Aggregates of RNA Binding Proteins and ER Chaperones Linked to Exosomes in Granulovacuolar Degeneration of the Alzheimer’s Disease Brain

Antibodies

All primary and secondary antibodies and their dilutions used in this study are listed in Supplementary Table 1. Many of these antibodies have been used by us in previous published studies (see references in Supplementary Table 1). Rabbit polyclonal VAPB antibody was custom made and used by us in previous studies [31–34].

Human postmortem brains

Human postmortem brain samples fixed in buffered formalin (n = 12 AD patients, n = 9 age-matched controls; Table 1) were obtained from the archives of the Institute of Neuropathology, RWTH University Hospital, Aachen (Germany) and from the Department of (Neuro) Pathology, Academic Medical Center (AMC), University of Amsterdam, The Netherlands (see Table 1). Both control and AD cases were selected from a retrospective searchable neuropathologic database, including cases with consent for postmortem brain autopsy and use of their brain material and medical records for research. Controls included in the present study were adult individuals without a history of neurological diseases and without evidence of cognitive decline, based on the last clinical evaluation. We included only cases in which an extensive neuropathological protocol was used (based upon the recommendations of the Brain-Net Europe consortium [35]). The AD cases included in the present study were individuals with a clinical diagnosis of probable mild to moderate AD [36, 37], more recently defined as “Alzheimer’s clinical syndrome” [38]. The selected cases were then re-assessed by neuropathologists using the routine updated protocol for neurodegenerative diseases [39, 40], and included cases with neuropathological verification and AD neuropathological changes and similar load of GVD based on the immunolabelling of GVBs in subicular neurons using the established marker CK1δ [17] (Table 1). We evaluated TDP-43 pathology in both AD and controls cases [41] using routinely used antibody against pTDP-43 (Cosmo Bio Co. LTD, TIP-PTD-M01). None of the control cases displayed TDP-43 pathology. TDP-43 pathology was, however, observed in 5 out of 12 cases, with mild to moderate TDP-43 pathology in the hippocampus (Table 1). We used these cases for co-labelling of pTDP-43 and GVBs. These cases did not show any apparent/significant differences compared to non-TDP-43 positive AD cases in the pattern of immunolabelling of other RBPs and of the proteostasis factors studied. We excluded cases with co-existing other neuropathological lesions (i.e., microvascular infarcts, hippocampal sclerosis, alpha-synuclein inclusions) to avoid confounding factors. The samples were used in compliance with the Declaration of Helsinki. The studies were approved by the Ethical Committees of the Medical Faculty, RWTH Aachen (EK 127/18), and of the Academic Medical Center, Amsterdam (W11_073). The postmortem tissues had been obtained within 6–30 h after death (see Table 1).

Transgenic (tg) mouse brains

CK1δ positive GVD bodies (GVBs)

Semi-quantitative assessment of GVD was performed in sections stained with DAB immunohistochemistry using CK1δ antibody as a marker for GVBs [17]. Three hippocampal sections each from the above-mentioned cases AD and controls were stained (see Table 1 for details).

Diaminobenzidine (DAB) immunohistochemistry

3–4μm paraffin sections were placed on poly-L-lysine coated slides and allowed to dry in an oven (37°C) overnight and then processed for immunohistochemistry described in detail elsewhere [33]. The sections were deparaffinized in xylene for 20 min and rehydrated in 100%, 96%, and 70% ethanol for 5 min each followed by endogenous peroxidase quenching (0.3% H₂O₂ in methanol) for 20 min. For antigen retrieval, sections were heated in citrate buffer, pH 6 (DAKO), for 20 min in a pressure cooker. After washing in PBS, sections were incubated with primary antibody (Supplementary Table 1) for 1 h at room temperature or at 4°C overnight. After washing in PBS, sections were incubated with appropriate polymeric HRP-linker secondary antibody (IL Immunologic, Duiven, The Netherlands) for 30 min at room temperature. DAB reagent (DCS Innovative Diagnostic System DAB kit) was used to visualize antibody binding. The sections were then counter-stained with 6% hematoxylin for 3 min. In the case of 82E1 antibody, we used Biotin secondary antibody conjugate (Vector Laboratories, Burlingame, CA, USA) followed by detection using the Vectastain ABC HRP kit (Vector Laboratories). All procedures were performed at room temperature.

