a central role in disease pathogenesis 28. In MSA, white matter tracts commonly display signs of myelin loss, axonal loss and gliosis 28, but it is currently unknown if the myelin‐related dysfunction is a primary event, which leads to axonal damage or a secondary event caused by loss of neurons and axons 13. Data provided by animal models of MSA with oligodendroglial overexpression of α‐synuclein and GCI‐like inclusions suggest that myelin degeneration precedes axonal degeneration 10, 29.

Multiple sclerosis (MS) is considered an inflammatory demyelinating disease of the central nervous system (CNS) and shows some similarities with MSA in that there is also evidence of oligodendroglial involvement, myelin degeneration, axonal loss and gliosis 24. Using an experimental rat model of inflammatory demyelination 17, it was shown that α‐synuclein expression is up‐regulated in neurons, astrocytes and oligodendrocytes in response to demyelination; a similar result was also seen in human MS tissue. The CNS has an innate ability for remyelination and this has been an area of intense research interest in the field of MS 6, 11. The cells responsible for this process are called oligodendrocyte precursor cells (OPCs) and they represent the fourth distinct class of glia in the CNS. OPCs have a stellate morphology with delicate, ramified processes and are evenly distributed throughout the CNS (grey and white matter) where they represent 5%–10% of all glia 25. Compared with mature oligodendrocytes, OPCs have more intense expression of Olig2 9, which is an oligodendrocyte lineage nuclear transcription factor 30. More specifically, OPCs can be identified by their cell surface expression of the chondroitin sulfate proteoglycan, NG2 and, therefore, are commonly called NG2 cells. Functionally, OPCs are proliferative and motile cells that, as their name suggests, give rise to mature myelinating oligodendrocytes not only during development but also during adulthood and in response to CNS injury or disease 6, 11. Most of the studies exploring the remyelinating capacity of NG2 cells have used experimental rodent models 15 as the NG2 protein is sensitive to factors such as post‐mortem delay, type of fixation and duration of fixation, which makes detection in human tissue more challenging 25. Despite these difficulties, NG2 cells have been identified in human brain in MS and are found in demyelinating lesions 11, 25.

The role of OPCs has not been previously studied in MSA; therefore, in this study, we have investigated the presence of NG2 cells and their relationship to the characteristic neuropathological features of MSA, including GCIs, myelin degeneration, axonal loss and gliosis. Along with normal controls, we also used Parkinson's disease (PD) as another α‐synucleinopathy with mainly neuronal inclusions and only sparse glial inclusions. In progressive supranuclear palsy (PSP), there are extensive neuronal, astrocytic and oligodendroglial inclusions composed of the microtubule‐associated protein tau 5. As oligodendroglial inclusions are a prominent feature in PSP, we chose this as a further disease control for comparison with MSA.

Materials and Methods

Case material

The brain tissue used in this study was donated between 1992 and 2006 to the Queen Square Brain Bank for Neurological Disorders located at UCL Institute of Neurology, London, according to the protocols approved by a London Ethics Committee and stored under license issued by the Human Tissue Authority (No. 12198). For each brain, one‐half was fixed in 10% buffered formalin and the other half was dissected, flash‐frozen and stored at −80°C. The formalin‐fixed half was used for routine pathological diagnosis according to the accepted diagnostic neuropathological criteria; normal controls had only mild age‐related changes.

Initially, we selected cases that had flash‐frozen tissue and those with a short (<12 h) post‐mortem interval (PMI), but in order to increase the sample size, some cases had a longer PMI (up to 50 h). As shown in Table 1, a total of nine MSA (subtypes = four mixed; three StrN; two OPC), six PSP, four PD and five normal cases (with only aging‐related changes) were selected for study. It is of note that a groupwise comparison showed no significant difference in PMI between groups (P = 0.14). Similarly, MSA and PSP cases were matched for disease duration (P = 0.81). For each case, 10‐μm frozen sections of the anterior frontal lobe, striatum at the level of the nucleus accumbens and cerebellum including the dentate nucleus were cut using a cryostat. Sections were mounted, briefly dried on a slide‐warmer (set at 40°C) and then stored at −20°C before use. Cerebellar tissue at the appropriate level was unavailable for three MSA cases and one PSP case.

Table 1.

