Blocking LRRK2 kinase may reverse tau buildup caused by the G2019S mutation

All housing, breeding, and procedures were performed according to the NIH Guide for the Care and Use of Experimental Animals and approved by the University of Pennsylvania’s Institutional Animal Care and Use Committee. LRRK2^G2019S knock-in mice (C57BL/6-Lrrk2^tm4.1Arte, RRID: IMSR_TAC:13,940) have been previously described [21]. Mice were bred as heterozygotes, and homozygous (+ / + or G2019S/G2019S) littermates were used for experiments. Both male (n = 30) and female (n = 35) mice were used. They were 3–4 months old at the time of injection.

MLi-2 was synthesized at Merck & Co., Inc. (Boston, MA). Diet was prepared at Research Diets (New Brunswick, NJ), and formulated to deliver a dose of 0, 10, or 60 mg/kg MLi-2 per day, based on an average daily food consumption of 4 g per 30 g mouse. The base for each diet was the vehicle diet (D01060501; Research Diets).

Archival tissues from four mice treated with the glucocerebrosidase inhibitor CBE in a previous study [22] were used in this study. The slides were stained using standard immunofluorescence as described below.

All procedures were done in accordance with local institutional review board guidelines of the University of Pennsylvania. Written informed consent for autopsy and analysis of tissue sample data was obtained either from patients or their next-of-kins. All cases used for extraction of paired helical filament (PHF) tau (Additional file: Table S1) were Braak stage VI and were selected based upon a high burden of tau pathology by immunohistochemical staining. The cases used for extraction were balanced by sex (female = 2; male = 1) and collected an average of 14 h post-mortem. After extracting all three cases and confirming the seeding capacity of tau, cases were pooled for injection and all mice were injected with the same AD PHF pool. 1

Frozen postmortem human frontal or temporal cortex tissues containing abundant tau-positive inclusions were selected for sequential extraction after immunohistochemical examination of these samples as described [23] using previously established methods. These brain tissues were sequentially extracted with increasing detergent strength as previously described [24], and a published protocol is available (10.17504/protocols.io.q26g7p9y8gwz/v1). After thawing, meninges were removed and gray matter was carefully separated from white matter. Gray matter was weighed and suspended in nine volumes (w/v) of high-salt (HS) buffer (10 mM Tris–HCL (pH 7.4), 800 mM NaCl, 1 mM EDTA, 2 mM dithiothreitol (DTT), protease and phosphatase inhibitors and PMSF) with 0.1% sarkosyl and 10% sucrose, followed by homogenization with a dounce homogenizer and centrifugation at 10,000×g for 10 min at 4 °C. The resulting pellet was re-extracted with the same buffer conditions and the supernatants from all extractions were filtered and pooled.

Additional sarkosyl was added to the pooled supernatant to reach a final concentration of 1% and the supernatant was rocked for 1 h at room temperature. The samples were then centrifuged at 300,000×g for 60 min at 4 °C. The pellet, which contains pathological tau, was washed once with PBS and resuspended in 100 μl of PBS per gram of gray matter followed by passing through a 27G/0.5 inch needle. The pellets were further suspended by brief sonication (QSonica Microson™ XL-2000; 20 pulses; setting 2; 0.5 s/pulse). The suspension was centrifuged at 100,000×g for 30 min at 4 °C. The pellet was suspended in one-fifth to one-half of the pre-centrifugation volume, sonicated briefly (60–120 pulses; setting 2; 0.5 s/pulse) and centrifuged at 10,000×g for 30 min at 4 °C. The final supernatant was utilized for all studies and is referred to as AD PHF tau. All extractions were characterized by Western blotting for tau, sandwich ELISA for α-synuclein and Aβ_1–40 and Aβ_1–42, and validated by immunocytochemistry in primary neurons from non-transgenic mice. For the extractions used in this study, tau constituted 9%–11% of the total protein, while α-synuclein and Aβ constituted 0.008% or less of the total protein.