Immunofluorescence

Single and double immunofluorescence staining were performed as described by us already elsewhere [33, 42]. In brief, deparaffinized tissue sections were heated in citrate buffer, pH 6 (Dako), for 20 min in a pressure cooker. Sections were then blocked (to avoid non-specific bindings) with ready to use 10% normal goat serum (Life Technologies, MD, USA) for 1 h at room temperature before incubating with primary antibody 4°C overnight. After two times washing in TBS-T for 10 min, the sections were incubated with Alexa conjugated secondary antibody (1 : 500 in PBS) at room temperature for 2 h. Sections were washed in TBS-T (2 x 10 min) and stained for 10 min with 0.1% Sudan Black in 80% ethanol to suppress endogenous lipofuscin auto-fluorescence. Finally, the sections were washed for 5 min in TBST and mounted with Vectashield mounting medium (Vector Laboratories) containing DAPI.

Image acquisition

Images of the DAB-stained sections were taken with a Zeiss Axioplan microscope equipped with a 40x objective and an Axio Cam 506 color camera (Zeiss). Images from immunofluorescence labelled sections were taken with a Zeiss LSM 700 laser scanning confocal microscope using 20x, 40x, and 63x objectives (Zeiss). Images were acquired by averaging 4 scans per area of interest resulting in an image size of 1024x1024 pixels. The laser intensity was kept constant for all the samples examined. Captured confocal images were analyzed using Adobe Photoshop CS5 and ZEN (Blue edition) 2009 software.

Semi-quantitative analysis of co-immunolabelling (human tissue)

Average pixel intensity of the target proteins (SigR1 and VAPB; Fig. 1B, F)/field of view were quantified in three representative sections from each of six AD patients (Braak stages V-VI) and in three representative sections from each of three controls (Braak stage 0) in 8–10 random fields, at low magnification (20x), capturing approx. 30-40 neurons/field of view. Average background pixel intensity was subtracted. Average pixel intensity of the target proteins (SigR1, GRP78, VAPB; Fig. 1C, D, F) in neurons were quantified (E, F) using one representative section each from six AD (Braak stages V-VI) patients and one representative section each from three controls (Braak stage 0). Approx. 20–30 neurons / field of view in 8-10 random 20x fields were found, examining around 100–150 neurons per case by measuring average pixel intensities of the regions of interest drawn around each neuron. We used the unpaired Student’s t-test for comparison between two sample groups. Values represent the mean±standard deviation (SD). Differences between values were regarded as significant when=*p<0.05.

OPEN IN VIEWER

Fig.1

A) Left panel: Representative H&E stained subicular neurons of age-matched control brain compared to an AD case. Arrows: GVBs. Right panel: CK1δ immunoreactive GVBs (arrows) in AD subicular neurons. Scale bars: 20μm. B) Increased levels of the ER chaperone SigR1 in subicular neurons of AD patient compared to the age-matched control. Scale bars: 20 μm. Values represent the mean±SD (**p<0.05). C-F) Double immunofluorescence labeling for CK1δ and for the ER chaperones SigR1 (C) GRP78 (D) and for the UPR co-factor VAPB (F). Note the overall increased levels of SigR1, GRP78 and VAPB in AD subicular neurons (white arrows), but reduced levels of SigR1, GRP78 (C, D, E) and VAPB (F) immunoreactivity in subicular neurons harboring GVBs (red arrowheads, quantification). GVBs harboring neurons are represented as CK1δ (+ve) and neurons without GVBs are denoted as CK1 (–ve) (Scale bars: 20 μm. Values represent the mean±SD (**p<0.05). G) Double immunofluorescence labeling of SigR1 and pTau (AT8). Note the Co-localization of increased SigR1 immunoreactivity with granular pTau (arrow) but decrease of SigR1 immunoreactivity (red arrowheads) in neurons containing mature pTau tangles in AD subicular neurons. Scale bars: 20 μm. H) Immunoblot analysis of hippocampal tissue homogenates from AD cases and their age matched controls showing an increase in ER chaperons SigR1, GRP78 and proteotoxic stress marker HSP70 in AD.