Demographic data for cases used in this study. Abbreviations: F = female; M = male; MSA = multiple system atrophy; NA = not available; OPC = olivopontocerebellar predominant; PD = Parkinson's disease; PMI = post‐mortem interval; PSP = progressive supranuclear palsy; StrN = striatonigral; –— = not applicable

Pathological diagnosis	Sex	Age at death (years)	Duration (years)	PMI (h)
MSA (mixed)	M	61	6	14
MSA (mixed)	F	67	6	8
MSA (mixed)	F	50	7	4
MSA (mixed)	M	69	10	14
MSA (OPC)	F	57	5	26
MSA (OPC)	F	82	5	40
MSA (StrN)	M	63	4	12
MSA (StrN)	F	58	7	12
MSA (StrN)	M	53	12	6
PSP	F	60	5	8
PSP	M	65	6	7
PSP	F	81	11	10
PSP	M	68	12	5
PSP	M	70	8	14
PSP	M	59	3	19
PD	M	88	26	3
PD	M	78	9	3
PD	M	71	23	5
PD	F	71	NA	50
Normal	F	77	—	8
Normal	F	78	—	31
Normal	M	88	—	32
Normal	M	71	—	39
Normal	F	76	—	30

Immunohistochemistry (IHC)

Frozen sections were thawed and dried at room temperature (RT) before washing in phosphate‐buffered saline (PBS); all incubations were performed at RT and wash steps/dilutions used PBS, unless otherwise stated. IHC was performed using the primary antibodies summarized in Table 2. In brief, sections were treated with 0.3% H₂O₂ for 10 minutes to eliminate endogenous peroxidase activity before washing. To block non‐specific binding, sections were incubated with normal goat serum (1:20; Vector Laboratories, Burlingame, CA, USA) for 30 minutes. Primary antibodies were applied for 60 minutes. For secondary detection of the NG2 antibody, the Dako REAL^TM EnVision^TM detection system (Dako, Cambridgeshire, UK) was used according to the manufacturer's instructions. For the other primary antibodies, sections were washed and incubated with either biotinylated goat anti‐mouse IgG or goat anti‐rabbit IgG (1:200; Vector Laboratories) secondary antibodies for 30 minutes. After washing, sections were incubated in avidin‐biotin complex (ABC) solution (Vector Laboratories) for 30 minutes, followed by washing and development using 3,3′‐diaminobenzidine as the chromogen. Sections used for NG2, α‐synuclein and AT8 staining were counterstained with hematoxylin before cover‐slipping.

Table 2.

Primary antibodies used for immunohistochemistry and immunofluorescence

Antibody	Epitope/Antigen	Species	Dilution	Source
NG2	Human NG2	Mouse	1:200	R&D Systems
α‐Synuclein	α‐Synuclein (amino acids 111–131)	Rabbit	1:800	Abcam
Olig2	Oligodendrocyte transcription factor 2	Rabbit	1:5000	Abcam
AT8	PHF‐Tau (Ser202)	Mouse	1:800	Autogen Bioclear
Tau	Tau (amino acids 243–441)	Rabbit	1:800	Dako
GFAP	Glial fibrillary acidic protein	Rabbit	1:1000	BioGenex
Iba‐1	Ionized calcium binding adaptor 1	Rabbit	1:1000	Wako Chemicals
MBP	Myelin basic protein	Mouse	1:4000	Dako
NFC	Neurofilament cocktail	Mouse	1:100	MP Biomedicals

Double immunofluorescence (DI) staining

DI staining was performed using NG2 and Olig2, glial fibrillary acidic protein (GFAP), Iba‐1, tau or α‐synuclein antibodies (Table 2). Following thawing, drying and NG2 antibody incubation steps described earlier, sections were incubated with biotinylated goat anti‐mouse IgG (1:200; Vector Laboratories) for 30 minutes, followed by washing and incubation in ABC solution for 30 minutes. After washing, sections were treated with fluorophore tyramide working solution from the TSA Plus Fluorescein System kit (PerkinElmer Life and Analytical Sciences, Shelton, CN, USA) for 15 minutes. After washing the second primary antibody was applied and incubated overnight at 4°C. Washed sections were then incubated with Alexa Fluor® 568 labeled goat anti‐rabbit IgG secondary antibody (1:500; Invitrogen, Paisley, UK) for 30 minutes, followed by washing. Sections were counterstained using Vectashield mounting medium with DAPI (Vector Laboratories) and cover‐slipped.

Qualitative analysis for co‐localization

A comprehensive qualitative analysis of DI‐stained sections was used to investigate: (i) co‐localization of NG2 with Olig2, GFAP and Iba‐1 staining; and (ii) co‐localization of NG2 with tau and α‐synuclein staining. The former was analyzed in the frontal lobe and cerebellum of a subset of control and PD cases. The latter was analyzed in the selected regions of all the MSA, PSP and PD cases, where the entire tissue section was visually and systematically scanned for co‐localization using confocal fluorescence microscopy (Leica DM5500 microscope and AF6000 3D deconvolution software, Leica Microsystems, Milton Keynes, UK); NG2 staining was green, other antibody staining was red and cell nuclei were blue.

Image analysis of pathological variables

Image analysis was performed on IHC‐stained sections to quantify: (i) the density of NG2‐positive cells, α‐synuclein‐positive and tau‐positive oligodendroglial inclusions; (ii) the percentage area of GFAP, Iba‐1 and neurofilament cocktail (NFC) immunoreactivity; and (iii) the degree of abnormal myelin basic protein (MBP) immunoreactivity using a semiquantitative scoring scheme. Blinded to pathological diagnosis, these analyses were performed in the cerebellar white matter and frontal white matter of all MSA, PSP, PD and normal cases.

