Axonal degeneration in the anterior insular cortex is associated with Alzheimer’s co-pathology in Parkinson’s disease and dementia with Lewy bodies

Formalin-fixed postmortem human brain tissues from 27 donors with PD, PDD, and DLB (range = 60–93 years) were collected in close collaboration with the Netherlands Brain Bank (NBB, Amsterdam, The Netherlands; www.​brainbank.​nl). In compliance with ethical and legal guidelines, all donors or next-of-kin had provided written informed consents for donation of brain tissue and access to their clinical and neuropathological reports for research purposes. The brain donor program of the NBB was approved by the local medical ethics committee of the VU University Medical Center, Amsterdam. The main inclusion criteria for the current study were: (i) clinical diagnosis of PD, PDD, or DLB according to revised MDS and McKeith diagnostic criteria [26, 27] and (ii) neuropathological confirmation of diagnosis [28]. Retrospective chart reviews were carried out for all included donors and the latest clinical dementia rating (CDR) global scores were retrieved. The CDR global score is a scoring algorithm designed to assess the overall severity of dementia. It is based on cognitive and functional deficits in the domains of memory, orientation, judgement, problem-solving, hobbies, personal care, and community affairs. Ratings are assigned on a 5-point scale ranging from 0 (normal) to 3 (severe) [29]. Subjects were excluded if they had a long history of neuropsychiatric disorders or concomitant neurological disorders, infarcts and other abnormalities within the insular cortex (Table 1).

Brain dissection was performed according to international guidelines of Brain Net Europe II (BNE) consortium (http://​www.​brainnet-europe.​org) and NIA-AA by an experienced neuropathologist (NBB: AMR). The entire insula was dissected into 0.5–1-cm thick blocks according to its anatomical borders with surrounding brain regions. Using the central sulcus as a landmark separating the anterior from the posterior insula, the anterior insular cortex (AIC) was then dissected further [30]. Tissue blocks were cryo‐protected with 30% sucrose, frozen and sectioned using a sliding microtome into 60-μm thick sections and these were stored at − 30 °C until further processing. Sections were first incubated in various gradients of alcohol followed by xylene to induce de-fatting, and then stained with Nissl. This allowed the identification of the agranular and dysgranular sub-regions based on the absence or presence of layers II and IV by two experienced researchers (YF and LJ) as previously described [19, 31].

For neuropathological diagnosis and staging, 6-μm paraffin sections from brainstem, limbic and neocortical brain tissue blocks of all donors were immunostained for α‐synuclein (clone KM51, 1:500, Monosan Xtra, The Netherlands), amyloid-β (clone 6F/3D, 1:500, Dako, Denmark), phosphorylated tau (p-tau, clone AT8, 1:500, Thermo Fisher Scientific, Rockford, IL), haematoxylin and eosin (H&E), TDP‐43 and congo red according to current diagnostic guidelines of BrainNet Europe [28]. Pathological staging protocols were based on Braak α‐synuclein (Braak α‐syn 0–6), Braak staging for neurofibrillary tangles (Braak NFT 0–6), Thal phase for amyloid-β (0–5), and CERAD score for neuritic plaques [32‐36]. Glial tauopathy such as age-related tauopathy of the astroglia (ARTAG) and primary age-related tauopathy were assessed primarily in the temporal cortex, olfactory cortex and amygdala by an experienced neuropathologist (AMR) [37, 38]. Clinical symptoms and diagnosis ante-mortem, along with neuropathological scores, were used to provide a final pathological diagnosis. DLB diagnosis was made based on the revised criteria for the clinical diagnosis of probable and possible DLB using patient’s ante-mortem clinical history, neurological, and psychiatric assessments [26]. Furthermore, using neuropathological assessment, DLB diagnosis was confirmed when there was concomitant AD pathology, cerebral amyloid angiopathy (CAA), and cortical Lewy pathology [39](Table 1).

