Aberrant Rab12 phosphorylation in DLB

We sought to determine whether phosphorylation of LRRK2-associated Rabs, including Rab12 and Rab10, is altered in neurodegenerative diseases with tau and/or α-synuclein pathology. We first investigated Rab12 phosphorylation; to replicate previous findings that LRRK2 is the primary kinase for Rab12 phosphorylation in the brain, the cortex was dissected from WT and Lrrk2 KO mice and whole cell lysates were subjected to quantitative immunoblotting for pS106-Rab12 and Rab12. Consistent with previous reports, we observed that the level of pS106-Rab12 in cortex derived from Lrrk2 KO mice was significantly decreased compared to WT (Supplementary Fig. 3a, b) [62, 68, 70, 109, 110, 123]. However, since about 45% of the pS106-Rab12 signal remained, consistent with previous work, this suggests that LRRK2 is not the only kinase that phosphorylates Rab12 in mouse cortex, though whether this differs across brain regions is unknown [59, 62]. Further, it should be noted that LRRK2-dependent phosphorylation of Rab12 in mouse brain may not necessarily reflect which kinases phosphorylate which Rab proteins in the human brain.

We first focused on LBD cases with a clinical diagnosis of DLB, which despite being characterized by the presence of Lewy bodies, often displays moderate to severe 3R/4R tau pathology post-mortem and therefore is a relevant disease to interrogate α-synuclein and tau co-pathologies [17, 42, 49, 105]. The hippocampus and temporal cortex were chosen for analysis because they display high levels of tau pathology and manifest tau pathology early in disease progression, both in DLB and AD [12, 13, 20, 122]. Brains from subjects with DLB with a range of AD co-pathology levels, and unaffected controls, were subjected to quantitative immunoblotting for pS106-Rab12 and Rab12 (Table 1, Supplementary Table 1). Positive and specific pS106-Rab12 signal was observed in the soluble, cytosolic Dounce-homogenized (Dounce) fraction, and the membrane-enriched Triton X-100-extracted (Triton) fraction (Fig. 1a-d) [71]. Only non-specific signal was observed in the less-soluble SDS-extracted fraction (Supplementary Fig. 4). We discovered that levels of pS106-Rab12 relative to total Rab12 were significantly elevated ~ tenfold in Dounce and ~ sevenfold in Triton hippocampus fractions derived from subjects with DLB, when stratified by Braak neurofibrillary tangle stage (Braak > 3), relative to controls (Fig. 1e). We observed ~ twofold higher pS106-Rab12 signal (standardized to Hsp70 loading control) in DLB cases in the Triton hippocampus fraction, which was not statistically significant, although there was significantly higher hippocampus pS106-Rab12 signal in the Triton fraction compared to the Dounce fraction (Fig. 1f). Interestingly, there was a significant decrease in total Rab12 in both hippocampus fractions in DLB cases, partially explaining the increase in pRab12/Rab12 (Fig. 1g). Rab12 phosphorylation relative to total Rab12 was also elevated ~ 20-fold in temporal cortex fractions of DLB cases (Braak > 3) when compared to either control or DLB cases (Braak ≤ 3) (Fig. 1h). However, we did not observe significant elevation in temporal cortex pS106-Rab12 signal (standardized to Hsp70 loading control) in DLB cases (Braak > 3) compared to controls (Fig. 1i). Total Rab12 levels were significantly decreased in both Dounce and Triton fractions in the temporal cortex in DLB cases (Braak > 3) compared to controls (Fig. 1j).