The density measures were performed on a light microscope fitted with an automated XY stage and controlled by Image Pro 6.2 software (Media Cybernetics, Rockville, MD, USA). This software enabled the digital delineation of the region of interest at low magnification (40×), followed by the random placement of a grid that contained a variable number of point frames; for NG2‐stained sections, a minimum of 80 point frames were sampled, whereas for the more numerous oligodendroglial inclusions, a minimum of 50 point frames were sampled. Each point frame was then observed on‐screen at higher magnification (400×) and the number of profiles counted within the boundaries of a smaller count frame (100 μm × 100μm). The total number of profiles in the count frames was divided by the number of point frames sampled to give a density measure in 100 μm². NG2‐positive cells were distinguished from NG2‐positive blood vessel staining based on size and morphology.

The percentage area measurements were performed by macroscopically marking six dots within the same area of interest described above and for each dot taking one microscopic digital image (200×) adjacent to the dot, avoiding large blood vessels. Using ImageJ software Version 1.44p (http://rsb.info.nih.gov/ij/), these images were converted to 8‐bit and used to detect the number of positive pixels in the image using a set threshold (GFAP = 0–163; Iba‐1 = 0–150; NFC = 50–150). This yielded a percentage areal fraction for each image, which was used to calculate a mean percentage areal fraction for each region and case.

Similar to the procedure described above, six microscopic images of MBP staining (200× magnification) were acquired for each region and case. Because of variations in the intensity of MBP immunoreactivity that were unrelated to any disease processes, MBP‐stained sections were not suitable for areal fraction measurements. Instead, a four‐point scoring scheme was devised to depict the increasing patchiness of MBP staining (Supporting Information Figure S1; + = normal; ++ = mild; +++ = moderate; ++++ = severe). This scoring scheme was applied to each digital image and a mean score calculated for each region and case.

Statistical analysis

For pairwise comparisons between disease groups, Mann–Whitney U‐tests were performed, whereas for comparisons of more than two groups, Kruskal–Wallis rank sum tests were used. Associations between two quantitative variables were investigated using Spearman's rank correlation coefficient. In this exploratory study, statistical significance was set at P < 0.05 and results were not adjusted for multiple comparisons. These statistical analyses were performed using GraphPad Prism 5 (GraphPad Software, La Jolla, CA, USA). Repeatability of the density measures (NG2 and GCI density) and the semiquantitative MBP scoring scheme were assessed by intra‐rater reliability testing using an intraclass correlation coefficient (ICC). At least 10% of cases were re‐counted by the same rater and the ICC was greater than 0.84, indicating very good repeatability. ICC was performed using SPSS PAWS 18 (IBM Corporation, Armonk, NY, USA).

Results

Identification and validation of NG2‐positive OPCs

Historically, immunohistochemical identification of NG2 cells in human post‐mortem brain tissue has been difficult because of the lability of the proteins that label them specifically 25. Using flash‐frozen tissue sections, NG2 IHC identified small stellate cells with delicate branched processes consistent with the morphology of NG2‐positive OPCs (Figure 1). Blood vessel pericytes were also labeled by NG2, as previously reported 25. In all cases, NG2‐positive cells were numerous and widespread in the cerebellar white matter, frontal white matter and internal capsule at the level of the nucleus accumbens, but were harder to detect in grey matter structures because of low level background staining of the neuropil. Optimal staining was achieved in frozen sections that were unfixed, although NG2 cells could also be detected in frozen sections briefly fixed (10 min) in acetone or 4% paraformaldehyde, but not in formalin‐fixed paraffin‐embedded tissue sections with or without various antigen retrieval techniques (data not shown).

Figure 1.

Identification and validation of NG2‐positive oligodendroglial precursor cells (OPCs).  NG2‐specific immunohistochemistry using flash‐frozen post‐mortem human tissue (A) identified cells with a stellate morphology and multiple long delicate processes (B and C) suggestive of OPCs; blood vessel staining (A, arrow head) was also evident. Double immunofluorescence staining using NG2 (green) and cell‐specific markers (red), such as Olig2 (D), glial fibrillary acidic protein (GFAP, E) and ionized calcium binding adaptor 1 (Iba‐1, F), demonstrated strong expression of nuclear Olig2 in NG2‐positive cells (D, arrow), whereas astrocytes (E) and microglia (F) were NG2‐negative. Arrow heads show NG2‐positive blood vessel staining. Immunofluorescence images represent transparent projections of confocal Z‐stacks and nuclei are labeled with DAPI (blue). Supporting Information Figure S2 shows individual fluorescent channels. Scale bars: 60 μm (A), 30 μm (B,C), 25 μm (D–F).

