TREM2 Limits Progression of Deficits and Spreading of Tau Pathology in Mice


 Background. Amyloid-β (Aβ) and tau form pathogenic lesions in Alzheimer’s disease (AD) brains. As ΑD clinically progresses, tau pathology propagates in a very distinct pattern between connected brain areas. The molecular mechanisms underlying this tau pathology spread remain largely unknown. Genome-wide association studies have identified polymorphisms in triggering receptor expressed on myeloid cells 2 ( TREM2 ) as genetic risk factors for AD and regulators of Aβ pathology-dependent tau propagation. Whether TREM2 contributes to neuron-to-neuron spreading of pathological tau remains unknown.Methods. Here, we crossed Trem2- deficient mice with P301S tau transgenic TAU58 mice and subjected the mice to behavioral testing and assessed neuropathology. Microglial activation states were determined using cytometry by of flight (CyTOF) and quantitative PCR. Tau spreading was assessed in vivo using tracing of focal tau expression.Results. Trem2 depletion significantly aggravated tau-induced early-onset motor and behavioural deficits. Neuropathologically, Trem2 reduction increased the number of hyperphosphorylated tau lesions in young TAU58 brains and reduced disease-associated microglia. Direct assessment of inter-neuronal spread of tau in vivo revealed significantly enhanced propagation of tau in the absence of Trem2 , suggesting that microglial TREM2 limits the progression of tau pathology in disease.Conclusion. Taken together, our data suggests that reduced TREM2 function accelerates the onset and progression of functional deficits and tau neuropathology in tau transgenic mice, which is - at least in part - due to increased tau spreading. Therefore, reduced TREM2 function may contribute to early AD by augmenting tau toxicity and its inter-neuronal propagation.


Methods.
Here, we crossed Trem2-deficient mice with P301S tau transgenic TAU58 mice and subjected the 29 mice to behavioral testing and assessed neuropathology. Microglial activation states were determined using 30 cytometry by of flight (CyTOF) and quantitative PCR. Tau spreading was assessed in vivo using tracing of focal 31 tau expression.

37
Conclusion. Taken together, our data suggests that reduced TREM2 function accelerates the onset and 38 progression of functional deficits and tau neuropathology in tau transgenic mice, which is -at least in part -due 39 to increased tau spreading. Therefore, reduced TREM2 function may contribute to early AD by augmenting tau 40 toxicity and its inter-neuronal propagation. Alzheimer's Disease (AD) is the most prevalent form of dementia (1,2). Cognitive decline is often accompanied 49 by behavioural changes, anxiety, depression or sleep disturbances, together leading to full-time care dependency 50 and ultimately death (3,4). AD brains are characterised by two pathological hallmark changes, the extracellular 51 deposition of amyloid-β (Aβ) in plaques and the formation of intracellular neurofibrillary tangles (NFT), 52 composed of hyper phosphorylated microtubule-associated protein tau (5). Fibrillary tau pathology (without Aβ 53 plaques) is furthermore characterizing brains in different forms of frontotemporal dementia (FTD), the second 54 most common form of dementia in individuals age 65 or younger (6) and coining the term 'tauopathies'. Studies 55 into familial FTD identified pathogenic mutations in the tau encoding MAPT gene, highlighting its role in TREM2 reduction has been observed (49), however, a detailed functional analysis of Trem2-deficient tau 87 transgenic mice has not been done to date. Furthermore, the effects of reduced TREM2 function on tau 88 propagation has not been assessed.

89
In the present study, we crossed the established tau P301S transgenic line TAU58 (51-54) with Trem2knockout 90 mice, generated by CRISPR/Cas9-mediated gene targeting. We found that TREM2 reduction worsened motor 91 deficits of TAU58 mice and aggravated behavioural deficits together with accelerated tau neuropathology in 92 young but not older mice. Interestingly, numbers of active, disease-associated (vs. resting/homeostatic) 93 microglia were reduced early-on in TREM2-deficient TAU58 mice. Injecting mice with an adeno-associated

125
Group sizes for each test are provided in Table S1.

127
RotaRod To assess motor function mice were placed on the RotaRod (UgoBasile, Gemonio VA, Italy) as 128 described elsewhere (58). Briefly, mice were placed on a rotating rod with accelerating speed (acceleration 129 mode 5-60rpm over 120 seconds). The latency to fall or until mice did one full revolution was recorded. Mice 130 were trained for 3 consecutive days, 5 trials each day and the best trial of the last day was used for analysis.