Characterization of α-synuclein, Aβ_1–40, and Aβ_1–42 from AD PHF preparations by sandwich ELISA has been described previously [24], and a published protocol is available (10.17504/protocols.io.261ged83ov47/v1). Assays were performed on 384-well MaxiSorp clear plates (ThermoFisher, Waltham, MA). The plates were coated with well-characterized capture antibodies (α-synuclein: Syn9027 (Center for Neurodegenerative Disease Research, Philadelphia, PA); Aβ_1–40 and Aβ_1–42: Ban50 (Fujifilm Wako, Richmond, VA, Cat# 013-26873, RRID: AB_3076195)) at 4 °C overnight, washed, and blocked with Block Ace (AbD Serotec, Raleigh, NC) overnight at 4 °C. The AD PHF preparations were diluted at 1:100 and added to plates. Meanwhile, serial dilutions of recombinant α-synuclein, recombinant T40, or peptides for monomeric Aβ_1-40 and Aβ_1-42 were used as the standards. The plates were incubated overnight at 4 °C, then washed and incubated with reporter antibodies (tau: BT2 (Thermo Fisher Scientific, Cat# MN1010, RRID: AB_10975238) + HT7 (Thermo Fisher Scientific, Cat# MN1000, RRID: AB_2314654); α-synuclein: MJF-R1 (Abcam, Cat# ab138501, RRID: AB_2537217); Aβ_1–40: BA27 (Fujifilm Wako, Cat# 014-26923, RRID: AB_3076198); Aβ_1–42: BC05 (Fujifilm Wako, Cat# 010-26903, RRID: AB_3076199) overnight at 4 °C. Plates were washed and incubated with HRP-conjugated secondary antibodies for 1 h at 37 °C. Color was developed with a 1-Step Ultra TMB-ELISA Substrate Solution (Thermo Fisher Scientific) for 10–15 min. The reaction was quenched with 10% phosphoric acid and the plate was read at 450 nm on a plate reader (M5, SpectraMax, Molecular Devices, San Jose, CA).

All surgery experiments were performed in accordance with the protocols approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Pennsylvania. AD PHF tau from individual extractions was vortexed and diluted with DPBS to 0.4 mg/ml. Tau was sonicated in a cooled bath sonicator at 9 °C (Diagenode Bioruptor®; 20 cycles; setting medium; 30 s on, 30 s off). Mice were injected when 3–4 months old. The mice were deeply anaesthetized with ketamine/ xylazine/ acepromazine and injected unilaterally by insertion of a single needle into the right forebrain (coordinates: -2.5 mm relative to Bregma, + 2.0 mm from midline) targeting the hippocampus (2.4 mm beneath the skull) with 1 µg tau (2.5 µl). The needle was then retracted to 1.4 mm beneath the skull, targeting the overlaying cortex and another 1 µg tau (2.5 µl) was injected. The needle was left in place for 2 min following the injection. Injections were performed using a 25-µl syringe (Hamilton, NV) at a rate of 0.4 µl/min. After 3–6 months, the mice were perfused transcardially with PBS, and brains were removed and underwent overnight fixation in 70% ethanol in 150 mM NaCl, pH 7.4. After perfusion and fixation, the brains were processed into paraffin via sequential dehydration and perfusion with paraffin under vacuum (70% ethanol for 2 h, 80% ethanol for 1 h, 95% ethanol for 1 h, 95% ethanol for 2 h, 3 times 100% ethanol for 2 h, xylene for 30 min, xylene for 1 h, xylene for 1.5 h, 3 times paraffin for 1 h at 60 °C). Brains were then embedded in paraffin blocks, cut into 6-µm sections and mounted on glass slides. Slides were then stained using standard immunohistochemistry as described below.

Mice were perfused transcardially with PBS. Lung and kidney tissues were collected and post-fixed in 10% neutral-buffered formalin (NBF). The lung tissues were first inflated by injection with 10% NBF. A published protocol for staining is available (10.17504/protocols.io.kxygx3q9zg8j/v1). After perfusion and fixation, both types of tissues were processed into paraffin via sequential dehydration and perfusion with paraffin under vacuum (70% ethanol for 2 h, 80% ethanol for 1 h, 95% ethanol for 1 h, 95% ethanol for 2 h, 3 times 100% ethanol for 2 h, xylene for 30 min, xylene for 1 h, xylene for 1.5 h, 3 times paraffin for 1 h at 60 °C). Following fixation and paraffin processing, both lung and kidney tissues were stained with hematoxylin and eosin dyes for morphological examination. Sections were immersed in Harris Hematoxylin (Fisher, Cat# 67-650-01) for 1 min, rinsed 2 times in distilled water, differentiated in 0.1% acid alcohol solution for 4 s and rinsed in tap water for 15 min. Sections were then immersed in eosin for 1 min, briefly rinsed in tap water, then dehydrated and mounted with Cytoseal Mounting Media (Fisher, Waltham, MA, Cat# 23-244-256).