Fig. 1

pS106-Rab12 quantitative immunoblotting in DLB hippocampus and temporal cortex. a–d Representative western blot of a Dounce-homogenized hippocampus derived from control (n = 3) and DLB (n = 7) subjects, b Dounce-homogenized temporal cortex derived from control (n = 4) and DLB (n = 8) subjects, c Triton-extracted hippocampus derived from control (n = 3) and DLB (n = 7) subjects, and d Triton-extracted temporal cortex derived from control (n = 5) and DLB (n = 11) subjects assessed for pS106-Rab12 (pRab12), Rab12, and Hsp70 as a loading control. Braak tangle stage (1–6) of each control or DLB subject is shown. Arrows denote the band analyzed for pRab12 and Rab12. e–j Quantification of e–g hippocampus and h–j temporal cortex western blots for e, h pRab12/Rab12, f, i pRab12/Hsp70, and g, j Rab12/Hsp70 showing Rab12 phosphorylation is elevated in both the hippocampus and temporal cortex in DLB (Braak > 3) cases compared to controls. Rab12 phosphorylation is also elevated in temporal cortex in DLB (Braak > 3) cases compared to DLB (Braak ≤ 3) cases. Data are presented relative to control cases (% control). Data are presented as mean ± SEM, and each point represents an individual subject. *p < 0.05 via unpaired Mann–Whitney U test, **p < 0.01 via Kruskal–Wallis H test with Dunn’s post hoc multiple comparisons test

Consistent with a relationship between Rab12 phosphorylation and tau pathology, we found that Rab12 phosphorylation was positively correlated with AD co-pathology measures, shown in Table 2. pS106-Rab12 (relative to total Rab12) was positively correlated with Thal amyloid phase, Braak neurofibrillary tangle (Braak tangle) stage, and CERAD neuritic plaque score in both the Dounce and Triton hippocampus fractions of control and DLB cases. However, after adjusting for all three pathologies together through multiple linear regression analysis, only CERAD neuritic plaque score was correlated with Rab12 phosphorylation in the Dounce fraction, and none were positively correlated with Rab12 phosphorylation in the Triton fraction. However, it should be noted that multicollinearity between these three pathologies was high, impairing the ability to determine the individual effect of each variable.

Table 2 Spearman correlation and multiple linear regression analyses of hippocampus and temporal cortex pS106-Rab12/Rab12 with control and DLB subject characteristics

In the temporal cortex, summarized in Table 2, Rab12 phosphorylation was positively correlated with Braak tangle stage, but not Thal amyloid phase or CERAD neuritic plaque score, in the Dounce fraction. However, after multiple linear regression analysis, all three pathologies were associated with Rab12 phosphorylation, suggesting a conditional relationship where an association of Thal amyloid phase and CERAD neuritic plaque score with Rab12 phosphorylation is dependent on Braak tangle stage. In the temporal cortex Triton fraction, Rab12 phosphorylation was positively correlated with Thal amyloid phase and Braak tangle stage, but not with CERAD neuritic plaque score, in control and DLB cases. The association with Braak tangle stage remained significant when adjusting for Thal amyloid phase and CERAD neuritic plaque score through multiple linear regression analysis, suggesting a correlation of Braak tangle stage with Rab12 phosphorylation that is independent of Thal amyloid phase and CERAD neuritic plaque score. Rab12 phosphorylation was not significantly correlated with age or PMI in either fraction in the hippocampus or temporal cortex.

pS106-Rab12 localizes to GVBs in DLB and AD

To further investigate the elevated Rab12 phosphorylation described in DLB (Fig. 1), we next determined the localization of pS106-Rab12 by IHC in the hippocampus, adjacent temporal cortex (comprising both entorhinal cortex and temporal cortex), and frontal cortex from DLB and control cases (subject characteristics can be found in Supplementary Table 1). There was no significant difference in age or PMI between groups (Supplementary Table 2). We observed pS106-Rab12 labeling cytoplasmic puncta highly reminiscent of GVBs across subjects (Fig. 2a), with significantly higher pS106-Rab12 GVB-like cell density in subjects with DLB when stratified by Braak neurofibrillary tangle stage (Braak > 3), relative to DLB (Braak ≤ 3) or control cases (Fig. 2a, b).

