Progression of alpha-synuclein pathology in multiple system atrophy of the cerebellar type

Autopsy cohort

Of 47 patients with a clinical and confirmed neuropathological diagnosis of MSA [2,7,31,32] who were followed longitudinally to autopsy in the Center for Neurodegenerative Disease Research (CNDR) at the University of Pennsylvania between 1989 and 2013, we included all cases with a clinical picture of MSA-C and a neuropathological picture of OPCA (n = 10, 1 female, 9 males; mean age at onset ± SD: 58 ± 7.7 years, range 51–74 years, mean disease duration ± SD: 6.1 ± 3.3 years, range 3.5–14 years) as well as 24 controls age-matched to the MSA-C group that did not show clinical or neuropathological signs of neurodegenerative disease (7 female, 17 male, mean age at death 63.9 ± 6.0 years). Informed consent for autopsy was obtained for all patients or from their next of kin. The study was approved by the University of Pennsylvania Institutional Review Board. Detailed clinical characteristics (gender, age at onset, age at death, clinical symptoms of onset, disease duration) were ascertained from an integrated autopsy database, as described previously [33,34] and by retrospective chart review of clinical visits within the University of Pennsylvania Health System (Table 1).

Tissue preparation, staining and immunohistochemistry

Pathology was examined in 20 regions of the CNS: the middle frontal gyrus, gyrus rectus and orbital gyri, the agranular motor cortex (Brodmann areas 4 and 6), somatosensory cortex, superior or middle temporal gyrus, angular gyrus, precuneus (Brodmann area 7), visual cortex (Brodmann areas 17 and 18), anterior cingulate gyrus, hypothalamus, amygdala, hippocampal formation, striatum and pallidum (the striatum block included anterior portions of the caudate nucleus and putamen plus the accumbens nucleus, where the caudate nucleus and putamen merge), the thalamus, midbrain [including the substantia nigra (SN) and red nucleus (RN)], upper pons at the level of the locus coeruleus, lower pons (including the motor nucleus of cranial nerve VII), the medulla oblongata at the level of the hypoglossal nucleus (XII) and inferior olive (IO), the cerebellum (including both a portion of the cerebellar cortex and dentate nucleus), and the cervical spinal cord. After overnight fixation of thin 0.5–1-cm-thick tissue blocks to ensure uniform penetration of 10% neutral buffered formalin, all tissue samples were embedded in paraffin using standardized cassettes, sectioned at 6–7 µm, and stained with haematoxylin and eosin (HE) and for immunohistochemistry (IHC) as previously described [17,35]. Briefly, IHC was performed with antibodies to α-syn (monoclonal antibody (mab) Syn303; 1:4000, generated in CNDR) [36], hyperphosphorylated tau (mab PHF1; 1:1000, gift from Dr. Peter Davies), pTDP-43 (rat antibody p409/410, 1:1000, gift from Dr. Manuela Neumann) [37] and amyloid-β (mab NAB228; 1:15 000; generated in CNDR) [38], whereas histochemical staining with HE as well as thioflavin S were performed as reviewed recently [39]. Neurofibrillary tangle stages and CERAD neuritic plaque scores are shown in Table 1.

To study each of the CNS regions in greater neuroanatomical detail, additional sets of 70 µm sections were prepared as described previously [17,30] from the same paraffin blocks as those used above. This thick section technique is performed on free-floating sections and permits recognition of specific cell types and cytoarchitectonic units owing to the visualization of large numbers of biological structures in a more three-dimensional manner than is possible with conventional 6-µm-thick sections [17,30]. The sections were stained for analysis of α-syn pathology and to provide for a more topographical overview and to assess neuronal loss (NL), we combined α-syn IHC with a pigment-Nissl stain (PN) for lipofuscin pigment (aldehyde fuchsin) and basophilic Nissl material (Darrow red) as described [30]. In addition, Campbell-Switzer silver staining (CS) was performed as previously described [40]. Severity of α-syn pathology and NL were assessed according to a 4-point semiquantitative rating scale (0, absent or not detectable; 1, ≤2 aggregates per region + mild; 2, moderate ++; 3, severe/numerous +++) [28,41] and was adjusted for NL. Double-labelling IHC was performed on selected 70 µm sections using the α-syn antibody described above together with the Vector Labs, Burlingame, CA, USA, SG (SK-4700; Vector) blue chromogen and the SMI-311 mab (1:1000; Covance, Princeton, NJ, USA), the rabbit polyclonal antibody (pab) for oligodendrocyte-specific protein (1:400; Abcam, Cambridge, UK) and the anti-oligo2 antibody (1:400; Millipore, Billerica, MA, USA).