131
Beam To further assess the fine-motor performance, mice were subjected to the Beam traversing challenge (51).

132
For this challenge mice need to cross a thin elevated wooden beam. The house of their home cage was placed at 133 the end of the beam, on a platform the same height as the beam. Mice were trained in the morning and tested in 134 the afternoon of the same day. The testing session was recorded, and mice performed two consecutive trials. The 135 videos were analysed manually, by an experimenter blinded to the experimental groups. The time to cross as 136 well as the number of foot slips was counted. The two trials were averaged for analysis. If mice fell off the beam 137 in one trial, the successful trial was used for analysis. Mice that fell in both trials were excluded from the 138 analysis.

139
Morris Water Maze To assess spatial memory, mice were subjected to Morris Water Maze (MWM) testing 140 (59). The test apparatus consisted of a large tank (1.4 m diameter), a Perspex platform (10cm diameter, 40cm 141 high) and four different visual cues. Each visual cue was placed equidistant from one another around the tank.

142
The platform was positioned roughly 20 cm from the inside wall of the tank. The tank was filled with water to a 143 height of approximately 0.5 -1 cm above the platform with white non-toxic acrylic-based paint diluted in the 144 water to conceal the platform. Additionally, the lights in the testing room were dimmed. Mice were acclimatised in the room for one hour prior to testing. On days 1-5, mice were trained to find the hidden platform. To achieve 146 this, mice had 4 swims per day, each trial starting from a different position. This was done to ensure mice 147 learned the location of the platform rather than the swimming trajectory. Additionally, the order of the positions 148 was changed every day. Mice were given 1 minute to locate the platform and then had to sit on the platform for 149 one additional minute once they found it. When mice did not find the platform within 1 minute they were guided 150 to it and had to stay on the platform for a minute. Two to three mice were alternated per round, to avoid fatigue 151 of the mice. The order of the mice and the number per round was kept consistent for the whole duration of the 152 test. Each trial was recorded and the latency to find the platform was noted. On day 6 mice were subjected to the 153 probe trial for 30 seconds. The hidden platform was removed from the maze and mice were placed into the tank 154 from position 1 and recorded. The video was analysed using AnyMAZE software (Stoelting). For the analysis, 155 the tank was divided into four quadrants (the target quadrant, the right quadrant, the left quadrant and the 156 opposite quadrant) and the platform zone (a circle corresponding to the size and position of the platform). The 157 software counted the number of entries into the platform zone, the latency for the first entry into the platform 158 zone, the time spent in the platform zone as well as the time spend in each of the quadrants and the overall swim 159 speed and distance.

160
A visual cue test was performed after the probe trial to ensure all mice used have sensorimotor abilities and 161 motivation. The platform was placed back into the same position as before, with a flag placed on top of it. Mice 162 had four swims from positions 1 through 4. The latency to reach the platform/ flag was noted and mice were 163 taken out of the maze as soon as they reached the platform. The four trials were averaged and an exclusion 164 criteria of an average time greater than 20 seconds was applied. 167 each) and an inner area where all four arms meet (5.5cm x 5.5cm) and is elevated 60cm off the ground. Mice 168 were acclimatised in the room for one hour prior to testing. At the start of the test mice were placed in the inner 169 area, facing the open arm, and allowed to explore for 5min while being recorded. The time mice spend in the 170 different areas was analysed using the AnyMAZE software (Stoelting).

171
Open Field Activity and exploration behaviour was assessed using the Open Field test (OF), as described prior to the test. At the start of the test, mice were placed in the top right corner of the arena and were allowed to explore it freely for 10 minutes. Mice were video recorded while performing the test and the videos were later 176 analysed using AnyMAZE software (Stoelting). For the analysis, the arena was divided into an inner zone (17.5 177 cm x 17.5 cm square in the middle) and an outer zone. The software tracked how much time animals spent in 178 each respective zone as well as the total distance travelled.