A published protocol for this method is available (0.17504/protocols.io.5jyl89m9dv2w/v1). Slides were de-paraffinized with 2 sequential 5-min washes in xylenes, followed by 1-min washes in a descending series of ethanols: 100%, 100%, 95%, 80%, 70%. Slides were then incubated in deionized water for one minute prior to microwave antigen retrieval in the BioGenex EZ-Retriever System. Slides were incubated in antigen unmasking solution (Vector Laboratories, Newark, CA; Cat# H-3300) and microwaved for 15 min at 95 °C. Slides were allowed to cool for 20 min at room temperature and washed in running tap water for 10 min. Slides were incubated in 7.5% hydrogen peroxide in water to quench endogenous peroxidase activity, washed for 10 min in running tap water and 5 min in 0.1 M Tris (diluted from 0.5 M Tris made from Tris base and concentrated hydrochloric acid to pH 7.6), and then blocked in 0.1 M Tris/2% fetal bovine serum (FBS) for 15 min or more. Slides were incubated with primary antibody in 0.1 M Tris/2% FBS in a humidified chamber overnight at 4 °C.

To stain pneumocytes, the slides with microwave antigen retrieval as described above, were incubated with anti-prosurfactant protein C antibody (proSP-C, Millipore Sigma, Cat# AB3786, RRID: AB_91588) at 1:4000. After washing with 0.1 M Tris for 5 min, the slides were incubated with biotinylated goat anti-rabbit (Vector, Cat# BA1000, RRID: AB_2313606) or horse anti-mouse (Vector, Cat# BA2000, RRID: AB_2313581) IgG in 0.1 M tris/2% FBS (1:1000) for 1 h. The slides were washed with 0.1 M Tris for 5 min, then incubated with avidin–biotin solution (Vector, Cat# PK-6100, RRID: AB_2336819) for 1 h, and washed for 5 min with 0.1 M Tris. Color was developed with ImmPACT DAB peroxidase substrate (Vector, Cat# SK-4105, RRID: AB_2336520) and counterstained briefly with Harris Hematoxylin (Fisher, Cat# 67-650-01). The lung ProsP-C data were grouped by object area (μm²) and plotted versus relative frequencies.

For phosphorylated tau, pS202/T205 tau (AT8, ThermoFisher, Cat# MN1020, RRID: AB_223647) was used at 1:10,000. The primary antibody was rinsed off with 0.1 M Tris and slides were incubated in 0.1 M Tris for 5 min. Primary antibody was detected using the BioGenex Polymer detection kit (Cat# QD440-XAK) according to the manufacturer’s protocol as outlined below. The slides were incubated in 50% Enhancer solution in 0.1 M Tris/2% FBS for 20 min, washed with 0.1 M Tris, incubated in 0.1 M Tris for 5 min, and incubated in 0.1 M Tris/2% FBS for 5 min. Slides were then incubated with 50% Poly-HRP in 0.1 M Tris/2% FBS for 30 min. The Poly-HRP was rinsed off with 0.1 M Tris, and slides were then incubated for 5 min with 0.1 M Tris. Color was developed with ImmPACT DAB peroxidase substrate (Vector, Cat# SK-4105) for 10 min. DAB was rinsed off with 0.1 M Tris, the slides were incubated in distilled water for 5 min and then counterstained briefly with Harris hematoxylin (ThermoFisher, Cat# 6765001).

The slides were washed in running tap water for 5 min, dehydrated in ascending ethanol for 1 min each (70%, 80%, 95%, 100%, 100%), then washed twice in xylenes for 5 min and coverslipped in Cytoseal Mounting Media (Fisher, Cat# 23-244-256). The slides were scanned on an Aperio AT2 microscope using a 20 × objective (0.75 NA) into ScanScope virtual slide (.svs) files. Digitized slides were then used for quantitative pathology.