Fig. 2

pS106-Rab12 IHC and IF in DLB and AD brains reveal pS106-Rab12 labeling of GVBs. a Representative 60 × DAB IHC images and insets of the hippocampus from a control, DLB (Braak ≤ 3), DLB (Braak > 3), and AD case. Tissue was incubated with an antibody against pS106-Rab12 (pRab12). Black arrowheads denote pRab12 GVB-positive cells. b Quantification of the pRab12-positive GVB cell density in the hippocampus, temporal cortex, and frontal cortex of control (n = 7), DLB (Braak ≤ 3, n = 9), DLB (Braak > 3, n = 7), and AD (n = 5) cases. c Quantification of pRab12-positive GVB cell density in CA3/4, CA2, CA1 and subiculum sub-regions of control, DLB (Braak ≤ 3), DLB (Braak > 3), and AD cases. d Representative 63 × IF images and insets of the hippocampus from a control, DLB (Braak ≤ 3), DLB (Braak > 3), and AD case. Tissue was incubated against pRab12 (green), Ck1δ (magenta), and neuronal marker MAP2 (yellow). e Representative 63 × single-plane image of DLB (Braak ≤ 3) hippocampus pRab12 and Ck1δ, and f fluorescence intensity profile of the yellow line drawn in (e) showing fluorescence overlap between pRab12 and Ck1δ. g Quantification of pRab12 and Ck1δ puncta co-localization in hippocampus of DLB (Braak ≤ 3, n = 3), DLB (Braak > 3, n = 5), and AD (n = 4) cases. Data are presented as mean ± SEM and each point represents an individual subject. *p < 0.05, **p < 0.01, ***p < 0.001 via Kruskal–Wallis H test with Dunn’s post hoc multiple comparisons test. Scale bar in (a) = 20 µm, inset scale bar in (a) = 5 µm, scale bar in (d) = 10 µm, inset scale bar in (d) = 2 µm, scale bar in (e) = 2 µm

To determine if this pS106-Rab12 labeling is specific to DLB or common across neurodegenerative diseases, we also analyzed hippocampus, temporal cortex, and frontal cortex derived from AD subjects. All AD cases had Braak neurofibrillary tangle pathology stages of 5 or above. We found that like DLB, pS106-Rab12 GVB-like cell density was significantly higher in AD cases compared to DLB (Braak ≤ 3) and control cases (Fig. 2a, b). Interestingly, we observed regional differences in pS106-Rab12 GVB-pattern labeling. The highest density of GVB-like labeling was observed in the hippocampus, with the lowest density in the frontal cortex (Fig. 2a, b, Supplementary Fig. 5a). In addition, within the hippocampus, higher GVB-like labeling was observed in the CA1 and CA2 compared to CA3/4 and subiculum, although this did not reach statistical significance for any sub-region comparison (Fig. 2c, Supplementary Fig. 5b). These patterns recapitulate regional differences in GVB density and tau pathology in AD and DLB [12, 13, 20, 112, 122]. Two additional raters independently observed similar group and regional differences in pS106-Rab12 GVB-like labeling (Supplementary Fig. 5c). Over all regions and cases analyzed for pS106-Rab12 GVB-like cell density across cohorts, inter-rater reliability between the three raters was high (intraclass correlation coefficient = 0.83).

Consistent with the link between GVBs and tau pathology, we observed a significant and positive correlation of pS106-Rab12 GVB-like cell density with Thal amyloid phase, Braak tangle stage, and CERAD neuritic plaque score in the hippocampus of control, DLB, and AD cases (Table 3). After adjusting for all three pathologies together through multiple linear regression analysis, Braak tangle stage, but not Thal amyloid phase or CERAD neuritic plaque score, was significantly associated with hippocampus pS106-Rab12 GVB-like cell density, suggesting an independent association of Braak tangle stage with pS106-Rab12 GVB-like cell density.

Table 3 Spearman correlation and multiple linear regression analyses of hippocampus pS106-Rab12-positive GVB cell density with control, DLB, and AD subject characteristics