To confirm the presence of NG2‐positive cells, DI labeling using NG2 and other cell‐specific markers, such as Olig2 (oligodendroglial lineage), GFAP (astrocytes) and Iba‐1 (microglia), was performed in the frontal region and cerebellum of PD and normal controls (Figure 1). As previously reported 9, 25, NG2‐positive cells with a stellate and branched morphology were also positive for the oligodendrocyte lineage marker, Olig2, but were negative for astrocytic and microglial markers. Olig2 immunoreactivity in NG2‐positive cells had a nuclear localization and was consistently more intense compared with nuclear‐Olig2 in NG2‐negative cells, the latter of which is likely to represent mature oligodendrocytes. Although not detected specifically, neurons were consistently NG2‐negative (data not shown).

Co‐localization of NG2 and disease‐associated markers

It is well established that mature oligodendrocytes contain α‐synuclein‐positive GCIs 8, but it was unknown if NG2‐positive OPCs were also vulnerable to GCI formation in MSA or tau accumulation in PSP. To investigate this, DI staining using NG2 and α‐synuclein or tau was performed in the frontal lobe, striatum and cerebellum of MSA and PSP cases, and assessed by confocal microscopy for co‐localization (Figure 2). Although oligodendroglial inclusions associated with MSA (GCIs) and PSP (coiled bodies) were found alongside NG2‐positive cells, α‐synuclein or tau immunoreactive lesions did not occur in NG2‐positive cells in any of the brain regions examined in MSA and PSP cases, respectively, suggesting that such inclusions were restricted to more mature oligodendrocytes. In addition to coiled bodies, PSP is also characterized by tau‐positive tufted astrocytes 5 which have an astrocytic morphology, but are negative for the astrocytic marker, GFAP 22. As NG2 cells morphologically resemble astrocytes, this raised the possibility that tufted astrocytes in PSP might represent tau accumulation in OPCs rather than astrocytes; however, tufted astrocytes in the striatum and frontal cortex of PSP cases were consistently NG2‐negative (Figure 2).

Figure 2.

Oligodendroglial precursor cells (OPCs) do not contain inclusions in multiple system atrophy (MSA) and progressive supranuclear palsy (PSP). Double immunofluorescence labeling using NG2 (green) and disease‐associated markers α‐synuclein and tau (red) demonstrated that glial cytoplasmic inclusions in MSA (A, arrows heads) and that coiled bodies (B, arrows heads) and tufted astrocytes (C, arrow head) in PSP are NG2‐negative. Furthermore, NG2‐positive OPCs (arrows) do not accumulate α‐synuclein in MSA (A) or tau in PSP (C). NG2‐positive blood vessels (B, arrow) were also observed. Nuclei are labeled with DAPI (blue) and yellow co‐localization (A) represents autoflourescent lipofuscin pigment. These images represent transparent projections of confocal Z‐stacks. Brain regions = putamen (A), cerebellar white matter (B) and caudate (C). Scale bars = 20 μm.

Quantification of NG2‐positive OPCs and other disease‐associate pathologies

OPCs have the ability to proliferate and respond to tissue injury 6, 11. To further elucidate the role of OPCs in the pathogenesis of MSA, the following pathological variables were quantified: NG2 cells (OPCs), oligodendroglial inclusions (GCIs in MSA and coiled bodies in PSP), myelin loss (MBP score), axonal density (NFC burden), astrogliosis (GFAP burden) and microgliosis (Iba‐1 burden). These variables were analyzed in the cerebellar and frontal white matters, which in MSA are thought to represent severe and mildly affected regions 19, respectively. This analysis was restricted to the white matter structures as they are particularly vulnerable to GCIs and myelin degeneration in MSA. As no statistical differences were detected between PD and normal cases, these groups were combined as controls. The results of this analysis are summarized in Table 3 and Figure 3.

Table 3.

Quantification data represented as mean values for pathological variables analyzed in the cerebellar and frontal white matters of multiple system atrophy (MSA), progressive supranuclear palsy (PSP) and control cases. Abbreviations: GFAP = glial fibrillary acidic protein; Iba‐1 = ionized calcium binding adaptor 1; MBP = myelin basic protein; NFC = neurofilament cocktail

Pathological variables	Cerebellar white matter				Frontal white matter
	MSA	PSP	Controls	P‐value	MSA	PSP	Controls	P‐value
NG2 cell density	0.21	0.16	0.13	0.06	0.16	0.19	0.20	0.34
Oligo. inclusion density	1.39	0.11	0.00	<0.01	0.38	0.07	0.00	<0.01
MBP score	1.92	1.15	1.07	0.02	1.60	1.25	1.11	0.04
NFC burden (%)	6.52	4.77	5.19	0.15	3.78	2.90	3.87	0.26
GFAP burden (%)	16.73	15.89	10.58	0.01	10.59	11.93	10.14	0.47
Iba‐1 burden (%)	2.70	2.70	1.31	0.02	2.18	2.45	0.78	0.01

Controls group includes idiopathic Parkinson's disease and normal control cases. Kruskal–Wallis rank sum tests were used for statistical testing. Numbers in italics represent the P‐value for the corresponding three numbers to the left of the P‐value.