A published protocol for this method is available (0.17504/protocols.io.5jyl89m9dv2w/v1). Slides were de-paraffinized with 2 sequential 5-min washes in xylenes, followed by 1-min washes in a descending series of ethanols: 100%, 100%, 95%, 80%, 70%. The slides were then incubated in deionized water for one minute prior to microwave antigen retrieval in the BioGenex EZ-Retriever System. Then the slides were incubated in antigen unmasking solution (Vector Laboratories; Cat# H-3300) and microwaved for 15 min at 95 °C. The slides were allowed to cool for 20 min at room temperature, washed for 5 min in 0.1 M Tris, and blocked in 0.1 M Tris/2% FBS for 15 min or more. Then the slides were incubated with primary antibody in 0.1 M Tris/2% FBS in a humidified chamber overnight at 4 °C. The primary antibodies used were rat anti-glial fibrillary acidic protein (GFAP) (2.2B10, ThermoFisher, Cat# 13-0300, RRID: AB_2532994, 1:1000), rabbit anti-Iba1 (Wako, Cat# 019-19741, RRID: AB_839504, 1:500), and pS202/T205 tau (AT8, ThermoFisher, Cat# MN1020, RRID: AB_223647, 1:1000). Primary antibodies were rinsed off with 0.1 M Tris, and the slides were incubated in 0.1 M Tris for 5 min, in 0.1 M Tris/2% FBS for 5 min, then with fluorescently-labeled secondary antibodies diluted in 0.1 M Tris/2% FBS for 3 h. Finally, the slides were washed 2 × 5 min with 0.1 M Tris and mounted with coverslips in ProLong gold with DAPI (Invitrogen, Cat# P36931). Fluorescent slides were imaged at 20 × magnification on a Zeiss AxioScan 7 microscope.

To quantify Iba-1-positive cells, annotations of the visual cortex were manually drawn guided by the Allen mouse brain atlas CCFv3. Utilizing the cell classification feature in QuPath, thresholds for control slides and experimental slides were established separately due to the intensity differences that limit the utility of a common cell detection algorithm. The minimum intensity thresholds were established at 2500 and 750, respectively. The cellular detections were then subject to object classification, which successfully differentiated cell bodies from staining artifact and cell processes. A common object classifier was used across control and experimental slides. The sizes and fluorescence intensities of individual cell bodies were averaged across all detections.

LRRK2 and pS935 LRRK2 were assessed using the Meso Scale Discovery (MSD) as previously reported [25]. Kidney lysates were prepared in 10 µl/mg MSD lysis buffer (R60TX-2) supplemented with 1 × HALT phosphatase and protease inhibitor (Thermo Fisher, Cat# 78442). Tissue was placed in a 2-ml round bottom tube (Eppendorf) with a steal bead (Biospec Products, 6.35 mm, Cat# 11079635C) and homogenized for 1.5 min at 30% amplitude in a Qiagen TissueLyser at 40 °C. Lysates were centrifuged at 13,000×g for 20 min and supernatants were collected for LRRK2 analysis. The total protein level was determined with a Micro BCA Protein Assay according to the kit protocol (Thermo Scientific, Cat# 23235). LRRK2 pS935 and total LRRK2 protein levels were quantified from lysates utilizing respective R-PLEX MSD Antibody Sets (LRRK2pS935, Cat# F211Q-3; total LRRK2, Cat# F211P-3). All incubation steps were performed on a shaker. Streptavidin-coated MSD plates (Cat# L45SA-2) were coated overnight at 4 °C in biotinylated anti-LRRK2pS935 or anti-LRRK2 antibody diluted 1:5 in Diluent 100 (Cat# R50AA-3). Plates were washed three times with 1 × MSD wash buffer (Cat# R61AA-1). The standard curve samples and the supernatants were added to the plate and incubated for 1 h at room temperature. Plates were washed as previously described. SULFO-Tag anti-LRRK2 pS935 (MSD, Cat# F211Q-3) or anti-LRRK2 (MSD, Cat# F211P-3) antibodies were diluted 1:100 in Diluent 100, applied to plate and incubated for 1 h at room temperature. Plates were washed three times, MSD Gold Read Buffer (Cat# R92TG-2) was applied, and plates were immediately read on an MSD plate reader.

Plasma was diluted 1:3 in acetonitrile, vortexed for 2 min, and centrifuged at 4000 RPM for 10 min at 40 °C. Supernatants were collected and assayed for compound levels using liquid chromatography and tandem mass spectrometry analyses as previously described [20].

Procedures of section selection, registration, and quantification were all done blinded to treatment.