Modified Bielschowsky silver staining [40] was used for the evaluation of axonal degeneration in the anterior insula. Prior to staining, all glassware was thoroughly rinsed with purified water for several hours or overnight and metals were avoided. All incubations were performed at cold temperature (5 °C) [40]. Free-floating 60-μm tissue sections (interval: 1 in 20) were rinsed in distilled water followed by Tris-buffered saline (TBS, pH 7.6) and incubated in 20% silver nitrate solution (AgNO₃; Sigma, CAS No: 7761-88-8) for 30 min at 5 °C in the dark. Subsequently, non-evaporated ammonium hydroxide 30%–33% (NH₃) solution was slowly added to the sections until the color turned black, and then incubated for 30 min at 5 °C in the dark (Honeywell, CAS No: 1336-21-6). Sections were then rinsed with evaporated ammonia-water for 10 min (100 µl ammonia in 50 ml distilled water). Subsequently, a developer solution was freshly prepared, which consisted of 100 ml distilled water, 20 ml 37% formalin (Sigma, CAS No: 50–00-0), 1 drop nitric acid (Fisher, CAS No: 7697-37-2), and 0.5 g citric acid (Sigma, CAS No: 77-92-9). Sections were incubated in the developer solution at 5 °C in the dark for visualization of the reduced metallic silver. Staining was monitored carefully and the incubation was stopped within 10–15 min. This was followed by rinsing in Hypo (sodium thiosulfate 5% in distilled water; Honeywell, CAS No: 7772-98-7) to stabilize the developing reaction and remove free AgNO₃ deposits from the tissue. Sections were then rinsed, mounted with 0.3% gelatin (Oxoid), dried and cover-slipped using Entellan (Sigma).

For α‐synuclein immunostaining, consecutive free‐floating 60-μm thick sections of the anterior insula were pre-treated with 98% formic acid (Sigma‐Aldrich, Darmstadt, Germany) and incubated with primary antibody mouse anti‐α‐synuclein (syn-1, 1:2000, 610786, BD Biosciences, Berkshire, UK) in TBS and 0.5% TritonX-100 solution overnight at room temperature, as previously described by Braak and colleagues [41]. For immunostaining with antibodies against p-tau and amyloid-β, free-floating sections were rinsed with TBS and pre-treated with citrate buffer (pH 6.0) in a steamer (95 °C) for antigen-retrieval, followed by 80% formic acid for amyloid-β. Non-specific staining was then blocked with 0.3% H₂O₂ and 0.1% sodium azide in TBS followed by 2% normal goat serum. Subsequently, sections were incubated in primary antibody mouse anti-p-tau (1:1000, MN1020, AT8, Thermofisher-scientific, Netherlands) and adjacent sections were incubated in mouse anti-amyloid-β (1:1000, M08720, clone 6f/3d, Dako, Denmark) diluted in TBS and 0.1% Triton-X overnight at 4 °C. All sections were incubated with biotinylated secondary antibody IgG (1:200, Vector Laboratories, Burlingame, CA) followed by standard avidin‐biotin complex (1:200, Vectastatin ABC kit, Standard; Vector Laboratories) in TBS for 2 h. Tissue samples were incubated in 3,3′‐diaminobenzidine to visualize staining and were mounted and counter‐stained with thionin (0.13%, Sigma‐Aldrich, Darmstadt, Germany). Immunofluorescent double staining of α-synuclein and glial fibrillary acid protein (GFAP) was performed as described elsewhere [19].

Load of α-synuclein, p-tau and amyloid-β pathology was calculated as percentage immunoreactivity per surface area in each region of interest (ROI) using area fraction plugin in ImageJ (1.52n) [42]. Anterior insular sub-regions (agranular and dysgranular) were identified on adjacent Nissl-stained sections, and the same borders were used to define ROIs on sections immuno-stained for pathological aggregates [19].