Figure 3.

Quantification of NG2 cells and other disease‐associated pathological variables in the cerebellar and frontal white matters of multiple system atrophy (MSA), progressive supranuclear palsy (PSP) and control cases. Horizontal lines represent mean values and dotted lines represent median values. Kruskal–Wallis rank sum tests were used for statistical testing. Oligo = oligodendroglial, GFAP = glial fibrillary acidic protein; Iba‐1 = ionized calcium binding adaptor 1; MBP = myelin basic protein.

A groupwise analysis of NG2 cell density in the cerebellum showed that their density was higher in MSA, although this was just short of statistical significance (P = 0.06). A pairwise comparison between MSA and controls was significantly different (P = 0.02), whereas a similar analysis between MSA and PSP or PSP and controls was not. In the frontal white matter, the mean density of NG2 was similar to that in the cerebellar white matter, but was not significantly different between groups (P = 0.34). Mean NG2 cell density in the cerebellar and frontal white matters did not differ in the different MSA subtypes (data not shown).

Consistent with the cerebellum being more severely affected than the frontal lobe in MSA, GCI density was significantly greater (P = 0.01) in the cerebellar white matter compared with the frontal white matter. In PSP, no significant difference in coiled body density was seen between frontal and cerebellar white matters. Pairwise comparisons confirmed that the density of oligodendroglial inclusions was greater in the cerebellum (P = 0.004) and frontal lobe (P = 0.02) of MSA when compared with PSP. Consistent with their pathological diagnosis, no oligodendroglial inclusions were observed in PD or normal control cases.

The paucity of MBP staining assessed using a semiquantitative scoring scheme showed a significant difference between groups in cerebellar (P = 0.02) and frontal (P = 0.04) white matters. In the cerebellum, pairwise comparisons showed a significant difference (P = 0.01) in MBP score between MSA and controls (Supporting Information Figure S3), but only a trend (P = 0.07) with PSP. In the frontal white matter, pairwise comparison showed a significant difference (P = 0.02) between MSA and controls, but not PSP. Qualitatively, the cerebellar white matter in most MSA cases had patchy MBP staining, which was less prominent in the frontal white matter (Figure 4). In PSP, such changes were milder in frontal white matter and absent in the cerebellar white matter, whereas they were altogether absent in control cases, suggesting that these changes in MBP relate to myelin loss. The density of axons, determined by burden analysis of NFC‐stained sections, showed no significant differences between groups in the frontal (P = 0.26) or the cerebellar (P = 0.15) white matters.

Figure 4.

Myelin basic protein (MBP) specific immunohistochemistry. A–C represent the white matter of the cerebellum and D–F the frontal lobe regions of a representative multiple system atrophy (MSA, A and D), progressive supranuclear palsy (PSP, B and E) and control (C and F) case. The cerebellar white matter of MSA cases was characterized by moderate to severe patchiness in MBP staining (A), which was less prominent in the frontal white matter (D). The cerebellar and frontal white matters of PSP cases (B and E) and control cases (C and F) were normal in comparison. Scale bar = 50 μm.

Astrogliosis determined by the mean GFAP burden (%) was significantly different between groups in the cerebellar white matter (P = 0.01), but not in the frontal white matter (P = 0.47). Pairwise comparison showed an increase in cerebellar astrogliosis in both MSA (P = 0.03) and PSP (P = < 0.001) compared with controls. In contrast, microgliosis determined by the mean Iba‐1 burden (%) was significantly different between groups in the cerebellar (P = 0.02) and frontal (P = 0.01) white matters. In both regions, the microglial burden was significantly higher in MSA compared with controls (P = < 0.01); microglial burden was also higher in PSP compared with controls, but this did not reach statistical significance (P = 0.06). Examples of such gliosis in MSA and controls are shown in Supporting Information Figure S3.

Correlations between pathological and clinical indices

Correlation analysis was performed to determine if the density of NG2 cells was influenced by the pathological variables described earlier; this analysis was restricted to MSA and PSP disease groups that were pooled. In the cerebellar white matter, the NG2 cell density showed a significant high correlation with oligodendroglial inclusion density (P = 0.04; r = 0.63), suggesting that NG2 cells increase as the number of oligodendroglial inclusions increase. No significant correlations were observed in the frontal white matter. When MSA and PSP were analyzed separately, the correlation coefficients remained positive (r = 0.31 and 0.60, respectively), but were not as strong or statistically significant.