To quantify the size of type II pneumocytes, we implemented an object classifier in the HALO software with a minimum optical density of 0.225. Sizes of individual cells were measured and the overall distribution of sizes was calculated to determine if there was an overall shift in the size of these cells or a shift in the average size per mouse.

To quantify tau pathology and register brain sections to the Allen Brain Atlas CCFv3 (RRID: SCR_017001), we used a modified version of the QUINT workflow [26]. Adaptations are available in a published suite of protocols (10.17504/protocols.io.kqdg3xbkzg25/v1, RRID: SCR_023856). Our workflow utilizes a suite of open access programs for segmentation and spatial registration of histological images of the mouse brain—QuPath [27] (RRID:SCR_018257) for image segmentation, QuickNII [28] and VisuAlign (RRID:SCR_017978) for image registration, Qmask for hemispheric differentiation and Nutil [29] (RRID:SCR_017183) for a combination of segmentation and registration and subsequent quantification. Twelve coronal sections per animal were quantified which contain the majority of regions with pathology. To segment pS202/T205 tau, individual sections were segmented using the pixel classification feature of QuPath. The optical density threshold was set at 0.03 to quantify pathology. All pixels above this threshold were quantified, with obvious artifacts removed. The resulting segmentations were exported at full resolution. Down-sampled brain sections were imported into QuickNII and aligned in 3-dimensional space to the 2017 Allen mouse brain atlas CCFv3. Sections were then imported into VisuAlign where anchor points were generated in the atlas and moved to the corresponding location on the section of interest via non-linear warp transformation. In addition, from the spatial coordinate information derived from QuickNII, the masking tool Qmask was used to generate masks over each brain hemisphere so ipsilateral and contralateral brain regions could be analyzed independently. We then used the quantifier feature in Nutil to align the segmentation from QuPath, the registration from VisuAlign, and the hemispheric masks from Qmask to generate percentage area occupied measures for each brain region. The resulting outputs include pixel quantification with spatial location in the brain mapped to 844 brain regions. The resulting output of each of the 12 coronal sections, was then averaged across brain regions to produce single percent area occupied values for each region.

Data were modeled under the assumption that tau proteins spread in both anterograde and retrograde directions along the brain’s structural connectome. Structural data were obtained from the anterograde viral tract-tracing experiments [30]. The high-resolution structural connectome was constructed by estimating the projection strength between each pair of 100-µm-wide voxels within the Allen Institute CCFv3 whole-brain parcellation [31]. Connections were normalized by dividing the edge strength between each source and target region by the size of the source region [31].

The equation used to model bidirectional pathology spread is as follows:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat y\left( t \right) = {\beta_0} + \;{\beta_a}{\log_{10}}{e^{ - {c_a}{L_a}t}}{x_0} + {\beta_r}{\log_{10}}{e^{ - {c_r}{L_r}t}}{x_0} + \varepsilon \left( t \right)$$\end{document}y^t=β0+βalog10e-caLatx0+βrlog10e-crLrtx0+εtwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta}_{0}$$\end{document}β0 is an intercept, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta}_{a}$$\end{document}βa is a weight for the importance of anterograde spread, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta}_{r}$$\end{document}βr is a weight for the importance of retrograde spread, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${c}_{a}$$\end{document}ca and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${c}_{r}$$\end{document}cr are time constants representing the global speed of anterograde and retrograde spread, respectively, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${L}_{a}$$\end{document}La and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${L}_{r}$$\end{document}Lr represent the out-degree graph Laplacian for each direction, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t$$\end{document}t is time, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${x}_{0}$$\end{document}x0 is a vector containing ones in the injection site regions (DG, CA1, CA3, VISam, and RSPagl) and zeros elsewhere, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varepsilon$$\end{document}ε is an error term[18]. The “optim” function in R was used to identify the combination of anterograde and retrograde time constants that yielded the strongest correlation between actual and predicted pathology for each group.

To compare model parameters across treatment groups, data from mice in each group were resampled 500 times and each resampled dataset was fit using the bidirectional spread model.

We used previously described approaches[18] to evaluate the specificity and performance of our model. To assess the dependence of the model fit on the spatial location of the seed sites, we randomly selected 500 sets of seed regions which had the same average distance from one another as the actual seed sites (1.96 ± 0.196 mm). The bidirectional model described above was then fit for each set of random seed sites, such that \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${x_0}$$\end{document}x0 contained ones in the randomly-selected seed regions and zeros elsewhere.