We included multi-labelling immunofluoresecent staining with axonal and myelin markers to visualize the axonal morphology and cytoskeletal abnormalities in 3D with CSLM. Adjacent 60-μm thick sections from PD(D) and DLB cases were rinsed in TBS, pre-treated with EDTA-Tris (pH 9.0) in a steamer (95 °C) and incubated in a cocktail of the following primary antibodies: (1) mouse  anti-myelin proteolipid protein (PLP, 1:500, MCA839G, plpc1, Bio-Rad, Netherlands) and (2) chicken anti-neurofilament heavy chain (NfH, 1:500, AB5539, heavy-chain, Millipore) diluted in TBS, 2% normal donkey serum, and 0.5% Triton-X. Immunostaining was performed for 48 h at 4 °C followed by incubation with secondary antibody goat anti‐chicken coupled with Alexa Fluor 594 (1:400; Molecular Probes, Waltham, MA), donkey anti‐mouse coupled with Alexa Fluor 647, and diamidino‐2‐phenylindole [4, 6] dihydrochloride (DAPI; Sigma) for nuclear staining for 2 h in the dark. Tissue samples were subsequently rinsed in TBS and blocked in 5% normal mouse serum for 1 h followed by incubation with Alexa488-conjugated mouse anti-phosphorylated-Serine129 (pSer129) α‐synuclein antibody (1:100, 11A5, gift from Prothena Biosciences Inc., San Francisco, CA) for 2 h at 4 °C in the dark. The tissue sections were then mounted on glass slides and cover‐slipped with mowiol-DABCO as a mounting medium (4‐88 Calbiochem).

Digital images of the immunostained slides were made with a photomicroscope (Leica DM5000) equipped with a color camera (DFC450), Leica LASV4.4 software and 63 × oil objective lens. Immunofluorescent labelling was visualized using CLSM LEICA TCS SP8 (Leica Microsystems, Jena, Germany). Adjacent Nissl-stained sections were used for delineation of subregions and the ROIs were superimposed on the images of the immunofluorescent stained sections. This was followed by sampling of all axons (Bielschowsky staining) in the superficial and deep layers of each sub-region. Image acquisition was done using 100 × /1.4 NA objective lens, 405 nm diode, and pulsed white light laser (80 Hz) with excitation wavelengths of 405, 499, 598, and 653 nm, scanned using frame/stack sequential mode. For optimal resolution, z-step size was calculated for each scan based on the Nyquist-Shannon sampling theorem [43] and line accumulation/averaging were used as deemed appropriate per channel. Deconvolution of image stacks was performed using Huygens Professional software (Scientific Volume Imaging, Hilversum, the Netherlands). Images are shown as maximum projections of all channels combined and all figures were composed using Adobe Photoshop (Adobe Systems Incorporated). Colocalization between GFAP and α-synuclein was assessed using Imaris 8.3 (Bitplane, South Windsor, CT).

Axonal length density, based on Bielschowsky silver impregnated sections, was quantified using the stereological space balls probe from stereoinvestigator software (MBF Bioscience, Williston, VT) and Leica microscope DMR HC (v2019.01.4) [44‐46]. Serial sections (1:20, Range of sections: 3–8) were used for quantification, and counting parameters were chosen to allow counting ≥ 200 axon intersections for each sub-region. A sphere (radius = 10 μm) was used along with a sampling grid 2700 μm × 2700 μm. ROIs corresponding to the grey matter of the agranular and dysgranular sub-regions were drawn at a low magnification (using a 2.5 × objective) and axonal quantification was completed at higher magnification (100 × oil objective). Only nerve fibers were counted and when a counting frame/sphere contained tangles or other pathologies, they were not counted and the following counting frame was used. Moreover, a fiber was counted only when it fully intersected with the sphere at least once. Tissue thickness was measured manually at each sampling frame (mean, 22 ± 1.42 μm) and coefficient of error (CE) was calculated for each sub-region (mean CE agranular = 0.13 ± 0.06; CE dysgranular = 0.12 ± 0.05).

To calculate estimated axonal length (L), total number of intersections of fibers with space balls (Qi) throughout all sections was multiplied by the volume of sampling frame. The volume (volume = grid X × grid Y × section thickness) is divided by the surface area of sphere (a = \(4\times \pi \times {r}^{2}\)) multiplied by the reciprocal of sampling fraction of the section (1:20) [44, 46]. To calculate density, total axonal length (L) was divided by sampled reference volume per ROI. Reference volume was derived through planimetry, calculated as a measure of total area of the ROI multiplied by section height [44, 46].