This type of analysis also detected a correlation between the density of inclusions and MBP score in the cerebellar white matter of MSA cases (r = 0.84). Short of statistical significance (P = 0.06), it suggested that higher numbers of GCIs were associated with decreased MBP immunoreactivity. Although not statistically significant, the cerebellar and frontal white matter of MSA cases showed a negative correlation between MBP score and Iba‐1 burden (r = −0.75, P = 0.10 and r = −0.62, P = 0.12, respectively), suggesting that cases with less myelin staining also had decreased microgliosis. In contrast, these regions in PSP showed a positive correlation (r = 0.67, P = 0.11 and r = 0.88, P = 0.05, respectively), suggesting that paucity of MBP staining was associated with increased microgliosis.

The relationship between case demographics and pathological variables was also analyzed. No significant correlations were observed between clinical variables such as disease duration or age at death and pathological variables described earlier. Importantly, NG2 cell density did not correlate with PMI in the cerebellar (r = −0.25, P = 0.28) or frontal (r = −0.09, P = 0.67) white matters, suggesting that the increase in NG2 cells observed in the cerebellum of MSA cases was unrelated to experimental factors such as PMI.

Discussion

This is the first study to investigate NG2‐positive OPCs in MSA and related disorders. In summary: NG2 cells (i) were successfully identified in flash‐frozen post‐mortem tissue and validated using cell‐specific markers; (ii) did not accumulate α‐synuclein in MSA or tau in PSP; (iii) were increased in the cerebellar white matter of MSA cases; and (iv) showed a positive correlation with the severity of oligodendroglial inclusions in MSA and PSP cases combined.

Identification of OPCs in human brain tissue

OPC research, especially in the human CNS, has been hindered by the lability of markers that can identify them 25. To detect OPCs in human tissue, we used a commercially available human NG2‐specific antibody (Table 2). Additional NG2 antibodies and other OPC markers, such as platelet‐derived growth factor receptor α (PDGFRα) and A2B5 25, were also tested, but were inferior to the NG2 antibody used (data not shown). Fixation was also an important factor as NG2‐positive cells were only detected in flash‐frozen post‐mortem tissue and not in formalin‐fixed paraffin‐embedded (FFPE) tissues. Valid NG2 staining has been shown in FFPE tissue, but only when the tissue has been fixed for less than 24 h 25, making most diagnostic or brain banking tissue unsuitable. As previously reported 3, 25, the best NG2 staining was achieved using tyramide‐based signal amplification kits. Although the majority of the cases used in this study had a PMI of less than 14 h, we were able to demonstrate NG2‐positive cells in cases with a PMI as long as 50 h, despite rodent studies showing that NG2 staining is reduced 12 h post‐mortem and altogether eliminated at 24 h 20. Our results are consistent with a human MS study that demonstrated NG2‐positive cells in cases with PMIs up to 48 h 3. Furthermore, the density of NG2 cells did not correlate with PMI in our study, which suggests that factors other than PMI might be influencing NG2 labeling; such factors may include the time interval to refrigeration of the body or agonal state of the patient. Despite the difficulties associated with NG2 labeling, we confirmed that the NG2‐positive cells were indeed OPCs, by double labeling with oligodendroglial, astrocytic and microglial cell‐specific markers. NG2 cells were also positive for the oligodendrocyte lineage nuclear transcription factor, Olig2, and negative for astrocytic and microglial markers, confirming the presence of OPCs. Consistent with a recent report 9, Olig2 immunoreactivity was increased in NG2‐positive cells compared with NG2‐negative/Olig2‐positive cells, the latter is thought to represent mature oligodendrocytes. This raises the possibility that the intensity of Olig2 labeling could be used in the future to distinguish between OPCs and mature oligodendrocytes, especially given that Olig2 IHC is reliable in brain‐banked FFPE tissues 1.

OPCs are not vulnerable to α‐synuclein or tau accumulation

Having reliably identified NG2‐positive OPCs, we demonstrated that these cells were not vulnerable to the pathological process, resulting in inclusion formation characteristic of MSA and PSP. NG2‐positive cells were not immunoreactive for α‐synuclein or tau nor were the cells containing GCIs or tau inclusions NG2‐positive, despite being observed alongside each other. This would suggest that the process that leads to the accumulation of α‐synuclein to form GCIs in MSA and aggregation of tau to form coiled bodies in PSP is restricted to more mature (NG2‐negative) oligodendrocytes. With respect to MSA, our results are in keeping with in vitro and in vivo studies showing that OPCs do not express high levels of α‐synuclein 17, 21. Although α‐synuclein is described as a synaptic protein that is expressed primarily in neurons 4, α‐synuclein mRNA and protein have also been detected in cultured rat oligodendrocytes where its expression appears to be developmentally regulated, being low in OPCs, increasing to maximal levels after 2–3 days in culture and then declining to low levels by 28 days 21. Similarly, α‐synuclein was not detected in OPCs in an experimental rat model of inflammatory demyelination, which showed increased neuronal and glial expression of α‐synuclein 17. There is still a possibility that the formation of such inclusions in OPCs could adversely affect the expression of NG2, making it difficult to identify them, similar to the way that tufted astrocytes in PSP are negative for the astrocytic marker GFAP 22. The current study has shown that tufted astrocytes, despite some morphological resemblance, are NG2‐negative. Although not investigated in this study, it is possible that α‐synuclein and tau might be accumulating in terminally differentiated, premyleinating oligodendrocytes; these represent a maturation stage in the oligodendrocyte lineage intermediate between OPCs and myelinating oligodendrocytes, and are NG2‐negative 15. Understanding the functional differences between OPCs and more mature oligodendrocytes, which makes one resistant and the other vulnerable to abnormal protein accumulation, will provide insight into the currently unknown mechanism of oligodendroglial inclusion formation in these diseases. It would also be interesting to see if OPCs are affected in transgenic mouse models where disease‐associated proteins, such as α‐synuclein, are overexpressed using an oligodendroglial lineage promoter, which has been shown to reproduce oligodendroglial inclusions as seen in human disease 10, 29.