To evaluate the performance of the bidirectional spread model, we compared fits obtained from models accounting for spread in only one direction (anterograde or retrograde) or based solely on the distance between regions. For each model type, data from mice of both genotypes and treatment groups were pooled and randomly resampled to obtainandfor each time point. The model fit was performed onand the fit withwas then evaluated. This process was repeated 500 times for each model type to obtain distributions of out-of-sample fits. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${y_{train}}\left( t \right)$$\end{document} y train t \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${y_{test}}\left( t \right)$$\end{document} y test t \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${y_{train}}\left( t \right)$$\end{document} y train t \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${y_{test}}\left( t \right)$$\end{document} y test t

All primary data are available here: 10.5281/zenodo.10630752. The number of animals analyzed in each experiment, the statistical analysis performed, as well as the P-values for all results < 0.05 are reported in the figure legends. In vivo pathological spread data were analyzed and all computations were performed in R (https://www.R-project.org/↗, RRID: SCR_001905) [32] as described. Type II pneumocyte size was statistically assessed via two distinct methods, the two-way ANOVA (Additional file 1: Fig. S1), and the linear mixed-effects model [33] (Fig. 2) to account for individual mouse variation. A linear mixed-effects model was fit with a random intercept for each mouse and a random coefficient for genotype. Genotype and treatment, as well as their interaction, were included as fixed-effects. Due to the large number of samples per animal, we used the Nelder-Mead optimizer. Area was square-rooted to improve normality of residuals.

For analyses assessing differences in percent area occupied, data were stratified based on brain hemisphere (ipsilateral/contralateral), brain region, MPI, and either genotype or treatment depending on the specific hypothesis test (i.e. stratified on treatment if testing genotype differences and vice-versa). Regions with zero variance across all treatments and genotypes were filtered out. To maintain flexibility and avoid reliance on strong parametric assumptions, robust linear regressions via the MASS package, with the ranked percent area occupied as the outcome and either genotype or treatment as the explanatory variable, were used (https://cran.r-project.org/web/packages/MASS/index.html↗, RRID: SCR_019125). All models were adjusted for sex and daughter regions when the number of daughter regions was 2 or greater. Effect sizes, including confidence intervals were estimated using the R package emmeans (R package version 1.8.3, URL: https://CRAN.R-project.org/package=emmeans↗, RRID: SCR_018734). Significance was determined using second-generation P values based on a null interval of ± 5% difference with 95% confidence intervals [34]. Only second-generation P values equal to 0 were considered significant.

Several previous studies have shown that sustained LRRK2 kinase inhibition can lead to morphological phenotypes in the kidney and lung [20, 38–40]. We therefore collected kidney and lung tissues and assayed each for morphological abnormalities. The kidney sections were stained with hematoxylin and eosin and showed no major morphological abnormalities (Additional file 1: Fig. S1a). The lung tissue was stained with hematoxylin and eosin and also immunostained for proSP-C, a marker for pneumocytes. We found that the mice treated with MLi-2, especially those treated with a high dose or for a longer time, showed enlarged pneumocytes (Fig. 2d, e). Quantitative assessment of pneumocyte size based on proSP-C staining revealed that the size distribution of pneumocytes in mice treated with MLi-2 was significantly increased in all groups, except for 3 MPI LRRK2^G2019S mice (Fig. 2f–i). We assessed the impact of MLi-2 treatment on individual mice, first with a linear mixed-effects model. Results showed that MLi-2 did not have a significant effect at 3 MPI, but increased pneumocyte size at 450 mg/kg in WT mice (P = 0.0498) and at 75 and 450 mg/kg in LRRK2^G2019S mice at 6 MPI (P = 0.015 and 0.019, respectively). We also found that the average pneumocyte size was increased in LRRK2^G2019S mice treated with 75 or 450 mg/kg MLi-2 (Additional file 1: Fig. S1b, c). These results are consistent with previous literature and suggest that a higher dose or more sustained LRRK2 inhibition will result in a stronger pneumocyte phenotype in the lung.