Semi-quantitative analysis of co-immunolabelling (mouse tissue)

Average pixel intensity of the target proteins (SigR1 and VAPB; Supplementery Figure 2D, E) / field of view were quantified in one representative section from each of three APP/PS1 (50-week) transgenic mice and one representative section each from three age-matched littermates. 8–10 random fields in low magnification (20x) were examined, capturing approximately 20–30 neurons/field of view. Average background pixel intensity was subtracted. We used the unpaired Student’s t-test for comparison between two sample groups. Values represent the mean±standard deviation (SD). Differences between values were regarded as significant when *p<0.05.

OPEN IN VIEWER

Fig.2

A) Left panel: Representative H&E stained CA1- subicular neurons of 104-week-old-pR5 tg mice compared to an age-matched wt mouse. Arrows: GVBs. Right panel: CK1δ immunoreactive GVBs (arrows) in pR5 tg mice CA1- subicular neurons. Scale bars: 15μm. B) Increased levels of the ER chaperone SigR1 in CA1- subicular neurons compared to the wt mice. Scale bars: 15μm. C, D) Double immunofluorescence labeling for CK1δ and for the ER chaperones SigR1 (C) and GRP78 (D). Note the reduced levels of SigR1, GRP78 immunoreactivity in neurons harboring GVBs (red arrowheads). White arrows: Increased levels of SigR1 and GRP78 in neurons without GVBs. Scale bars: 15μm. E) Double immunofluorescence labeling of SigR1 and pTau (AT8). Note the Co-localization of increased SigR1 immunoreactivity with pTau: Scale bar: 15μm.

Western blot analysis

Western blots of brain tissue were performed as described previously [33]. Briefly, human and mouse brain tissues were homogenized in ice-cold radioimmunoprecipitation assay (RIPA) lysis buffer, with protease inhibitor cocktail (Roche Life Science, Penzberg, Germany). The crude lysates were briefly centrifuged and then processed for the bicinchoninic acid protein assay (BCA; Thermo Scientific) according to the manufacturer’s protocol. Equal amounts of protein were boiled in Laemmli sample buffer for 5-10 min and processed for SDS-PAGE. The protein gels were transferred onto polyvinylidene fluoride (PVDF) membranes. The blots were then blocked with 4% skimmed milk in TBS-T for 30 min and incubated with specific primary antibody (Supplementary Table 1) overnight at 4°C under agitation. Thereafter, the blots were washed three times in TBS-T for 10 min each and incubated for 1 h with appropriate horseradish peroxidase (HRP)-conjugated secondary antibody (Thermo Scientific). Immunoreactive protein bands were visualized by exposing the blots on an X-ray film (Fuji Films). Quantification of the band intensities was performed after normalizing to tubulin levels using Adobe Photoshop CS5. Values represent the mean±SD (**p<0.05).

ER stress and aberrant association of ER chaperones with GVBs

GVD is predominantly present in brain regions that are most affected by AD pathology such as hippocampus (CA1, CA2, CA4, and subiculum) [29]. GVBs can be detected as membrane bound vesicular structures by routine hematoxylin-eosin (Fig. 1A) or by silver staining. We used the established GVB marker, Casein kinase 1 isoform delta (CK1δ) [29] to label GVBs in subicular neurons (Fig. 1A; Supplementary Figure 1A).

ER chaperones are critical regulators in maintaining proteotoxic effects and/or autophagy overload due to Aβ and pTau accumulations in AD brain [43]. Consistent with this, we observed significantly elevated overall immunoreactivity for the ER chaperone SigR1 in subicular neurons of the AD brains (Fig. 1B). However, those neurons which were affected by GVD, which comprise about 25% of the total neuronal population in the cases examined by us, showed lower SigR1 levels, and there was only minor co-localization of SigR1 with GVBs (Fig. 1C). Similarly, immunoreactivity for GRP78, a prime initiator of the UPR, was reduced in subicular neurons harboring GVBs (Fig. 1D, E), and VAPB, a UPR co-factor [44], was found to be generally increased in subicular neurons of AD brains. Those neurons which harbored GVBs showed a less pronounced VAPB buildup (Fig. 1F). GVBs showed only minor, if any colocalization with any of these ER chaperones (Fig. 1C, D, F). Similarly, human AD neurons harboring small granular pTau deposits showed elevated SigR1 immunoreactivity, and these pTau aggregates were regularly co-localized with SigR1 (Fig. 1G; Supplementary Figure 1B), suggesting that SigR1 chaperone is involved in clearing such small granular pTau accumulations. Neurons with mature NFTs lacked both the chaperone buildup and significant GVBs (Fig. 1G; Supplementary Figure 1C). On the other hand, moderately affected dentate gyrus neurons [45] which carried only a minor load of pTau and GVBs (Supplementary Figure 1G) and showed particularly high levels of SigR1 (Supplementary Figure 1E) and GRP78 (not shown). The overall increased levels of ER chaperones were confirmed by western blot analyses of AD patient and age-matched control hippocampi (Fig. 1H).