OPCs are increased in the cerebellar white matter of MSA

OPCs are proliferative and motile cells that can give rise to mature myelinating oligodendrocytes throughout adulthood and are responsible for remyelination in the CNS during injury and disease 6, 25. Analysis of myelin integrity in our study using IHC revealed a patchy and/or reduced pattern of myelin staining in MSA, which was largely absent in PSP and controls. Such abnormal myelin staining is indicative of myelin loss and has previously been reported in MSA 13, 18, 23. The cerebellar white matter was most severely affected, but the frontal white matter also showed evidence of myelin loss in some MSA cases. It is of note that the density of GCIs also followed a similar regional distribution pattern, being more numerous in the cerebellar white matter compared with the frontal white matter. Together, these features support our original assumption that the cerebellar white matter represented a more severely affected region than the frontal white matter in MSA 19. Together with myelin loss and increased GCI density, the cerebellar white matter also showed an increase in the density of NG2 cells when compared with controls, while in contrast, their density in the frontal white matter of MSA was comparable to that in PSP and controls. Given the biological role of OPCs, these data raise the possibility that NG2‐positive OPCs might be attempting to remyelinate areas with myelin loss by proliferating, differentiating and maturing into myelinating oligodendrocytes in regions that are severely affected in MSA. This would be consistent with numerous experimental in vivo studies showing that OPCs are amplified following a demyelinating event and give rise to myelinating oligodendrocytes; a response that is very robust 6, 11. Human studies investigating relative changes in the density of NG2 cells are rare and have focused almost entirely on MS, where their density has been shown to be variable depending on the study and the type of MS lesion analyzed 25. In MS, active lesions are characterized by a normal density of oligodendrocyte lineage cells, whereas chronic lesions have a stark paucity of mature oligodendrocytes; the observation that NG2 cells were found in both types of lesions has led researchers to suggest that OPC‐directed remyelination fails in chronic lesions and its consequences lead to the progression of MS 11.

Another indication that such remyelination might be occurring in MSA is the observation that axons that have been remyelinated have a thinner myelin sheath compared with myelin laid down during development 2, 12. In our study, our analysis was restricted to axonal density, which did not find significant differences between disease groups, but a previous study has shown that myelin sheaths around axons in MSA are significantly thinner compared with controls 23, suggesting these thinner myelin sheaths may be a result of remyelination. It is of note that inflammatory signals from microglia and astrocytes are thought to be responsible for the recruitment and proliferation of OPCs 11. In our study, there was an increase in microgliosis and astrogliosis in the cerebellum of both MSA and PSP cases, yet NG2 cells were only significantly increased in MSA, suggesting that specific inflammatory signals or a threshold of inflammation might be needed to recruit OPCs. Recent studies have indicated that NG2 cells can also give rise to other cells, notably protoplasmic astrocytes 15. Although it is tempting to speculate that the increase in NG2 cells in the cerebellum of MSA is part of the increased astrocytic response, the latter feature was also shown in PSP without any increase in NG2 cells.

Despite the potential relationship between NG2 cells and myelin loss in the cerebellar white matter of MSA cases, statistical analysis failed to detect a correlation between these variables. A correlation indicating that NG2 cells in the cerebellar white matter increase as the number of oligodendroglial inclusions increase was statistically significant only when MSA and PSP cases were pooled in the analysis, highlighting the need for further studies utilizing a larger sample size for this type of correlation analysis. Although short of statistical significance, a correlation suggesting that MSA cases with severe myelin loss had decreased microgliosis is consistent with a previous report 7. Ishizawa et al also reported that regions with very severe myelin loss (grade III) had fewer GCIs 7, whereas a correlation between these variables in our study indicated (although short of statistically significance) that increasing myelin loss was associated with higher numbers of GCIs. In retrospect, the myelin loss seen in our MSA cases was not as severe as that described as grade III by Ishizawa et al, which might explain the discrepancy. Nevertheless, these studies highlight myelin degeneration and formation of GCIs as important steps in the pathogenesis of MSA.