To investigate the effect of pharmacological inhibition of LRRK2 on the development and spread of tau pathology, we used an established seed-based model of tauopathy that induces the misfolding of endogenous tau into hyperphosphorylated tau inclusions following injection of AD brain-derived tau into mice [24]. In the present study, we injected LRRK2^G2019S knock-in and WT littermates with AD brain-derived PHFs and then aged them to either 3 or 6 months MPI. AD PHF tau induces misfolding of endogenous mouse tau and subsequent progression of phosphorylated tau pathology throughout the mouse brain [18, 24]. Brains were sectioned, and 12 representative sections were selected and stained for phosphorylated tau pathology (AT8, pS202/T205 tau). To quantify the phosphorylated tau pathology throughout the brain, we implemented a segmentation and brain registration platform adapted from the QUINT workflow [26] (Additional file 1: Fig. S2). Segmentation was performed in QuPath using an optical density pixel threshold. In parallel, tissue sections were placed in a three-dimensional atlas using QuickNII [28], and nonlinear warp transformations were performed in VisuAlign. Segmentations were aligned to registered anatomical regions and quantified in the Nutil [29] software. This strategy enabled us to produce highly-reproducible measures of tau pathology burden in 844 anatomical regions across the brain (Additional file 1: Fig. S3).

Previous work using computational modeling of intracellular pathology progression has demonstrated that anatomical connectivity is a major constraint on pathology progression [16, 18, 22]. While this type of modeling is useful for understanding how pathology progresses through the brain and the factors underlying regional vulnerability, we hypothesized that it could also be useful for understanding the impact of compound administration on network dynamics of pathology spread. We therefore implemented linear diffusion modeling on the brain-wide tau pathology data collected in this study.

However, it is unclear how LRRK2^G2019S changes tau pathology progression. In addition to cell-autonomous explanations, there is growing evidence of the involvement of microglia in LRRK2-related phenotypes [41] and in tau pathology progression and neurodegeneration [42]. LRRK2 is present in microglia and upregulated in response to inflammatory stimuli [43]. Microglial morphology and inflammatory phenotypes have been reported to be similar between wild-type and LRRK2^G2019S mice, but upon stimulation, the LRRK2^G2019S microglia have increased reactivity [44, 45]. A recent study showed that LRRK2 inhibits microglial lysosomal degradative activity through transcriptional regulation of lysosomal gene expression (46). We therefore sought to determine whether there were histological changes in microglia in wild-type or LRRK2^G2019S mice treated with LRRK2 kinase inhibitor. We characterized visual cortex at 6 MPI where the tau pathology was increased in LRRK2^G2019S mice (Fig. 3) and reversed with LRRK2 kinase inhibition (Fig. 5). As a positive control, archival tissue from mice injected with the glucocerebrosidase inhibitor conduritol-β-epoxide (CBE) was included. We have previously demonstrated a robust cortical gliosis in these mice [22].

Mice were stained for GFAP, microglial marker Iba-1, and pS202/T205 tau. In contrast to the CBE-injected mice which had enlarged microglial cell bodies and increased GFAP signal, there was no dramatic differences in GFAP or Iba-1 signal in wild-type or LRRK2^G2019S mice (Additional file 1: Fig. S10a). The fluorescence intensity and average size of microglial cell bodies were determined through cell detection with intensity thresholding and object classification within QuPath to remove cell processes. Overall, wild-type and LRRK2^G2019S microglia had similar cell body size and Iba-1 intensity, with a slight trend to lower cell body size and higher Iba-1 intensity in the 450 mg/kg MLi-2-treated animals (Additional file 1: Fig. S10b, c). As a comparator, the Iba-1 intensity was elevated to approximately two folds in CBE-injected mice (Additional file 1: Fig. S10c). These experiments do not provide strong support of a gliosis phenotype in the mice studied here, but there may be more subtle changes in gliosis missed by this experiment.

Additional file 1. Table S1: Characterization of PHF preparations from AD brains. Figure S1: Long-term MLi-2 impacts lung, but not the gross kidney morphology. Figure S2: Quantitative pathology workflow. Figure S3: Quantitative pathology analysis from all mice. Figure S4: Sex differences in tau pathology in wild-type mice. Figure S5: Representative staining from 3 MPI mice. Figure S6: Wild-type compared to LRRK2G2019S mice at 3 MPI. Figure S7: Tau pathology compared by treatment group at 3 MPI. Figure S8: Representative staining from 6 MPI mice. Figure S9: Examination of linear diffusion fits by hemisphere. Figure S10: Microglia quantification in caudal cortex of 6 MPI mice.