For double-labelling immunofluorescence (IF), 15 µm frozen sections were microwaved in 10 mM citrate buffer, pH 6.0, for 15 min; blocked in 2% FBS with 0.1 M Tris and incubated with the mouse mab syn-506 (specific for α-synuclein, amino acids 1–89; 1:10 000; CNDR [42]) and rabbit pab Olig2 (specific for oligodendrocytes; 1:250; Millipore), rat mab TA51 (specific for phosphorylated heavy and medium molecular weight neurofilament; 1:10; CNDR [43]), rabbit pab NFL1/2 (specific for low molecular weight neurofilament; 1:1000; CNDR [44]) or rabbit pab 17028 (a microtubule associated protein 2 specific antibody; 1:500; CNDR, unpubl. data), followed by a goat anti-mouse Alexa Fluor^® 546-labelled IgG (H+L) antibody, a goat anti-rabbit Alexa Fluor^® 588-labelled IgG (H+L) antibody and a goat anti-rat Alexa Fluor^® 588-labeld IgG (H+L) antibody (1:500; Invitrogen, Carlsbad, CA, USA) in 2%FBS in 0.1 M Tris. Slides were cover-slipped using Vectashield mounting media with DAPI (Vector) and digitalized images were obtained using a Leica TCS SPE-II scanning laser confocal microscope.

Targeted next-generation sequencing

Genomic DNA was extracted from frozen brain tissue using commercial reagents (DNA mini kit; Qiagen, Valencia, CA, US). To exclude mutations in COQ2 which have been associated with MSA, targeted sequencing was performed using a neurodegenerative disease-focused panel that targets the coding exons of 45 genes, including COQ2. The library of the target regions was prepared with Haloplex enrichment kit (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s protocol. Sequencing was performed on an Illumina, San Diego, CA, USA, Mi-Seq using 150 base paired-end reads. Following quality control procedures, alignment of sequence reads and variant calling was assessed using SureCall software (Agilent).

Statistical analysis

Data analysis was performed using SPSS (version 17.0; SPSS Inc., Chicago, IL, USA). The average (and range) of data on patient characteristics were estimated by calculating the median (and 25th–75th percentiles). Differences between two clinical subgroups were compared using Wilcoxon Mann–Whitney test. To compare raw data of multiple subgroups, Kruskal–Wallis analysis of variance on ranks was applied and in case of significance, by Dunn’s method. Trend analysis was conducted using the Mantel–Haenszel chi-square test. All correlations were studied using Spearman’s rank order correlation coefficient. Bonferroni-correction for multiple testing was applied when contrasts were not driven by a specific hypothesis. For all other tests, P-values < 0.05 were considered significant. All statistical tests were two-sided.

Morphology of α-syn pathology in MSA

Four phases of α-syn pathology in MSA-C

On the basis of the different types of α-syn inclusions described above, we proceeded to analyse phases of α-syn pathology in MSA-C according to an increasing overall burden of pathology (Table 1).