Are OPCs involved in the formation of GCIs in MSA?

If we consider the increase of NG2 cells in areas of the MSA brain that also have severe myelin loss and increased GCI density, and take into account our current knowledge of OPC function and α‐synuclein expression in oligodendrocyte lineage cells, we can piece together a potential hypothesis for the formation of GCIs and the pathogenesis of MSA. In the event of an unspecified injury to the white matter causing myelin loss, NG2‐positive OPCs can respond by proliferating at the site of injury, begin to differentiate and mature into myelinating oligodendrocytes in order to replace those that are lost 6, 11. In vitro, this maturation process in oligodendrocyte lineage cells has been shown to be associated with a transient increase in the cellular expression α‐synuclein; expression is low in OPCs and mature oligodendrocytes, but highest in the intermediate stages of maturation 21. If this differentiation/maturation process is somehow delayed or incomplete [as has been suggested in MS 6, 11], the prolonged expression of α‐synuclein could lead to its eventual cellular aggregation and the formation of GCIs 10, 29. Such aggregation has been shown to be detrimental to the survival of oligodendrocytes 29, and therefore, loss of myelinating oligodendrocytes would not only amplify this whole process, but may also lead to a lack of trophic support to non‐myelinated axons 11, potentially resulting in downstream neuronal α‐synuclein aggregation and neurodegeneration. Although highly speculative at this juncture, our hypothesis would explain (i) how α‐synuclein accumulates in more mature oligodendrocytes that have been shown to have no or very little endogenous α‐synuclein expression 14, 21; (ii) why NG2‐positive OPCs do not accumulate α‐synuclein in MSA (as shown in this study); and (iii) the higher density of GCIs in areas with significant myelin loss. Our hypothesis would also concur with a recent study that suggests myelin loss and the re‐localization of a myelin‐associated protein, p25α, occurs prior to the formation of GCIs 23. To further test this hypothesis, it would be important to confirm if α‐synuclein expression in oligodendrocyte lineage cells is increased in vivo during human/mammalian brain development and if α‐synuclein accumulates in oligodendrocytes at intermediate stages of maturation; confirmation of the increase in NG2 cells in a larger cohort of MSA would also be important. Mouse models of MSA, with overexpression of α‐synuclein in oligodendroglial cells 10, 29, might also be useful to investigate our hypothesis.

Conclusion

In conclusion, this is the first study to investigate the role of NG2‐positive OPCs in neurodegenerative diseases that affect oligodendrocytes, namely MSA and PSP, whereas previous studies in humans have been limited to MS. This relatively recently recognized class of neuroglia have been traditionally difficult to identify, but validated IHC staining was achieved by using a select cohort of tissue that was optimally fixed and reagents that provided increased sensitivity for the NG2 antigen. Although OPCs were resistant to the deleterious accumulation of disease‐associated proteins in MSA and PSP, their relative numbers were increased in the white matter region of the MSA brain, which is also affected by significant numbers of oligodendroglial inclusions and myelin degeneration. Given the role of OPCs in remyelination by proliferating, differentiating and replenishing myelinating oligodendrocytes, the increase in OPCs in MSA identified in this study might represent an attempt to repair disease‐associated damage, which, according to our speculative hypothesis, might itself lead to the formation of GCIs in MSA and progression of the disease. Therefore, as has been suggested in MS 6, 11, endogenous OPCs may represent a viable target for future therapies intended to enhance remyelination in MSA patients. Further studies on NG2 cells in the context of MSA will rely on the current and future markers available to identify them in human post‐mortem brain tissue.

Conflicts of Interest

The authors declare that they have no conflict of interest.

Supporting information

Figure S1.  Semiquantitative scoring scheme used to assess myelin basic protein specific immunohistochemistry. Scores describe the increasing degree of staining patchiness. 1+ = normal, 2+ mild, 3+ = moderate, 4+ = severe. Magnification = 200×.

Figure S2.  Double immunofluorescence staining using NG2 (green) and cell‐specific markers (red) for oligodendrocytes (Olig2), astrocytes (GFAP) or microglia (Iba‐1). Arrows highlight NG2‐positive cells and arrowheads show NG2‐positive blood vessels. Images represent transparent projections of confocal Z‐stacks. GFAP = glial fibrillary acidic protein, Iba‐1 = ionized calcium binding adaptor 1. Scale bars: 25 μm.

Figure S3.  Examples of ionized calcium binding adaptor‐1 (Iba‐1), glial fibrillary acidic protein (GFAP) and myelin basic protein (MBP) specific immunohistochemistry in the cerebellar white matter of a representative MSA and control case. These markers demonstrated quantitative differences between MSA and control groups. Scale bars: 50 μm.