179
AAV production Packaging of recombinant AAVs with AAV1 capsid (Penn Vector Core -Gene Therapy

180
Program, University of Pennsylvania) was performed as previously described (60   195 paraformaldehyde prior to removal of the brain and the whole brain was used for immunohistochemistry.

196
Briefly, the brain was post-fixed in cold 4% paraformaldehyde overnight and changed to 70% ethanol the next 197 day. Tissue was placed in histology cassettes and processed overnight in the Excelsior tissue processor 198 (ThermoFisher, Waltham, MA, USA). Processed tissue was embedded in paraffin and sectioned either coronally 199 at 3µm or horizontally at 5μm (whole brain). Sections were allowed to air dry over night before being baked at 200 65°C for two hours. To remove any residual paraffin, slides were placed in xylene for 20 minutes and then re-201 hydrated by immersing the slides in decreasing concentrations of ethanol. Heat-mediated antigen retrieval was 202 performed in a Milestone Histology microwave (Milestone, Sorisole, Italy) using citrate buffer. Sections were 203 stained using sequenza racks (ThermoFisher, Waltham, MA, USA). Slides were blocked in 100µl blocking buffer (BB, containing 3% heat inactivated Goat Serum 2% Bovine Serum Albumin in PBS) for one hour at 205 room temperature. 100µl primary antibody diluted in BB was added and slides were incubated overnight at 4°C 206 (

208
were diluted in BB, and 100µl were added on to the slides and incubated for one hour at room temperature.

211
Quantification of images was performed using either the Zeiss ZEN 2.6 blue edition software or ImageJ.

212
Inkscape was used to create the figures and construct the heat maps. For the heat maps counts were divided into 213 14 different areas and averaged across mice per genotype, per depth. Depth is given relative to bregma. 214 215 RNA Extraction. Mice for RNA extraction were sacrificed, brains were removed and sub-dissected into 216 hippocampus and cortex. Tissues were snap frozen in liquid nitrogen. RNA was extracted according to 217 manufacturer's protocol for TRIzol (ThermoFisher Scientific). Briefly, tissue was homogenised in TRIzol using 218 a plastic pestle. Lysates were centrifuged briefly to remove debris. Chloroform was added to the supernatant to 219 precipitate proteins. After a short incubation period, tubes were centrifuged to separate the aqueous from the 220 phenol phase. The aqueous phase was transferred to a new tube and 70% ethanol added. After gentle mixing,

221
RNA purification was performed by using the Qiagen RNeasy Kit according to manufacturer's instructions,

222
including the optional DNase treatment. RNA concentrations were measured using a Nanodrop (ThermoFisher 223 Scientific).

224
qRT-PCR First-strand cDNA was synthesised from the extracted RNA by using the SuperScript VILO 225 MasterMix (ThermoFisher Scientific) according to manufacturer's instructions. 1µg of RNA was used in the 226 cDNA reaction and resulting cDNA was diluted 1:10 for qRT-PCR. Standards were made from pooled samples.

227
A master mix consisting of 2x Power SYBR Green (ThermoFisher Scientific), DEPC treated water and the 228 respective primers was made for each gene of interest for qRT-PCR analysis. Primers were designed specifically 229 for qRT-PCR and ordered from Macrogen (South Korea) (Primer list Table S3). qRT-PCR was run on a ViiA 7

231
Brain single cell suspensions. Following perfusion with 1X PBS, brains from adult mice were collected and 232 placed in ice-cold Hibernate A medium (A1247501, Thermo Fisher Scientific). The brain was transferred to a petri dish on ice without media and cut into 2mm 2 pieces with a scalpel blade. Tissue dissociation was 234 performed with Digestion Media containing 1 mg/mL Collagenase D (Sigma) dissolved in Hibernate A Medium 235 in a GentleMACS C tube (Miltenyi Biotec). The tissue was incubated at 37°C for 30 min with stirring using 236 program 37C_ABDK_01 on the GentleMACS Octo Dissociator (Miltenyi Biotec). Tissue suspensions were 237 further mechanically dissociated by gentle trituration using a 5 mL serological pipette for twenty strokes. The 238 cell suspension was then passed through a pre-wetted 70 µm cell strainer with a syringe plunger and washed 239 with ice-cold 1X PBS without calcium and magnesium. After centrifugation at 600 x g for 4 min at 4°C, the 240 supernatant was discarded and the cell pellet resuspended in 10 mL of 1X PBS. The cell suspension was 241 transferred to a 15 mL conical tube and mixed with 5 mL of isotonic Percoll (9 parts Percoll: 1 part 10X PBS) 242 for a final 30% (vol/vol) Percoll gradient. This was centrifuged at 1500 x g for 20min at 4°C with maximum 243 acceleration and no brakes during deceleration. The cloudy white myelin overlay was removed by aspiration, 244 and the remaining supernatant and cell pellet were resuspended and transferred to a new 15 mL conical tube.