Clinicopathological correlations and genetic results

We next analysed clinicopathological correlations for the entire cohort of MSA-C cases. The patterns of α-syn pathology described above correlated significantly with duration of disease (ρ = 0.83, P < 0.001) but were not related to age at onset (ρ = −0.13, P = 0.71) or age at death (ρ = 0.3, P = 0.38). Furthermore, severity of α-syn pathology increased significantly with duration of disease across many of the regions analysed here (Table S1). In contrast, we observed no significant correlation between the regional severity of α-syn inclusions and tau or Aβ pathology (P > 0.05 and ρ < 0.04 each). Myelin loss in pons and cerebellum correlated significantly with severity of GCI pathology, and increased with disease duration (ρ > 0.4, P < 0.05 each).

Although mutations in COQ2 have been associated with an increased risk of MSA, the overwhelming majority of MSA cases present as sporadic disease [45]. We screened the MSA-C cases in this study for genetic variations in COQ2. No pathogenic mutations were identified.

Morphology of α-syn inclusions in thick sections of MSA-C

We first analysed the morphology of α-syn aggregates in 70-µm-thick sections. As noted above and reported elsewhere [17,30,40], we emphasize illustrations of 70-µm-thick sections because they allowed us to view the histology and pathology in a more three-dimensional manner than what is possible by examining conventional 6-µm-thick sections. We also examined 6-µm-thick sections for this study, but these provide a shorted depth of viewing that result in objects having a more two-dimensional appearance. In line with previous evidence [2,7,12,13,39,46–53], we found the main α-syn pathology to consist of white matter GCIs. GCIs were reported to be partially argyrophilic [54], especially in cases with longer duration of disease (Figure 2E). Moreover, similar to Lewy bodies in PD and DLB, GCIs also show properties of amyloids including positivity for thioflavin S and other amyloid binding dyes [18]. Consequently, argyrophilia and positivity for amyloid binding dyes are likely to be a characteristic of more mature α-syn inclusions in MSA-C. NIs and NRIs were less frequent than GCIs, which is also in accordance with previous studies of MSA [55–57].

Importantly, our study indicates that neurons in MSA-C manifest other abnormalities that variably include cytoplasmic and nuclear swelling, intraneuronal loosening of lipofuscin pigment (Figure 1B, H), and dendritic morphological changes prior to the development of NI (Figure 2K–L), which we speculate may reflect a form of cellular stress or incipient neurodegeneration. In addition, our study indicates that nuclear α-syn aggregates in neurons are more frequent than previously reported [49]. Although α-syn is mainly a presynaptic protein, it was also detected at the inner nuclear membrane [58]. Such nuclear α-syn aggregates could have potentially severe functional consequences in cells that do not show the prominent cytoplasmic aggregates observed in other synucleinopathies [16,28].

Cerebellum and its projections as early foci of pathology in MSA-C

On the basis of our analysis of α-syn inclusion morphology, we proceeded to look for evidence of a possible spreading of α-syn pathologyin MSA-C. We focused on MSA-C as we expected the spread of α-syn in MSA-P to be different and to require a separate study. Our intention was not to propose a new grading system of MSA-C [7,12,46] but, rather, to determine where α-syn pathology might begin and how it progressively spreads in autopsy-confirmed MSA-C cases.

Cases with the lowest overall burden of pathology (phase 1, Table 1) showed prominent GCIs in the cerebellar white matter and in medullar and pontine white matter cerebellar projections, which is in agreement with previous neuropathology and neuroimaging studies [2,7,12,32,46,59–61]. GCIs and NRIs were associated with myelin loss in the pontocerebellar fibres (Figure 2D, E) [62,63], which has been linked with the characteristic ‘cross-bun sign’ detectable in magnetic resonance imaging of living MSA-C patients [64,65].