Next, to directly correlate pS106-Rab12 labeling to pathological tau and α-synuclein accumulation in control, DLB and AD cases, we quantified hippocampus AT8-labeled neurofibrillary tangle density (Supplementary Fig. 6a-c) and α-synuclein-labeled Lewy body density (Supplementary Fig. 7a, b); information of subjects analyzed can be found in Supplementary Table 1. Consistent with the use of these markers for tau and α-synuclein pathological staging, hippocampus neurofibrillary tangle and Lewy body densities were elevated in cases with higher Braak tangle stage and LBD pathology, respectively (Supplementary Fig. 6, 7). We found hippocampus pS106-Rab12 GVB-like cell density was positively correlated with hippocampus neurofibrillary tangle density, but not hippocampus Lewy body density (Table 3). Interestingly, after multiple linear regression analysis, both hippocampus neurofibrillary tangle density and Lewy body density were associated with hippocampus pS106-Rab12 GVB-like density, suggesting an association between Lewy bodies and pS106-Rab12 GVB labeling that is conditional upon, or mediated by, the presence of neurofibrillary tangles. Severity of amyloid-β pathology was not quantified, although amyloid-β PET imaging shows strong correlation of amyloid-β plaque load with Thal amyloid phase, with elevated amyloid-β plaque load in hippocampal regions found at later Thal amyloid phases [85]. To determine the potential relationship between pS106-Rab12 GVB labeling and cognitive status, MMSE scores available from a subset of control, AD, and DLB cases were analyzed (Supplementary Table 1). Hippocampus pS106-Rab12 GVB-like cell density trended towards a negative correlation with final MMSE score in these cases but was not significant (Table 3). Finally, pS106-Rab12 GVB-like cell density was not significantly correlated with age or PMI (Table 3).

Co-localization analysis demonstrates that pS106-Rab12 overlaps with the canonical GVB-specific marker casein kinase 1δ (Ck1δ, Fig. 2d-g) [127], and the pS106-Rab12 signal was specific (Supplementary Fig. 8a). Of note, we found Ck1δ- and pS106-Rab12-positive GVBs only in DLB and AD cases, but not controls, consistent with minimal GVB formation in the absence of disease (Fig. 2d) [127]. The majority of pS106-Rab12 and Ck1δ signal showed strong overlapping fluorescence, and over 80% of observed puncta in GVB-positive neurons were double-positive for pS106-Rab12 and Ck1δ in DLB (Braak ≤ 3), DLB (Braak > 3) and AD cases (Fig. 2e-g).

To further demonstrate Rab12 localizes to GVBs, we also performed IF using an antibody for total Rab12 in one control case (Braak tangle stage 2) and one DLB case (Braak tangle stage 6). No overt difference in Rab12 signal was observed in neurons between the cases (Supplementary Fig. 8b). Further, within GVB-positive neurons, Rab12 signal in GVBs was not higher than other Rab12 signal within the neuron (Supplementary Fig. 8c). However, at the single-plane level, we observed Rab12 labeling in a subset of GVBs, shown by overlapping immunoreactivity with Ck1δ (Supplementary Fig. 8d-f). Notably, this Rab12 signal in DLB was not observed when the Rab12 primary antibody was not added to the slide during the IF procedure (Supplementary Fig. 8 g).

pS106-Rab12 labels GVBs in both genetic and sPD

We next performed IHC for pS106-Rab12 in both the hippocampus and adjacent temporal cortex of a cohort of PD cases that harbor the G2019S pathogenic LRRK2 mutation (LRRK2GS PD), compared to sPD cases and unaffected controls. In this cohort, two out of seven LRRK2GS PD subjects and six out of ten sPD subjects had PDD (Supplementary Table 3). AT8 labeling was used for Braak neurofibrillary tangle staging and stratifying LRRK2GS PD and sPD cases into (Braak ≤ 3) and (Braak > 3) groups (Supplementary Fig. 9, all subject characteristics can be found in Supplementary Table 3). There was no significant difference in age or PMI between groups (Supplementary Table 4).