245
The cells were centrifuged at 600 x g for 4 min at 4°C, the supernatant was discarded and the cell pellet  112Cd, 113Cd, 114Cd, and 116Cd were conjugated to anti-mouse CD45.2 antibodies according to 252 manufacturer's instruction using the MCP9 Antibody Labeling kit (Fluidigm). A seven choose three barcoding 253 matrix was generated and barcoding reagents were titrated to achieve optimal labelling. Cells were stained in 254 96-well plates. To each sample a unique combination of exactly three metal cell barcoding reagents diluted in 255 50uL PBS, 0.5% BSA was added and then incubated for 20 min at room temperature. Cells were washed with 256 150uL FACS buffer at 4°C. Barcoded cells were then combined in a single 1.5ml tube and washed with FACS 257 buffer. Surface staining mix was prepared by diluting all primary antibodies in FACS buffer. Cell suspension 258 was centrifuged at 600g for four minutes at 8°C and the supernatant was removed. Cells were re-suspended in 259 surface stain mix (Table S4) and incubated at room temperature for 20 minutes. Cells were washed in FACS 260 buffer and staining was repeated with secondary antibodies at room temperature for 20 minutes. Cells were 261 washed in FACs buffer and incubated over night with 250nM iridium intercalator (Fluidigm) in Maxpar cell 262 fix./perm. buffer (Fluidigm) to label cellular DNA. Subsequently, cells were washed with PBS followed by distilled water and resuspended in 10% EQ beads (Fluidigm) in distilled water. Mass cytometry acquisition was 264 performed on a CyTOF2.1 (Helios) mass cytometer (Fluidigm). After acquisition, data was normalized by using 265 the bead standard and the executable MATLAB normalizer application, dead cells and beads removed and de-266 barcoded using Boolean gating using FlowJo software (BD) (65). All analyses on CyTOF data were performed 267 after arcsinh (with cofactor equal to 5) transformation of marker expression. Data was analyzed in R using our  to obtain TAU58/Trem2 Δ/Δ , TAU58/Trem2 Δ/+ and TAU58/Trem2 +/+ mice ( Figure 1B) and subjected them to 283 functional testing at different ages. Unless otherwise specified, littermate controls refer to Trem2 Δ/Δ , Trem2 Δ/+ 284 and Trem2 +/+ mice that do not carry the P301S tau transgene. We have previously reported learning deficits in 285 4-month-old TAU58 mice when subjected to Morris water maze (MWM) testing for spatial memory formation 286 (53). To study mice at pre-symptomatic stage, we tested young 3-month-old mice and found no learning 287 performance deficits in TAU58/Trem2 +/+ and TAU58/Trem2 Δ/+ compared to Trem2 +/+ , Trem2 Δ/+ and Trem2 Δ/Δ 288 littermate controls ( Figure 1C). In contrast, 3-month-old TAU58/Trem2 Δ/Δ mice displayed significantly delayed 289 learning in the MWM acquisition phase while the probe trial was unaffected. We have furthermore previously 290 reported motor deficits in 3-month-old TAU58 mice, which were phenotypically still normal at 2 months of age 291 (51,52). Accordingly, TAU58/Trem2 +/+ mice performed like littermate controls during Rotarod testing at 1 and 2 months of age ( Figure 1D). While TAU58/Trem2 Δ/+ and TAU58/Trem2 Δ/Δ mice performed normally at 1 293 month of age, they presented significant deficits as compared with littermate controls at 2 months of age, with 294 TAU58/Trem2 Δ/Δ mice significantly impaired compared to TAU58/Trem2 +/+ mice. At 3 months of age, 295 TAU58/Trem2 +/+ have deteriorated and presented similar deficits as TAU58/Trem2 Δ/+ and TAU58/Trem2 Δ/Δ 296 mice compared to littermates. These deficits progressed further by 6 months of age. We have previously shown 297 that TAU58 mice present with increased foot slips while crossing a narrow beam at 3 months of age (51). In the 298 present cohort, significant differences in the number of foot slips began in 1 month-old TAU58/Trem2 Δ/+ mice, 299 which progressed to include TAU58/Trem2 Δ/Δ mice by 2 and 3 months of age (Figure 1E). No significant 300 differences in the number of foot slips between the three genotypes was detected in the 6 months cohort.