Cerebellar involvement was highly selective: Subcortical white matter showed severe GCIs already in cases with a low burden of pathology/short disease duration (Table 1). In contrast, deep white matter and the granular layer were affected to a much lesser degree, and the molecular layer remained free of any α-syn inclusions (Table 1, Figure 2A–C). The dentate nucleus showed NIs only in cases with a long disease duration (Table 1), which coincides with previous observations [7,46,66]. Importantly, GCIs showed a close spatial relation to myelinating axons (Figure 2F–H), thereby indicating that GCIs are α-syn lesions that occur in myelinating oligodendroglia (and not in satellite cells). In line with this and previous findings [46,54], we found severe myelin loss in regions characterized by early and severe GCI formation, such as cerebellar subcortical white matter or pontocerebellar fibres (Figure 2D). We hypothesize that this highly selective pattern of cerebellar involvement in MSA-C could be partially explained by the association of GCI with myelinated fibres. Thus, we found that the cerebellar molecular layer, which contains the unmyelinated parallel fibres of granular layer neurons, remained free of GCI pathology even in cases with the longest disease duration [67]. Similarly, the distal parts of PC axons, which penetrate the cerebellar deep white matter surrounding the dentate nucleus, have been reported to be unmyelinated [68,69], which could explain the relative sparing of the cerebellar deep white matter by GCI pathology (Figure 2A).

With an increasing burden of α-syn pathology (phase 3, Table 1), GCIs, Nis, and NRIs became detectable in the neocortex (Figure 3H). In accordance with previous studies, neocortical NIs developed only gradually and initially affected chiefly lower cortical layers [70,71]. Although cortical involvement in MSA is reportedly rare [61,70], some studies have reported NL in the motor and supplementary motor cortex [72] and have suggested a relationship with striato-nigral involvement [73]. In the midbrain, GCIs and NRIs prominently involved the crura cerebri, which contains striatonigral ‘comb’ fibres [74], the involvement of which might be linked to NL in the SN [12,46,75]. Similarly, the presence of severe GCIs and NRIs in the internal capsule (Figure 3G) could contribute to NL in the putamen and SN. In line with previous studies, we observed loss of hypothalamic neurons in cases of MSA-C (Table 1) [46, 76, 77]. The hypothalamus is of primary importance for thermoregulation, glucoregulation, osmoregulation, stress responses and immunomodulation [78]. Thus, its involvement could contribute to autonomic dysfunction in MSA-C [76]. Finally, MSA-C cases with a further increasing burden of pathology (phase 4, Table 1), were characterized by involvement of the entorhinal cortex, hippocampus (Figure 3I–L) and the basolateral subnucleus of the amygdala.

Mechanisms of NL in MSA-C

Remarkably, NL in MSA-C frequently affected groups of neurons that showed only mild α-syn aggregates (including neurons of the SN and the locus coeruleus, whereas the pathology is characterized by massive involvement of oligodendroglia [7,46,66,75]. Furthermore, previous studies on cases of minimal change MSA with NL restricted to the substantia nigra and locus coeruleus, but more widespread GCI pathology, suggested that oligodendroglial pathology could lead to neuronal dysfunction and clinical symptoms before overt NL in MSA [79,80]. We hypothesize that neuronal dysfunction and NL could be caused by GCI along myelinated axons at a considerable anatomical distance from the neuronal cell body. The mechanisms by which α-syn accumulation in GCIs could possibly lead to axonal dysfunction and neuronal death are still unknown, but both were seen in an earlier transgenic mouse model of MSA induced by α-syn overexpression [81]. There is increasing evidence that oligodendrocytes provide essential metabolic support to neurons by transferring glycolytic intermediates [82,83]. Furthermore, oligodendrocytes are responsible for maintaining brain lipid homeostasis, and myelin instability was recently suggested to precede α-syn pathology in MSA [84,85]. In addition, altered myelin protein composition including changes in the cellular interactions between MBP and p25α have been suggested to contribute to MSA pathology [20].

Finally, it is unclear if the concept of ‘prion-like’ propagation attributed to other synucleinopathies can be extended to MSA [22,23,25,27,86–88]. Our findings point to a stereotypical spreading pattern of α-syn-containing lesions similar to what has been described for PD and DLB [28,41]. A limitation of our study is the comparatively low number of cases with MSA-C available. It should therefore be regarded as a pilot-study that should entail a multi-centre follow-up effort to validate our findings and establish stages of α-syn pathology in MSA-C.