We observed pS106-Rab12 labeling of GVBs in the hippocampus and temporal cortex across subjects (Fig. 3a, Supplementary Fig. 10a). Significantly elevated pS106-Rab12 GVB cell density was detected in LRRK2GS PD cases when stratified by Braak neurofibrillary tangle stage (Braak > 3) compared to controls (Fig. 3a, b). pS106-Rab12 GVB cell density was increased to a similar magnitude in sPD when stratified by Braak neurofibrillary tangle stage (Braak > 3) relative to controls (Fig. 3a, b). pS106-Rab12 GVB cell density in sPD (Braak > 3) cases was significantly elevated compared to controls when only hippocampus was analyzed (Fig. 3b), though the comparison between sPD (Braak > 3) and controls did not reach statistical significance when all brain regions were considered. Consistent with GVB staging in AD, pS106-Rab12 GVB labeling was less elevated in the temporal cortex in LRRK2GS PD and sPD compared to the hippocampus (Fig. 3a, b, Supplementary Fig. 10a) [112]. Within the hippocampus, pS106-Rab12 GVB labeling was higher in CA1 and CA2 compared to CA3/4, which is also consistent with GVB staging in AD (Fig. 3c, Supplementary Fig. 10b) [112]. Subiculum pS106-Rab12 GVB labeling was not significantly different from other hippocampus sub-regions (Fig. 3c, Supplementary Fig. 10b). Similar group and regional differences in pS106-Rab12 GVB labeling were independently reported by two additional raters (Supplementary Fig. 11). Importantly, we observed a significant positive correlation of hippocampus pS106-Rab12 GVB cell density with Braak tangle stage, but not Thal amyloid phase or CERAD neuritic plaque score across control, LRRK2GS PD and sPD cases (Table 4). This association of Braak tangle stage with pS106-Rab12 GVB cell density remained significant when adjusting for all three pathologies together through multiple linear regression analysis, suggesting a correlation of Braak tangle stage with Rab12 phosphorylation that is independent of Thal amyloid phase and CERAD neuritic plaque score. Surprisingly, we observed a significant positive correlation of hippocampus pS106-Rab12 GVB cell density with age, but not with PMI (Table 4).

Fig. 3

pS106-Rab12 GVB labeling in LRRK2GS PD and sPD. a Representative 60 × images and insets of the hippocampus from a control, LRRK2GS PD (Braak ≤ 3), LRRK2GS PD (Braak > 3), sPD (Braak ≤ 3), and sPD (Braak > 3) case. Tissue was incubated with an antibody against pS106-Rab12 (pRab12). Black arrowheads denote pRab12 GVB-positive cells. b Quantification of the pRab12 GVB-labeled cell density in the hippocampus and temporal cortex of control (n = 6), LRRK2GS PD (Braak ≤ 3, n = 4), LRRK2GS PD (Braak > 3, n = 3), sPD (Braak ≤ 3, n = 5), and sPD (Braak > 3, n = 5) cases. c Quantification of pRab12-positive GVB cell density in CA3/4, CA2, CA1 and subiculum sub-regions of control, LRRK2GS PD (Braak ≤ 3), LRRK2GS PD (Braak > 3), sPD (Braak ≤ 3), and sPD (Braak > 3) cases. Data are presented as mean ± SEM and each point represents an individual subject. *p < 0.05 compared to controls, #p < 0.05 for hippocampus compared to control hippocampus, ***p < 0.001 via Kruskal–Wallis H test with Dunn’s post hoc multiple comparisons test. **p < 0.01 between regions via unpaired Mann–Whitney U test. Scale bar in (a) = 20 µm, inset scale bar = 5 µm

Table 4 Spearman correlation and multiple linear regression analyses of hippocampus pS106-Rab12-positive GVB cell density with control, LRRK2GS PD and sPD subject characteristics

We also explored the LRRK2 L1165P variant which is exceedingly rare and detected in a single early-onset PD case with unclear causality, as well as a LRRK2GS subject with Schizophrenia and no PD diagnosis [21, 41]. The LRRK2L1165P PD subject and LRRK2GS Schizophrenia subject both had Braak neurofibrillary tangle stages of 1–2 (Supplementary Table 3). While the schizophrenia subject that harbors the LRRK2GS mutation showed no observable Lewy pathology, in contrast, the PD case with the LRRK2L1165P mutation had Lewy-related α-synuclein pathology in limbic regions, including the hippocampus and amygdala [41]. We found hippocampus pS106-Rab12 GVB labeling in the LRRK2L1165 PD case, but not the LRRK2GS Schizophrenia case (Supplementary Fig. 12).