ER chaperones and GVBs in pR5 Tau tg mice

The pR5 mouse strain is a well-established tauopathy model that develops NFTs due to overexpression of the longest human tau isoform harboring the P301 L mutation [46, 47]. We confirmed that 104-week-old pR5 mice displayed CK1δ immunoreactive GVBs coinciding with NFT pathology [8, 9] in hippocampus (Fig. 2A; Supplementary Figure 1A). These GVBs were morphologically similar to the GVBs present in human AD; they have already been shown to be immunoreactive for several UPR markers [9]. We found that, similar to the situation in human AD brain, CA1-subicular neurons of pR5 mice consistently showed overall increased levels of the ER chaperones SigR1, GRP78, and VAPB (Fig. 2B-D; Supplementary Figure 1F). Again, those neurons that harbored GVBs showed decreased SigR1 and GRP78 levels, and there was only minor, if any co-localization of SigR1 with GVBs (Fig. 2C, D). Also consistent with the changes observed in human AD brain, granular, pre-tangle pTau aggregates co-localized with SigR1 immunoreactivity (Fig. 2E) and some of them were also found to be co-localized with GVBs (Supplementary Figure 1H).

CK1δ-positive incipient GVBs and ER chaperones in APP/PS1 transgenic mice

APP/PS1 tg mouse brains showed prominent extracellular Aβ plaque pathology (Supplementary Figure 2A), but their hippocampal pyramidal neurons, in contrast to the pR5 mice, harbored only rather small, probably incipient accumulations of GVB-like, CK1δ-immunoreactive granular structures (Supplementary Figures 1A, 2C). They were distinct from mature human GVBs in many respects, such as smaller size, frequent presence of granular cores without outer layer morphology, and recurrent appearance in distinct groups, but not scattered over the neuronal cytoplasm (Supplementary Figures 1A, 2B, C). Like human AD and pR5 mouse neurons, SigR1 and VAPB were found at overall elevated levels in APP/PS1 mouse neurons; however, in contrast to the human AD patient and pR5 mouse GVBs, the CK1δ-immunoreactive granules in APP/PS1 mouse neurons were SigR1- and VAPB-positive (Supplementary Figure 2D, E). Similarly, these structures were strongly ubiquitin immunoreactive (Supplementary Figure 2F). Consistent with the immunohistochemical results, immunoblot analysis of total APP/PS1 mouse brain homogenates revealed a significant age-dependent increase in the proteotoxic stress markers SigR1, GRP78, Hsp70, and Hsp27 as well as ubiquitin conjugates (Supplementary Figure 2G).