Statistical analyses were performed using SPSS (version 26.0, IBM) and graphs were made using graph prism, version 9. When tissue or staining quality was insufficient, cases were omitted from analysis. Demographics of controls, PD, PDD and DLB were compared using chi-square test for categorical data and ANOVA, with age corrections, for numerical data. To analyze differences in axonal length density between insular sub-regions (agranular and dysgranular) across all three patient groups (PD, PDD, DLB), an ANCOVA, with adjustment for age, and post-hoc paired t-test were used. Quantitative scores for p-tau and α-synuclein pathology load showed non-normal distribution and were log-transformed followed by parametric analysis with ANCOVA using age as a covariate followed by post-hoc paired t-test for group differences. For amyloid-β, due to the presence of multiple zero scores from pathology quantification, scores were dichotomized (0 or 1) and analyzed using logistic regression. To assess the pathology and group effects on axonal density, variables were pooled into the model and a nested linear mixed model analysis was performed using backward elimination. Axonal length density was the dependent variable, area% load of p-tau, α-synuclein, and amyloid-β were main effects and age was a covariate. Spearman’s rank correlation was computed to assess the relationship between CDR scores and age, group, disease duration, axonal density, p-tau, α-synuclein, and amyloid-β scores. Analyses were then re-computed per gender. Correction for multiple comparisons was not performed. Statistical significance was set at P < 0.05.

All included donors had moderate to severe disability (estimated Hoehn and Yahr scale, 4–5). DLB donors were mostly males (DLB = 78% vs PDD = 54% males), significantly younger than PDD donors (71 ± 6.1 years vs 81.4 ± 8.4 years; P = 0.02) and had a significantly shorter disease duration (5 ± 2.4 years) compared to both PD and PDD (18 ± 4 years and 13.4 ± 5 years, respectively; P = 0.001). All disease groups had advanced Braak α-synuclein stages, whereas Braak NFT stages were most advanced in PDD and DLB, and accompanied by glial tauopathy (ARTAG). The DLB group also had advanced amyloid-β Thal phases (range: 3–5) except for one younger patient (DLB_5, Thal phase = 0). CAA was more frequent (66% of cases) in the DLB group compared to PDD (25%) and PD (20%) and included both type I and type II CAA. The demographics, age-of-onset, disease duration, CDR scores, and neuropathological staging of all donors included in this study are summarized in Table 1.

CDR global scores (n = 22) showed a significant positive correlation with group as well as p-tau scores, indicating that worse cognitive performance (higher scores) correlates with DLB group (r(42) = 0.73; P < 0.001) and higher p-tau load (r(38) = 0.37, P = 0.017). We also found a significant negative correlation between CDR scores and disease duration, indicating a shorter disease duration correlates with worse cognitive performance (r(42) =  − 0.4; P = 0.008). A significant negative correlation was also found between CDR scores and axonal length density within the dysgranular insula (r(17) =  − 0.5; P = 0.04). Correlations computed based on gender showed significant results only for the female gender, including a significant negative correlation between CDR scores and disease duration (r(16) =  − 0.74; P < 0.001) as well as axonal density (r(16) =  − 0.52, P = 0.02), and a significant positive correlation between CDR scores and p-tau (r(15) = 0.82, P < 0.001) as well as α-synuclein scores (r(10) = 0.93; P < 0.001).

We observed Lewy bodies (LBs), mostly in pyramidal neurons of deeper layers (V-VI), and Lewy neurites (LNs) including bulgy LNs (Fig. 1a–c) across all layers in both subregions. Cytoplasmic granular α-synuclein deposits (Fig. 1d–e), string-shaped and circular α-synuclein deposits (Fig. 1f–g) as well as astrocytic synucleinopathy (Fig. 1h–j) were observed, particularly in the agranular insula of PD and PDD donors. The anterior insula in DLB donors showed more severe α-synuclein pathology with extensively dense LNs and astroglial star-shaped synucleinopathy (Fig. 1k–m). In one PDD donor, vascular synucleinopathy was observed in the anterior insula (Fig. 1i–j). Furthermore, by staining astrocytes, α-synucleinopathy was found within the GFAP-positive astrocyte cell bodies and surrounding processes in both sub-regions (Fig. 1n–p). Colocalization of GFAP and α-synuclein is also shown (Fig. 1q–s).