pS106-Rab12 GVBs accumulate with age in the PS19 tauopathy mouse model

To extend our human brain post-mortem findings to a mouse model with predictable pathological progression of tau, we used the well-studied PS19 strain of mice, which overexpress the P301S pathogenic mutant human tau in neurons and recapitulate hallmarks of tauopathy, including progressive tau pathology, cognitive dysfunction, and neurodegeneration [115, 131]. Consistent with previous studies, the PS19 mice display tau hyperphosphorylation as observed in tauopathies, evidenced by increased IF of AT8 in hippocampus and entorhinal cortex in aged (9–10 month-old) PS19 mice compared to WT (Supplementary Fig. 13a, b). AT8 IF signal was significantly higher in the entorhinal cortex compared to the hippocampus of PS19 mice (Supplementary Fig. 13b). Further, we performed quantitative western blotting and observed increased AT8, at molecular weights representing human hyperphosphorylated tau, in entorhinal cortex derived from both young (3–4 month-old) and aged (9–10 month-old) PS19 mice compared to WT (Supplementary Fig. 13c, d). The increase in hyperphosphorylated tau was associated with a concomitant increase in total tau detected by the antibody Tau5 in both young and aged PS19 relative to WT mice (Supplementary Fig. 13e).

In hippocampus and entorhinal cortex derived from aged PS19 mice, we observed neuronal, pS106-Rab12-positive GVB labeling, shown by overlapping immunoreactivity with GVB marker Ck1δ, that was not observed in WT mice (Fig. 4a-d). This pS106-Rab12 signal in PS19 mouse brain was not observed when the pS106-Rab12 primary antibody was not added to the slide during the IF procedure (Supplementary Fig. 13f). Quantification of pS106-Rab12- and Ck1δ-labeled puncta in these neurons demonstrated high co-localization, with 85% of puncta in GVB-labeled neurons displaying labeling of both pS106-Rab12 and Ck1δ (Fig. 4e). Significantly more GVB-labeled neurons were observed in aged PS19 compared to aged WT mice, in which Ck1δ labeling was not detectable (Fig. 4f). These pS106-Rab12 GVBs accumulated with age, as GVB-positive neuron density was significantly higher in aged PS19 compared to young PS19 mice (Fig. 4f, Supplementary Fig. 13 g). Similar to regional differences in AT8 signal, greater GVB neuron density was observed in the entorhinal cortex compared to hippocampus in aged PS19 mice (Fig. 4f), suggesting a correlation between hyperphosphorylated tau accumulation and GVB-labeled neuron density. Notably, while young PS19 mice displayed almost no detectable GVB labeling (Fig. 4f, Supplementary Fig. 13 g), the rare GVB-positive neurons observed displayed pS106-Rab12-positive GVBs (Supplementary Fig. 13 h).

Fig. 4

pS106-Rab12 labels GVBs in aged PS19 mouse brain. a Representative 63 × images of Hoechst (blue), pS106-Rab12 (pRab12, green), Ck1δ (magenta), and neuronal marker NeuN (yellow) in hippocampus and entorhinal cortex of aged WT and PS19 mice. b Zoomed-in 63 × representative images of aged PS19 entorhinal cortex showing overlap of pRab12 and Ck1δ puncta. c Representative 63 × single-plane images of aged PS19 entorhinal cortex pRab12 and Ck1δ, and d fluorescence intensity profile of the yellow line drawn in (c) showing fluorescence overlap between pRab12 and Ck1δ. e Quantification of pRab12 and Ck1δ puncta co-localization in aged PS19 entorhinal cortex (n = 8). f Quantification of Ck1δ-positive GVB neuron density in young (3–4 month-old) and aged WT (n = 5 young, 6 aged) and PS19 (n = 6 young, 8 aged) hippocampus and entorhinal cortex. g, h Representative 63 × images of Hoechst (blue), pRab12 (green), AT8 (magenta), and NeuN (yellow) in aged PS19 entorhinal cortex, showing pRab12 puncta in neurons with (g) low and (h) high AT8 signal. i Quantification of pRab12 puncta number in aged PS19 mouse (n = 9) pRab12-positive entorhinal cortex neurons with low or high AT8 signal. Data are presented as mean ± SEM and each point represents an individual animal. ***p < 0.001 via (f) two-way ANOVA with Bonferroni’s post hoc multiple comparisons test or (i) unpaired Student’s t test. Scale bar in (a) = 5 µm, scale bar in (b) = 2 µm, scale bar in (c) = 1 µm, scale bar in (h) = 2 µm

Consistent with previously published work showing pS106-Rab12 IF specifically detects a phospho-epitope, phosphatase (PPase) treatment in PS19 brain slices before staining abolished pS106-Rab12 signal, while preserving the Ck1δ signal, which does not detect a phospho-epitope (Supplementary Fig. 14a, b) [14]. PPase treatment similarly abolished the signal of AT8, which detects S202/T205 phospho-epitopes of tau (Supplementary Fig. 15a, b).