RNA binding protein homeostasis and GVBs

pTau can alter autophagy at multiple steps [48] and can trigger the formation of GVBs in vivo as well as in vitro [25]. Similarly, DPR aggregates, which alter autophagy pathways [26], were found within GVBs in neurons of ALS/FTD patients with C9orf72 expansion [15]. RBPs such as TDP-43 form aggregates in AD patient neurons and quite often granular pTDP-43 aggregates co-localize with GVBs in AD patient hippocampal neurons [29]. Furthermore, RBPs associated with SGs progressively accumulate together with tau in mouse models of tauopathy, as well as in human AD and FTLD-Tau brain tissue [49]. Therefore, in an attempt to identify the trigger for GVD, we next examined RBPs in subicular neurons affected by GVD. Subicular neurons from AD patients contained several NFTs, resembling pTDP-43-immunoreactive aggregates and many granular or vesicular aggregates (20–30%), some of which were co-localized with GVBs as described previously by others [29, 50] (Fig. 3A, lower panel). Consistent with this, pR5 mouse hippocampal neurons also frequently showed globular and vesicular pTDP43-immunoreactive structures (Fig. 3B) which were co-localized with the GVB marker CK1δ (Fig. 3B, lower panel). Even though pR5 mouse neurons displayed abundant pTau-positive NFTs, larger tangle-like pTDP-43 aggregates were absent (Fig. 3B). To decipher whether the co-localization of pTDP-43 with GVBs is indicative of a more general buildup of RBPs in AD neurons, we examined AD brains with antibodies against Matrin 3 and FUS as well as SG proteins used by us previously when studying ALS pathomechanisms [42, 52]. Matrin 3 is of prime importance because it regulates the transcription and stability of RBPs including TDP-43 and FUS [53]. Moreover, MATR3 mutations lead to ALS and distal vacuolar myopathy in patients [54–56]. We found nuclear accumulation and cytoplasmic mis-localization of Matrin 3 in those neurons which contained GVBs. These globular or vesicular cytoplasmic Matrin 3-immunoreactive structures co-localized with GVBs (Fig. 3C). In contrast, neurons in control cases showed basal levels of nuclear Matrin 3-immunoreactivity and no cytoplasmic Matrin 3 (Fig. 3C, upper panel). Of note, Matrin 3 levels were unchanged and there was no obvious cytoplasmic mis-localization of Matrin 3-immunoreactivity present in pR5 mouse hippocampal neurons (Fig. 3D). Consistent with the pattern of Matrin 3-immunoreactivity, cytoplasmic mis-localization and co-localization of FUS together with GVBs (Fig. 3E) was observed in human AD neurons. However, there was no clear-cut link between GVD pathology and reduced or increased intensity of nuclear FUS staining. Interestingly, even though pR5 hippocampal neurons showed significant granular cytoplasmic FUS labelling, FUS immunoreactivity was not co-localized with GVBs (Fig. 3F; lower panel).

OPEN IN VIEWER

Fig.3

A) DAB immunohistochemistry (upper panel) for pTDP-43 showing immunostaining associated with NFT-like structures and double immunofluorescence labeling (lower panel) demonstrating co-localization of pTDP-43 immunoreactivity with GVBs in human AD subicular neurons. Scale bars: 15μm. B) DAB immunohistochemistry for pTDP-43 showing diffuse nuclear immunoreactivity in wt mice CA1-subicular neurons (upper panel) and cytoplasmic granular accumulations (middle panel, white arrows) and rare globular accumulations (middle panel, black arrow) in pR5 tg mice. Double immunofluorescence labeling (lower panel) demonstrating co-localization of pTDP-43 immunoreactivity with GVBs in pR5 tg mice CA1- subicular neurons. Scale bars: 15μm. C) Matrin 3 and CK1δ double immunofluorescence: Increased levels of Matrin 3 in the nucleus and vesicular immunoreactive structures co-localizing with GVBs in human AD subicular neurons (middle and lower panels). Note the normal nuclear Matrin 3 levels in a neuron from the control case (upper panel). Scale bar: 20μm. D) Matrin 3 and CK1δ double immunofluorescence showing no cytoplasmic mis-localization of Matrin 3 as well as no co-localization with GVBs in pR5 tg mice CA1-subicular neurons. Scale bar: 15μm. E, F) FUS and CK1δ double immunofluorescence: Strong co-localization of cytoplasmic FUS accumulations with GVBs in human AD patient subicular neurons (E), while there is no co-localization of cytoplasmic FUS with GVBs in pR5 tg mice CA1-subicular neurons (F). Scale bars: 15μm. G) CK1δ-immunoreactive GVBs often co-stained with G3BP. Note the loss of nuclear G3BP immunoreactivity (white arrowheads) in the neurons containing GVBs. Scale bars: 15μm. (H) Loss of S6-kinase nuclear immunoreactivity (white arrowhead) and increased cytoplasmic accumulation in subicular AD neurons, which consistently showed CK1δ labeling (White arrow). Note the S6-Kinase-immunoreactive NFT-like structure (red arrowhead). Scale bars: 20μm. I) Increased cytoplasmic accumulations of S6-kinase in CA1-subicular neurons of pR5 tg mice, which were consistently co-localized with CK1δ labeling (arrow). Scale bar: 15μm.