The anterior insula of the PD, PDD and DLB donors showed various degrees of severity for p-tau and amyloid-β pathology. NFTs, ghost tangles, glial tauopathy (including ARTAG and subpial thorny astrocytes), and neuritic plaques were observed in both sub-regions (Fig. 2a-l). We also observed fragmentation of NFTs in the agranular insula along with argyrophyllic grain disease (AGD) deposits in DLB (Fig. 2e-l). p-Tau pathology was also observed in a fork cell in deeper layers of the agranular insula (Fig. 2b). Neuropil threads, ghost tangles, and neuritic plaques were most prominent in DLB. Amyloid-β plaques were observed in the dysgranular insula and comprised of diffuse plaques as well as dense core plaques. Meningeal amyloid-β, CAA type I and II, and dyshorric amyloid (i.e. perivascular amyloid deposits) were most prominent in the dysgranular insula in DLB patients (Fig. 2m–u).

Neurodegenerative features such as localized axonal thickening, thinning, spheroids and myelin abnormalities were observed in the agranular insula in all donors (Fig. 3). In cases with more advanced Thal phases, amyloid-β plaques surrounding axons and bulbous axonal swellings were observed in the vicinity of the plaque (Fig. 3h–j). In PD donors, early NFT formation (pre-tangles) was seen associated with dystrophic neurites (Fig. 3k–m). NFTs maturing to ghost tangles were particularly present in DLB donors, where severe axonal degeneration and AGD deposits, seen as comma shaped deposits, were commonly seen (Fig. 3o–s).

To assess cytoskeletal changes in the anterior insular subregions in PD, PDD and DLB donors, we studied morphological changes in the pattern of NfH and PLP (myelin) immunoreactivity. We observed blister-like myelin formations around thickened axonal segments in both sub-regions, not associated with pSer129 α-synuclein aggregates, in all disease groups (Fig. 4a, j). In DLB, severe myelin changes were observed, including swellings (containing multiple vacuoles), demyelination, and myelin blisters with disrupted axon-myelin continuity (Fig. 4a, j, k). NfH features consisted of fragmentation, extracellular debris and spheroids with single axonal connections, indicating axonal transection (Fig. 4d–g). Extracellular encagement of LBs and donut-shaped NfH aggregates were also observed (Fig. 4c, h). Bulgy LNs were seen in myelinated fibers as well as non-myelinated fibers in the agranular and dysgranular insular sub-regions (Fig. 4b, l).

To evaluate whether there were differences in the burden of α-synuclein and co-occurring AD pathology across groups and sub-regions, we quantified the load of protein pathology using area fractionation on free-floating 60-μm thick insular sections. Comparisons between the PD, PDD and DLB groups showed significant differences in the load of α-synuclein (n = 23) pathology in the dysgranular insula (F(2,18) = 4.9, P = 0.019) and a trend for the agranular insula (F(2,18) = 3.5, P = 0.054). The DLB group showed significantly higher scores compared to PD and PDD in dysgranular subregion and compared to PD only in agranular subregion (P = 0.016 and 0.015 for dysgranular region; P = 0.02 for agranular region; Fig. 5). No difference was found between PD and PDD (P = 0.86). Between the agranular and dysgranular insula, α-synuclein pathology load was significantly higher in the agranular insula in all groups (mean α-synuclein % area, 3.25% in agranular and 1.55% in dysgranular; t(21) = 5.3, P < 0.001). For p-tau (n = 26), we also observed a significant difference between disease groups for both the agranular (F(2,19) = 5.5, P = 0.01) and dysgranular insula (F(2,19) = 7.1, P = 0.005). The DLB group showed a significantly higher load of p-tau compared to PD and PDD groups for both sub-regions (agranular: P = 0.01 compared to PD and P = 0.006 compared to PDD; dysgranular: P = 0.004 compared to both groups). Comparing p-tau pathology within insula sub-regions across groups, the agranular insula showed significantly higher p-tau load compared to the dysgranular insula (mean = 18.0% and 7.8%, respectively; t(19) = 5.1, P < 0.001). There were no group differences for amyloid-β pathology (n = 26; χ²(2) = 3, P = 0.22 for agranular sub-region; χ²(2) = 2.6, P = 0.26 for dysgranular sub-region). However, comparing amyloid-β pathology within the anterior insular sub-regions across groups, the dysgranular insula (mean = 5.0% ± 4.8%) showed significantly higher amyloid-β load compared to the agranular insula (mean = 3.0% ± 8.3%; t(25) =  − 2.2, P = 0.037) (Fig. 5).