We next sought to investigate the relationship between hyperphosphorylated tau levels and pS106-Rab12 GVB labeling within each neuron in PS19 mice. We found that pS106-Rab12 puncta co-localized with AT8 in aged PS19 entorhinal cortex neurons where AT8 levels were low and displayed puncta-like labeling of AT8 (Fig. 4g). We also discovered that the number of pS106-Rab12 puncta found in these pS106-Rab12-positive neurons is significantly higher in neurons with high AT8 labeling (i.e., “pre-tangle” tau pathology) compared to neurons with comparatively low AT8 labeling (Fig. 4g-i) [46].

Pathological tau accumulation-initiated pS106-Rab12-positive GVBs are lysosomal structures

Previous work has demonstrated that GVBs are lysosomal structures with a dense core (labeled by Ck1δ) surrounded by an outer membrane that is labeled by lysosomal membrane markers like lysosomal-associated membrane protein 1 (LAMP1) [55, 128]. Therefore, we performed IF co-labeling of LAMP1 with pS106-Rab12 and Ck1δ in aged PS19 mouse entorhinal cortex. We found that, consistent with GVBs, Ck1δ labeled the GVB dense core, which was surrounded by a LAMP1-positive membrane as indicated by ring-like LAMP1 fluorescence around Ck1δ puncta (Fig. 5a, b). pS106-Rab12, consistent with its co-labeling of Ck1δ, labeled the dense core of GVBs (Fig. 5a, b). Consistent with quantification described in Fig. 4e, 79% of observed pS106-Rab12 and Ck1δ puncta in neurons with Ck1δ-positive GVBs were triple positive for pS106-Rab12, Ck1δ, and LAMP1 (Fig. 5c).

Fig. 5

pS106-Rab12-positive GVBs are lysosomal structures that also contain mitophagy protein markers. a Representative 63 × single-plane images of Hoechst (blue), pS106-Rab12 (pRab12, green), LAMP1 (magenta), and Ck1δ (yellow) in aged PS19 entorhinal cortex. b Fluorescence intensity profile of the yellow line drawn in (a) showing fluorescence overlap between pRab12, LAMP1 and Ck1δ, including the ring-like fluorescence intensity profile of LAMP1 surrounding pRab12- and Ck1δ-positive puncta. c Quantification of pRab12, Ck1δ, and LAMP1 co-localization in aged PS19 entorhinal cortex (n = 5) showing the majority of pRab12 and Ck1δ puncta in neurons with GVBs are triple positive for pRab12, Ck1δ and LAMP1. d Representative 63 × images of Hoechst (blue), pS65-Ub (green), Ck1δ (magenta), and NeuN (yellow) in aged PS19 entorhinal cortex showing overlap of pS65-Ub and Ck1δ puncta. e Representative 63 × single-plane images of aged PS19 entorhinal cortex pS65-Ub and Ck1δ, and f fluorescence intensity profile of the yellow line drawn in (e) showing fluorescence overlap between pS65-Ub and Ck1δ. g Quantification of pS65-Ub and Ck1δ co-localization in aged PS19 entorhinal cortex (n = 5). h Schematic showing that in tauopathies, GVBs form in neurons with early hyperphosphorylated tau accumulation, and the dense core of these GVBs contain pRab12 and pS65-Ub. Data are presented as mean ± SEM, and each point represents an individual animal. Scale bar in (a) = 1 µm, scale bar in (d) = 2 µm, scale bar in (e) = 1 µm. Schematic created with BioRender.com

In addition to displaying lysosomal markers, GVBs have been previously reported to co-localize with the mitophagy marker phospho-Ubiquitin S65 (pS65-Ub), and pathogenic LRRK2 mutations disrupt mitophagy [45, 46,