Because aggregation of RBPs proceeds through the SG pathway, we stained SGs together with GVBs. Using antibodies against the SG marker G3BP as previously described by us and others [42, 57], we found that many human AD as well as the pR5 mouse neurons develop cytoplasmic accumulations of G3BP, along with reduced nuclear G3BP immunoreactivity (Fig. 3G). These cytoplasmic accumulations were often CK1δ-positive (Fig. 3G). SGs including processing bodies (P-bodies), which are also cytoplasmic accumulations of non-translating messenger ribonucleoprotein complexes (mRNPs), are cleared by autophagy [58]. Compared to the controls we found increased granular cytoplasmic immunoreactivity and reduced nuclear immunoreactivity for the P-body marker S6 kinase in AD brain subicular neurons (Fig. 3H; Supplementary Figure 3A); some of the granular S6 kinase staining co-localized with CK1δ-positive GVBs (Fig. 3H, lower panel). Similarly, increased granular cytoplasmic immunoreactivity of S6 kinase co-localized with CK1δ-positive GVBs in pR5 hippocampal neurons (Fig. 3I). Taken together, abnormal cytoplasmic RBP aggregates co-localize with GVBs, suggesting that RBP aggregates possibly trigger GVB formation and that GVBs might actually be involved in the removal of such aggregates. They are in line with a recent study showing that DPR aggregates contribute to GVD pathology in C9orf72-related FTLD/ALS [15].

Endosomes, exosomes, and GVD

Recent studies suggested that exosomes from AD brain contain Aβ oligomers and that exosomes may act as vehicles for the neuron-to-neuron transfer of toxic Aβ oligomers in cell culture models [59, 60]. Blocking the formation, secretion or uptake of exosomes was found to reduce both the spread of oligomers and their toxicity in these cell culture systems [59, 60]. Considering that most exosomal vesicles originate from endosomes [61], we therefore examined endosomes in neurons affected by GVD. Staining with antibodies against Rab11 and Mannnose-6-phosphate confirmed that these late endosomal markers co-localize with GVBs in AD patient subicular neurons (Supplementary Figure 3D), as shown already by others [17]. Additional immunohistochemistry revealed that GVBs also co-localize with the early and intermediate endosomal markers EEA1 and Rab7 (Supplementary Figure 3D) and with the Golgi marker GM130 (Supplementary Figure 3C). Of note, in a recent study we showed that the Golgi protein GolginA4, that was found to be elevated in cerebrospinal fluid of AD patients, is a major constituent of GVBs [62] (Fig. 4A). Thus, we hypothesized that GVBs might actually undergo exocytosis.

OPEN IN VIEWER

Fig.4

A) Co-immunofluorescence labeling of CK1δ-immunoreactive GVBs together with Golgin A4 in AD patient subicular and pR5 tg mice CA1-subicular neurons. Scale bars: 15μm. B) Increased size and density of Flotillin-1-immunoreactive vesicles in AD patient subicular neurons and in APP/PS1 tg mice and pR5 tg mice CA1-subicular neurons compared to the control and to wt mice. Upper right panel: DAB immunohistochemistry showing increased Flotillin-1-immunoreactive vesicles associated with GBVs (white arrows). Scale bars: 15μm. C, D) Co-immunofluorescence labeling of CK1δ-immunoreactive GVBs, together with Flotillin-1 in AD brain subicular neurons (C) and in pR5 tg mice CA1-subicular neurons (D). Note that while many GVBs are co-localized with Flotillin-1 (white arrows); few actually appear to be completely enclosed by Flotillin-1 labeled vesicles (yellow arrows). Scale bars: 10μm. E, F) Co-immunofluorescence of GolginA4 together with Flotillin-1, in AD brain subicular neurons (E) and in pR5 tg mice CA1-subicular neurons (F). Note that Golgin A4-labeled granules often co-localize with Flotillin-1 labeled vesicles (arrows). Scale bars: 10μm.