301
Analysing the time taken to cross the narrow beam, we found significant differences only in 1 month-old 302 TAU58/Trem2 Δ/+ mice compared to TAU58/Trem2 +/+ mice, with TAU58/Trem2 Δ/+ mice taking longer to cross 303 the beam ( Figure 1E). The analysis of the 3-and 6-month-old cohorts remained incomplete as a large number 304 of transgenic mice fell of the beam during testing. Notably, 16% of 3-month-old TAU58/Trem2 Δ/Δ and 305 TAU58/Trem2 Δ/+ mice fell off the beam during both trials, while all TAU58/Trem2 +/+ mice completed at least 306 one trial successfully. 40% of 6 months-old TAU58/Trem2 Δ/Δ mice fell off the beam during both trials and an 307 additional 20% fell during one of the trials, while 66% of TAU58/Trem2 Δ/+ mice fell during both and 22% 308 during one of the trials compared to 30% (both trials) and 20% (one trial) of TAU58/Trem2 +/+ mice ( Figure   309 1F). In addition to motor deficits, TAU58 mice present with an early-onset and progressive disinhibition

320
Increased tau phosphorylation upon Trem2 loss in young TAU58 mice. TAU58 mice are characterized by 321 progressive tau hyperphosphorylation with brain region-specific differences in onset and progression of pathology (51-53). Phosphorylation of tau at serine 214 (pS214) is known to be an early marker in the disease 323 process, while phosphorylation at serine 422 (pS422) is described to be present later in disease course (68). We 324 have previously reported the presence of tau pS214 in TAU58 amygdala as young as 3 months of age with the 325 late-stage marker tau pS422 appearing from 6 months of age (52). Here, we found that pS214 occurred in the 326 amygdala as early as 1 month of age (Figure 2A). By immunostaining of brain sections, there was no notable 327 difference in levels of pS214 between TAU58/Trem2 +/+ , TAU58/Trem2 Δ/+ and TAU58/Trem2 Δ/Δ mice. For 328 comparison, pS214 showed a similar pattern in the motor cortex ( Figure S3A). pS422 labelling was sparse at 1 329 and 2 months of age with no differences between TAU58/Trem2 +/+ , TAU58/Trem2 Δ/+ and TAU58/Trem2 Δ/Δ 330 mice ( Figure 2B). However, at 3 months of age, TAU58/Trem2 Δ/Δ mice showed significantly more pS422  (Table S4).

386
In the present study, we show that reducing Trem2 accelerates phenotypes of P301S tau transgenic TAU58 mice 387 in early disease stages. These deficits were accompanied by increased levels of tau phosphorylated at Serine 388 422, a late stage pathology marker (71), in young TAU58/Trem2 Δ/Δ mice, as well as by reduced microglial 389 activation compared to TAU58 mice. Using an independent in vivo model for tau spreading, we showed that loss 390 of Trem2 leads to an increased inter-neuronal propagation of pathological tau.

391
TREM2 limits tau pathology and associated functional deficits. Our study supports the emerging consensus 392 that reduced TREM2 function/levels are associated with increased tau pathology in transgenic mice (46,49,50).

393
Experimental differences (e.g. mouse lines, genetic backgrounds and ages used) may explain the deviating 394 findings of heterozygous Trem2-deficient mice augmenting tau phosphorylation in tau transgenic mice in one 395 study, although even in these mice brains were protected from atrophy (48)

442
In summary, our data support a disease-limiting function of TREM2 on tau pathology due to directly limiting 443 neuron-to-neuron transmission of tau proteins, decreasing pathological tau phosphorylation, and, therefore,