To account for gender differences, analyses were repeated per gender. For α-synuclein load, there were siginificant differences between groups only for females (F(2,14) = 49.7, P < 0.001). Post-hoc analysis showed significant differences between DLB and PD and PDD (P < 0.001). Similarly, for p-tau, females showed significant differences between groups (F(2,19) = 12.68, P < 0.001). The significant differences were found between DLB and PD (P = 0.002) as well as PDD (P < 0.001). Males also showed a trend of group differences (F(2,21) = 3.46, P = 0.046), which was due to significant difference between DLB and PDD only (P = 0.014).

In summary, across groups, the agranular insular showed more severe α-synuclein and p-tau, but not amyloid-β pathology which was significantly higher in the dysgranular sub-region. Furthermore, α-synuclein pathology in the dysgranular insula significantly differed between DLB and PDD, while p-tau was higher in DLB compared to both PD and PDD. For amyloid-β, no differences between groups were observed.

Using modified Bielschowsky silver impregnation and stereology, axonal length and axonal length density (n = 25) were assessed within the anterior anterior insular sub-regions across all groups. Taking all groups together, the mean total axonal length density for the agranular and dysgranular insula were (8.00 ± 2.14)  ×10⁻⁹ m/m³  and (10.6 ± 2.86) × 10⁻⁹ m/m³, respectively. The axonal length density was significantly higher in the dysgranular compared to the agranular insula among all groups (t(23) =  − 2.7, P = 0.01; and t(23) =  − 5.7, P < 0.001). A significant effect of disease group on axonal length density was only observed in the dysgranular sub-region (F(2,23) = 3.7, P = 0.04; agranular: F(2,23) = 2.5, P = 0.11). Specifically, a significant reduction was found in DLB (8.66 ×10⁻⁹ m/m³) compared to PD (11.75 ×10⁻⁹ m/m³) and PDD donors (11.94 ×10⁻⁹ m/m³) (P = 0.04 and 0.02, respectively) (Fig. 5).

Next, we aimed to understand the role of each pathology on axonal degeneration and loss. Using a linear mixed model including all pooled variables (axonal density, disease group, age, p-tau, amyloid-β, and α-synuclein pathology load), p-tau load showed the highest association with axonal length density in both anterior insular subregions (β =  − 3.45 × 10^–5; F(1,42) = 4.2; P = 0.046). Post-hoc analysis showed a significant effect for p-tau on axonal length density in the DLB group (β =  − 5.04 × 10^–5; t(24.2) =  − 3.24, P = 0.003) compared to PD (β =  − 1.5 × 10^–5; t(40.6) =  − 0.42, P = 0.67) and PDD (β =  − 6.4 × 10^–5; t(41.3) =  − 0.7, P = 0.5) as well as for the agranular compared to dysgranular insula (β =  − 5 × 10^–5; t(42.3) =  − 3, P = 0.004). Testing the effect of other pathology variables on axonal density showed a significant effect for α-synuclein (β =  − 3 × 10^–4; t(29) = -3, P = 0.006) but no significant effects for amyloid-β (P = 0.77). Post-hoc analysis for group and sub-region effects on axonal length density showed significant effects for the DLB group (β =  − 3.34 × 10^–4; t(45.7) =  − 3.3, P = 0.002) compared to PD (P = 0.4) and PDD (P = 0.7) and agranular compared to dysgranular sub-region (β =  − 3 × 10^–4; t(36.6) =  − 2.7, P = 0.009) (Fig. 5).