Even though not specific, universally applicable exosomal markers have evolved so far. Tetraspanins such as CD9, CD63, and CD81 and components of the endosomal sorting complex required for transport (ESCRT) including TSG101, Flotillin 1, and Alix [59, 63] are currently being used to label exosomal vesicles. We found a consistent pattern of immunolabelling of vesicles at low density in subicular neurons of human control cases and of control mice using both a monoclonal (BD Biosciences) and a polyclonal antibody (Cell Signaling) against Flotillin 1 (Fig. 4B). In contrast, incubation with the monoclonal antibody against CD9 (antibody online, Germany) yielded abundant cytoplasmic immunoreactivity in many other cells in both human AD and control and mouse brains (Supplementary Figure 3B). Based on these observations, we decided to focus on the monoclonal Flotillin 1 antibody (BD Biosciences) for our subsequent analyses. The same monoclonal antibody has also been used for the staining of human autopsy specimens in the recent study by Sardar Sinha et al. on exosomal mechanisms in AD [60]. We found that Flotillin 1-positive vesicles were markedly increased in number and size in subicular neurons of AD patients and in CA1-subicular neurons of pR5 and APP/PS1 tg mice; in many of these pyramidal neurons, Flotillin-1 immunoreactivity was co-localized with CK1δ-positive (Fig. 4C, D) and Golgin A4-positive GVBs (Fig. 4E, F).

Prompted by these observations, we next examined whether these Flotillin-1-positive vesicles also contain pTau and RBP aggregates, as detected by us in GVBs (see above and Fig. 3, Supplementary Figure 1). We observed frequent and prominent co-localization of Flotillin-1-immunoreactive vesicles with granular pTau in human AD subicular neurons and in pR5 mouse hippocampal neurons (Fig. 5A). Some of these Flotillin-1-immunoreactive vesicular structures actually appeared to enclose smaller granular pTau immunoreactive species (Fig. 5A; arrows). Similarly, pTDP-43-immunoreactive aggregates were found to be co-localized with or even enclosed by Flotillin-1-positive vesicles (Fig. 5B) in both human AD and pR5 mouse neurons. In line with a recent report showing co-localization of 82E1-immunoreactive intracellular Aβ oligomers with Flotillin-1-positive granular structures in human AD brain [60], we found (albeit minor) co-localization of Flotillin-1-positive vesicles with 82E1-positive intracellular deposits in human AD subicular neurons (Fig. 5C) and in hippocampal APP/PS1 mouse neurons (Fig. 5D). However, only very minor, if any, co-localization of Aβ-(81E1-) immunoreactive intraneuronal material with CK1δ-positive GVBs was observed by us in AD patient subicular neurons (not shown). There was no detectable co-labelling of extracellular 82E1-positive Aβ in AD patient and APP/PS1 mouse hippocampi with our Flotillin-1 antibody (not shown).

OPEN IN VIEWER

Fig.5

A, B) Double immunofluorescence for Flotillin-1 together with pTau- (A) and pTDP-43 (B), demonstrating that these aggregates are often co-localized with and/or are enclosed by Flotillin-1 labeled vesicles (arrows) in AD patient subicular and in pR5 tg mice CA1-subicular (lower panels), neurons, maximum intensity projections (below) to show the co-localization. Scale bars: 15μm. C) Double immunofluorescence of Aβ oligomer (82E1 antibody) and Flotillin-1. There was a rather rare, somewhat arbitrary coincidence of Flotillin-1-labeled structures with 82E1 immunoreactivity within human AD subicular neurons (arrows) in the vicinity of 82E1-labeled Aβ plaques (P). Scale bars: 15μm. D) In APP/PS1 tg mice 82E1 immunoreactive intracellular material is often co-localized with Flotillin-1-labeled vesicles (arrows) within subicular neurons in the vicinity of Aβ plaques (P). Scale bars: 10μm.

OPEN IN VIEWER

Fig.6

Schematic representation: pTau and RBPs can potentially trigger GVD in AD neurons. GVD affected neurons seldom contain pre-tangle pTau and show an activated UPR. SigR1 and other ER chaperones are elevated to minimize the misfolded protein stress as a protective mechanism. Persistent UPR could potentiate further tau phosphorylation and leads to mature tangle formation and promotes neurodegeneration. On the other hand, neurons with higher load of GVBs do not contain mature tangles. Some GVBs might undergo exocytosis.